Unsupervised Approach to Extracting Problem Phrases from User Reviews of Products
Elena Viktorovna Tutubalina, Vladimir Vladimirovich Ivanov · 2014
This paper describes an approach to problem phrase extraction from texts that contain user experience with products.In contrast to other works, we propose a straightforward approach to problem phrase extraction based on syntactic and semantic connections between a problem indicator and mentions about the problem targets.In this paper, we discuss (i) grammatical dependencies between the target and the problem indicators and (ii) a number of domain-specific targets that were extracted using problem phrase structure and additional world knowledge.The algorithm achieves an average F1-measure of 77%, evaluated on reviews about electronic and automobile products.