Workshop description
Automatic Assessment of Atypical Speech (AAAS) explores the assessment of pronunciation and speaking skills of children, language learners, people with speech sound disorders and methods to provide automatic rating and feedback using automatic speech recognition (ASR) and large language models (LLMs). Automatic speaking assessment (ASA) is a rapidly growing field that answers to the need of developing AI tools for self-practising second and foreign language skills. This is not limited to pronunciation assessment, but the AI tools can also provide more complex feedback about fluency, vocabulary and grammar of the recorded speech. ASA is also very relevant for detection and quantification of speech disorders and for developing speech exercises that can be performed independent of time and place. The important applications of non-standard speech also include interfaces for children and elderly speakers as an alternative to using text input and output. The topic is timely, because the latest large speech models allow us now to develop ASR and classification methods for low-resourced data, such as atypical speech, where annotated training datasets are rarely available and expensive and difficult to produce and share. The goal of this workshop is to present the latest results in ASA and discuss the future work and collaboration between the researchers in Nordic and Baltic countries.
Topics
In particular we would like to invite students, researchers, and other experts and stakeholders to contribute papers and/or join the discussion on the following (and related) topics:
- Automatic speaking assessment (ASA) for L2 (second or foreign language) pronunciation
- ASA for spoken L2 proficiency
- ASA for speech sound disorders (SSD)
- Automatic speech recognition (ASR) for L2 learners
- ASR for children and young L2 learners
- ASA and ASR for Nordic and other low-resource languages and tasks
- Spoken L2 learning and speech therapy using games
- Automatic generation of verbal feedback for spoken L2 learners using LLMs
Important dates
- First CFP: 04 October 2024
- Submission DL for papers: 16 December 2024
- Notification of acceptance: 24 January 2025
- Camera-ready DL: 3 February 2025
- Workshop: 5 March 2025
Organizers
- Mikko Kurimo (chair), Aalto University, mikko.kurimo@aalto.fi
- Giampiero Salvi, NTNU
- Sofia Strömbergsson, Karolinska Institutet
- Sari Ylinen, Tampere University
- Minna Lehtonen, University of Turku
- Tamas Grosz, Aalto University
- Ekaterina Voskoboinik, Aalto University
- Yaroslav Getman, Aalto University
- Nhan Phan, Aalto University
Invited speakers
Speakers will be announced later.
Program
AAAS-2025 will be a full day workshop (9-17) on 5 March 2025 (Wednesday) with invited talks and oral and poster presentations. Details TBA.
Paper submission
The NoDaLiDa/Baltic-HLT 2025 formatting templates will be used for the workshop. The reviews will be double-blind. Details TBA.
You can access the template on the NoDaLiDa/Baltic-HLT 2025 website.
Proceedings
TBA.
Presenter instructions
TBA.
Registration and fees
Register via the NoDaLiDa/Baltic-HLT 2025 registration website. Details TBA.
Venue
The workshop will take place in Tallinn after the NoDaLiDa/Baltic-HLT 2025. We plan to provide hybrid access for registered participants who can not attend physically. Details TBA.