Suicidal-ideation-detection

To categorize tweets according to the severity level of suicidal intentions expressed in tweets’ text.

Description

This work aims to categorize tweets according to the severity level of suicidal intentions expressed in tweets’ text.

Dataset

Tweets texts having word suicide in them

Preprocessing

In preprocessing following operations are done:

• Repetitive neighbor words removal
• Repetitive characters removal and spelling correction
• Negative words replacement
• Stop words removal
• Symbols and numbers removal
• Manual cleaning of some parts causing processing overhead

Labels

• 0

Tweet content is not suicidal

• 1

Tweet content is nearly related to suicidal ideation

• 2

Tweet content indicating a user potentially committing suicide

Folders and files

• Tweets Data Retrieval

  • Tweets_Data_Retrieval: Retrieving suicidal tweets

• Training and Testing

  • antonyms_list: File containing words and their antonyms for negative words replacement
  • prepro_lib: Contains files for preprocessing operations
  • Data_preprocessor: Preprocessing of data
  • Training_Tweets_cleaner: Symbols and other minor cleaning
  • Training_Testing: Training and testing the model

• Datset

  • Raw_suicide_tweets: Tweets' texts
  • Labelled_tweets: Training data
  • test: Test data
  • tweets_words: Words extracted from tweets data after preprocessing

Requirements

• Python 3.6 or above
• Tweepy
• NLTK
• Yaml
• Enchant
• Scikit Learn
• Pandas
• Numpy

Platform

Python