Automated Speech Scoring in Czech

The aim of the project is to design and implement software applications for the automatic evaluation of spoken texts in Czech created by students of Czech as a foreign language who are preparing for certified language exams. It is difficult for foreign students to create a comprehensive, fluent and understandable speech in Czech suitable for a given communication situation. It is a more demanding communication skill than, e.g., creating individual sentences. At the same time, it is a skill that is essential for understanding basic communication situations and existence in everyday life.

The creation of a coherent speech in Czech should be supported especially in the teaching process in language schools or courses. At the same time, it is necessary to have suitable tools for speech evaluation – continuous speaking is, e.g., part of certified exams in Czech for foreigners who apply for permanent residence or citizenship in the Czech Republic. Due to the current workload of the institutions that provide these exams, automated speech scoring would be a useful tool. In addition to educational institutions and testing centers, it can also be used by individual students during their preparation for certified exams. The developed computer tools (intended for specific certified exams) will also support the objectivity of evaluation in education and will be accessible to the professional public free of charge.

The results of the project will include linguistically processed databases containing spoken texts of foreign students focused on the levels of language examinations needed to obtain permanent residence and grant citizenship in the Czech Republic. 

An important goal of the project is to contribute to the improvement of mutual understanding between people, to support the national language and the culture of spoken communication.

Project results:

Rysová Kateřina, Novák Michal, Rysová Magdaléna, Polák Peter, Bojar Ondřej: Database of Spoken Czech as a Foreign Language (Permanent Residency in the Czech Republic). Institute of Formal and Applied Linguistics MFF UK, Prague 2024. Available from WWW https://lindat.mff.cuni.cz/services/teitok-live/evaldio/en/index.php?action=db_residency.

Novák Michal, Polák Peter,  Rysová Kateřina,  Rysová Magdaléna,  Bojar Ondřej: Towards Automated Spoken Language Assessment: A Study of ASR Transcription of Examinations for Non-Native Speakers of Czech. The Prague Bulletin of Mathematical Linguistics, 2024, 122 (2), ISSN 1804-0462. [in press]

Acknowledgement:

The research is funded by the Programme to Support Applied Research in the Area of the National and Cultural Identity for the Years 2023 to 2030 (NAKI III) of the Ministry of Culture of the Czech Republic within the project Automated Speech Scoring in Czech (DH23P03OVV037).