Viewing morphology as an inference process
SIGIR '93 Proceedings of the 16th annual international ACM SIGIR conference on Research and development in information retrieval
Approximating a collection of frequent sets
Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining
TSP: Mining top-k closed sequential patterns
Knowledge and Information Systems
Summarizing itemset patterns: a profile-based approach
Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining
Generating semantic annotations for frequent patterns with context analysis
Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining
Hi-index | 0.00 |
Discovering a representative set of theme patterns from a large amount of text for interpreting their meaning has always been concerned by researches of both data mining and information retrieval. Recent studies of theme pattern mining have paid close attention to the problem of discovering a set of compatible top-k theme patterns with both high-interestingness and low-redundancy. Since different users have different preferences on interestingness and redundancy, how to measure the attributes of the users' preferences, and thereby to discover "preferred compatible top-k theme patterns" (PCTTP) is urgent in the field of text mining. In this paper, a novel strategy of discovering PCTTP based on users' preferences in text mining is proposed. Firstly, an evaluation function of the preferred compatibility between every two theme patterns is presented. Then the preferred compatibilities are archived into a data structure called theme compatibility graph, and a problem called MWSP based on the compatibility graph is proposed to formulate the problem how to discover the PCTTP. Secondly, since MWSP is proved to be a NP-Hard problem, a greedy algorithm, DPCTG, is designed to approximate the optimal solution of MWSP. Thirdly, a quality evaluation model is introduced to measure the compatibility of discovering theme patterns. Empirical studies indicate that a high quality set of PCTTP on four different sub text sets can be obtained from DBLP.