This project provides a method for the military to assess the mental health of soldiers and take responsible actions for them if they in a dark place.
According to recent studies, PTSD is the most common mental health problem faced by returning troops. The symptoms include difficulty concentrating, lack of interest/apathy, feelings of detachment, loss of appetite, hypervigilance, exaggerated startle response, and sleep disturbances.
The dataset used in this project is obtained by combining 3 seperate datasets.
1. emotions dataset from hugging face.
2. text_emotion dataset from kaggle.
3. Suicide_Detection dataset from kaggle.
Note: The final dataset is not uploaded here, but it can be generated by downloading each dataset from the given links and running the scripts
The final dataset contains about 16 emotions which are condensed into 3 for better performance.
Linear SVM is used in this project for emotion detection as it gave the highest accuracy( 92.7% )
The person is asked a series of questions and their responses are tokenized into sentences.
The sentences are passed into the model to identify the emotions in each and a percentage composition is calculated for each emotion.
The generate_report function creates a report according to the percentage values.
Here is a sample response and the report generated.
