Revolutionizing Patent Preclassification with AI-enabled Automation
AI technology has transformed the French Patent Department's preclassification process, reducing manual tasks and improving efficiency. From manual dispatching to AI-driven one-step dispatching, the new methodology includes text processing, model construction using machine learning, and application on a training set for team prediction. Results show increased reliability and reduced error rates, with ongoing improvements in progress for more automation and efficiency.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
Pre-classification and AI French Patent Department Godefroy Lem nager
AI-enabled Patent Preclassification Previously: Manual dispatching of patent applications to examination teams 2 steps : 450 new applications per week Predispatching: Based on the title and text of each application Service Managers Dispatching to examiners: Based on: Text and claims structure Workload of each examiner / IPC evolution / examiners experience Team Managers S1 S2 S3 P24 P11 P12 P13 P23 P22 P21 P33 P31 P32 Examiners 2 /
AI-enabled Patent Preclassification Today 450 new applications per week One-step dispatching: Predispatching (Service Managers): Distribution : Team managers dispatch patent applications in their poles S1 S2 S3 P24 P11 P12 P13 P23 P22 P21 P33 P31 P32 Examiners 3 /
Methodology 1. Text processing Text Mining : lowercase, punctuation removal, stopwords removal, lemmatisation, stemming, 2. Model construction Machine learning model: FastText In the future, maybe TF-IDF and neural networks 3. Model application on the training set (ca. 20 000 applications) 4. Comparison with the correct team 5. Graphic interface R-Shiny to apply the model on new applications 6. Export of the predicted team 4 /
Results Before April 2019 After April 2019 8 teams New organisation with 10 teams Reliability: 85 % of the training set Reference P1 P2 P3 P4 P5 P6 P7 P8 New reliability: 75 % 1385 51 61 20 49 35 6 49 P1 P2 P3 P4 P5 P6 P7 P8 36 1503 84 15 39 26 39 140 56 84 2160 36 50 61 12 15 Prediction 21 15 38 2219 35 24 47 106 42 51 66 1326 36 15 34 136 Error rate with manual distribution : 5 to 10 % 14 47 29 25 1589 45 45 104 7 45 9 23 10 47 1908 76 33 56 21 46 43 54 59 1398 5 /
Work in progress The new organisation implies for the IA tool another learning session Reduce the time spent by service managers in file distribution (1d/w) Next step : IPC Proposal for examiners 6 /
