Speech recognition on-line services

Our on-line services allows for speech-to-text conversion and monitoring TV and radio broadcast channels.

The web-based speech-to-text conversion QuickSten/WebSten tool allows for editing words synchronously with sound. The experimental service on web allows for converting media data (files or links) to text and make word corrections while listening to the respective speech segment. We can see below a piece of speech-to-text conversion results for the record of a politician press-conference. Total duration of the record is 917 seconds, WER=25.8%. For majority of erroneously recognized words the user needs to correct just one symbol. The web-interface allows for correcting single word at once as well as merging two running words.

Channel monitoring service MediaAudit allows for finding the broadcast episodes by keywords.
Below you can see the search result for a set of keywords specified as академі*, which covers words academy, academic and academician in different cases. Normally, most of found broadcast episodes are relevant to your search. You can click a highlighted word to check the episode relevance and analyze the spotted information.

Please contact info@cybermova.com for more information.