Automatic Recognition of Spontaneous Speech for Access to Multilingual Oral History Archives

Abstract

Abstract-Much is known about the design of automated systems to search broadcast news, but it has only recently become possible to apply similar techniques to large collections of spontaneous speech. This paper presents initial results from experiments with speech recognition, topic segmentation, topic categorization, and named entity detection using a large collection of recorded oral histories. The work leverages a massive manual annotation effort on 10 000 h of spontaneous speech to evaluate the degree to which automatic speech recognition (ASR)-based segmentation and categorization techniques can be adapted to approximate decisions made by human annotators. ASR word error rates near 40% were achieved for both English and Czech for heavily accented, emotional and elderly spontaneous speech based on 65-84 h of transcribed speech. Topical segmentation based on shifts in the recognized English vocabulary resulted in 80% agreement with manually annotated boundary positions at a 0.35 false alarm rate. Categorization was considerably more challenging, with a nearestneighbor technique yielding F = 0 3. This is less than half the value obtained by the same technique on a standard newswire categorization benchmark, but replication on human-transcribed interviews showed that ASR errors explain little of that difference. The paper concludes with a description of how these capabilities Manuscript

Adaptation of machine translation for multilingual information retrieval in the medical domain

“…This technique is used to adapt language models [23,24] as well as translation models [25,26] or their combination [27]. Similar approaches to domain adaptation are also applied in other tasks, e.g., automatic speech recognition [28].…”

Matching meaning for cross-language information retrieval

“…The ASR process (with a 38% measured word error rate on held out data) was optimized for the interviewee rather than the interviewer (by automatically detecting and then consistently using only the interviewee's microphone) and was trained using 200 hours of in-domain held out data along with other other standard ASR training resources (Byrne et al, 2004). This resulted in the text contained in the ASRTEXT2004A field of each segment in the test collection.…”

Proceedings of the third workshop on Searching spontaneous conversational speech

“…Planned speech, such as broadcast news, is typically organized into topical segments, for example, news stories, that can be treated as documents within the retrieval system. Spontaneous, conversational speech is less well structured and topic can change spontaneously [2]. Within the speech stream topic boundaries are challenging to identify and may not be well defined.…”