README.md

metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: title
      dtype: string
    - name: assignee
      dtype: string
    - name: inventor/author
      dtype: string
    - name: priority date
      dtype: timestamp[ns]
    - name: grant date
      dtype: timestamp[ns]
    - name: result link
      dtype: string
    - name: representative figure link
      dtype: string
    - name: abstract
      dtype: string
    - name: description
      dtype: string
    - name: ita_only
      dtype: int64
  splits:
    - name: train
      num_bytes: 8798146201
      num_examples: 105195
  download_size: 3443662872
  dataset_size: 8798146201
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

AI Patents Dataset

This dataset contains data about 105k patents obtained filtering by the following keywords from Google Patents:

• artificial intelligence
• neural network
• LLM
• large language model
• machine learning
• deep learning
• natural language processing

Data filtered from European countries and United States patents from January 2000 to October 2024 (grant date).

Italian-only data

Google Patents does not have granted Italian patents for the selected keywords. Therefore, we included non-granted patent applications in Italian, 1.6k in total. These are identified by ita_only==1.

We added the following keywords to the query:

• intelligenza artificiale
• rete neurale
• reti neurali
• apprendimento automatico