Product Feature-Extraction And Useful-Recommendation For Online Comments

In recent years, the no-structured text data, the online comments ushered in explosive growth with the rapid development of the Web2.0 era. The data contains a lot of useful information, it affects the users' decision and provides the chief gauge for the research and development of the product. However, due to the huge amount of data and different concerns from consumers, how to select high quality online comments, and how to extract information on all aspects of product features from the data, are particularly important.In order to find the product information which is concerned by customers, this paper proposes a product features extraction algorithm based on mutual self-extension mode. The core idea of the algorithm is to achieve the desired results through the computer self-learning. Combined with FP growth algorithm is an improvement of the model. During the iteration, the confidence coefficient of the extracted-word and the extracted-mode can insure a high precision, avoid deviating theme at the same time. At last, this method find Combination Extracted-Word by Similarity Extracted-Mode. It can reduce many feature extraction mistakes caused by word segmentation technology and part-of-speech tagging technology, and get a high precision with reducing little recall rate. Finally, the experimental results show that the algorithm has an average precision of 78.50% and an average recall of 79.81% and an average F-score of 78.97% in Chinese reviews. The product features extraction effect in English, precision is 80.22%, the recall is 72.28%, F-score is 76.04%. Precision is greatly improved compared to other similar studies in the literature.Established the features database based on the features extraction algorithm above, and usually, customers have an expectation when they read the online comments. Different products, or even the same products of different brands, these expectations are different. If the comment can meet the customer's expectations, it will be considered useful by users. Based on this consideration, using support vector machine method to achieve a text classifier, proposed nine kinds of feature vectors, and finally, achieve the comment of the usefulness machine recognition. Experimental results show that the recognition of the usefulness of the comments precision is 90.67%. The results show that the algorithm is efficient, and it proves that the user's expectations of product features is a great impact on the usefulness recognition.The research results of this paper can provide decision support for the customers of online shopping, provide a theoretical basis for the manufacturer to improve products and services and can also promote the e-commerce platform by improving online review system. The research provide great theoretical and practical significance.