2019/09/01 by Giannis Karamanolakis, Karamanolakis, Giannis, Daniel Hsu +3
Computer Science · #Advanced Text Analysis Techniques #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Sentiment Analysis and Opinion Mining #Text and Document Classification Technologies #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1909.00415
openalex publication_date 2019/09/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
User-generated reviews can be decomposed into fine-grained segments (e.g.,\nsentences, clauses), each evaluating a different aspect of the principal entity\n(e.g., price, quality, appearance). Automatically detecting these aspects can\nbe useful for both users and downstream opinion mining applications. Current\nsupervised approaches for learning aspect classifiers require many fine-grained\naspect labels, which are labor-intensive to obtain. And, unfortunately,\nunsupervised topic models often fail to capture the aspects of interest. In\nthis work, we consider weakly supervised approaches for training aspect\nclassifiers that only require the user to provide a small set of seed words\n(i.e., weakly positive indicators) for the aspects of interest. First, we show\nthat current weakly supervised approaches do not effectively leverage the\npredictive power of seed words for aspect detection. Next, we propose a\nstudent-teacher approach that effectively leverages seed words in a\nbag-of-words classifier (teacher); in turn, we use the teacher to train a\nsecond model (student) that is potentially more powerful (e.g., a neural\nnetwork that uses pre-trained word embeddings). Finally, we show that iterative\nco-training can be used to cope with noisy seed words, leading to both improved\nteacher and student models. Our proposed approach consistently outperforms\nprevious weakly supervised approaches (by 14.1 absolute F1 points on average)\nin six different domains of product reviews and six multilingual datasets of\nrestaurant reviews.\n