SPA Template

SPA - SPeech Analytics

From SPA

The original motivation for project SPA - "Speech Analytics" was to answer the many requests we received for transcribing audio/video files in European Portuguese. Our intention was to provide an interface easy to use by a non-expert, that would allow manual editing of the automatic subtitles.

However, requests have multiplied and diversified, namely in terms of languages (Spanish, English, etc.) and varieties covered (European, Brazilian, ...). On the other hand, although the majority of the users is only interested in the automatic transcripts, other technology partners showed interest in obtaining info about characteristics (male/female) and state (angry, neutral) of the speakers, the language or variety, and all the other metadata that could be retrieved from the speech signal.

The SPA platform has recently been redesigned from scratch to include a large number of speech recognition and metadata extraction modules, rather than a service which grew in a bottom-up way from a solution for a single type of requests.

The large majority of users may use the SPA service with all the default options. There are several modules available:
The automatic labels and transcripts are only adequate for the domain in which the respective module has been trained.