Simple and Efficient Approach to the Aspect Extraction from Customers’Product Reviews

摘要

In this paper we present an approach to the one of the most popular natural language processing tasks of automatic aspect extraction from product reviews. Our approach is based on using clustering of word embeddings, morphological features, information about syntax dependencies and word frequencies. We use these features in the well-known machine learning method - the Decision tree. The primary evaluation of our method quality for the task of identifying the explicit aspects from the reviews demonstrate good performance in the precision and recall for cross-domain aspect extraction task.