Abstract Science and technology management departments are important users of science and technology intelligence. Actively understanding the intelligence needs of science and technology management departments has become a vital aspect of providing accurate intelligence services in the era of big data. The user portrait approach enables and simplifies this process. Through multi-source data collection and analysis, intelligence users are tagged with labels to describe their characteristics and needs, and recommendations are generated accordingly. This paper uses five methods related to natural language processing to generate labels and extract keywords from text: direct extraction, word pair matching, subject word extraction, generation scheme based on TF-IDF, and combinatorial word processing. After generating labels, a word forest table is used to analyze their association and similarity. Collaborative filtering, common sense, tag association, and other recommendation algorithms are then employed to recommend labels for different users, and preliminary user portraits are constructed. This study’s empirical research and findings show that this set of methods can effectively outline the information needs and characteristics of science and technology management departments. Furthermore, the recommended content is illuminating for science and technology intelligence work.
Zhao Hui,Hua Bolin,He Hongwei. User Profile Tag Generation and Information Recommendations for Science and Tencnology Intelligence[J]. 情报学报, 2020, 39(11): 1214-1222.
