Project Overview: In this project, I developed a sentiment analysis model to enhance chatbot interactions by determining user satisfaction levels and identifying when human intervention was needed. The goal was to improve customer support efficiency by ensuring chatbots could escalate conversations at the right moment, reducing frustration and enhancing user experience.
Key Contributions & Approach:
- Data Collection & Preprocessing:Worked with large-scale chatbot conversation logs containing sentiment-labeled text.Used NLP techniques in R (tidytext, syuzhet, text2vec) for text tokenization, stopword removal, and sentiment classification.
- Sentiment Analysis & Machine Learning: Applied Bing and NRC lexicons to analyze user emotions and detect sentiment shifts. Built and optimized machine learning models (Naïve Bayes, SVM, and LSTM networks) to predict escalation points where human intervention was required.
- Improved sentiment classification accuracy by 15%, leading to a significant reduction in chatbot escalation errors.
- Real-Time Monitoring & Visualization: Created an interactive Tableau dashboard to monitor chatbot engagement trends and sentiment patterns. Integrated SQL databases for efficient chatbot data storage and retrieval.
- Business Impact:Reduced chatbot misclassification errors, ensuring seamless customer experiences.
- Enabled proactive customer service improvements, enhancing chatbot efficiency and reducing unnecessary escalations.
Technologies Used:
- R (tidytext, tm, quanteda, ggplot2, dplyr)
- Python (NLTK, Scikit-learn, pandas, NumPy, Matplotlib)
- SQL (Chatbot data storage & retrieval)
- Tableau / Power BI (Sentiment tracking & visualization)
Skills Utilized:
- Natural Language Processing (NLP)
- Sentiment Analysis & Machine Learning
- Chatbot Performance Optimization
- Data Preprocessing & Feature Engineering
- Data Visualization & Reporting