Application Of Ingwersen’s Cognitive IR Theory

In the field of cognitive information retrieval, Peter Ingwersen, who is an information scientist in the royal school of library and information science in the University of Copenhagen in Denmark, has done a series of research of high systemic and integrity. His research can be thought as began in the 1990s, followed with 30 years of research from the view of cognitive to information retrieval interaction to cognitive information retrieval interaction. Finally his team established Integration of Information Seeking and Retrieval in Context model in 2005, made the cognitive information retrieval theory move towards perfect, thereby holistic cognitive view has been reached all-round development. In the study of cognitive information retrieval theory, polyrepresentation is regarded as an important theoretical achievement. Ingwersen and his team have been committed to the experimental research of the cognitive theory, with highly attention to polyrepresentation. At the same time, scholars of information retrieval, geography information system also conducted a series of experimental research related to the theory of polyrepresentation.Due to the application of theory of polyrepresentation is still in its preliminary stage, there is still huge development space in the query expansion based on polyrepresentation of user’s information need, and in the technique innovation research area. This paper summarized the origin, development and application of the cognitive theory of information retrieval through literature review, and then discussed the concept, characteristics, application and development of the theory of polyrepresentation. Under Ingwersen’s solid cognitive information retrieval theory and the theory of polyrepresentation, this paper presents an information retrieval experiment of query expansion based on polyrepresentation of users’ information need from the cognitive space of user’s perspective. Experimental environment sourced from Lemur project, which is the joint project of natural language technology institute at Carnegie Mellon University and the University of Massachusetts intelligent information retrieval center. This experiment selected 50 retrieve themes from TREC 2013 Web Track as research subjects, using 50 topics and describes as 50 benchmark queries. Relevance judgment provided by TREC was used to evaluate the effect of query expansion. Five cognitive categories of users’information need are used in this query expansion experiment, that is, objective, motivation, background knowledge, the ideal answer, and keywords.20 users are emailed to obtain the answers of these five aspects of each subject. According to the users’ answers, this experiment extracted keywords for query expansion from these five aspects, permutation and combination of extracted keywords from five aspects were used for query expansion respectively. Three smoothing methods are used in this experiment, they are Twostage, Dirichlet and Jelinek-Mercer. Evaluation tool ireval is used for analyze retrieval results of query expansion, evaluation index includes MAP, NDCG, Bpref, P@topN etc. Finally, this paper gives the evaluation of different query expansions. This experiment shows that Twostage smoothing method gives the best retrieval performance. Experimental results also show that if we want to maximum retrieval performance, five cognitive categories of users’ information need to be used in pair with smoothing method, that is, target with Dirichlet smoothing method, the ideal answer with Jelinek-Mercer smoothing method, target or user keywords with Twostage smoothing method.