A TEXT FEATURE WORD EXTRACTION METHOD APPLIED TO ENTERPRISE COMPETITIVE INTELLIGENCE SYSTEM

To acquire industry feature words, professionals need to collect and analyze the feature word sets from industry sites according to their experience and finally merge the feature word sets at different sites to form industry feature word sets. This method is characterized by a large workload, the difficulty in ensuring the accuracy of text classification, and the necessity to adjust feature words by repeating the above process. To solve the above problems, a universal feature word extraction scheme and a system framework were first proposed in this study based on the actual requirements of the enterprise competitive intelligence system. Then, the key problems involved in the process of feature word extraction were elaborated in detail. Finally, the traditional feature word weight was improved on basis of predecessors' results, and the Sogou lexicon was introduced to correct the word frequency and part of speech. A good classification effect was achieved through experiments with the KNN classifier, which was verified using the vector space model (VSM), and a high average F1 value was acquired.