You Ouyang

An Integrated Multi-document Summarization Approach based on Word Hierarchical Representation

This paper introduces a novel hierarchical summarization approach for automatic multi-document summarization. By creating a hierarchical representation of the words in the input document set, the proposed approach is able to incorporate various objectives of multi-document summarization through an integrated framework. The evaluation is conducted on the DUC 2007 data set.

A Novel Composite Kernel Approach to Chinese Entity Relation Extraction

Relation extraction is the task of finding semantic relations between two entities from the text. In this paper, we propose a novel composite kernel for Chinese relation extraction. The composite kernel is defined as the combination of two independent kernels. One is the entity kernel built upon the non-content-related features. The other is the string semantic similarity kernel concerning the content information. Three combinations, namely linear combination, semi-polynomial combination and polynomial combination are investigated. When evaluated on the ACE 2005 Chinese data set, the results show that the proposed approach is effective.

Learning Similarity Functions in Graph-Based Document Summarization

Graph-based models have been extensively explored in document summarization in recent years. Compared with traditional feature-based models, graph-based models incorporate interrelated information into the ranking process. Thus, potentially they can do a better job in retrieving the important contents from documents. In this paper, we investigate the problem of how to measure sentence similarity which is a crucial issue in graph-based summarization models but in our belief has not been well defined in the past. We propose a supervised learning approach that brings together multiple similarity measures and makes use of human-generated summaries to guide the combination process. Therefore, it can be expected to provide more accurate estimation than a single cosine similarity measure. Experiments conducted on the DUC2005 and DUC2006 data sets show that the proposed learning approach is successful in measuring similarity. Its competitiveness and adaptability are also demonstrated.

LDA-Based Topic Formation and Topic-Sentence Reinforcement for Graph-Based Multi-document Summarization

In recent years graph-based ranking algorithms have attracted much attention in document summarization. This paper introduces our recent work on applying a topic model, namely LDA, in graph-based summarization. In the proposed approach, LDA is used to automatically identify a set of semantic topics from the documents to be summarized. The identified topics are then used to construct a bipartite graph to represent the documents. Topic-sentence reinforcement is implemented to calculate the salience scores of topics and sentences simultaneously. By incorporating the information embedded in the topics, the sentence ranking result can be improved. Experiments are conducted on the DUC 2004 data set to evaluate the effectiveness of the proposed approach.