<TITLE: Conceptual Modelling Languages
ACADEMIC DOMAIN: technology
DISCIPLINE: information technology
EVENT TYPE: doctoral defence discussion
FILE ID: UDEFD040
NOTES: continuation of UDEFP040

RECORDING DURATION: 109 min 11 sec

RECORDING DATE: 22.3.2004

NUMBER OF PARTICIPANTS: unknown

NUMBER OF SPEAKERS: 3

S1: NATIVE-SPEAKER STATUS: Finnish; ACADEMIC ROLE: research student; GENDER: male; AGE: 31-50

S2: NATIVE-SPEAKER STATUS: German; ACADEMIC ROLE: senior staff; GENDER: male; AGE: 51-over

S3: NATIVE-SPEAKER STATUS: Finnish; ACADEMIC ROLE: senior staff; GENDER: male; AGE: 51-over

SS: several simultaneous speakers>



<PRESENTATION UDEFP040 by S1>

<S2> i have not (realised that i) have to play the role of mephistopheles erm during this defence i never played this role before for this reason er but i like mephistopheles quite a lot in goethe's faust at least er for this reason let me try erm to begin first with a story erm in er 99 erm i was joining a team which was trying to recover er er the schema of S-A-T-R three and er when i was introduced to this team i was also invited by the head office of S-A-T and the main guy told me conceptual modelling is out it's not of importance at all and nobody's interested anymore et cetera er i think er this was a very good starting point er for work for lasting for about half a year in conceptual modelling er S-A-T-R three schema it was a very nice statement and i think er er you are trying now to do almost the same task and for this reason i believe it's er pure PhD is hard to do er hard to judge and never completed <SS> @@ </SS> er and let me explain first why i think so erm conceptual modelling is a term which is er somehow appearing about 20 years ago when people tried er to understand what kind of databases they have been developing er it had been already before er coming up with er let's say some (pre-relational) schemata with child's words or something like that er maybe concentrating on how to er demon- how to er show what this er relation contain er you will call it now er extensional er modelling or extensional er representation er at that time it was completely different it was understanding what an engine is really re- representing and what kind of understanding you should have it was not supposed to be er er erm er a tool for er communicating between people er to understand what is in the database it came later or it came er let's say e- even er ten years later after er (semi-) (xx) models appear er i think er in about 83 84 we had er at least er this notion of conceptual modelling and then they had er development er which split up er quite rapidly in differen- into different schools let me begin maybe this includes the users in scandinavian school the scandinavian school had a little bit more thought on intensions what kind of intensions are underneath and what kind of understanding you had really er er this modelling some (parts) as a goal , it was not really intended to be implemented and this is is also for me valid for er kauppi's work (for instance) they've erm are very much (relying on that) this er scandinavian school is still lasting and continuing the work and i think in our age er whenever we are talking now about er communication and er people have to understand (and to) communicate erm using their models of the world er becomes really important er somehow to continue this er research unfortunately it is not done but it is not your fault it's a fault of somebody else , er another school appeared er let's say er by er fundamentalists of constructivism er i would call the (john) school i would call also people er somewhere working in america mainly concentrating on how to construct how to do constructively the er somehow er er modelling not having any kind of methodology but maybe having some kind of constructors to construct the model and then to say okay fine now we are constructing anything what we need to know (xx) continues e- even (xx) everybody is constructing anything er he thinks and then you have this kind of construction , the problems (only with) this school er dep- er natural languages er are not really completely entirely constructive of course i think everybody likes to know that these schools they have a grammar underneath the language and you have to learn the grammar the funny thing is only er whe- whenever we deliver er the language to a child er grammar is not taught by- i hope you we're not going first er in the age of two years to learn grammars or something like that we learn the language by using it and by er using as it as tool and this is one of the very big faults of this constructive school and i think er to compare a constructive school with a little bit more intensional school er it becomes a bit also already unfair for this reason i believe you are never in the shape er to complete your PhD degree <S1> mhm-hm </S1> maybe today we will have some kind of let's say finalising remarks about it but er you have no (chance to) and the you have the cell school er which is even more important er but not (worth) considering not (worth) (xx) i would call it the asian school in the asian societies you first we are thinking about relations among humans and how these relations could be expressed unfortunately these guys never created a model except one guy er this was peter (scherman) which er was taking er the approach and putting it in th- er constructive environment later he was of course claiming he er maybe had er the chinese language in mind but he developed this kind of er drawings the problem is er at that time er that everybody was thinking constructively and for this reason er everybody was saying okay fine somebody was drawing now we can understand (what's the guy's thought) , my understanding of the notion of a bridge is completely different of the understanding of the notion of a bridge from my dentist and er maybe other people have also some other notions of bridges maybe you could construct engineer for this reason we have first to state what kind of semantics we are using and this is the fault of all these er modelling conceptual modelling languages they want to model concepts but nobody is talking first what's a concept what you mean really of (this) now we have content management systems and of course we deliver content er to everybody not knowing what is content er not knowing what are the concepts behind and what are the topics we are taking er with this these content er th- this is concepts for this reason i think er this large variety of languages has er some merits some advantages and a lot of disadvantages and now you made a try to compare the languages i think any kind of comparison is -s in almost all cases unfair unless you state and you never do it in the PhD how you are comparing what are the quality criterias and on which er purpose you are somehow er measuring er the models you have chosen three models three models which ware were very very different you were going for the IFO model and the IFO model was an outcome of the constructivist school then you went over er to er the scandinavian models er COMIC and kauppi's er theory er you are using er you mention it here as well er HIT semantics HIT semantics is the outcome schools of the braur- er brauerian school er on constructive er logics or mathematics and it is a very tough outcome er and that is a reason why it is not used by many you er find a book <NAME>'s book mi- maybe er is a very nice book maybe er or <NAME> has the same impression but er if you look into the citation for the er er who is citing this book you'll find there are maybe three four people many come from czech republic that's (xx) <S1> mhm-hm </S1> er it is the most rigid model but not used there should be a reason for that and now we come er to our age in our age we have er of course er we're trying to speed up er speed up means er we have to develop er some er means to er communicate , currently i i distinguish for conceptual modelling among linguistics different kinds of languages to do this the first of course natural language the problem with natural language is everybody understands er any word in a different way and of course nobody expresses what kind of semantics he has behind because we don't express all relationships we have er among humans , sometimes you er also have (xx) some classical human interface problems but anyway er natural language is a very nice tool because it is very expressive and er you can use it any needs you have languages er which er they are trying to pull at- anything on an easier representational way er these languages like U-M-L are thinking that they express anything by diagrams and drawings , of course these diagrams and drawings can be taken very differently er very very difficult to understand and are not at all a possible way to understand each other because if you have no way to express semantics underneath except you put it there you mention description logic in your PhD i found it er really nice to see er that you mention description logic there is a common er there is a folklore misbelief er and er a lot of people are repeating it description lang- er description logic is expressive enough to do conceptual modelling it's not expressive enough descriptive logic is a very small minor er a very small variant of er predicate logics in order er to cope with difficult questions like derivation er like er computation and for this reason you are restricting the language and if you are restricting the language you have not the expressive power any (I-S-O) language and for this reason er this er let's say semantics underneath like description logics don't really support la- languages like (xx) you can go with (xx) first you can go to XML in XMLH er you can express anything by notions or by words and you think they understand each other that's a very good thing i learnt er that even slavyanic languages which has the same wording er use er the word in a very different sense once er it's a very er er harassing sexual term and in the other case it is a compliment er then in this case maybe er you should be very careful with using er er this kind of wording this i think is a er lang- is a part which er maybe er has to be refined and which has to get some (informant) opinions , a (survey) of er conceptual modelling languages er maybe is er somehow er also er express through these IFO or these entity-relationship models with entity-relation model i had difficulties it took about 20 years to get a formal basics as