Language ID Container

2.2.1

This release increases the language converage to 44 languages: Arabic, Bashkir, Belarusian, Bulgarian, Catalan, Mandarin, Czech, Welsh, Danish, German, Greek, English, Spanish, Estonian, Basque, Finnish, French, Galician, Hindi, Croatian, Hungarian, Indonesian, Italian, Japanese, Korean, Lithuanian, Latvian, Mongolian, Marathi, Malay, Dutch, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Swedish, Tamil, Thai, Turkish, Ukrainian and Vietnamese. Additionaly the new expected_languages config option allows to restrict the languages that will be predicted.

The output now includes predicted_language and error fields, indicating success or failure reason for identification that considers amount of speech and prediction confidence.

A new speech detection algorithm ensures that we only identify parts of the audio that contain speech and return an error if no speech can be found in the file.

Known Issues

Speechmatics is aware of CVE-2018-20225 Python Pip vulnerability in the Container. Speechmatics does not use pip within the container in a way that would expose the CVE (CVE-2018-20225), customers who choose to extend the container by installing new packages are responsible for using extra-index-url securely.

1.0.0

This is the first release of Speechmatics' Language Identification (ID) Container. Language Identification currently identifies the predominant language spoken in a media file and helps with automating the process of selecting which language pack to use for transcription. 12 languages are supported.