Textual consumer input is crucial in the automotive sector for locating and fixing car problems as well as raising customer happiness. The manual analysis has become infeasible due to the overwhelming volume of feedback. We propose an innovative topic modelling approach utilizing a W&G-BERT model for automotive entity recognition. This approach provides an advantage as we exclusively cluster only based on entities that hold significance within the given use case, therefore the presence of spelling, semantic, and punctuation errors within sentences does not impact the computation. For validation, a semi-quantitative evaluation was carried out by involving six experts specializing in automotive product quality. We achieved a cluster correctness score of 90 % based on their expert analysis.
Leveraging Entity Recognition for Automotive Customer Feedback Topic Modeling
2024-03-22
1640506 byte
Conference paper
Electronic Resource
English
A Hybrid Method for Automotive Entity Recognition
British Library Conference Proceedings | 2021
|A Hybrid Method for Automotive Entity Recognition
SAE Technical Papers | 2021
|Transportation Research Record | 2024
|MOVING ENTITY RECOGNITION METHOD, MOVING ENTITY RECOGNITION SYSTEM, AND ENTITY RECOGNITION EQUIPMENT
European Patent Office | 2016
|