a model 20 years is a lot of time and er 20 years of usage this model of using this model is a lot of misunderstanding and a lot of mistreatment and er usually er whenever you are opening a book you have first to capture what kind of semantics an author is using i know very good modelling books i don't er maybe er but i should not name the guys where in each chapters the guy is changing semantics that means you first have to capture what semantics he is now using and he is using and see there's 11 chapters to the book er 11 different semantics that's a good thing for training students er to become skilled (and think) but there's a problem this er really er whenever you want to computer to do a work properly you should have an understanding what the computer will really do with your notion and for this reason you should have a formal semantics underneath the computer is a (boring) stupid bit-based bit- er yeah bit-based er er device it can nothing do then er you er told the computer and for this reason you have the problem in er er let's say mapping down er thoughts everybody has , there is another notion er which is becoming er far too much more formal er which er relates back let's say er er to the <NAME> er trials to formalise mathematics , fortunately he never finished er he could never finish that's a good thing er otherwise we would have now engines er teaching er mathematics to students to understand what an engine should do but er this formal notion er was also not communicatable among humans and for this reason it was not useful really for er conceptual modelling for this reason you don't see er predicate logic underneath here er (we) should do it maybe er having a very nice well-defined semantics you are using in your er PhD as i counted er six different semantics maybe i'm wrong maybe i missed one but er er this is already a good variety of er semantics and an- and whenever you want to map you should also give the mapping and er provide the mapping in some case you did and in some case i missed it er maybe er i must read line by line your PhD maybe i didn't get the right place but er the problem is er this is er different semantics (as if) you should have a means to express er in a certain way what cannot be expressed in that language , then there are other possibilities to er model the world to to model the world er through er let's say constructs er you are er er pulling er you are defining first constructions and then then you say okay fine now you can construct your reality that is bottom-up approach er to modelling usually not taught in universities very useful whenever you want to teach the er engine how to do these modelling (parts) this er let's say for me the er bunch of languages er you have er for modelling er i would count maybe hundred hundred fifty different languages you have and now you can compare all the er hopefully under your pension you can succeed er youd- you decided to er choose three but er for me there is another problem with the PhD er modelling for me is more than conceptua- er than structural modelling modelling means also how to use a language in order er to perform the operations with the language that means you have to model also functionality , er operating these systems this is not done by classical modelling languages there are a very few exceptions where this language has to er has been extended in a proper way so you can reason on behaviour of systems of course it is a very big old mission in computer science , this is i think er there is (<NAME>) 95 modelling of structure and functionality now we come er to this internet age which is a very nice thing that we have now the internet and er er and then you have er <SIGH> very strange things to do we have to communicate this units that is really very but hard and therefore any human can now interact with a computer now we need to state what is really interaction and what is really interaction means you has this er the model now the user as let's say an agent communicating with the system and users have their relationships users have their unstated and sometimes stated sometimes even er somehow contracted er collaborations and then you have to model er these er let's say communications these collaborations in a proper way that means now we add another dimension to modelling this is good for interactivity and now i think er we are in the age er to change our computing paradigms when i was raised i was taught er in computer science programming (is small) write a small program hopefully it runs later er i was taught that maybe er programming is large (if we think) the problem is only er programming is large nobody (really did) so far in a proper form nobody er has a com- er has developed so far a semantics performing (the large) now we see already er another kind of programming coming up i would er call it constructive programming or component based programming or how it now it called model (addressing) er programming that means in this case you are trying to do what engineers already er learned hundred fifty years ago while assembling all these er mechanical tools around , they are first standardising er our er tools here and then they could er assemble thing and now we are learning how to assemble and the next what will come now in this internet age is er programming in the world (that is called) programming in the world means everybody is contributing something and you never know what you get you get viruses or worms or something else or you get even useful programs you have to communicate it you have to (rely on) it this would be er for me er a completely different stage of er modelling this is for me a very er small survey er er or some thoughts about the topic of your PhD i think i should ask you also some more er let's say er or less general questions and er for this reason maybe i should er go back again to your PhD , er i think i have questions you told me for hour and half well okay fine i i have questions for four hours <SS> @@ </SS> erm if you don't kill me <SS> @@ </SS> but maybe i should restrict myself a bit and er let's begin with a er very nice starting point er this is a starting point which is only er difficult for the scandinavian school of modelling er to er distinguish intensions and extensions , and you have made a classification of extensional language and of these intensional languages which seems to be right but you have to prove , it and let's then begin with this starting point and then i have other questions how you can prove what is in- intensional language </S2>
<S1> well erm to give a very er short and shallow answer i would say that er it's a language where the relation of intensional containment is being used without actually fixing the semantics or meaning of the intensional containment as such however er in my opinion erm that the intensional containment relationship it can't be un- er interpreted in a way that er er it is just left alone as a calculus instead it would really need to have er some sort of erm interpretation and those ones that we very often er come in terms with are those ones of er the is-a relationship , however of course if you rely on kauppi's er theory or let's say kangassalo's er interpretation of kauppi's theory relationship of intensional containment can cover many other er relationships that actually are formally the same meaning they er give a parse ordering in the er set of concepts </S1>
<S2> now i could (xx) there's a trap now i can (place (xx) mephistopheles) really problem <S1>  mhm-hm </S1> er in your er PhD there is an- is a very nice place er where you talk about (semantics) i would like to er er make this distinction er to have syntax or syntactics <S1> mhm-hm </S1> semantics and pragmatics distinguished and if i use this triangle then this case i can maybe better define what means intensional and what means extensional <S1> mhm-hm </S1> language er could you correct your PhD in this direction </S2>
<S1> excuse me can you first [(xx)] </S1>
<S2> [oh could you] correct the statements er you made now or er your PhD er the statements in your PhD in this direction more properly </S2>
<S1> erm i'll bite the bullet i'll see what's coming up so erm erm please er point out the exact er formulation problem (you made me) answer </S1>
<S2> okay let's let's er you define some languages er to be intensional <S1>  mhm-hm </S1> you define other languages to be extensional <S1> mhm-hm </S1> er and you define some languages are hybrid <S1> mhm-hm </S1> because you cannot really classify those languages , er COMIC is hybrid i think yeah <S1> well </S1> er <S1> yes </S1> and now i can ask you why COMIC is hybrid <S1> mhm-hm </S1> why E-R is extensional <S1> yes </S1> why IFO is extentional <S1> mhm-hm </S1> why HIT is intensional <S1> mhm-hm </S1> er i could have a different opinion <S1> yes </S1> and now i want to know the reasons why <S1> yes </S1> and this answer you gave now </S2>
<S1> well the answer i gave is simply the basis of the categorisation but you are right it doesn't really go as deep as you want basically you start with er the er different types for different examples of languages first of all considering E-R i would rather say that the erm flavours of E-Rs that we know quite well they have a rather extensional emphasis for instance i erm in er chapter two where we discuss the er erm working model of the categorisation i never actually say that i'm using E-R over here as a erm as an ex- er example language since the er we can define the er semantics of E-R model more or less the way we want to and that was in the lectio as well the erm the reason why i referred to er the terminological basis like we can actually erm call anything we want er intensional or extensional if we don't have any erm proper basis for that any proper definitions erm i would rather say that er the language that i have er defined in chapter two page 19 where we mainly talk about named set and named relations it is very thoroughly extensional it doesn't have any er notions of concepts as such so simply by its simplistic semantics that's kind of extensional there are E-R models that resemble this language it has to be er er it can be shown and er on the other hand er the er intensional approach languages in the terminology and hopefully in the semantics as i'm trying to implicate in chapter seven they actually er establish the intensional relation erm between concepts so that it could be interpreted as being something truly intensional for instance something that could be constructed using the HIT model or shown by se- set theory in er possible worlds semantics </S1>
