From online reviews to smartwatch recommendation: An integrated aspect-based sentiment analysis framework

Abstract

In the current landscape, smartwatches have gained popularity as wearable devices thanks to their fitness tracking and health monitoring capabilities. However, the abundance of features and options has made it challenging to select the right alternative. In this regard, we propose a text analytics-based product recommender system that leverages online reviews as peers' recommendations and creates a shortlist of available alternatives based on existing users’ perceptions. It uses a pre-trained transformer-based aspect-level sentiment analysis algorithm, InstructABSA, to quantify consumer sentiments expressed in textual reviews, which are analysed using the integrated House of Quality (HoQ) and Preference Ranking Organisation Method for Enrichment Evaluation-II (PROMETHEE-II) to construct a relative performance index for the selected manufacturers. The proposed framework may assist potential customers in making well-informed purchase decisions and help manufacturers understand their relative position in the market. It also helps customers compare the alternatives concerning selected features and associated consumer perceptions. In addition, manufacturers may use it to discover their perceived strengths and weaknesses. The proposed framework is tested on a review dataset pertaining to 12 smartwatch manufacturers, and their relative ranks are proposed.