Extracting Product Features from Online Reviews for Sentimental Analysis

For elaborately understanding what product features the reviews focuses on, we propose an approach based on patterns to extraction features (titles). Trough setting length, upper and lower limit probability and frequency thresholds, we extract patterns of POS tags and features from the training corpus. To enhance adaptability of the pattern set, we merge some fundamental patterns into a new fuzzy pattern. Then a pattern matching algorithm is applied to extract the titles and opinion words from the reviews. We conducted a platform to extract features from product reviews automatically, the result of our experiments shows that our approach is effective.