<S2> erm , i have some again some problems <S1> mhm-hm  </S1> let me discuss it with er your erm one of the really fancy examples you have yours er erm er the example on i don't remember the page but anyway on female ghosts and female president of the united states <S1> mhm-hm </S1> it's a very nice @example@ @@ </S2>
<S1> er actually it was not female ghost but ghost </S1>
<S2> okay <S1> mhm-hm </S1> or female presidents and fe- er they were ghost women or er [anyway] </S2>
<S1> [it doesn't matter what] the gender of [ghost is] </S1>
<S2> [erm] i think er whenever we are modelling we are making er somehow a decision what kind of world we want to see and what kind of world we want to represent erm maybe never documenting what kind of world we have and for this reason we have to choose different languages one er would be er er the set of let's say possible semantics it is not possible world but instead of possible er semantics and another one would be with er certainly pre-defined or well-defined semantics and this would be for me the distinction between intensional and extentional <S1> mhm-hm </S1> er languages that you have to use er well-specified semantics er er and you cannot change it whenever you talk about (intergrational) models er when i was trying to summarise for myself what is an integrational model i got about erm 60 different variants semi- (xx) model the problem was only er out of the 60 55 were contradictory <S1> mhm-hm </S1> and for this reason (it gives you) very small subset again er for the choice but again er you had a possibility to choose between pointer semantics or set semantics <S1> mhm-hm </S1> or even multi set semantics and in this case er here again you make a choice bet- among different semantics and for me it would be now the case (xx) (define) an extensional language never has a concept of ghost it's not defined you cannot consider it <S1> mhm-hm </S1> nobody has seen a ghost <S1> mhm-hm </S1> a female president er could be a concept currently it's empty and empty set is a best set you can have any property is (valid) and for this reason er you have different er possible ways to express things i think this er let's say er differentiation would make your hybrid languages or er <S1> mhm-hm </S1> classification er let's say er you can omit hybrid languages </S2>
<S1> but er in what way erm this kind of entity or the expression of ghost is different or of from the female president of the united states in the 20th century can't we simply see them both as empty sets 'cause you know in in other words er worlds for instance in hamlet there are ghosts so it just happens that over here there are not </S1>
<S2> let's say erm er whenever we are doing some work we have a target <S1>  mhm-hm </S1> er money to complete it as fast as possible or a- to represent what we see or what we want to see all we are educated in er an environment where we can let's say er live this in our beliefs <S1> mhm-hm </S1> i er state that er ghosts don't belong to my sets of concepts i know that there are people which are claiming there are ghosts erm i have some let's say er medical er explanations for that but er i don't believe in ghosts <S1> mhm-hm </S1> therefore er the ghost is for me a concept er which is fancy which can be used at anytime somebody (xx) with a ghost the female president is er a merge between two concepts it is er the concept of females <S1> mhm-hm </S1> which we er er is somehow well-defined and er of a president let's say president of a certain state <S1> mhm-hm </S1> in order to define it a lit- a little bit properly these are two concepts i se- choose to belong to my beliefs <S1> mhm-hm </S1> if i go now with all the possibilities to model my language becomes too rich for my purposes <S1> mhm-hm </S1> therefore i want to optimise er my er modelling efforts <S1> mhm-hm </S1> and for this reason i think okay fine let's omit the ghosts let's omit the female ghosts anyway i cannot detect what is the difference between the male and the female ghost <S1> mhm-hm </S1> and er let's er talk about things i have in my mind and this means in this case you are clarifying the first application error <S1> mhm-hm </S1> and then in this case they say fine extensional language or intensional languages are not so important to distinguish they are only some let's say possible ways to express in early stage or late stage er what i want er to represent of the world </S2>
<S1> mhm-hm so as to s- as for the summarisation you would kind of fix first the basic vocabulary of the world and you would consider for instance that ghost er ghost is not there whereas you would probably have both president and female and by that sense you can see the combination <S2> yeah </S2> that sounds like a very feasible approach i'll have to say clearly erm erm i'm sure it has erm obvious limits </S1>
<S2> er and er then er i would say that , after your agreement after i heard you would agree er with me (several things) erm er let me change a bit erm my questions into different area you are choosing three languages three modelling languages and the first question is why these three why not a fourth one </S2>
<S1> erm if we talk about chapter two to start with there the question of intension- intensionality is sort of introduced in connection with er the modelling languages erm those ones erm i have to admit that i first thought er the E-R model to be something very obviously er extensional as it is er presented by peter chen 1976 however after er a colleague er marko junkkari pointed out that he can't find anything so obviously extensional in this approach i re-read the article and noticed that his terminology er would basically allow us to use er anything like is- anything for instance is-a relationship in his model 'cause er he simply in a footnote says that this would be duly possible so in that case erm why the like to answer the pa- first part of the question why the example of an extensional language over here is the one where we used named set and names named relationship so that's why it is there , the other question er or the other part why when we talk about intensional languages i selected something that has er concepts and intensional containment erm in the set of concepts and then some cardinali- co- erm constructs er the reason for that is er erm of er rather pragmatical , when taking a look of the er er erm the description of natural language that i wanted to model i realised that that must be the minimal set that you can deal with in order to make the language at least rich enough more or less rich enough to er make something out of it er of the example that is so that's er er the second language as for the third the er description logic erm i was er in my mhm engineering career that is er where i work there are er at S-E-R-N that is there are many people interested in creating applications with computer aided design the relationships like part-of er very often er come about and description logic is a well-known approach of dealing with that kind of a subject for instance erm the er language that i use over here is er er influenced by er er er er a swedish er doctoral thesis of er description logic that was specifically er designed to handle part-of relationship and that's why that's an example of the third language now as for why i didn't feel like having more of each of the categories er i think that would have been erm complicating the matter and why i didn't define more categories i didn't feel the actual need for that </S1>
<S2> well i- i was told to complain (about) anything and erm er i have one observation in your er PhD er when you compare these languages the youngest language won er because of er it was more mature language and erm description logics is for me a (xx) predicate logics <S1> yes </S1> in (xx) predicate logics you can express certain things other things can be somehow expressed by very complex formulas and other things can be never expressed of course you have er in each language (phrase) stating why this language is the most important and the best language and never i- you should never use another language methodology or something like that <S1> mhm-hm </S1> er and er of course you have also (xx) of description logics and this (xx) states you can express anything with you (xx) world so that's true <S1> mhm-hm </S1> because of this (cognitive) constraints you have holes in er er erm <SIC> satisfactionate </SIC> for this reason er er description logics cannot express these holes because there is no negative statement er to be made er for this reason er we have to be very careful with er the language if you go now into something like er COMIC if you go now er into other languages you have different er expressive means and for this reason you have to compare first what is the quality criteria for the selection of the language <S1> mhm-hm </S1> and what was your quality criteria </S2>
<S1> that's is i would say enough granularity er enough means to er erm express er typical example but i completely agree that for instance er it would have been rather interesting to find the er quality criteria in a manner that people actually use the language and how understandable it is for them erm other than the formal criteria but for that kind of a empirical er erm survey i simply didn't have enough time </S1>
