Topic extraction with multiple topic-words in broadcast-news speech

Abstract

This paper reports on topic extraction in Japanese broadcastnews speech. We studied, using continuous speech recognition, the extraction of several topic-words h m broadcast-news. A combination of multiple topic-words represents the content of the news. This is a more detailed and more flexible approach than using a single word or a single category. A topic-extraction model shows the degree of relevance between each topic-word and each word in the article. For all words in an article, topic-words which have high total relevance score are extracted. We trained the topicextraction model with five years of newspapers, using the frequency of topic-words taken f " headlines and words in articles. The degree of relevance between topic-words and words in articles is calculated on the basis of statistical measures, i.e., mutual information or the X2-value. In topicextraction experiments for recognized broadcast-news speech, we extracted five topic-words h m the IO-best hypotheses using a X2-based model and found that 76.6% of them agreed with the topic-words chosen by subjects.

A spoken dialogue system for document information retrieval utilizing topic knowledge

“…These were calculated by using Eqs. (11) and (12), respectively. For the WCR, W all represents the total number of words, W sub the number of substitution errors, and W del the number of deletion errors.…”

“…(3) Use the estimated frequency m ij of w i in t j , which is calculated by using Eq. (1), to obtain the χ ij 2 value having a negative value [11] according to Eq. (2).…”

“…This flow is also used in our previous system. Recently, researches on topic estimation and topic tracking have been conducted in many places for the purpose of automatically indexing broadcast news speech [9][10][11]. If these topic estimation techniques can be applied to an information retrieval system and the topic information in the dialogue that is currently ongoing can be used to narrow down the documents, the retrieval process can advance more efficiently.…”

Smarter Blogroll: An Exploration of Social Topic Extraction for Manageable Blogrolls

“…Indeed, a wide variety of topic extraction (e.g., [1,11]) and summarization (e.g., [2]) techniques for non-technical language have been demonstrated in the research literature, many of which would be applicable to this project. Some past research has applied topic analysis to email corpora [14].…”