DoSSAD - Leveraging domain specific semantics in aspect detection for product design
Publication Date
2-1-2020
Document Type
Conference Proceeding
Publication Title
2020 IEEE 14th International Conference on Semantic Computing (ICSC)
DOI
10.1109/ICSC.2020.00052
First Page
249
Last Page
252
Abstract
Accurately identifying customer requirements is an essential activity in product design and development. Particularly, the development lifecycle of mechanical products is long-term, and enhancements or upgrades are challenging to execute. The growth of online product reviews paves way for a data-driven approach to capture customer experiences with existing products and their sentiments towards specific features of products, thereby enabling faster product development cycles. Aspect Detection is becoming an increasingly popular way to identify the feature of a subject expressed by a segment of text. Through this work, we propose to study Aspect Detection as an essential step towards the qualitative analysis of customer reviews. Product design and development is often a focused domain, where one cannot set generic product guidelines. Therefore, our study focuses specifically on eco-friendly products as an example of a product category. Existing research in aspect detection has often leveraged huge volumes of text data with a wide vocabulary to provide generic aspect detection. Through this paper, we study the impact of domain-specific semantics on aspect detection, with a focus on product development and design as an application. We present an extensive evaluation of aspect detection models trained using datasets with varying amounts of domain-specific data (i.e., product-category-specific data) during the training phase.
Keywords
Aspect extraction, Aspect-based sentiment analysis, Natural language processing, Word embedding
Department
Computer Engineering
Recommended Citation
Aashay Mokadam and Mahima Agumbe Suresh. "DoSSAD - Leveraging domain specific semantics in aspect detection for product design" 2020 IEEE 14th International Conference on Semantic Computing (ICSC) (2020): 249-252. https://doi.org/10.1109/ICSC.2020.00052