Datasets
Patents

Patents

Fixpoint indexes patents and lets you search them, turn them into LLM-ready plain-text or RAG chunks, or do data extraction on them. Fixpoint can also resolve all of the people and companies involved in the patent.

You can search by a variety of keyword filters, such as the company that owns the patent, or keywords within the patent. Fixpoint returns metadata about each retrieved patent, and also lets you do extra data extraction on the patent itself.

Making a patent search request

Example request:

curl --location 'http://localhost:8000/extractions/patent_extractions' \
--header "x-api-key: $FIXPOINT_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
    "filters": {
        "keywords": ["natural language"],
        "after_date": "20240101",
        "assignees": ["Google"]
    }
}'

Example response:

{
    "patent_extractions": [
        {
            "patent_info": {
                "patent_number": "US20240370666A1",
                "inventors": "Markus Freitag",
                "assignees": "Google Llc",
                "filing_date": "2024-07-15T00:00:00Z",
                "publication_date": "2024-11-07T00:00:00Z",
                "title": "Automatic post-editing model for generated natural language text",
                "summary": "APE model 206 can be trained to process natural language text to generate edited text correcting word translation error(s), gender error(s), etc. In a variety of implementations, APE model 206 can be trained for editing text translated from a specific source language to a specific target language."
            },
        },
        {
            "patent_info": {
                "patent_number": "US20240296313A1",
                "inventors": "Samy Bengio",
                "assignees": "Google Llc",
                "filing_date": "2024-05-13T00:00:00Z",
                "publication_date": "2024-09-05T00:00:00Z",
                "title": "Generating natural language descriptions of images",
                "summary": "processing the alternative representation of the input image using a neural network to generate a sequence of words in a target natural language relating to content of the input image, including selecting at least one word for inclusion in the sequence of words by conditioning the neural network on …"
            },
        },
    ]
}

Feature status

  • keyword search - ✅
  • patent metadata - ✅
  • semantic search - in progress ⚠️
  • PDF data extraction - in progress ⚠️
  • parse PDF to plain text or RAG chunks - in progress ⚠️
  • entity resolution - in progress ⚠️
Company DataExtraction (Single Site)