<S2> yeah of course erm you have still enough lifetime of er under your pension er <S1> mhm-hm </S1> to tackle it but er @@ er whenever i'm er looking for a quality criteria for a language <S1> mhm-hm </S1> could be er er feasible er er redundant <S1> mhm-hm </S1> let's say er i can manage er redundant somehow it should be some possible way to express consistency maybe this would be also some kind of criteria maybe it could be also criteria er to express anything in a short (space) <S1> mhm-hm </S1> in this case er natural languages are (thinkable) <S1> mhm-hm </S1> because we can use un- sub-sentences no of these languages you have considered are using sub-sentences <S1> mhm-hm </S1> you are using very simple maybe S-P-O sentences and nothing else <S1> mhm-hm </S1> it is then therefore er natural language is the best candidate to express anything (this is) your argument @@ <S1> mhm-hm </S1> but er i think er the main thing is really i think you ought to have let's say a good (mutability) or let's say er a good er flexion property and then you have to define what is ever flexion property <S1> mhm-hm  </S1> er you have a possible way to er to declare your contracts your personal contracts by modelling <S1> mhm-hm </S1> er to to declare your theory what kind of meaning you have this is for me the conceptual modelling gap too er er since er usually you are documenting only the schema then you go away and then the other can read what you have been thinking about <S1> mhm-hm </S1> that's a good thing er to keep them busy and to go er er er to continue the work but er the problem is you have to state first what you want really to do and (xx) languages you have , some kind of criteria <S1> mhm-hm </S1> i would choose description logics but i wa- i want to be er more let's say er logically based i claimed that er if you put a good schema in description logics and you're using description logics let's say a typical schema of the size of thousand (xx) operation types this is description logics nobody will understand nobody has even a chance <S1> mhm-hm </S1> to to somehow to trace what is there in the schema but <S1> [but is there] </S1> [only you] should have some criterias </S2>
<S1> yes but in in your (xx) example in your example isn- the way isn't it a fact that nobody could trace what goes in in that er erm description logic schema it- isn't it simply because there is no alternative formalism for instance a graphical formalism fors- for the same thing actually that wouldn't be so hard to create on top of the er textual er er presentation </S1>
<S2> no , that would hardly (xx) presentation at all <S1> mhm </S1> the funny thing is still you should be clear @@ what are concepts and what are properties <S1> mhm-hm </S1> and i think on your slides you make the @same mistake@ <S1> mhm-hm </S1> er firstly , it can be sometimes a concept usually it is be a property <S1> mhm-hm </S1> and i think this er distinction (xx) first made in description logics you don't have an expressive means for that therefore we have to add a logics on top of the description logics and always to map the er graphical presentation , otherwise you will get er nonsense schema so the birthday is the same concept as the birth <S1>  mhm-hm </S1> and as a (dog) @@ <S1> mhm-hm </S1> or something like , then in this case you have first er to add expressive power @before you@ state what you want then you can change the representation </S2>
<S1> i i don't completely see the point for instance er in the example of the description logic er even though this is extremely simple of course page 26 to 27 erm indeed erm , i see your point partially that er these erm it might be difficult to er express er the er variety or the essence of the presentation in a graphical format but i don- i didn't get the idea of why you would need the er difference erm between property and concept if you basically only have some syntactic items like in the presentation over here </S1>
<S2> let's say erm er i- most of the examples i like very much er erm after er 26 27 i want to go for to open a bottle of wine and @@ <S1> yes yes yeah </S1> er let's look what you you have there and you have a wine bottle <S1> [mhm-hm] </S1> [okay fine] it is for me erm maybe er a concept to be represented then you have a concept of er age <S1> mhm-hm </S1> is it a concept or is it a property </S2>
<S1> well in this case of course it's just a lexical item er and oveth- and over there <S2> mhm-hm </S2> i know that in in many graphical er formalisms you need to make a difference if it is er an attribute or an entity but er isn't it basically er how it appears in the er in the er combined expressions like wine bottle that the slot names indicate er its er role </S1>
<S2> erm again i'd like you er even (xx) more er stupid (trap) but erm <S1> mhm-hm </S1> er now we have properties er age is a derived property <S1> [ye-] </S1> [and er] is a property which is depending either on other properties or it is derived and computable <S1> mhm-hm </S1> it means you should model its derivation mechanism in this case you have to be very careful er with distinguishing what is a real world concept what you need really to represent and what is a property erm later you map into your relational database (xx) (roots) <S1> mhm-hm </S1> you can map it in a different way you can map it into (derived) (xx) or you er can map it even er to some kind of computation- er to some functions you put somewhere er for the interface therefore er you have a lot of different possibilities to express the same thing and in order to be er er in a union representation you should be able to </S2>
<S1> but i still fail to see i mean i i agree completely of this er these semantical questions 'cause those are the ones that i'm er truly interested in myself but in er in connection of er graphical representation of description logics is it really relevant </S1>
<S2> it depends er on (xx) error <S1> mhm-hm </S1> it depends what you want to represent and what kind of decisions you made before modelling <S1> mhm yes </S1> er for this reason sometimes it is er important customers are coming and asking for the age of the wine <S1> mhm-hm </S1> before paying the bottle in other cases it's not of importance at all therefore er you have first to put this metasemantics <S1> mhm-hm </S1> on top of the model before we are modelling </S2>
<S1> how do you define metasemantics </S1>
<S2> the same way as philosophers did it <S1> mhm-hm </S1> hundred years ago unfortunately they died already therefore we cannot ask them what they thought but er reading their books i found that they are able to say okay fine this is the assumption we are making <S1> mhm-hm </S1> this is the restriction we are making this is the language er we would like to use for this reason they have to follow a set of constrictions <S1> mhm-hm </S1> and (sit- since) our er er association er for communicating and then we are er using this kind of agreement before er modelling <S1> mhm-hm </S1> this for me also is an achievement of the scandinavian school they never stated it erm er maybe i have missed one of the most important papers but i will (continue on) , anyway er er this statement has be- has been made before modelling <S1> mhm-hm </S1> i mean nobody will tell you the documentation er because if you want to keep your secrets but er er you will ha- have to make such assumptions before <S1> mhm-hm </S1> but let me come er to one of the statements er i'm again not (xx) er agree with you , say as you said page 26 27 is a very er good example how you can express something in description logics i claim usual database application or information system application <S1> mhm-hm </S1> er starts with s- something like let's say er if we take it as a small maybe 300 (xx) (operation types) and maybe one and a half thousand attribute types <S1>  yes </S1> er if you put it into description logics way and i give you five minutes and er i will m- make with you a bet er six bottles of champagne free and good champagne if you can express me within five minutes what is there <S1> mhm-hm yeah </S1> er i think i will win it and for this reason i will get drunk afterwards <SS> @@ </SS> but er the problem is er really er to graphically to er represent erm (survey) to have a good survey until a certain point until the graphics becomes again not (xx) <S1> mhm-hm </S1> and there was at some point where you cannot get (before) you need some kind of abstraction mechanism on top of that some kind of component er modelling on top of that therefore there's there are several layers on top of that er for this reason i'm claiming er okay fine so small tiny textbook examples <S1> mhm-hm </S1> or in a lecture students want to sleep you want to go away and er for a lecture you can have small examples but in reality you have large examples and large examples un- are (exists) (xx) </S2>
<S1> well the same of course would apply to er the other language presentations in the same chapter because they are the same example and erm i suppose everybody has seen entity-relation diagrams that wouldn't er fit on a spreadsheet 'cause they are simply get so large and but er of course we are playing along with the rather limited example in the over here </S1>
<S2> since i have too much questions er i would like to change again the discussion a bit differently er er different erm path , you stated and this is a statement er which is something like a religious belief er the containment is one of the major issues you have to tackle er during modelling , er i can read it er or i can state it as meph- me- mephistopheles along i don't have to believe now you have to make me to convince me or really the containment is really the important property and the only property you are so much tackling , why </S2>
<S1> i suppose you refer to the comment on page er 47 intensional representation of er concepts is er essential and so on recently in knowledge representation processing of er transitivity and problems related have become an important field of study erm neither of the comments i find exactly religious but er i agree that er this er thesis as such has a rather strong basis of er or erm is strongly influenced by the approach that uses intensional containment a lot erm why is it important erm first of all i think erm erm well as in computer science if there are lots of researchers er on some field you are more likely to get the grant so maybe that's why you make some things kind of important to get the grant but er on the other hand erm if we think of those er er typical interpretations of this these relations if they would be called the intensional containment they are those ones that come er about in practical modelling situations really often er like er part-of er is-a even has attribute er in in some ways , however i kind of er have er feeling that professor kauppi herself would not have er welcomed er part-of examples erm in er in her work er she would not probably considered them as good example of the intensional containment but that of course we can only guess however if we think of er the er erm corpus of work that nowadays goes into the processing of transitivity for instance in simply this interpretation of the intensional erm containment is quite substantial and we know that er even though the part-of relationship er is supposed to be kind of a well-behaved it can't be considered transitive in in er all possible cases </S1>
