QuickPat

https://bulkdata.uspto.gov/data/patent/application/redbook/fulltext/2020/

Some commands

• pip install beautifulsoup4

• pip install lxml

• pip install openai

• python read.py

Project Scripts

processPatents.py

Processes the raw patent file (xml) into a pandas dataframe with the fields we're interested in.

indexPatents.py

Reads back the dataframe, encodes each abstract, then indexes into a vector db. The vector db is FAISS.

generateAbstracts.py

Generates some abstracts using GPT4 and writes it to a csv file.

evaluateAbstracts.py

Evaluates the generated abstracts from previous step using cosine similarity and euclidean distance.

rag.py

Searches the index for other abstracts that match the search query abstract.

generateClaims.py

Based on an input claim and claims of similar-abstract applications, generates additional claims using GPT4 and writes to a csv file.