Browsing by Author "Rodrigues, Vasco Veiga Martins"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Aplicação de técnicas de Clustering ao contexto da Tomada de Decisão em GrupoPublication . Rodrigues, Vasco Veiga Martins; Marreiros, Maria Goreti CarvalhoNowadays, decisions made by executives and managers are primarily made in a group. Therefore, group decision-making is a process where a group of people called participants work together to analyze a set of variables, considering and evaluating a set of alternatives to select one or more solutions. There are many problems associated with group decision-making, namely when the participants cannot meet for any reason, ranging from schedule incompatibility to being in different countries with different time zones. To support this process, Group Decision Support Systems (GDSS) evolved to what today we call web-based GDSS. In GDSS, argumentation is ideal since it makes it easier to use justifications and explanations in interactions between decision-makers so they can sustain their opinions. Aspect Based Sentiment Analysis (ABSA) is a subfield of Argument Mining closely related to Natural Language Processing. It intends to classify opinions at the aspect level and identify the elements of an opinion. Applying ABSA techniques to Group Decision Making Context results in the automatic identification of alternatives and criteria, for example. This automatic identification is essential to reduce the time decision-makers take to step themselves up on Group Decision Support Systems and offer them various insights and knowledge on the discussion they are participants. One of these insights can be arguments getting used by the decision-makers about an alternative. Therefore, this dissertation proposes a methodology that uses an unsupervised technique, Clustering, and aims to segment the participants of a discussion based on arguments used so it can produce knowledge from the current information in the GDSS. This methodology can be hosted in a web service that follows a micro-service architecture and utilizes Data Preprocessing and Intra-sentence Segmentation in addition to Clustering to achieve the objectives of the dissertation. Word Embedding is needed when we apply clustering techniques to natural language text to transform the natural language text into vectors usable by the clustering techniques. In addition to Word Embedding, Dimensionality Reduction techniques were tested to improve the results. Maintaining the same Preprocessing steps and varying the chosen Clustering techniques, Word Embedders, and Dimensionality Reduction techniques came up with the best approach. This approach consisted of the KMeans++ clustering technique, using SBERT as the word embedder with UMAP dimensionality reduction, reducing the number of dimensions to 2. This experiment achieved a Silhouette Score of 0.63 with 8 clusters on the baseball dataset, which wielded good cluster results based on their manual review and Wordclouds. The same approach obtained a Silhouette Score of 0.59 with 16 clusters on the car brand dataset, which we used as an approach validation dataset.