Using Proximity in Query Focused Multi-document Extractive Summarization

The query focused multi-document summarization tasks usually tend to answer the queries in the summary. In this paper, we suggest introducing an effective feature which can represent the relation of key terms in the query. Here, we adopt the feature of term proximity commonly used in the field of information retrieval, which has improved the retrieval performance according to the relative position of terms. To resolve the problem of data sparseness and to represent the proximity in the semantic level, concept expansion is conducted based on WordNet. By leveraging the term importance, the proximity feature is further improved and weighted according to the inverse term frequency of terms. The experimental results show that our proposed feature can contribute to improving the summarization performance.