Retrieval of Spoken Information from Large Audiovisual Archives

The Speech Retrieval product of ALL offers the possibility to search through large audiovisual archives for speech just in the same manner as the search for texts can be realized in usual information retrieval systems.
The system encompasses two self-contained modules: Preprocessing and Retrieval.
Preprocessing is a speech recognition application, by the use of which the archive administrator can initiate the preparation of the speech segments of the archived files for retrieval. The preparation process includes detecting the boundaries of the speech segments within the input file, recognizing at each speech segment what was said, and indexing the speech segment by the recognized words.
The other module (Retrieval) provides an adequate interface for the end-users to form queries, find and rehear the sought speech segments. The module accepts queries in written form and collects the identifying items of the matching speech segments into a hitlist. The speech segment to hear can be chosen from the hiltlist.
The Preprocessing module has an English version and a Hungarian one. The Retrieval module is language-independent.
The preparation of the speech retrieval service can easily be extended to the management of archived textual documents. This includes converting (temporarily) the textual documents of the archiving file to txt format, then using the indexing component of the Preprocessing module. The extended services offer operating an information retrieval system that enables its users to retrieve spoken or written information from mixed digital repositories, which preserve sound/video files and text documents (txt, doc, pdf, etc. files) also. The hitlist for a query may contain references to speech segments as well as textual documents.
The Speech Retrieval system of ALL was designed and developed within the scope of the EASAIER (Enabling Access to Sound Archives through Integration, Enrichment and Retrieval http://www.elec.qmul.ac.uk/easaier/) EU FP6 project. The complete EASAIER system has been deployed at the British Library and the Caledonian University, Spoken Words. (The latter manages the audiovisual archives of BBC.)