Aditu.eus:

Automatic transcription and subtitling

Overview

Aditu is Elhuyar’s speech recognition web service. Aditu offers the possibility to automatically transcribe, subtitle, translate and dub audio and video using Orai technology. Audio recordings and videos can be automatically transcribed or subtitled in six languages (Basque, Spanish, Catalan, Galician, English and French), or in two language combinations (Basque-Spanish and Basque-French). In addition, via an advanced editing interface, it is possible to automatically correct the result. Furthermore, the transcription can be translated into six languages using machine translation. It is also possible to correct translations in the interface. Using TTS technology, Aditu has also become a dubbing platform: it allows segments of audio translated into another language to be dubbed or voiced over. It is also possible to use a system that imitates voices or generates generic voices. Finally, it allows the result to be downloaded in different formats. 

Challenge

Automatic transcription technology is a challenge in itself, and even more so for a poorly-resourced language like Basque; in particular under special conditions or in special cases: local speech forms, informal registers, noisy environments, poor audio quality, personalisation (to correctly recognise local words, place names and proper names), children’s voices, speaker identification, etc. This tool integrates various technologies developed by Orai in the field of natural language processing.

Collaboration

The project was developed by Elhuyar and Orai.

Result

Orai transcription technology achieves very good results under suitable conditions in Basque, Spanish, Catalan, Galician, English and French, and in language combinations. It is integrated into Aditu, which has many users and customers, mainly in the fields of education and communication. We are continuously improving our resources through research and are incorporating these improvements into Aditu.