A behavioural mode research on user-focus summarization

Different persons often choose different contents in multi-document as summary. To optimize summarization, we will focus on the selection of content and seeking their valuable features. Statistical methods for automatic summarization are very important. In this paper, we research the correlation between the eigenvalue of content unit in the original document cluster and the probability of the content unit to be selected as a human summary based on a statistical method. When a Basic Element and word are considered as a content unit, we draw conclusions, in user-focus summarization. It is excellent that the BE is regarded as content unit granularity, and it is proved that the frequency eigenvalue of the BE is more suitable to embody content units’ weightiness than the TFIDF value. Moreover, the paper reveals that the given topic on user-focus summarization is helpful for the selection of content unit and quality of summarization. They often choose those content units as a summary in which the emerging frequency is relatively high in the sentences including the content unit of a given topic and neighboring sentences. Through researching potential behavioural modes about manual summary, we will put these effect factors of summarization quality into the process of content unit selection and summary generation to optimize automatic summarization.