Team Doggo Ipsum
Singapore Polytechnic
Diploma in Applied AI & Analytics
Project Good Vibes
Yu Hoe | Tingxiao | Kritchanat | Choon Wei | Kah Shin
https://youtu.be/KhF4W_aLjhY
Anxiety
Depression
Suicidal Ideation
About 10% of people are affected by some form of Anxiety Disorder in Singapore.
The most common mental illness in Singapore. During the pandemic, about 13% percent of over 1,000 participants in a study reported symptoms of depression.
In 2021, there was a total of 378 reported suicides in Singapore, and is the leading cause of death for those aged 10-29.
Our app solves the social problem of wellbeing
Application purposes & aims
The purpose of this app is to provide an online chat room platform with the help of AI to promote a friendly environment where users can chat anonymously with strangers based on our mood matching algorithm to encourage a positive vibe
By leveraging digital innovations and AI to create mental health solutions, our application aims to support digital mental wellbeing for the whole-of-Singapore in response to the Smart Nation Initiative
Digital Mental Health | Smart Nation :https://together.smartnation.gov.sg/working-adults/free-webinars/supporting-digital-mental-health/
https://good-vibes-chat.netlify.app
PUT QR CODE HERE
Our application
AI #1: Sentiment Analysis
1,600,000 Tweets
Our model trained over 1.6 mil tweets that were extracted using Twitter API
Source: Sentiment140 dataset
Sentiment analysis
To analyse the sentiment and output a metric to identify whether the sentiment is positive, neutral or negative
AI #2: Hate Speech Detection
🤬
😡
Severe Toxic
Toxic
🥵
Obscene
Multi-Label Classification
Insult
Identity Hate
Threat
😠
🤥
🤐
Outputs a set of binary probabilities (Sigmoid) of each class (Shown on the left)
To prevent any hateful and abusive remarks to be sent to other users in our application. As we are promoting a positive and friendly online environment
Why this AI?
BERT base model (uncased) Architecture
Uses a method of masked language modeling (MLM) to have a fixed meaning independent of its context.
Contains attention mechanism. Transformers are non-sequential processing, so sentences are processed as a whole (parallelization) rather than word by word when compared to LSTM.
Every element in a sequence pays attention to every other element
Technology Stack used in our Web Application
Front-End Stack
Tailwind CSS
Client
Styling/Design
Back-End Stack
Socket.io
AI Stack