search

TAHIR0110 / ThereForYou

ThereForYou: Your mental health ally. Kai, our AI assistant, offers compassionate support. Track your mood trends, find solace in a secure community, and access crisis resources swiftly. We're here to empower your journey towards improved well-being, leveraging technology for a brighter tomorrow.
Other
79 stars 97 forks source link

Sentiment Analysis (not the basic one) #64

Open Cralsic123 opened 2 months ago

Cralsic123 commented 2 months ago

Is your feature request related to a problem? Please describe. Some of the most appealing causes of suicide are depression, anxiety, and feeling competitively lost.

Describe the solution you'd like Building an advanced sentiment analysis model, will analyze the sentiment of the user and accordingly provide the level of chatbot. If a single chatbot is made that is available for all moods such as suicidal, depressing, or anxiety it would be hard. So I think the AI chatbots should be divided into a few parts that will handle the different moods. My analysis model will provide the most probable mood case for a user and accordingly, the chatbot will be assigned to the user.

Additional context

Cralsic123 commented 2 months ago

@TAHIR0110 @sanjay-kv I would love to contribute in this issue and I can guarantee my contribution will be effective.

TAHIR0110 commented 2 months ago

@Cralsic123 assigned!

Hireath08 commented 2 months ago

Hi @TAHIR0110 , I am interested in NLP based tasks as I have done text summarizer, language translator projects in it, I would love to work on this issue, If you assign me one.

Hireath08 commented 2 months ago

Describe the feature

Data Collection: Text reviews, social media posts etc. Data Preprocessing: Remove noise, lowercase, normalize. Data Labelling Mark each text as positive, negative, neutral (manual or crowdsourcing). Building lexicon: Positive, negative, neutral words and relevant phrases. Finding n-grams: Look for 2-3 word sequences matching your lexicon. Score sentiment: Add up the sentiment value of matched words/n-grams in each sentence. Model Evaluate: Train a simple model (Naive Bayes etc.) and assess performance on unseen data. Refine: Improve lexicon, n-grams based on evaluation and consider negation handling.

This method builds upon basic sentiment lexicons.

psyuktha commented 2 months ago

hey @TAHIR0110 can you assign me this?

sanjay-kv commented 2 months ago

Admin the labels followed in the repo is wrong.

it should be without space eg: level1, level2, level3

Samik123Mit commented 1 month ago

pls assign me the issue , i have worked with the task model.