sannketnikam / emotion-detection-in-text

This project employs emotion detection in textual data, specifically trained on Twitter data comprising tweets labeled with corresponding emotions. It seamlessly takes text inputs and provides the most fitting emotion assigned to it. This app has more than 400 visitors!

Emotion Detection in Text using Natural Language Processing

Introduction

Emotion detection in text data involves identifying the emotions expressed in textual data. This can be a challenging task since emotions are often expressed in complex and subtle ways. Natural language processing (NLP) techniques can be used to analyze text data and identify the emotions expressed in it.

The aim of this project is to develop a model that uses NLP techniques to accurately detect emotions in text data. The model can be used for sentiment analysis, customer feedback analysis, and social media monitoring. The model is trained on a dataset of text data that has been labeled with the corresponding emotions expressed in it.

Dataset

The dataset used for this project contains text data labeled with one of eight emotions: anger, disgust, fear, joy, neutral, sadness, shame and surprise. The dataset contains a total of 34795 rows.

Methodology

• The methodology used for this project involves the following steps:

1. Preprocessing the text data: The text data is preprocessed by removing stop words, punctuation, user handles and converting all text to lowercase.
2. Model training: A machine learning model is trained on the extracted features to predict the emotions expressed in the text data. The model used for this project is a Logistic Regression and MultinomialNB.
3. Model evaluation: The trained model is evaluated on the test data to measure its accuracy in detecting emotions in text data.

Results

The Logistic Regression achieved an accuracy of 62% on the data.

Installation

1. Clone the repository to your local machine:

git clone https://github.com/SannketNikam/Chatbot.git

2. Install the 'requirements.txt':

pip install -r requirements.txt

3. To run this project :

streamlit run app.py

4. It'll automatically open the Streamlit app in your default browser.

emotion-detection-in-text's Issues

Hi, I'm learning machine learning sentiment analysis and your code is a great reference, but I'm running into some issues that I can't resolve

From the code it appears that train_loss and val_loss contain only a single sentiment category label, not the full loss value. This may be due to problems with data processing or saving.
train_acc and val_acc also contain only a single sentiment category label instead of the full accuracy value. This may also be due to problems with data processing or saving. How can these problems be solved, and how do I retrain the model to improve it so that I can draw specific loss curves and accuracy curves during model training? Thank you very much for your answer.

Translated with www.DeepL.com/Translator (free version)

Training part

could you please provide training part?