Litcius/Paper detail

Topic Modeling on Customer Feedback from an Online Ticketing System using Latent Dirichlet Allocation and BERTopic

Charmaine S. Ponay

202210 citationsDOI

Abstract

This project aims to use Topic Modeling on Customer Feedback from an Online Ticketing System using Latent Dirichlet Allocation and BERTopic. The experiment started with analysis of the Topics generated from a base LDA model and computing its coherence score and fine-tuning the LDA model and comparing the coherence score with the base Model. It was found that the fine-tuned LDA model increased the cohesion score by 8.33%. This project also explored the use of BERTopic, using the multilingual model and found that it was able to extract Tagalog words from the data composed of English and Tagalog. As such, BERTopic can be further experimented on using data with more relevant Tagalog words. Also, using human judgements or “eye-balling” techniques, it was able to evaluate that the topic distribution generated by both the LDA and BERTopic model was able to capture the information needed by the organization.

Topics & Concepts

Latent Dirichlet allocationComputer scienceTagalogTopic modelCohesion (chemistry)Natural language processingArtificial intelligenceCoherence (philosophical gambling strategy)Dirichlet distributionInformation retrievalStatisticsLinguisticsMathematicsChemistryBoundary value problemMathematical analysisOrganic chemistryPhilosophyAdvanced Text Analysis TechniquesCustomer churn and segmentationTechnology Adoption and User Behaviour
Topic Modeling on Customer Feedback from an Online Ticketing System using Latent Dirichlet Allocation and BERTopic | Litcius