Questions_in_Table

For any excel file, BERt will reply for the question asked.

To start the application:

---

1. pip install streamlit
2. pip install streamlit-chat
3. pip install transformers

To run the application:

---

on terminal, in this folder, type: python -m streamlit run table_chat.py --server.maxUploadSize=1024

-> if the file to be loaded is bigger than 1 GB, change it in the maxUploadSize argument.

Basic information about the UI:

---

after running the application

1. in the setup, load the file (in excel format only) and click on the button: 'load the file'.
2. then in chatbot tab, if you doesn't select the columns, then the whole table will be used.
3. After the question is typed in, enter is pressed, its necessary to click on the button: 'start search' as well.

Note: selecting the columns ensures good performance. When the table is big (I tested on 10 MB file), even on GPU, the model doesn't work. But with 2 columns preselected, I was always able to get a reply from the model.