Photo: Antti Yrjönen
Kielipankki – The Language Bank of Finland offers a comprehensive set of resources, tools and services in a high-performance environment. Kalle Lahtinen tells us about his research on linguistic affect and speech emotion in the context of spontaneous Finnish.
I’m Kalle Lahtinen, a doctoral researcher at the Signal Processing Research Centre at Tampere University. Prior to my postgraduate studies, I studied computer science, signal processing, acoustics, and mathematics. My doctoral thesis is supervised by Professor Okko Räsänen and Senior University Lecturer Liisa Mustanoja from Tampere University, as well as Senior University Lecturer Juraj Šimko from the University of Helsinki.
The topic of my doctoral thesis is Human and Machine Perception of Affect in Speech. I investigate linguistic affect and speech emotion in the context of spontaneous Finnish. By spontaneous, I refer to unacted, unconstrained expression in (everyday) speech. I use both spoken and written language in my work. I examine how the semantic content of speech (what was said) and the corresponding acoustic signal (how it was said) vary, interact, and shape the interpretation of affective expression as a whole.
In my research, I analyze large language corpora using signal processing and machine learning methods. My doctoral thesis produces new data and approaches for analyzing and modeling linguistic affect in a way that considers both individual and situational variation in spoken language. The findings contribute to advances in both linguistic research and the developments of speech technology applications.
My work is part of the interdisciplinary CONVERGENCE of Humans and Machines project at Tampere University, funded by the Jane and Aatos Erkko Foundation.
FinnAffect, a corpus of spontaneous emotional speech in Finnish compiled during the first phase of my doctoral research, will be published in the Language Bank of Finland. The corpus is compiled from the Donate Speech, Helpuhe, and Tampuhe datasets (which are already available or will soon be available through the Language Bank). The FinnAffect corpus consists of 1,474,728 transcribed utterances lasting 1–20 seconds, of which 12,000 have been annotated for emotional valence and arousal. The corpus contains speech from thousands of speakers and is therefore the first of its kind in Finland in the field of linguistic affect research. The compilation and publication of the FinnAffect corpus was made possible by the extensive spoken-language resources already deposited in the Language Bank of Finland and, of course, by the expert support they provided to me there.
Lahtinen, K., Vaaras, E., Mustanoja, L., Räsänen, O. (2025). Investigating affect mining techniques for annotation sample selection in the creation of Finnish affective speech corpus. Interspeech 2025, 3958-3962. DOI: https://doi.org/10.21437/interspeech.2025-687
Lahtinen, K., Mustanoja, L., Räsänen, O. (2025). FinnAffect: An affective speech corpus for spontaneous Finnish. Speech Communication 175, 103327–. DOI: https://doi.org/10.1016/j.specom.2025.103327
The FIN-CLARIN consortium consists of a group of Finnish universities along with CSC – IT Center for Science and the Institute for the Languages of Finland (Kotus). FIN-CLARIN helps the researchers of Social Sciences and Humanities to use, refine, preserve and share their language resources. The Language Bank of Finland is the collection of services that provides the language materials and tools for the research community.
All previously published Language Bank researcher interviews are stored in the Researcher of the Month archive. This article is also published on the website of the Faculty of Humanities of the University of Helsinki.