<S2> erm . since i have to play the role of meph- mephistopheles properly i should not er refer to kauppi , i should refer to peirce peirce had a very nice paper er when he was stating er we have firstness secondness and thirdness properties and er he was stating firstness is what should be modelled what should be er somehow er detected secondness is relationship and er let's say my interp- in my er let's say brief er erm er explanation (should) (xx) more but er and in this case containment would be a concept of secondness in order to call things er how they are contained in each other you would like to make some kind of let's say er forest order or some graphical order or maybe even hierarchical order <S1> mhm-hm </S1> for this reason you er invent containment in order to order your items but in this case you would say okay fine er then let me first develop concepts and then say okay fine now i want to call concepts since i must somehow er have some efficiency in in the represent- er in representing er properties of concepts let's say (shorter way) in this case containment would be not the main property the main property would be to find the proper concept , or </S2>
<S1> yes [but er] </S1>
<S2> [er] maybe maybe peirce is wrong </S2>
<S1> his ideas or this kind of ideas have been utilised in the er connection of er erm professor kauppi's concept theory by er doctor hautamki whose er book logical point of view actually explicitly uses the kauppi system when we are talking about er what's the proper word actually next page er 48 er detern- determinables and their values so er even though i know that the starting point is quite different the er he kind of forces the structure to be the same kind of a similar this has kauppi from the er point of view of er primary and secondary properties </S1>
<P:05>
<S2> okay i buy this erm then er er you say that containment has something to do er with is-a relationships <S1> that's actually [one of the] </S1> [er could] you clarify this </S2>
<S1> that's actually one of the few examples that er professor kauppi really gives in her book er and that is in a case well for instance is-a er a dog is a mammal for instance er that's something that er i can't actually i can't remember if i think er professor kauppi used to use more like er mathematical examples like triangle er being a figure and that i find very much like the is-a relationship </S1>
<S2> what is is-a relationship for you </S2>
<S1> wa- er [well] <S2> [what] is your definition of a is-a relationship </S2> er it depends really completely <S2> [@@] </S2> [on the on the semantics] of course of course if we want to establish it using er possible worlds that's simply wonderful er i like that it's er economical on the other hand if we want to see it as a special case of intensional containment then okay a circular d- a circular definition but then again we wanted in some approach for instance in kauppi's approach to use the intensional containment as a relationship that doesn't have any interpretation or definition is just given </S1>
<S2> is er then- sub-typing expressible er through is-a relationships </S2>
<S1> are there er any other <S2> er sub-typing er </S2> sub-typing <S2>  yeah if it's expressible through is-a relationships </S2> i don't know erm [let-] <S2> [you] didn't get into trouble </S2> ye- <S2> [@@] </S2> [erm] sub-typing and er how about inheritance </S1>
<S2> er inheritance is a notion erm i like (particularly in teaching) a lot because it keeps students busy thinking er what er kind of concepts should we properly develop <S1> mhm-hm </S1> er erm er inheritance is one of them , erm let let's not er open er the pandora's box erm er this way but er let's er first er what is is-a relationship for you </S2>
<S1> erm er if i have to n- now take the card i would still say that i would like to interpret it as er something in my slides for instance even though a bit later </S1>
<S2> yeah why , you had a lot of slides probabaly you could (use them) </S2>
<S1 LOOKING FOR A SLIDE, P:05>
<S1> oh pardon erm i have to go er back a bit 'cause er this thing is slightly , alright so for instance this one erm different the subset relationship in different accessible worlds that would be the is-a er presen- or the er possible worlds presentation of the is-a relationship and er i i take this card </S1>
<S2> then in this case it would be fine it's not a sub-relationship it's er a sub-model relationship er in the (task) sense then this case i buy this <S1> well yes </S1> otherwise er you get really into the traps er of co-variant and contra-variant and all these stupid things er er which are not er very much for the is-a relationship okay fine you take this model as your (representation) you're alright <S1> mhm-hm </S1> so and now i come back to the containment this is a er er this containment is-a relationship expressible to- through is-a relationships </S2>
<S1> containment is supposed to be more like er th- erm even though of course we wouldn't know is-a once again i suppose it's it's still the only example that kauppi ever gave but i er we could erm let's see something else erm <S1 LOOKING FOR ANOTHER SLIDE> <S2> @@ </S2> this might be a bit more trickier to find but since you really explicitly asked i suppose so we have to (answer) er so here's an extract of an article that explains what is-a is er what this card so this is the second card erm and it's er i think 1980 already or 1983 er by brachman erm sorry about the bad typesetting but this is er simply a copy paste so over there he gives the following examples of what the is-a relationship is so it would be generic generic like superset superset and generalisation specialisation and so on so there are awfully many different er interpretations and i can kind of find er something that looks very much like a er er kauppi's erm intensional containment among them i think you'll find it <GOING THROUGH SLIDES> there conceptual containment , there erm brachman however m- even though it's a popular article of course it's awfully well-written erm he discusses how to analyse these different is-a erm relationships in connection with this other and he basically says you don't </S1>
<S2> yeah that's this it's in fact erm fate of computer science articles erm i think it was good that you were choosing er kauppi's theory because of er this was a little bit more thought er before writing down something er (xx) and specialisation i find that people never thought about the difference of these two things and er nobody si- er are able to distinguish these to er these two things in lang- in n- natural language we do so but they are not able to and er brachman's brachman is a very nice guy but or was a very nice guy but erm sorry for erm sometimes yeah it's good to forget what articles other people wrote </S2>
<S1> you didn't like his approach over there or uh-huh </S1>
<S2> er the problem in- n- er in- here again identifying too much things in one , concept er sometimes you need to be sharp while (xx) and for this reason you have to er erm apply or you have to state what is really erm is-a and is enough in his approach in is-a relationship it's not enough to add an attribute er to a sub-type <S1> mhm-hm </S1> er i would have difficulties with my er belief in economy of thinking or expressing something <S1> mhm-hm </S1> if i have er let's say a person a student i add to the sub-type student of a lot of (properties) and i would like to have , in brachman i would not have , i would have to add a role er and okay fine that's erm the story of development </S2>
<S1> a- are you sure it was in this article 'cause i can't remember that kind of a example </S1>
<S2> that's <P:05> er article written for the I triple-E 89 ne- e- 81 </S2>
<S1> probably yes erm we can of course see the references erm <GOING THROUGH SLIDES, P:06> sorry about this technical </S1>
<S2> mhm yeah but we did not agree of the slide , you have to show [@@] </S2>
<S1> [@@] yeah <READING ALOUD> what is-a is and is it er er an analysis of taxonomic links </READING ALOUD> and it is indeed I triple-E in [1983] <S2> [mhm-hm] okay fine </S2> but this erm the detail that you mentioned that is actually quite crucial that i failed to notice </S1>
<S2> to er and then we come er back to containment here , erm i think whenever you want to talk about containment there's a good expressive means er to be able to be more economical and not to repeat all the time er what we have already said <S1> mhm-hm </S1> er then in this case er you need also concepts er to express er we agreed er that this is not enough so now we have chosen three languages that express er let's say er what you want er to see in the world <S1> mhm-hm  </S1> and now i'm asking you what concepts in these languages you can use to express other properties of containment </S2>
<S1> ah in other languages <COUGH> </S1>
<S2> in- in the three languages you have chosen </S2>
<S1> yes erm what are the other like er if we do mapping for instance from these concept diagrams to another language er what are the relationships in other languages that would map through it this relation intensional containment in concept diagrams well that's indeed would be those of erm is-a in IFO in the form of generalisation and specialisation er then functions er in some ways er 'cause IFO doesn't have really attributes it's er functions and then erm er grouping of course </S1>
<P:05>
<S2> erm for the last one i i er wouldn't buy <S1> er grouping that is [the] </S1> [er] grouping as a er as a er possible way to express containment </S2>
<S1> yes </S1>
<S2> how </S2>
<S1> erm if i show you for instance one of the translations , that i'm afraid <LOOKING FOR A SLIDE> yeah for instance over there these of course are aggregates erm those ones with the er asterisk erm this one student being enrolled er for a class is a function and luckily this er er like combined attributes erm that's more or less erm everything that was there in the example and here of course it is an obvious is-a structure erm but now that you mention it er being once again boringly i have to w- find the page where i talk about the IFO mapping because there was a notion of the grouping </S1>
<S2> here in chapter six </S2>
<S1> er yeah page 77 <READING ALOUD> however in the scope of this presentation it seems impossible to have a semantic criteria in which er concept D intensional containment structures co- could correspond to fine- to f- finite sets of er the IFO </READING ALOUD> so er grouping is out erm erm the er compound attributes erm seem valid </S1>
