Group Decision-Making System with Sentiment Analysis of Discussion Chat and Fuzzy Consensus Modeling
Adilet Yerkin, Pakizar Shamoi
TL;DR
The paper tackles the problem of group decision-making when participants express preferences in natural language, proposing a fuzzy-consensus framework that combines explicit voting with sentiment and emotion analysis processed by a FIS to compute a Total Preference ($0$-$10$) for each alternative. It also introduces a human-consistent consensus measure, using expert agreement and confidence as inputs to a second FIS and quantifying overall agreement via the interquartile range (IQR). The methodology is validated on a restaurant-selection scenario, where longer discussions are not required; the results show a high-consensus outcome (IQR $=0.19$) and a robustly ranked top option reflecting group sentiment. The work advances practical, sentiment-aware GDM by bridging natural-language inputs with fuzzy reasoning, offering a scalable approach for small to mid-size decision problems and paving the way for adaptive weighting and larger-scale testing in future work.
Abstract
Group Decision-Making (GDM) plays a crucial role in various real-life scenarios where individuals express their opinions in natural language rather than structured numerical values. Traditional GDM approaches often overlook the subjectivity and ambiguity present in human discussions, making it challenging to achieve a fair and consensus-driven decision. This paper proposes a fuzzy consensus-based group decision-making system that integrates sentiment and emotion analysis to extract preference values from textual inputs. The proposed framework combines explicit voting preferences with sentiment scores derived from chat discussions, which are then processed using a Fuzzy Inference System (FIS) to compute a total preference score for each alternative and determine the top-ranked option. To ensure fairness in group decision-making, we introduce a fuzzy logic-based consensus measurement model that evaluates participants' agreement and confidence levels to assess overall feedback. To illustrate the effectiveness of our approach, we apply the methodology to a restaurant selection scenario, where a group of individuals must decide on a dining option based on brief chat discussions. The results demonstrate that the fuzzy consensus mechanism successfully aggregates individual preferences and ensures a balanced outcome that accurately reflects group sentiment.