AustraliaPublic sector2018 – ongoingIP Australia
Patent Auto Classifier for routing applications
An audit found IP Australia's patent classifier averaged 64.7% correct classifications
Problem
Patent applications needed classification by technical subject before reaching the appropriate examination team. Incorrect routing required examiners to redirect applications.
Approach
IP Australia implemented the Patent Auto Classifier in 2018. It processed application text with natural language processing and scored relevant patent-classification symbols. Applications went to the examination queue associated with the highest relevance, with staff correcting routing errors.
Results
Takeaway
Annual retraining coexisted with persistent classification gaps. The auditor identified an opportunity to analyse existing performance data for emerging problems.
Categories
- Industry
- Public sector
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Regulator, audit or court
Sources
Source summaries
ANAO examined deployed patent-processing AI and tested the classifier in detail. It reported correct, missing and incorrect classification shares and identified opportunities for better performance monitoring, without finding that the classifier had been withdrawn.
Full details are in the original sources.
