A Statistical Approach to Automatic Speech Summarization

Abstract

This paper proposes a statistical approach to automatic speech summarization. In our method, a set of words maximizing a summarization score indicating the appropriateness of summarization is extracted from automatically transcribed speech and then concatenated to create a summary. The extraction process is performed using a dynamic programming (DP) technique based on a target compression ratio. In this paper, we demonstrate how an English news broadcast transcribed by a speech recognizer is automatically summarized. We adapted our method, which was originally proposed for Japanese, to English by modifying the model for estimating word concatenation probabilities based on a dependency structure in the original speech given by a stochastic dependency context free grammar (SDCFG). We also propose a method of summarizing multiple utterances using a two-level DP technique. The automatically summarized sentences are evaluated by summarization accuracy based on a comparison with a manual summary of speech that has been correctly transcribed by human subjects. Our experimental results indicate that the method we propose can effectively extract relatively important information and remove redundant and irrelevant information from English news broadcasts

Consolidation-Based Speech Translation and Evaluation Approach

“…We tested whether the consolidation function in the summarization could contribute to the quality of machine translation in the IWSpS [14]. Five news stories, consisting of 25 utterances on average, of CNN broadcast news speech were transcribed and summarized at 40% and 70% extraction ratio [2]. The word accuracy of the ASR results was 78.4%.…”

“…We proposed a dependency score to avoid concatenating two words which do not have a dependency structure [2]. However, dependency detection of spontaneous speech is still challenging.…”

“…In the past, we have worked on spoken language processing system such as summarization of speech in meetings [1] and broadcast news [2] and machine translation (MT) of travel conversations in the C-star project [3], appointment negotiation in the Verbmobil project [4], and dialogue in ecommerce in the NESPOLE! Project [5].…”

“…We have proposed a statistical summarization approach which extracts words from transcriptions according to compression ratios by focusing on 1) extracting important content words, 2) excluding redundant and irrelevant phrases, and 3) concatenating words in this summarization approach to maintain original meanings [2]. This approach accomplishes important information extraction and speech consolidation simultaneously.…”

Topic and Stylistic Adaptation for Speech Summarisation

“…Variations of manual summarisation results are merged into a word network as shown in Figure 2, which is considered to approximately express all possible correct summarisations covering subjective variations. The word accuracy of automatic summarisation is calculated as the summarisation accuracy using the word network [5]:…”