An NTU-Approach to Automatic Sentence Extraction for Summary Generation

AN NTU-APPROACH TO AUTOMATIC SENTENCE EXTRACTION FOR SUMMARY GENERATION Kuang-hua Chert Language & Information Processing System Lab. (LIPS) Department of Library and Information Science National Taiwan University 1, SEC. 4, Roosevelt RD., Taipei TAIWAN, 10617, R.O.C. E-mail: khchen @ ccms.ntu.edu.tw Fax: +886-2-23632859 Sheng-Jie Huang, Wen-Cheng Lin and Hsin-Hsi Chen Natural Language Processing Laboratory (NLPL) Department of Computer Science and Information Engineering National Taiwan University 1, SEC. 4, Roosevelt RD., Taipei TAIWAN, 10617, R.O.C. E-mail: { sjhuang,denislin } @nlg.csie.ntu.edu.tw, hh_chen @csie.ntu.edu.tw Fax: +886-2-23638167 ABSTRACT Automatic summarization and information extraction are two important Internet services. MUC and SUMMAC play their appropriate roles in the next generation Internet. This paper focuses on the automatic summarization and proposes two different models to extract sentences for summary generation under two tasks initiated by SUMMAC-1. For categorization task, positive feature vectors and negative feature vectors are used cooperatively to construct generic, indicative summaries. For adhoc task, a text model based on relationship between nouns and verbs is used to filter out irrelevant discourse segment, to rank relevant sentences, and to generate the user-directed summaries. The result shows that the NormF of the best summary and that of the fixed summary for adhoc tasks are 0.456 and 0.447. The NormF of the best summary and that of the fixed summary for categorization task are 0.4090 and 0.4023. Our system outperforms the average system in categorization task but does a common job in adhoc task. 1. INTRODUCTION Towards the end of the 20 th century, the Internet has become a part of life style. People enjoy Internet services from various providers and these ISPs (Internet Services Providers) do their best to fulfill users' information need. However, if we investigate the techniques used in these services, we will find out that they are not different from those used in traditional Information Retrieval or Natural Language Processing. However, the cyberspace provides us an environment to utilize these techniques to serve more persons than ever before. The members under the leadership of Professor Hsin-Hsi Chen of Natural Language Processing Lab. (NLPL) in Department of Computer Science and Information Engineering, National Taiwan University have dedicated themselves in researches of NLP for many years. The research results have been reported in literature and received the reputation from colleagues of NLP field. Many systems for various NLP applications have been developed, especially for Chinese and English. Some systems could be accessed directly via WWW browsers. For example, an MT meta-server [1] provides an online English-to-Chinese translation service. (http://nlg3. csie. ntu.edu.tw/mtir/mtir.html) Language & Information Processing System Lab. (LIPS) in Department of Library and Information Science, National Taiwan University also devotes itself in researches of language, information and library sciences. Chen and Chen [2] proposed hybrid model for noun extraction from running texts and provided an automatic evaluation method. Chen [3] proposed a cowus-based model to identify topics and used it to determine sub-topical structures. 163 Generally speaking, we are capable of dealing with numerous NLP applications or apply NLP techniques to other applications using our …

Content truncated to 500 words