NER For Restaurants' Reviews

This project use Name Entity Recognition to get insights from Restaurants' reviews.

Requeriments

• python==3.8.5
• numpy==1.19.2
• pandas==1.2.0
• pytorch==1.7.1
• transformers==4.2.2
• streamlit==0.75.0

(*) For a complete revision, please check environment.yml file.

Virtualenv

Tested Operative System:

• Windows 10: OK
• All linux-based os: OK

Steps

1. Clone repository and move to directory:

   git clone https://github.com/dpalominop/NERForRestaurants.git && cd NERForRestaurants

2. Create and activate a virtual environment (I recommend to use conda):

   conda env create -f environment.yml

   conda activate ner

3. Run the web application:

   streamlit run app.py

4. Open a browser and write in url:

   localhost:8051

Only Development Mode

(*) These steps are intended only to pretrain or finetune a model from a previous one.

1. Open config.yml and change value of stage to devel:

   stage: "devel"

2. Install git-lfs to run long files:

   sudo apt-get install git-lfs

3. Select a model from https://huggingface.com/models and clone in your local directory:

   cd models && git lfs install && git clone https://huggingface.co/{user_name}/{model_name}

4. Set the pretrained model to use in src/config.py:

   BASE_MODEL_PATH = "../models/{model_name}"

5. Set the dataset to use in src/config.py:

   TRAINING_FILE = "../datasets/reviews.csv"

6. Train your custom model:

   cd ../src && python train.py

7. Use your new custom model to predict tags in a text:

   python predict.py

Demo

Temporarily, the web application will be hosted in https://f8e1b44954ff.ngrok.io

Help?

Please, contact me to: [email protected]