ESG ConsultBot

Group course work for ARIN7102 (HKU 2023)

Create Environment

conda env create -f environment.yml

Run API-Server with ES

1. run preprocessing

python preprocess.py --data_dir ./data/PDF --output_dir ./data/TXT --to_es true

2. start an elasticsearch single-node (with Docker)

cd elasticsearch
docker-compose up -d

3. create an index mapping, then start word embedding and the build local database

curl -X PUT -H "Content-Type: application/json" -d @mappings.json http://localhost:9200/test-index
python index.py en ../data/HSBC_HK_ESG_2022.tsv test-index

4. all set! start the qa-cli(server)

cd ..
python server.py 

Example Results

Q: What did Adidas do in greenhouse gas emissions?

A: Adidas' estimated environmental impact for 2021 is split equally among the various value chain segments. For example, sourcing and processing are both significant sources of greenhouse gas (GHG) emissions; however, raw materials management is a significant source of GHG. In order to effectively manage these multiple sources of CO2s, Adidas plans to invest roughly 90% of its annual revenue in "sustainable processes."

Create GPU environment in GPU-farm

conda create -n esg-chatbot python=3.8
conda activate esg-chatbot

pip install paddlepaddle-gpu