We are developing speech processing technologies such as core multilingual unlimited vocabulary speech recognizers for automatic speech transcription, audio indexing, and speech-text alignment.

The VoxSigma software suite provides audio segmentation and partitioning, speaker identification, language recognition, and large vocabulary speech recognition capabilities in many languages.
VoxSigma has been designed for professional users needing to process large quantities of audio and video documents such as broadcast data or call-center communications, either in batch mode or in real-time.

The Vocapia Research VoxSigma software suite for Linux offers state of the art performance for broadcast data and conversational data in many languages. The VoxSigma API includes Unix/Linux commands, C and C++ libraries, REST API, import and export in XML format. The VoxSigma software is available both via licensing and via our web service.
In addition the Vocapia Scribe3 GUI offer a convenient way to upload, process and post-edit audio and video documents.

[Voxsigma request form]

Spoken language identification (LID) is the process of recognizing the language spoken in an audio document (broadcast audio, podcast, or telephone). The standard VoxSigma language identification component can recognize one of 100 languages. By default, it is assumed that each document contains only one language. However, this can be adjusted by specifying the maximum number of languages and optionally providing a limited list of possible languages. The LID module includes a customization tool that allows users to tailor the models to their specific data, such as adding a missing language or dialect.

Speech transcription (also called speech-to-text, or STT) is the process of recognizing the words spoken in an audio document, while also providing a timecode, duration, and confidence score for each identified word. The speech transcription module can handle various audio types, including multilingual and dual-channel audio. It supports an unlimited vocabulary for multiple languages, including Arabic, Cantonese, Czech, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Italian, Latvian, Lithuanian, Mandarin, Pashto, Persian, Polish, Portuguese, Romanian, Russian, Spanish, Swahili, Swedish, Turkish and Urdu. The STT module includes a customization tool that allows users to tailor the models to their specific data or task.

Large vocabulary continuous speech recognition is a key technology that can be used to enable content-based information access in audio and video documents. Most of the linguistic information is encoded in the audio channel of audiovisual data, which once transcribed can be accessed using text-based tools. Via language identification, speech recognition, and speaker recognition, spoken document retrieval can support random access using specific criteria to relevant portions of audio documents, reducing the time needed to identify recordings in large audio/video databases. Some applications are data mining, news-on-demand, media-monitoring, and telephone speech analytics.

Speech-text alignment is the process of synchronizing a speech signal with a speech transcript or closely related text, providing time codes for words and sentences. The alignment process assigns timecodes to each word and each punctuation mark in the audio transcript and provides confidence scores to identify areas where the alignment may not be perfect in particular when the provided transcript differs from what has really been said. There are many uses of this technology, including audio books, language learning, and video subtitling.