Explainable Sentiment Analysis Application for Social Media Crisis Management in Retail

Abstract

Sentiment Analysis techniques enable the automatic extraction of sentiment in social media data, including popular platforms as Twitter. For retailers and marketing analysts, such methods can support the understanding of customers' attitudes towards brands, especially to handle crises that cause behavioural changes in customers, including the COVID-19 pandemic. However, with the increasing adoption of black-box machine learning-based techniques, transparency becomes a need for those stakeholders to understand why a given sentiment is predicted, which is rarely explored for retailers facing social media crises. This study develops an Explainable Sentiment Analysis (XSA) application for Twitter data, and proposes research propositions focused on evaluating such application in a hypothetical crisis management scenario. Particularly, we evaluate, through discussions and a simulated user experiment, the XSA support for understanding customer's needs, as well as if marketing analysts would trust such an application for their decision-making processes. Results illustrate the XSA application can be effective in providing the most important words addressing customers sentiment out of individual tweets, as well as the potential to foster analysts' confidence in such support.

An investigation into the deep learning approach in sentimental analysis using graph-based theories

“…Transparency: Our model is characterised by a transparent prediction generation process, this includes the earlier conceptual stages (i.e., Figs 12 and 13 ) followed by a visual data distribution and the impact of the proposed techniques on best adjusting the decision boundary for sentiment classification (Figs 31 , 32 and 33 ). As opposite to the classical classifiers [ 102 ], the proposed DNN structure allows different approximations of the problem (i.e., polarity, subjectivity, frequency, etc), that enables a global observation of the SA over all the news’ stations. The compliance of the backward selection method with backpropagation algorithm (see: “Features’ selection via attention scoring”, “Improving DNN performance via a deterministic backward walk”) does not require any additional training examples or hidden layers as the case in [ 103 ], which allowed the model complexity to be restricted to the embedded space.…”

Classificação Automática de Petições Iniciais Usando Classificadores Combinados

“…No presente estudo retiraram-se os espac ¸os em branco, pontuac ¸ões, símbolos financeiros, caracteres especiais e as stop words. A retirada das stop words é uma etapa importante, visto que reduz o espac ¸o de busca e melhora significativamente o processo de aprendizagem do algoritmo [Sousa et al 2019], e, até mesmo, a explicabilidade [Cirqueira et al 2020]. Para representac ¸ão dos dados optou-se pelo Term Frequency -Inverse Document Frequency (TF-IDF) dada a sua simplicidade e vasta aplicabilidade como em [Sousa et al 2019, Cirqueira et al 2020].…”

Análise de discursos em notícias sobre homofobia, racismo e sexismo em comentários de portais brasileiros de notícias

“…A análise de sentimentos efetua a classificação de um conteúdo textual de acordo com a polaridade que o mesmo representa diante do documento ou texto [7]. As polaridades estão divididas em três principais, negativo, neutro e positivo [30].…”

“…https://www.nltk.org/7 Figuras ou símbolos que caracterizam uma palavra ou ideia 8. https://polyglot.readthedocs.io/en/latest/…”