Usually patent data does not contain any unique identifiers for the patenting assignees or the inventors, as the main tasks of patent authorities is the examination of applications and the administration of the patent documents as public contracts and not the support of the empirical analysis of their data. An inventor in a patent document is identified by his or her name. Depending on the patent authority the full address or parts of it may be included to further identify this inventor. The goal is to define an inventor mobility index that traces the career of an inventor as an individual with all the job switches and relocations approximated by the patents as potential milestones. The inventor name is the main criteria for this identifier. The inventor address information on the other hand is only of limited use for the definition of a mobility index. The name alone can work for exotic name variants, but for more common names the problem of namesakes gets in the way of identifying individuals. The solution discussed here consists in the construction of a relationship network between inventors with the same name. This network will be created by using all the other information available in the patent data. These could be simple connections like the same applicant or just the same home address, up to more complex connections that are created by the overlapping of colleagues and co-inventors, similar technology fields or shared citations. Traversal of these heuristically weighted networks by using methods of the graph theory leads to clusters representing a person. The applied methodology will give uncommon names a higher degree of freedom regarding the heuristic limitations than the more common names will get.


Inventor disambiguation, patents, fuzzy search, heuristic method