Abstract—for information extraction and topic classification, this paper extracts topic words from the classified documents, and builds the topic lexicon according to the topic. Topic words are extracted from each document by pretreating the document and using the TF-IDF weight formula. The topic words are extracted by the size of weight proportionally. After processing each document uninterruptedly, the topic lexicon is built according to the topic. Experiments prove that it has good accuracy to extract topic words. The topic lexicon is easy to build, and it satisfies the needs of word segmentation of all kinds of documents. It is a new method in information extraction and text classification.
Index Terms—text classification, TF-IDF, topic lexicon, topic words.
Cite: Shouning QU, Jian Lu and Jing Li, "Research on Correlative Techniques of Building Specific Topic Lexicon," International Journal of Computer Theory and Engineering vol. 1, no. 4, pp. 394-397, 2009.