With nearly 67,000 patents applied in 2017, the German Patent and Trade Mark Office faces rising challenges in evaluating applications in time. Other than trademarks filed for application, the patent authorities are legally bound to check whether the invention is a novelty, an inventive step and if it matches industrial applicability. Especially the novelty factor is a challenging issue since no patent may be granted if it is already state of the art technology. This leaves the investigator to investigate through a plethora of granted patents to be checked.
When an innovation is applied for patent status it usually gets transfered to the section in charge. There, the examiner checks various criteria if the invention matches the requirements to gain patent status. Most of the time, the examiner works as a specialist in a certain field (electronics, optics, etc.) to assure that the applied invention is technically understood. As already mentioned, one key element is the novelty factor. Patents which only reproduce state of the art technology, will be rejected. The crux however is the description of the technical procedure. Very often patent specifications are intentionally held vague and lead to law suits claiming an infringement. For the common man, there is no difference in the terms "rotation" and "circular movement". What might look like legal sophistry can easily be the core of a legal battle devouring enormous sums. Those linguistic nuances turn the patent approval process into a time consuming task. Plus, the sheer amount of data challenges the examiner to carefully investigate, resulting in delayed approval. Needles to say that no human can possibly overlook hundred thousands of valid patents.
All the problems mentioned above led the German Patent and Trade Mark Office to the conclusion that a software might help to tackle the Sisyphean task. Conventional programming with loops, if/ then conditions etc. was no option. The tool needed some sort of intelligence to "understand" the equality of terms like "rotation" and "circular movement". The quest was to develop some sort of search engine that enables its user to insert a file as starting point and rummaging from there to related files.
With 15 years of experience in the fields of enterprise search and knowledge management, interface projects provided extensive know-how on how to connect and handle large amount of data and how to retrieve valuable information out of it. The tool needed the following initial setup:
- connectors to retrieve data from file servers, databases, etc.
- an OCR tool
- a central search option
- some machine learning method to "understand" content instead of just searching
- context analysis
- a phrase search
- rummaging option
This method has another advantage. Not just that you instantly find relations in topics only an intelligent human being could see (light and light bulb are usually different for a computer), you are also able to browse through data in foreign languages without any translation. By let's say converting the vectors from the English vector space into the German one, mathematically, you only need to match terms in the same area. This comes quite handy since patents are rarely restricted to one national authority.
Picture the following situation: you are an examiner at the Patent office and you get a patent application for a canopy that can be illuminated to work as a lamp. In its description, the applicant uses words like "canopy", "illuminating", "light", "table" and "emission". The file consists of 20 pages with 12,000 words and drawings, stored as a PDF.
First thing you do, is to enter one relevant term into the search engine - for instance "illuminating".
In our example, the system displays 90 patents which might be related to your search. While adding a relevant hit to your search card section, the content of the finding will be displayed in detail.
From there, you are now able to mark terms or phrases to start rummaging your investigation. The search is not just limited to simple words but whole paragraphs dramatically expanding the way you were researching.