chatbot-analysis 0.3

No project description provided

Sentiment and Query Intent Analysis Library

Overview

This library provides tools for sentiment analysis and query intent detection using a fine-tuned BERT-based model to improve the emotional intelligence of your chatbot. The library enables sentiment classification at both sentence and document levels, as well as intent categorization most suited for educational chatbot conversations.

Features

✅ Sentence-Level Sentiment Analysis - Classifies individual sentences as Positive, Neutral, or Negative.
✅ Document-Level Sentiment Analysis - Aggregates multiple sentences and determines the overall sentiment.
✅ Query Intent Detection - Identifies the intent behind a query, such as course information, class schedule, greetings ✅ Chat Summary Generation - Provides a summary table insights into chat sentiment and intent distribution.
✅ Conversation Conversion & Analysis - Converts chat logs initially in a json format into a csv format structured data with sentiment, intent, and response time analysis to be used to generate a dynamic dashboard on a website link: https://chatbot-dashboard-analysis.streamlit.app/ Keys needed in the chat log: _id,Person,stime_text,stime_timestamp,last_interact_text,last_interact_timestamp,llm_deployment_name,llm_model_name,vectorstore_index,overall_cost,overall_tokens,role,content,recorded_on_text,recorded_on_timestamp,token_cost,tokens,user_sentiment,query_intent,conversation_id,previous_query_intent,overall_chat,chat_sentiment,chatbot_response_time,overall_conversation_time

---

Installation

Ensure you have the necessary dependencies installed:

pip install torch transformers pandas tabulate

---

Usage

1️⃣ Sentence-Level Sentiment Analysis

Determine sentiment (Positive, Neutral, Negative) for a given sentence. This can return the sentiment of the user.

from sentiment_analysis import sentiment_sentence

sentence = "I love this course!"
sentiment = sentiment_sentence(sentence)
print(sentiment)  # Output: Positive

2️⃣ Document-Level Sentiment Analysis

Analyze sentiment for a collection of sentences or a document. This gives back the sentiment at class level. The input should be in a dictionary format where the keys are the sentence number and values are the content

from sentiment_analysis import sentiment_document

document = {"sentence1": "The class was great!", "sentence2": "I learned a lot."}
doc_sentiment = sentiment_document(document)
print(doc_sentiment)  # Output: Positive

3️⃣ Query Intent Detection

Classify the intent behind a query ("Course Overview and Information", "Course Assessment" , "Checking Announcement", "Request for Permission", "Learning Course Content", "Class Schedule", "Greetings", "Ending", "Casual Chat" , "No Query")

from sentiment_analysis import query_intent

query = "Can I get more details on the assessment criteria?"
intent = query_intent(query)
print(intent)  # Output: Course Assessment

4️⃣ Generate Chat Sentiment & Intent Summary

Summarize sentiment and intent for multiple sentences. Use chat_summary if you want query intent to be included.

from sentiment_analysis import chat_summary
from sentiment_analysis import sentiment_summary

document = {
    "sentence1": "I love this class!",
    "sentence2": "When is the next assignment due?"
}
summary = chat_summary(document)
summary = sentiment_summary(document)
print(summary)

5️⃣ Analyze and Convert Chat Logs

Convert chatbot logs into structured data with sentiment, intent, and response time analysis.

from sentiment_analysis import conversion

chatlog = {
    "chatlog": [
        {"_id": {"$oid": "1"}, "Person": "User1", "messages": [
            {"role": "user", "content": "I'm very happy with the course!"},
            {"role": "ai", "content": "Glad to hear that!"}
        ]}
    ]
}

df = conversion(chatlog, num_of_chats=1)
print(df.head())

---

Query Intent Categories

The query_intent function categorizes queries into the following:

Category	Description
Course Overview and Information	General course-related inquiries
Course Assessment	Questions about exams, assignments
Checking Announcement	Checking for latest updates
Request for Permission	Asking for special permissions not available in the bector database
Learning Course Content	Questions about materials or topics
Class Schedule	Checking class timings
Greetings	Saying "hello" or "goodbye"
Ending	Ending a conversation
Casual Chat	Non-academic casual discussions
No Query	No identifiable query present

---

Sentiment Classification Categories

The library classifies sentiment into:

• Positive 😊
• Neutral 😐
• Negative 😡

---

License

This project is licensed under the MIT License.

Author

Developed by Chalamalasetti Sree Vaishnavi. For questions or contributions, please contact [Your Email].