SpeechTech Archive is a system for creating and searching audiovisual archives. The purpose of the system is to provide users with the ability to quickly find events (occurrences of words or phrases) in a very large archives of video or audio recordings. For this purpose, the first the index files of all recordings in the archive is automatically generated, in which later the search is carried out. This procedure provides a much faster search found key phrases than can be achieved by direct application to search recordings on their own.
Scanning speed of direct recording (ie, only using the systems of speech recognition without the use of indexing), is currently in the order of comparable length of recordings of archived files themselves. Utilization of SpeechTech Archive, based on indexing of recordings, reaches significantly shorter times to search the extensive archives. So for example, the system can find in the 1000 hours of video the searched occurrences of phrases within tens of seconds.
When creating the archive there is applied a number of algorithms based on automatic speech recognition applied to input audio or audiovisual recordings. Each record is recognized and the results are recorded in the index file. During recognition not only the "best" recognized text is stored, but far more information and variations of recognition are stored to the resulting index. Later, when searching, it allows the user to find unusual or foreign words that are not included in the dictionary of the speech recognition system.
SpeechTech Archive is usually possible for customers to deploy without major modifications, but it can also be extended or adapted to specific requirements.