This is a demo of a Patent Search Engine web-app based on T5 and BERT-like large language models (LLMs) that allow to enter a query in Russian/English/Chineese and retrieve relevant data in the domain of English language patents (test data scraped from Google Patents).
Installation of this demo is fully automated with Docker.
[Advanced] Deployment of this web-app to remote server is also fully automated with the help of Ansible-playbook scripts and Docker Compose for Nginx and Cerbot.
[Currently expired!] Public access is granted by following www.patent-search.online
- Clone the repo and create a venv
- Install the required Python packages by running
pip install --ignore-installed -r requirements.txtOR simply run therun.shscript [windows]:bash .\run.shOR [unix/linux]:.\run.sh. - Configure model_config.yaml [optional]
- Download Hugging Face models by running
python3 model_utils.py - Run the Python Flask application by running
python3 app.py. - Open a web browser and click to some kind of
http://***.*.*.*:2000link in terminal to access the Patents&Suppliers Search Service. - Enter a query in the search box and click "Search" to retrieve relevant patent documents and potential suppliers.
- Clone the repo and
cd Patent-Search - Configure model_config.yaml [optional]
- Run
docker build -t patent-search .to create an image. - After building an image simply run
docker run patent-search - In the terminal follow one of the links to access Patents&Suppliers Search Service.
- Enter a query in the search box and click "Search" to retrieve relevant patent documents and potential suppliers.
Vladislav Gromadskii