Skip to content

VuBacktracking/Deep-Neural-Network-Vietnamese-Student-Feedback-Sentiment-Analysis

Repository files navigation

Vietnamese Student Feedback Sentiment Analysis Project


In this Project, I have used the Corpus named Vietnamese Student Feedback Corpus, which belongs to The UIT NLP Group and has more than 16,000 pieces of feedback from Vietnamese students, to solve the Sentiment Analysis task. I have trained some models and chose the best model with the result in the following table.\

Note: You can download the corpora in https://nlp.uit.edu.vn/

Model used:

Model Feature Precision Recall F1-Score Accuracy
LSTM Word2Vec 90.0 89.7 89.8 89.7

Word2Vec

workflow


Web Application

We have built a web application for Analyzing the Sentiment of Vietnamese Student's Feedback

Corpus Analysis

You can browse the corpus which contains feedback like 4 corpora in the Data/testForApp directory. The application will predict the sentiment for each feedback, and then display the data frame of the result and the chart for analysis.

workflow

workflow

workflow

Feedback's Sentiment Analysis

You can input Vietnamese feedback about schools and universities including lecturers, facilities, curriculum, etc, the application will predict the sentiment and display the polarity beside.

workflow

workflow


How to use?

Clone the project

git clone https://github.com/VuBacktracking/Deep-Neural-Network-Vietnamese-Student-Feedback-Sentiment-Analysis.git

Install required packages

pip install -r requirements.txt

Generate model (optional)

We already ran and saved the model lstm_model.h5 in models directory

#run this file to generate the models
lstm_create_model.ipynb
#run this file to check the result of model
lstm_result.ipynb

Now run

streamlit run app.py

Future Work