About Project

HearMePal is a web-based chatbot application designed to provide informative and supportive mental health support using Natural Language Processing (NLP) and Neural Network approaches.

This application is built using Flask as the backend, Tailwind CSS as the frontend, and NLTK for text data preprocessing. The dataset is specifically compiled with the help of psychology major graduates, so that the content delivered is relevant, ethical, and responsible.

HearMePal is not a substitute for a professional psychologist or psychiatrist. This application is intended as a medium for education and initial support related to mental health.

Main Features

Interactive NLP-based chatbot
Intent classification using Neural Network
Custom NLP preprocessing pipeline (NLTK)
Web-based application
Modern UI with Tailwind CSS
Lightweight backend using Flask
Custom psychology-based dataset

Tech Stack

Backend:

Python
Flask
NLTK

Frontend:

HTML
Tailwind CSS
JavaScript

Machine Learning & NLP:

Neural Network
Tokenization
Stemming
Bag of Words
Custom dataset class

Dataset

The HearMePal dataset was manually designed and curated using an intent-based chatbot approach, and was developed in collaboration with Yohana Eronika, a psychology graduate.

This approach ensures that:

Questions are psychologically relevant
Responses are educational and supportive
They are not diagnostic or clinical therapy

Dataset Format:

{
  "intents": [
    {
      "tag": "intent-name",
      "patterns": ["example user input"],
      "responses": ["chatbot response"]
    }
  ]
}

Dataset Components:

tag: Label intent used as the target class of the model
patterns: Variations in user input
responses: Chatbot response for that intent

Example Dataset:

{
  "tag": "definition-mental-health",
  "patterns": [
    "What is mental health?",
    "Define mental health for me."
  ],
  "responses": [
    "Mental health is a condition of a person that allows the development of all aspects of development..."
  ]
}

NLP Preprocessing Pipeline

Stages of text preprocessing:

Tokenization, Split sentences into single words.
Stemming, Change words to their base form in order to reduce word variation.
Remove Duplicate Words, Eliminates duplicate words in a single input.
Bag of Words (BoW), Converts text into a numerical representation.
Custom Dataset Class, The dataset is packaged in a custom class for model training.