Today, people first make their complaints and compliments on internet about a product which they use or a company they are a customer of. Therefore, when they are going to buy a new product, they first analyze the complaints made by other users of the product. These complaints play an important role in helping people make decision of purchasing or not purchasing product. It is impossible to analyze online complaints manually due to the huge data size. However, companies are still losing a lot of time by analyzing and reading thousands of complaints one by one. In this article, online text based customer complaints are analyzed with Latent Dirichlet Allocation (LDA), GenSim LDA, Mallet LDA and Gibbs Sampling for Dirichlet Multinomial Mixture model (GSDMM) and the performances of them are compared. It is observed that GSDMM gives much more successful results than LDA. The obtained topics of the complaints are presented to users with a mobile application developed in React Native. With the developed application not only the customers will be able to see the topics of complaint from the application interface but also the companies will be able to view the distribution and statistics of the topics of complaints.
Topic modelling latent dirichlet allocation gibbs sampling gibbs sampling for dirichlet multinomial mixture natural language processing
TÜBİTAK
1919B011902805
Thanks to TÜBİTAK for their support to the project numbered 1919B011902805 within the scope of TÜBİTAK-2209-A University Students Research Projects Support Program 2019/2.
1919B011902805
Primary Language | English |
---|---|
Subjects | Artificial Intelligence |
Journal Section | Araştırma Articlessi |
Authors | |
Project Number | 1919B011902805 |
Publication Date | July 30, 2021 |
Published in Issue | Year 2021 Volume: 9 Issue: 3 |
All articles published by BAJECE are licensed under the Creative Commons Attribution 4.0 International License. This permits anyone to copy, redistribute, remix, transmit and adapt the work provided the original work and source is appropriately cited.