<S2> okay erm erm there's a problem here also erm the containment can be expressed in different languages and all of these three languages are in some cases <S1>  yes </S1> er implicitly through the constructural use <S1> implicit- </S1> er implicitly through the constructural use <S1> mhm-hm </S1> a typical thing we could say you model a person and you model let's say a policeman as er a person belonging to police department <S1> yes </S1> so in this case you can er model policeman through er relationship type if you go to the entity-relations model you can do it in a different well i mean also in the same way it means you have implicit now er let's say is-a relationship in through another concept <S1> yes </S1> and then er you have the missing semantics er you didn't state that this is containment you said okay fine another relationship type and for this reason i can er now handle it so and so <S1> yes </S1> and er you missed er semantics during modelling <S1>  mhm-hm </S1> for this reason you have er also er hidden semantics behind and i think er this should be somehow stated er and stateable er and when you try it now er i would like to erm one of the chapters i was really liking er was er er the chapter where you try to develop er your er er let's say outcome of the comparison what should be a language of containment <S1> mhm-hm </S1> and er when i was looking at this i thought okay fine constructural er erm m- modelling er maybe this (er language) at least erm they didn't do it formally er so i could reason er so that i could really say okay fine this is a language but erm , this language has also a lot of hidden assumptions er you can make during modelling how your documents show this <S1> mhm-hm </S1> how your document will will a person developing a schema , er was thinking why erm we represent a certain part the reality that's (integrated) <S1> mhm-hm </S1> how you can (represent) in your language are you able to </S2>
<S1> er well of course it depends on the area of application and once again i'm sure that many of the languages er discussed in this thesis are completely eligible for er typical well modelling situations but would i be able to erm yes i hope so </S1>
<S2> you would be , and how you would do it </S2>
<S1> develop a language that actually meets the criteria <S2> yeah </S2> erm er if we talk about the criteria on page 82 i suppose erm a clear distinction between er the language that is expressions occurrences things in the world and conceptual levels that is the first one then a capability to explain er to express what is intensional and what is er d- er intensional and what is extensional and a capability to express relations between concepts as follows there should be a way to express sub-concept super-concept relationships there should be means to map the different uses of intensional containment into different relationship in domain of application and there should be a possibility to include concept theoretical aspects erm would i be able to do that erm <SIGH> i do hope that the variant of the er erm concept diagrams that i use in this chapter it manages to er express or fill meet a criteria the third criteria quite well for the first two ones erm of a clear distinction of expressions occurrences and a capability to express er truly what's intensional and what what's extensional for that i think the formalism would need some further work </S1>
<S2> er let me er deepen er the issue a bit erm wher- er i have some problems with all of these conceptual modelling languages whenever i want really to match the modelling process whenever i want really er to keep track what has been done (according to it) <S1> mhm-hm </S1> erm one reason why i have this er was observation i made in th- this er very nice german er team er S-A-T er we are er erm hunting for people er developing certain parts of software er there was no documentation available er almost everybody left already er the er group er where er which was developing er this (xx) part er nobody knew what what was there nobody could state what is- is there and er the outcome of that er strange er let's say mismanagement was that er S-A-T-R three schema this of er er redundancy rate about eight that means eight times more types are represented than needed er the information is already there <S1>  mhm-hm </S1> it is er of course a typical thing for a bad managed prog- process some engineering people er at the end of this they came er up er with some kind of (xx) procedure let's say you go to er some kind of (suffice) er a relation then you say okay fine first er you arrange the first model whenever a concept and ever any concept is learned fine <S1>  mhm-hm </S1> second you go er er you have to have er let's say a possible way or possible methodology er third is manageable fourth is er is predictable what is the outcome and therefore it's maybe later on even optimisable <S1> yes </S1> erm on which (xx) you would state our er the discoveries and the er er chapter six er seven in your PhD </S2>
<P:07>
<S1> mhm </S1>
<P:09>
<S1> you mean the capabilities of using the language as a tool or the </S1>
<S2> yeah er we are using these languages as tools for er expressing what we er discover in the world and for er mapping it er to computational facilities </S2>
<P:05>
<S2> that's </S2>
<S1> you know that's one of these these questions that y- er you always think you you are prepared to answer almost anything and <S2> @@ </S2> @then@ then somebody asks a question like that mhm </S1>
<P:12>
<S1> i think i have to refer to your er opening er statement 'cause over there you mentioned that er some of the er interesting uses of the model are that of the articulation of concepts then er some kind of er a formalism and then semantics erm those are of course the er actual cornerstones by which we can eh- play here in the field of modelling erm and in a way i think that er most things that we do with er any of these languages shown or specified over here for instance IFO they are erm really nice in formalism but they er fail short in communications use so combining them erm is a challenging task and h- however over here i don't think that i have really constructively created a language that could er er that could er er er approach all the different er ways all the different erm a- areas of modelling so over there i i have to say er the er erm how my tools what kinds of way- er methods my tools would provide to in a er software project or modelling project it it would be erm simply rather primitive </S1>
<S2> primitive is (understatement) but it's er i think erm erm er i- in order to reach er the first level of the <S1> mhm-hm </S1> language (xx) you have to have well-defined concepts and , what i would expect what i would like to er let's say request after this defence er sit er sit down and er write a small article basing your semantics let's say maybe on the HIT semantics or er something like that er in order to er state clearly what kind of things you can express and what kind of things you never can express <S1> mhm-hm  </S1> er i think this would be a very good outcome s- some kind of advice (another request) during your defence er during your defence you are in a ba- er erm in a bad corner erm so but er i think er it is necessary at least er to to to have clear understanding of this question , erm i'm looking at the clock and think it's really er amazing how quickly it's moving i have time <S1> @@ </S1> i have er okay fine then er then er let me er turn er to different corner er which is much more er difficult to express er in modelling er whenever we are modelling er we ex- er we have concepts er which er appear in any state of er the observed reality <S1> mhm-hm </S1> er they have er properties and they have er concepts which er appear from time to time in the relational world we are using (null) there is an overloading (null) they use as much as possible in order to express it maybe in some cases they appear maybe in that some cases it does not appear , erm going through <NAME> 's book erm on the HIT semantics er there is very nice chapter on optionality , i think in er modelling optionality is a very necessary concept , how you would express optionality through your concepts </S2>
<S1> optionality through concepts but er i would actually rather reserve the er er for instance er optional attribute or optional erm part-of er simply as the primitives of the language even though i know that it's a cheap solution and probably doesn't really erm address the problem but er erm that would be a kind of practical way of doing it that's not probably what you meant er what's the er erm what's your erm opinion or deep meaning of optionality that should be seen in the language </S1>
<S2> you should have a possible way to express in each case something er that is optional <S1> yes </S1> and how erm er optionality of different concepts i- is related to each other <S1> yeah </S1> and er it is also necessary for very simple thing if you look for name <S1> mhm-hm </S1> er sometimes you have er er the first name of a person or let's say you go to spain and they have a lot of additional names of er <S1> yes </S1> (brazil) or something like that and in this case er you have er a call optionality er which is er somehow related to other optionality <S1> mhm-hm </S1> you should have er represented first name unless a person got these (xx) <S1> yes </S1> yes a statement er which is very difficult to maintain in the databases <S1> yes </S1> er and er my er observation of all these er con- all these three languages you have compared is that optionality is not really reflected er and erm i would like to see it in your language </S2>
<S1> erm well erm that i can't provide i'd sh- kind of give some er vague comments rather for instance in the er one of the starting point languages that is the concept diagram language of the COMIC approach it is possible to have er concept structures with optional parts so that's er unluckily i don't have a slide about that but er that's one possibility of er erm erm well shedding light to that question erm , others , well erm that would be i think if we really want to think of about optionality as conceptual problem then er this kind of intensional language er concept parts are in reality optional that might be doable another possibility would be to use er functional relationships er by which you can actually express at least in the schema that there is a functional dependency of the er person being married and his or her name or then the third erm er an a <SIC> thirdier </SIC> er the <SIC> thirdier </SIC> possibility of simply using the optional attributes </S1>
