Commodity feature extraction for Chinese online comments based on semantic
Hongmei Nie, Jiaqing Zhou · 2017
Opinion mining based on Chinese online comments has been widely concerned, and its goal is to analyze user's attitude towards commodities' features from massive online comments. Commodity feature extraction is the basis of opinion mining. Most existing commodity feature extraction methods cannot achieve cooperating analysis of semantic rules and commodity feature extraction, or applying statistical methods to extract the commodity features, the feature seed weights are not reasonable, or the thresholds of the candidate features are not reasonable. For these reasons, this paper proposes a method to extract commodity feature based on Chinese online comments. In this method, the semantic rules of commodity feature are more perfectly defined, which significantly reduces the noise of candidate feature set, and simplifies the subsequent features' extraction. In this paper, we also set the seed weight reasonably, and obtain the changed threshold and seed set by iterative method. Experimental results show that the proposed method can effectively and quickly perform unsupervised learning. Moreover, compared with other methods, it has better recognition performance.