using RNN architecture(GRU, LSTM, etc.) for Multi Label Classification on Persian Sentences
The purpose of the project was calculate the similarity between two pair sentences and also finding the subject of two sentences.
Process of the project:
- Use Hazm library for parsing and extracting subjects
- Prepare sentences for feeding into model
- Create multiple LSTM and GRU models with Keras
The dataset consist of 3 columns. 2 columns are the sentences and the third one is the similarity score.
All of the subject of the sentences should be in one of 6 specified classes. The classes are as follows:
- Boy/Man
- Girl/Woman
- Child
- Animal
- Others(like people)
- Unkown
After finding subject, every pair-sentences were assigned to their own classes.
link of Dataset:
https://github.com/Ledengary/COPER/blob/main/Datasets/PerSICK.csv