Winning solution of competition held by Bridgei2i under InterIIT Tech Meet 2021 [video, slides]
Team Members | Vasudev,Mukund, Jayesh, Yadukrishnan, Tanay, Anirudh, Siddhant |
---|
We provide well-commented code for your reference. The overall directory structure is as follows:
Outputs for the evaluation data is located at ./predictions/
High level directory overview
├── text-cls
│ ├── phoneme.py # code conversion to phonemes
│ └── train_cls.py # training classification model
├── summarization
│ ├── train.py # fine mbart model on dataset (refer to training_utils/args.py)
│ └── evaluate.py
├── ner
| └── run.py # run NER + sentiment
├── assets
| └── ppt.pdf # brief solution description
├── preprocess.py # preprocessing script
└── README.md
Running the app for theme classification and summarization
streamlit run app.py
Running the app for NER (only for tweets because of dependency issues)
streamlit run app_ner.py
IMPORTANT
- To run the NER code (./ner/run.py) and corresponding app (./app_ner.py), please install transformer==2.5
- To run summarization (./summarization/train.py) and corresponding app (./app.py) please use transformers==4.4