2017/09/26 by Athanasios Giannakopoulos, Giannakopoulos, Athanasios, Diego Antognini +7
Computer Science · #Advanced Text Analysis Techniques #Computation and Language (cs.CL) #FOS: Computer and information sciences #Sentiment Analysis and Opinion Mining #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1709.09220
openalex publication_date 2017/09/26 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
Aspect Term Extraction (ATE) detects opinionated aspect terms in sentences or\ntext spans, with the end goal of performing aspect-based sentiment analysis.\nThe small amount of available datasets for supervised ATE and the fact that\nthey cover only a few domains raise the need for exploiting other data sources\nin new and creative ways. Publicly available review corpora contain a plethora\nof opinionated aspect terms and cover a larger domain spectrum. In this paper,\nwe first propose a method for using such review corpora for creating a new\ndataset for ATE. Our method relies on an attention mechanism to select\nsentences that have a high likelihood of containing actual opinionated aspects.\nWe thus improve the quality of the extracted aspects. We then use the\nconstructed dataset to train a model and perform ATE with distant supervision.\nBy evaluating on human annotated datasets, we prove that our method achieves a\nsignificantly improved performance over various unsupervised and supervised\nbaselines. Finally, we prove that sentence selection matters when it comes to\ncreating new datasets for ATE. Specifically, we show that, using a set of\nselected sentences leads to higher ATE performance compared to using the whole\nsentence set.\n