Implicit Feedback Mining for Recommendation

Social media creates valuable feedback, either explicitly or implicitly, which can be used to develop an effective recommendation. Explicit feedback, like rating, allows users to explicitly express their preference on items. However, the reluctance of users to provide explicit feedback makes it difficult to get sufficient and representative explicit feedback. In contrast, implicit feedback has the advantage of being collected at much lower cost, in much larger quantities, and without burden on users. Thus, we mine the implicit feedback, including tweets and tags, to provide virtual rating and user similarity for recommendation. Taking the factor that tweets reflect users' sentiment on some item into consideration, we use sentiment analysis score as the virtual rating, and propose a Weighted Semantic Tag Similarity Method (WSTSM) to get user similarity. Experimental on a real SINA microblog dataset demonstrates that our method outperforms the traditional PMF in terms of RMSE by 8.55% due to the informative implicit feedback embedded in tweets and tags.