Fee paying
Database description

Categorizer is an AI based tool for automated patent categorization. It generates a classification model using a list of training documents in the first step, and the model is used for categorizing a new set of patents in the second step.

The use cases of Categorizer are the following:
- Automate patent categorizations with the help of training documents
- Map new patent portfolios to known categories
- Manage patent portfolios and competitive analysis

The key benefits of Categorizer are the following:
- Users do not require domain expertise
- Complex categorizations can be completed in a matter of minutes
- Metrics allow users to identify primary, secondary, and tertiary categories for each patent
- Highly accurate classifiers are designed along with performance metrics
- Minimal and flexible input that includes tagged documents or category names and description

Categorizer is available in two versions, viz., generic and custom versions. In its generic version, Categorizer uses a two-step process. In the first step, an AI model is created using a set of patent documents tagged with category names. The model developed in the first step is used for categorizing a list of patent documents in the second step.
The generic AI model is ideal if you have an existing set of patents that are already categorized manually, and you need to categorize a new list within the next few minutes.
In the custom version, the AI model is created automatically at the backend, and the model is made available on the user’s account. It eliminates the first step required in the generic case.

Custom AI models are ideal for the following cases:
- Looking for the best possible categorization accuracy
- The user does not have tagged documents
- Needs quantitative metrics to identify primary, secondary, and tertiary categories
- Categorization is required for non-patent documents
- Requires detailed performance metrics of the model

Patent Collections
Data export
Data Export
Bibliographic data
Title and/or abstract

Bibliographic data – At least one of the following is available: publication number, filing number, grant number, publication date, filing date, grant date, applicant name, assignee name, inventor name, patent classification.
Partial text – At least one of the following (but not all) are available: title, abstract, description, claims.
Full text – All of the following are available (where applicable for the given jurisdiction): description, claims