Concept Extraction based on Association Linked Network

Text keywords at different semantic levels have different semantic representation abilities. Although words have been organized by semantic dictionaries (e.g. WordNet) with exact semantics, the dictionaries can not be constructed automatically by machine and there are still many words which are not included in the dictionaries. This paper proposes a novel method to automatically extract keywords of higher semantic level which named concept. According to the Association Linked Network (ALN) of webpages, the ALN of keywords (kALN) is constructed first which holds the keywords of a domain and the relations among these keywords. By analyzing graph characteristics of kALN, keywords are grouped into communities. Then drawing on Entropy and Mutual Information, concepts are extracted from each kALN community. Experimental results show that the proposed method of concept extraction is acceptable in accuracy and complexity.