Effective Complaint Detection in Financial Services through Complaint, Severity, Emotion and Sentiment Analysis

Sentiment analysis plays a crucial role in understanding customer feedback and opinions within the financial domain. This research addresses the problem of sentiment-aware complaint identification using the FINancial Complaint CORpus (FINCORP) dataset, which provides annotated complaints along with sentiment, emotion and severity labels. The dataset encompasses various sentiment classes including Negative, Neutral, and Positive, enabling a wide understanding of customer sentiment towards financial products and services. Leveraging natural language processing (NLP) techniques such as text classification and sentiment analysis, this study aims to develop models capable of accurately detecting and categorizing sentiment, emotion of customers and the severity level of financial complaints. Effectively analyzing these factors allows financial institutions to gain valuable insights into customer satisfaction levels, pinpointing areas for improvement and enhance both customer experience and service quality. This research contributes to the advancement of sentiment analysis methodologies in the financial domain, facilitating better decision-making and customer relationship management strategies. DistilBERT, Glove and RoBERTa has outperformed in classifying the sentiment, emotion and severity of complaints, achieving F1-scores of 0.71, 0.74, and 0.86, respectively using logistic regression and Bi-LSTM classifiers.