MA Configurations

MA Configuration I (Mercury)

MA Configuration I (Mercury) provides capabilities for the collection of meeting speech and basic discourse/dialog understanding. In this configuration, collected speech is transcribed and then aligned across participants. The Meeting Browser displays transcripts and the recognized speech from each meeting along with the time each utterance was spoken. Users can click any utterance to replay the collected audio from that time in the meeting, which is especially useful for listening to segments for which the automated transcription is of lower quality than desired.

The following figure shows the components of MA. Components in yellow are proprietary and are not included in the MA download. Connecting to an existing chat server is useful, but optional for participant collaboration. However, a speech-recognition server (e.g., SRI's Decipher ASR suite) is mandatory, but is not included. Other speech recognition engines can be substituted as long as they adhere to the established XML-RPC APIs. These components plus an ASR module create an end-to-end system that collects audio during a meeting, transcribes that speech, and then presents the meeting transcription and other artifacts in an easy-to-use web browser interface.

MA Configurations

Additional Notes on MA Framework Configuration I (Mercury)

MA Configuration II (Transcriber)

The MA Configuration II (Transcriber) enables transcribing an uploaded speech file using the same XML-RPC APIs for the speech-recognition software employed within Mercury. This capability enables the transcription of meeting recordings that were collected outside of the standard Mercury Client/Server modules.

Note: the Mercury Client/Server independently records each speaker with exact timing information for all speakers such that a set of speech files can be interleaved and displayed with the proper participant attribution in the Meeting Browser. The transcriber will not distinguish between multiple speakers on the same track. Also, if speech is uploaded with significantly different characteristics than the models over which the speech engine was trained, recognition quality will suffer.

Installation Guide

MA Server Installation Guide This document introduces the main components of MA and describes how to install MA Configuration I.

Transcriber Installation Guide  This document describes how to install MA Configuration II.