Speech Recognition APIs in the Context of Using English as a Second Language
Fragment książki (Rozdział w monografii)
MNiSW
20
Poziom I
| Status: | |
| Autorzy: | Czyż Karolina, Derkacz Michał, Smołka Jakub, Łukasik Edyta, Skublewska-Paszkowska Maria |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Strony: | 29 - 40 |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | TAK |
| Licencja: | |
| Sposób udostępnienia: | Witryna wydawcy |
| Wersja tekstu: | Ostateczna wersja opublikowana |
| Czas opublikowania: | W momencie opublikowania |
| Data opublikowania w OA: | 30 grudnia 2019 |
| Abstrakty: | angielski |
| Speech recognition systems are applied in many different solutions (e.g. web and mobile applicationsfor language learning or voice assistants). They are frequently used by non-native speakers. Speech recognition accuracy or its tolerance to pronunciation imperfections may be an important aspect. This article compares four APIs for speech recognition: Web Speech API, Microsoft Speech Service, Watson Speech to Text and Android SpeechRecognizer. The aim was to determine which API best recognises the speech of a person who uses English as his/her non-native language. The tests involved two groups of participants: (1) persons with modest language skills (level A1-A2), (2) people whose language level was at least B1. The participants read a set of sentences. Their speech was processed by each API included in the comparison. The results were assessed using (1) the percentage of incorrectly recognised words, (2) the word error rate, (3) the Levenshtein distance, (4) the number of incorrectly recognised words in a sentence. The best API for more advanced English speakers is the Watson Speech to Text service. The best API for non-fluent English speakers is Android SpeechRecognizer. |