<S2> erm you are right in your observation and i buy it er that er optionality erm binary optionality is expressible <S1> mhm-hm </S1> of course binary means one con- concept is optional er er according to another concept <S1> yes </S1> but optionality is often er related to each other like you had it with this marriage and and and first name <S1> mhm-hm </S1> in this case one er should not appear before the other appear <S1> yes </S1> that mean in this case you have a relationship among optional items <S1> yes </S1> i don't know how to express this concept (maybe it will occur me afterwards but) <S1> @@ </S1> i don't know </S2>
<S1> well i don't know either i would rather use the functional dependency or the er like constraint between those concepts but i don't know if that's orthodox </S1>
<S2> er i think er the nine people of the uar- (U-A-N) people developed some constructs that tried to somehow but that's the only model where you can er that which i know <S1> mhm-hm </S1> maybe er i missed something (xx) (model) (xx) this kind er express this kind of optionality and optionality even more erm er , in modelling er they are usually database people are let's say sometimes a bit erm er er tough in representing anything of the world we have to find a name er (xx) spelled (xx) er and nothing else <S1>  mhm-hm </S1> er i go to the library i i got down to the library er my name misspelled i go to different country then i have no er expression for the T-H and for this reason something else is taken and and so on that means we don't have such kind of sharp representation and narrow approximate er representation er whenever i would like to er let's say er or i seek to develop er such kind of er com- at least er modelling language which can at least tackle the structure i would like to acquire some kind of approximation or or approx- er approximative er property or fuzzy property or er let's say even some kind of modality er can be represented , how you would do it or maybe you would disagree and say fine let's only er put in the database what it's really (valid) </S2>
<S1> erm i , well yes erm fuzziness of course in some areas of modelling is n- nowadays once again erm a big topic but er in in the approach that er we have taken things tend to be very discrete erm and that's of course mainly or partially because of the academic tradition that you use lots of textbook examples where there is not so much like erm well there is some pre-understanding of er the modelling field and and basically no fuzziness but er if i was to add er even this fuzziness support to er modelling language are you sure that the computational er n- demands for the language would not simply explode </S1>
<S2> that's er er the usual trade-off erm as you want to pre- represent in a proper way and you have to pay more <S1> mhm-hm </S1> or or you want to er er pay less and then er don't represent anything <S1> mhm-hm </S1> or you want to have data a- a- that is high (quantity) and then er don't think about quality of data <S1> mhm-hm </S1> it's er that depends er on the assumptions you are making before modelling <S1> mhm-hm yeah </S1> they are choosing the proper language and the proper tools some time some parts immediately er according er to er er th- the common claims that er we have er we are sure about our data that is the same quote you read in textbooks or in university classes but natural- er in in in practical er let's say database applications er a typical example is (john) telephone directory <S1> mhm-hm </S1> if you take an object as the (telephone) er entry with all then 15 per cent of the objects are wrong 15 per cent <S1> yes </S1> not zero point something 15 per cent that's a lot i don't know about the finnish (xx) if it's much better but anyway it's <S1> i don't really think so </S1> er and for this reason you have to have er a means er er how to tackle approximation as well <S1> mhm-hm </S1> in modelling . er , you said me i have two hours and two hours i count from from something then er let me come back er to one er er er of the er first part , erm mathematicians er it took mathematicians about one hundred 150 years or something like that to discover that they have only three er main things they are tackling they call it algebra typology and some kind of generation facility or symmetry or something like <S1> mhm-hm </S1> er this er triad came out also of er if you look into areas like (xx) er (xx) -graphy or if you go to quantum er mechanics we have the same er triad all the time er and now we have let's say the area of conceptual modelling er what we do er maybe quite well is er to reason about structure , er we don't do so well in reasoning about collaboration er don't er do (really) er in er er th- let's say at least first er (sensifying) let's say er er stage er this evolution of systems but we need to integrate systems er we need er to er collaborate er about systems and we need er in this case to extend our facilities by mechanisms to collaborate and to integrate <S1> mhm-hm </S1> so and er i would like to challenge you now erm after we've clarified some parts , how you would er build on top of your comparison a model er integrating COMIC integrating description logics let's say integrating IFO or operational or anything else er er and somehow er be able er to er at least er to map items from each other how yo- how you would do it </S2>
<S1> well if you are thinking about simple er concept diagram to er er something algorithms they are more or less provided let's see for instance erm page 79 er where these concept diagrams are mapped into IFO and even though the algorithm is rudimentary it is er usable , as for the actual answer er to er your question erm it is quite close to the question i posed in the er slide if there is actually if there is a win-win situation in a way that we can have both the intuitiveness of er the COMIC approach and then the clear semantics of erm some other more traditional approaches that er erm i think erm if we think of point of collaboration i would rather see that the s- er especially if we are thinking of unifying some er conceptual er schemata of different organisations erm if the starting point is er a simple most intuitive er concept diagrams where the meanings and interpretations inside of the organisations of the basic concepts have been fixed then we are rather well-off then it is much more fruitful and possible to combine the er or at least compare the er er concept diagrams of each organisations and then hopefully come up with something that er can have er a rather formal presentation further whether it is IFO whether it is I-R erm whether it is even description logics that's basically er a matter of the er er wits er and erm erm and er er labour of the person who develops the algorithms i would recommend er translations to er IFO but i'm sure some people would find description logics er er better for for instance information exchange later </S1>
<S2> erm let me separate erm er the issue er we are talking now let's talk first about integration and er evolution erm of er systems erm in this case we need er set er er we need some let's say common language my problem is er maybe i have some problem with the (xx) or something like that er or er er maybe i (xx) too much that er er only one could er (have you develop) the word formula but not me unfortunately therefore er only one can ever use the model to express anything er and not description logics and not er IFO <S1> mhm-hm </S1> erm how you would in this case integrate systems or indicate er erm that at least model or schemata developed so far er so that you can er somehow er trade common data <S1> mhm-hm </S1> how you would do it in your approach </S2>
<S1> well erm first of all erm i don't know exactly erm what parts of the book you would find like my approach for instance in the er pure so-called intensional language that was er that is expressed over here in chapters four and five here we er actually i think even expressly write er of er on page 57 erm last sentence <READING ALOUD> moreover using a proper set of axioms we can provide the users with a structure that is useful in many activities of conceptual modelling these include er for instance finding new concepts locating concepts and linking together various concept systems </READING ALOUD> if you replace concept system with a conceptual schema er there's the answer but of course mhm once again over here er this paper is purely academical i don't think that we could have addresses er addressed all like er system evolution items that you referred to but er this er these functions that are sort of provided in in the chapter they would as well as the lega- legality checking they would er provide er some kind of basis for er schema integration </S1>
<S2> well now we can begin with the discussion we had erm 60 minutes ago no 65 minutes ago once more (xx) and we agreed er whenever we we develop a model of (xx) er schema <S1> mhm-hm </S1> er we need er also to underpin the model by some kind of formal semantics er sorts they have why developing these concepts <S1> yes </S1> and my belief is if you er want to integrate systems you should open the intensional parts and to model intensional parts er this was er in the agreement we had already with you er in a little bit more than one hour ago <S1> er [very much so] </S1> [maybe] we can agree again <S1> yes </S1> that er @@ that i mean in this case for integrating you have to open what is behind the model <S1> mhm-hm </S1> and then you have er to agree on what is behind first and then er to integrate the system <S1> mhm-hm </S1> the same would be my belief er for evolution in this case it is database what is the underpinning the formal underpinning i didn't represent the schema because i'm lazy i don't want <S1> @@ </S1> that (xx) er everybody what i've i'm thinking <S1>  mhm-hm </S1> er for this reason i have this er er parts not represented but i need to open this in order to er m- be able to develop the schema furthermore or database further , and now we have the more physical part , some er times er er let's say applications have to collaborate we have now er in industry the age of merges everybody's merging with everybody hopefully that the outcome is good and now er how we can merge let's say our database is developed some way somehow on top (made in) er maybe er these academical languages maybe in declarational maybe even-orientated er maybe let's say er polyvariational er language how can they now collaborate in a proper way how we can do that </S2>
