An Aspect-Sentiment Pair Extraction Approach Based on Latent Dirichlet Allocation

Abstract

Online user reviews have a great influence on decision-making process of customers and product sales of companies. However, it is very difficult to obtain user sentiments among huge volume of data on the web consequently; sentiment analysis has gained great importance in terms of analyzing data automatically. On the other hand, sentiment analysis divides itself into branches and can be performed better with aspect level analysis. In this paper, we proposed to extract aspect-sentiment pairs from a Turkish reviews dataset. The proposed task is the fundamental and indeed the critical step of the aspect level sentiment analysis. While extracting aspect-sentiment pairs, an unsupervised topic model Latent Dirichlet Allocation (LDA) is used. With LDA, aspect-sentiment pairs from user reviews are extracted with 0.86 average precision based on ranked list. The aspect-sentiment pair extraction problem is first time realized with LDA on a realworld Turkish user reviews dataset. The experimental results show that LDA is effective and robust in aspect-sentiment pair extraction from user reviews.

Automatic Detection of the Topics in Customer Complaints with Artificial Intelligence

“…The basic idea behind LDA is this: topics have probability distribution over a fixed vocabulary and documents are composed of random mixture of latent topics. While the input of the model is documents, its outputs are the topics, probabilities of words under these topics, topic for each word, and the topic mixture for each document [9].…”

A Survey on Opinion Mining: From Stance to Product Aspect

“…Therefore, This research topic attracts many researchers and has been extensively explored in last decade. Previous works could be generally divided into three groups which are corpus level aspect extraction [53]- [57], [61], [64]- [68], [129]- [134], corpus level aspect and opinion mining [51], [58]- [60], [62], [63], [69]- [71], [101]- [103], [113], [118]- [120], [135]- [137] and document/sentence level aspect and opinion tagging [72], [74]- [78], [80]- [82], [84]- [86], [88]- [90], [138], [139], [144], [145]. For the former two categories, rule-based methods and unsupervised based methods are commonly used, and various supervised learning models (e.g.…”

“…Methodologies for corpus level aspect and opinion mining could be categorized as frequency and relation based approaches [51], [58]- [60], [101], [103], [135], [137] and topic model based approaches [62], [63], [69]- [71], [102], [113], [118]- [120], [136]. Table 7 describes the approaches together with their core techniques and performances.…”

“…And many various of LDA [148] have the ability to perform corpus level aspect extraction and opinion mining in parallel. In Table 7, last eleven rows describe the topic model based approaches [62], [63], [69]- [71], [102], [113], [118]- [120], [136] for corpus level aspect and opinion mining.…”

“…Differs from mining approaches above, some efforts have been made to model aspect and opinion pair jointly, and the extracted topics are in the form of 'Electronic reviews'. Ekinci [69] incorporates a ranking mechanism into LDA and proposes an approach for aspect-sentiment pair extraction in Turkish language corpus. To use complementary information provided by different type online media, Zuo [71] propose a cross-collection topic modeling approach for aspect and opinion mining.…”