This repo aims to analyze a hindi conversation between an agent and customer
- Dividing the whole conversation into individual statements
- Counting frequencies of different words to get the most valuable statement
- For the purpose of sentiment analysis, positive and negative sets are created
- Weights of each word of each statement are analyzed with these sets to determine the sentiment of each statement
- For individual speaker, we analyze the change in sentiment
A separate anlaysis with the help of huggingface model named txlm-roberta-hindi-sentiment
is carried out in the sentiment_analysis.ipynb
notebook.