Geospatial Name Entity Recognition

This project is an NER application which can extract names of cities, states and countries (places) from a natural language sentence. This project was selected as top 5 in India amongst 220 teams for the Smart India Hackathon 2023, on a problem statement by the Indian Space Research Organization (ISRO).

Contributors

• Aryan Shukla
• Shreya Jain
• Aashray Gupta
• Aman Kumar

Deployment

To deploy this project, simply run the main.ipynb notebook keeping states.csv, cities.csv and countries.csv as well as the checkpoint for the re-trained model that we trained on our local machine in correct path as mentioned in the notebook.

• After running the main.ipynb notebook, it should look something like this:

• The streamlit version of the application can also be run by running the streamlit.ipynb notebook on google colab and uploading app.py

• After running the streamlit.ipynb notebook it should look something like this:

• Go to the link suggested by the your url and enter the ip address obtained after running !wget -q -O - ipv4.icanhazip.com

• After doing so, you will enter the Streamlit App,

Note:

• Make sure to replace the links of data files (from Data folder) and model checkpoint in the drive in the app.py file.
• The model also works for Hindi langauge sentences.