Automatic Keyphrase Extraction by Bridging Vocabulary Gap

Proceedings of the Fifteenth Conference on Computational Natural Language Learning, pages 135–144, Portland, Oregon, USA, 23–24 June 2011. c⃝2011 Association for Computational Linguistics Automatic Keyphrase Extraction by Bridging Vocabulary Gap ∗ Zhiyuan Liu, Xinxiong Chen, Yabin Zheng, Maosong Sun State Key Laboratory of Intelligent Technology and Systems Tsinghua National Laboratory for Information Science and Technology Department of Computer Science and Technology Tsinghua University, Beijing 100084, China {lzy.thu, cxx.thu, yabin.zheng}@gmail.com, sms@tsinghua.edu.cn Abstract Keyphrase extraction aims to select a set of terms from a document as a short summary of the document. Most methods extract keyphrases according to their statistical prop- erties in the given document. Appropriate keyphrases, however, are not always statis- tically significant or even do not appear in the given document. This makes a large vocabulary gap between a document and its keyphrases. In this paper, we consider that a document and its keyphrases both describe the same object but are written in two different languages. By regarding keyphrase extraction as a problem of translating from the language of documents to the language of keyphrases, we use word alignment models in statistical machine translation to learn translation proba- bilities between the words in documents and the words in keyphrases. According to the translation model, we suggest keyphrases giv- en a new document. The suggested keyphrases are not necessarily statistically frequent in the document, which indicates that our method is more flexible and reliable. Experiments on news articles demonstrate that our method outperforms existing unsupervised methods on precision, recall and F-measure. 1 Introduction Information on the Web is emerging with the development of Internet. It is becoming more and more important to effectively search and manage information. Keyphrases, as a brief summary of a document, provide a solution to help organize and ∗Zhiyuan Liu and Xinxiong Chen have equal contribution to this work. retrieve documents, which have been widely used in digital libraries and information retrieval (Turney, 2000; Nguyen and Kan, 2007). Due to the explosion of information, it is ineffective for professional human indexers to manually annotate documents with keyphrases. How to automatically extract keyphrases from documents becomes an important research problem, which is usually referred to as keyphrase extraction. Most methods for keyphrase extraction try to extract keyphrases according to their statistical prop- erties. These methods are susceptible to low perfor- mance because many appropriate keyphrases may not be statistically frequent or even not appear in the document, especially for short documents. We name the phenomenon as the vocabulary gap between documents and keyphrases. For example, a research paper talking about “machine transliteration” may less or even not mention the phrase “machine translation”. However, since “machine transliter- ation” is a sub-field of “machine translation”, the phrase “machine translation” is also reasonable to be suggested as a keyphrase to indicate the topics of this paper. Let us take another example: in a news article talking about “iPad” and “iPhone”, the word “Apple” may rarely ever come up. However, it is known that both “iPad” and “iPhone” are the products …

Content truncated to 500 words