The Innovation

This patent application concerns a computer-implemented method for managing an archive to determine approximate matches associated with strings in data records of a dataset. It involves a pre-processing phase to identify potential approximate matches and store them in an archive, and a run-time phase to determine actual matches during database operations. The method includes processing data records to find string representations, generating close representations as deletion variants, calculating frequency of occurrence, and storing potential matches based on these representations. The method aims to improve efficiency in database operations such as join or rollup by using approximate string matching​​.

Technical Contribution

The Board’s analysis focused on the technical contribution of the claimed features. The appellant argued that the method provided technical effects, such as updating data records and improving accuracy and execution speed of computation graphs. The board acknowledged the technical contribution of the archive as a data structure enabling efficient database operations. However, it was noted that the method did not contribute non-obviously over a general-purpose computer system or prior art document D1, which already disclosed similar concepts in approximate string matching. The Board also found that certain aspects, like calculating and storing a “significance value” for strings based on renormalized frequencies, did not contribute to the technical character of the invention, as they were abstract and not used in a technical context within the claimed method (Reasons nos. 10-18, 22-27)​​​​​​​​​​.

Key Findings

• Pre-processing data records to determine potential approximate matches – technical
• Generating close representations as deletion variants – technical
• Calculating and storing a “significance value” for strings based on renormalized frequencies – non-technical

Read the full decision here: T 1867/18 (Approximate string matching/AB INITIO) 05-06-2023​​.