<S1> you're really [asking from me] </S1>
<S2> [er er yeah] i'm i'm asking you now what is what would be your proposal er to collaborate er among applications </S2>
<S1> erm practical applications th- that's a that's really er interesting and far-reaching question because of course erm m- nowadays people are er interested in these things called ontologies and er erm and s- the semantic web and of course there people talk a lot about those kinds of things meaning in a way that er er the applications would share a so-called er well let's say a conceptual schema that er more or less defines the er contents of the information erm i have to say that's not exactly my expertise and i might not know enough about it er but as for as what i have seen the er agreements of this kind of simple ontologies they have always been extremely erm technical only they wouldn't really show anything about the deep meaning of any information and that's simply how it is computers are simply computers and even programmers are simply er programmers but i would say if you are really talking about applications er communicating with each other i take your bite but i'm not providing a conceptual schema i'm providing this kind of a simple ontology so that it might be established using the conceptual schema so that the programmers would understand what they're talking about </S1>
<S2> i was sure erm that this is er one of the difficult questions to ask you but erm i was also sure you are able to answer after reading your PhD or er this er erm <S1> [(xx)] </S1> [work here] because of i thought er er on the basis of that you can deduct <S1> [mhm-hm] </S1> [derive] what er could be done in this case er you refer and i think this is a very good move er to this chapter four and five where you er where you're using this er functional logics approach it's a very good an- and simple approach because of er we're cutting down anything to er the basic er elements that's maybe a little bit too heavy but in this case you have connected collaboration agreement first and then you take (xx) collaboration agreement now i can now say okay fine since this is meant to (xx) and then i have a part of my theory my intention (xx) er which is method and (xx) i thought er okay fine it's now er two hours working for you and for this reason maybe er er during er late night we will come up with the solution erm but er i think it is possible to develop er based on this approach <S1> mhm-hm </S1> , erm i promised to have four hours a questions i have er at least one question i want to answer er your answer erm to er and this is er whenever we are thinking about modelling er of course we can think about structures structures are sometimes very useful but er let's say (xx) (classes) er a lot of people preach normalisation is a very good weapon or <S1> mhm-hm </S1> they have a good operating st- the systems and you go to practise and then in practise they state to you oh (xx) missing er it's not running er because normalisation is stupid <S1> mhm-hm </S1> forget about it and what they told you in the university is completely wrong , both are right and er the question is now what you want to do with the structures that means in this case you should also (put) at least via modelling and meaning an operational meaning what is the purpose of your structuring mechanisms i can take er let's say your faculty and your students as er let's say some kind of (xx) maybe a generalisation or a a specialisation or some kind of let's say only rules i can er take whether i take this or that approach depends on what is the reason i need let' say to express in structures <S1> mhm-hm </S1> what are my expectations for behaviour <S1> yes </S1> how you can tackle this your appro- in these approaches er you have er somehow tackled here how is it tackled </S2>
<S1> well over here er luckily good heavens i can refer to an article by professor <NAME> er er in the er COMIC article that was er as i said one of the er motivations there is a kind of a idea of er combining er state machine in the er er in these er er concept diagrams i don't know i don't think that the idea over there is completely well developed because er it was more realised through er conditions but er i think it could contain some elements of the er som- some elements of the behaviour of er the actual well intended application however i don't know if that would be truly enough to kind of er solve the practical normalisation problem so that you would actually make the implementation more or less automatically efficient i think for that kind of a thing you would actually need a conceptual modelling language that is so close or so integrated with the programming language that er well that i see would be almost the onl- only possibility to solve that problem </S1>
<S2> i think erm erm let's say erm your er misbelief that this is possible er er is a misbelief which was claimed er long time but i think it's not true <S1> mhm-hm  </S1> er it is you can cluster element and contain never separate (tools) for that they have to run smoothly together <S1>  mhm-hm </S1> er you did it intentionally in this relations (xx) okay fine good (interrelations) must go together later (xx) came and then we have key relations <S1> mhm-hm </S1> at least we have some kernel operations going underneath it was a good idea at that time it means you have some kind of er construct keeping thing together and we need to have the same meta-construct , keeps things together that means in this case you need er okay fine er (person) and er the relationships in the university may be different for this reason we have faculty and student or something like that <S1>  mhm-hm </S1> and in this case you say fine they have universities or er some structural body or institutions er from one side we have persons on the other side and we have some er associations relationships and everything to do and in er in this case we have some meta-information <S1> yes </S1> behind and my belief is a good conceptual modelling language should have different levels of abstraction and er then in this case i would expect , er that you have proved <S1> mhm-hm </S1> to have some different levels of abstraction </S2>
<S1> that would be very much possible as well </S1>
<S2> so when we are dealing this er some kind of er let's say with behaviour er (trait) of behaviour we need also to separate our structuring mechanisms erm i think er we are (xx) also to database classes and to learn that is a multiway dependency <S1> mhm-hm </S1> it's a very nice er vehicle to amuse students or to confuse students depending on how you are doing it er this er multiway dependencies erm you have er er logical construct to explain what is the internal structure in th- in a schema unfortunately you have (given) sort of textbooks but er let us (xx) of these constructs then in this case you should say okay fine now let's say er er er a social security number belongs to a person er since i have to hav- i- identify with something like that a birthday is a property er which er in reality never changes and the database can er be changed into (presenting this state) <S1> mhm-hm </S1> sex is a property er which er yeah okay fine we can change erm from time to time but not too often not each months at least as i know then er salary and rank and all these things can be changed quite rapidly that means in this case you can model let's say somehow shells around er a person what is changed and what is not changed that means you have a model and you would have a lot of separation and i think that kind of principles you need to develop before doing modelling er how you would do it this let's say er three languages how you would er develop a methodology to use these three languages </S2>
<S1> well erm that i i would s- actually refer to the ISO report simply 'cause er i don't think that it has that kind of a thing separating er elements that are bound to change often and separating elements that are erm the from elements that are not likely to change that often erm that is something that er is really in modelling not really in language in language we can simply develop some kind of auxiliary support mechanism to point out that this thing is likely to change quite often but er erm in the ISO report we actually find as a principle of modelling that we should concentrate on the conceptual level and er how was it like er , traits of reality that are likely to remain erm er static or they are not likely to change too often <S2> mhm-hm </S2> so it's really in the er modelling methodolog- methodology not so much really in in languages </S1>
<S2> well it's how you are using the language how you learn this natural language how you use the nat- natural language <S1> mhm-hm </S1> your mother says you never use it this and that way (you use this and that way) <S1> mhm-hm </S1> it's typical thing you should have in modelling languages as well . okay er i er made my case now it's your er work maybe i should somehow er finish er with official er (bit) , er okay fine in this case er , i should propose that you are somehow succeeded in defending your PhD at least i realise you knew it better than me <SS> @@ </SS> er you knew it or the page all the time when i had er to look each chapter (of course) er i find the work really er sufficient i find the work nice erm it is one of the very few works and it is a good compliment er since i was a bit too critical maybe er during these two hours er which i read from cover to cover usually i don't spend too much time with a book reading it from cover to cover but er that was a really nice i think for this reason i propose the faculty to accept you and er the grade i should not say er that is something told you later okay fine and that would be , well you you can now wait for something else </S2>
<SS> @@ </SS>
<S1> thank you very much doctor professor <NAME S2> for your kind words and your questions that truly showed your expertise of the er subject and many other subjects as well so er erm if there are people in the audience who would like to ask further questions please ask the <FOREIGN> kustos </FOREIGN> for the floor </S1>
<P:20>
<S3> okay , thank you for for discussion and comments and er this discussion is closed </S3>
<S1> there will be cof- coffee served outside of these doors please join us for some coffee </S1>
