Spotle AI-thon - The AI Global Challenge had 7000+ participants from best campuses in India, Singapore worked on addressing the mental health challenge with AI. Top 10 teams from IIT Roorkee, CMI, NIT, IIM Indore, Charotar University, DIAT made it to the final round. This is a showcase Top 10 presentation from Team Temp, Chennai Mathematical Institute
Spotle AI-thon Top 10 Showcase - Analysing Mental Health Of India - Team Temp - CMI
1. Analysing the mental health
of India during Covid using
Twitter data
Team : temp
Swarnadeep Bhar, Debangshu Bhattacharya, Swaraj Bose
Chennai Mathematical Institute
#SpotleAIthon
2. 1
2
3
4
5
Problem Statement
Introduction
Build classifier ➔ Classify tweets
Analysis
Conclusion and future works
AGENDA
3. PROBLEM STATEMENT
• Happy
• Sad
• Angry
ANALYSE THE MENTAL
HEALTH OF INDIA
USING TWEETS
RELATED TO COVID-19
DATA DURING THE
TIME PERIOD 13-22
SEPTEMBER, 2020
CONSIDER AT LEAST
3 EMOTIONS
• Build the classifier
• Evaluate performance
CLASSIFY THE
TWEETS W.R.T
EMOTIONS
• Meaningful analysis on
the classified tweets
ANALYSIS
5. DATASET CLASSIFICATION MODEL EVALUATION
AIT 2018 dataset used for
training
• Multi-label classification
considered
• BERT for word embeddings
Transfer learning using neural
networks
79.3% accuracy
TWEET CLASSIFIER
Emotion AUC
Happy 0.9288
Sad 0.8064
Angry 0.8046
Fear 0.8015
6. WORLD INDIA
TREND ANALYSIS
• FEAR MOST DOMINANT EMOTION
• JOY LEAST EXPERIENCED EMOTION
• FEAR MOST DOMINANT EMOTION
7. METROPOLITAN CITIES
• Sadness in Mumbai
• Occurrence of
activity(fear):
• Mumbai, 16th
September
• New Delhi, 17th
September
• Overall pattern consistent
with earlier graphs
8. ANALYSING WORD CLOUDS
Image source : https://www.dreamstime.com/digitalmarketing-source-covid-coronavirus-word-
cloud-red-covid-words-grey-word-tag-world-map-background-abstract-image175797825
11. CONCLUSION
Fear is the most widespread emotion experienced
Sadness has gone down in Mumbai
Some events around 16-17th September may have caused some fluctuations from the trend in
Mumbai and New Delhi
Major takeaways from word cloud analysis:
❑ Launch of products or games causing much needed relief (even if temporary)
❑ Allegations and rumours causing further discontent and resentment amongst the masses
❑ Concern for elderly family members
Future scope:
❑ Dealing with sarcastic tweets
❑ Refining the model to not allow antithetical emotions to be classified together
❑ Analyse a further horizon of data