Article Preview
TopIntroduction
The overwhelming growth of textual data on the World Wide Web (WWW), both in scope as well as in depth, is leading to an unrestricted growth of the Cyberspace. Numerous websites are being created everyday and a number of pages are being added to them in a frequent manner. Moreover, the emergence of social media is vehemently proliferating user-generated content in the form of blogs, social networking sites, discussion boards and forums. It is greatly affecting the pace in which data are being accumulated on the Web. A trend of the blogosphere noticed by Bansal and Koudas (2007a) and Platakis et al. (2009) is that 175,000 new blogs including 1.6 million posts are being added to the WWW every day, and these figures are rapidly rising with time. The vast accumulation of textual data on the WWW is reflected by the large number of hits for a given user query. For example, the number of hits by Google for the query term “tag cloud generation” is about 95300, in which top 10 results were so closely related to the query that it became hard to distinguish their singularity. This huge accumulation of textual data on the Web has attracted researchers from variety of fields including data mining, information retrieval, natural language processing, and visualization. These researchers are interested in mining information components to conceptualize huge repository of text documents. However, the prime focus of all these tasks remained as summarizing textual data using various techniques ranging from keyphrase extraction (Abulaish & Anwar, 2012) to question-answering (Light et al., 2001) to sentiment analysis and opinion mining (Pang & Lee, 2008).
Comprising one or more words, keyphrases (a.k.a. keywords) of a document provide a brief summary of its content. Moreover, a quick random navigation through the keyphrases of a document leads to an overall grasp of the information contained in the document and makes the users free from the tedious task of manually exploring the complete content. Due to existence of large document collection in the form of digital libraries, text databases and textual data on the WWW, the value of such summary information has increased significantly. In addition to tag cloud generation, keyphrases are useful to various applications, such as document indexing and retrieval (Jones & Staveley, 1999), document summarization (Zha, 2002; Kupiec et al., 1995), thesaurus construction (Kosovac et al., 2000), and document categorization and clustering (Han et al., 2007; Jhones & Mahoui, 2000). Keyphrases are also very useful for digital libraries and web search engines. In digital libraries, the keyphrases of a scientific paper can help users to get a rough sense of the paper (Gutwin et al., 1999), whereas in web search the keyphrases can serve as metadata to index and retrieve webpages (Li et al., 2004). In addition, keyphrases can help users to comprehend the content of a collection without navigating through the pile of documents. Keyphrases are also helpful to expand user queries, facilitate document skimming by visually emphasizing important phrases, and they offer a powerful means of measuring document similarity that can be exploited to group documents into various categories.