The classifier will be launched in September, can you give us a brief description of this new module?
Andreas: It will make our lives a lot easier and help companies increase productivity as well as make better decisions. It will, in essence, allow you to create your own unique and custom version of IPRally’s cutting-edge graph neural network technology, and get full control over the process. It brings immediacy and control to not only the search and review process, which IPRally has done before, but also to strategic areas like portfolio categorization and benchmarking. It's ridiculously easy to use. It can be set up, calibrated and applied by the user in minutes. And it can be trained with unusually small data sets without compromising on accuracy.
Sakari: We have designed the classifier to do one task really well: distinguish technologies from each other, fast. The classifier allows combining IPRally’s AI with accumulated intelligence in your own organization to provide quick insights and add value to patent research projects and day-to-day work. It allows you to look at the patent space through your own lenses, in the way that you are used to and that makes sense in your context and your organization.
"Hundreds of hours of manual data processing that has been spent in landscaping or benchmarking projects can now be spent elsewhere."
Can you tell us a bit about the background and the pain points it addresses?
Andreas: For as long as I have been in this industry, companies have spent disproportionate amounts of time and resources on cleaning, categorizing and making sense of patent information. Or they have outsourced this, and spent a lot of money on it. With the launch of IPRally’s classifier, we can rely on AI to match technical attributes with custom patent categories. It can predict in which categories previously unknown patents from a technology area or a competitor portfolio belong, with surprisingly high accuracy. This means that search or monitoring results can be automatically tagged with customer-specific categories, so that the data can be assessed or distributed to the right people much faster and much more accurately. And the good thing is that the whole organization can benefit from it. Patent experts, naturally, but also non patent experts will be able to search for information using their own terminology, without being turned off by complex search methods like boolean. Hundreds of hours of manual data processing that has been spent in landscaping or benchmarking projects can now be spent elsewhere
What in particular makes the graph neural network (GNN) technology suitable for a patent classifier?
Sakari: GNN’s are universally known to be excellent for bringing structure to large datasets, and to allow complex relationships to be processed fast. They have been proven to provide high recall and precision in search tasks, which implies that they are able to focus on the right technologies. Now we use the same graphs and neural network models to classify data, not in a general way but trained with the customers’ historical data.
"A well-trained classifier is like a team member having superpowers and being constantly present."
From a general business perspective, what are the implications of this launch, can you give some examples?
Andreas: Just imagine having every result from a weekly patent alert already classified with your own specific classes, and what this does to the review and distribution process. A lot less time will be wasted on irrelevant data, and people will be much more engaged. But the classifier also means that IPRally can be used for more strategic projects, like portfolio segmentation, benchmarking and competitive intelligence. Imagine, for example, that you are looking to acquire a company and want to see where the overlap between their patent portfolio and your portfolio is, and having your product categories automatically applied to the portfolio you’re potentially acquiring. How would you do this currently, and what would it mean to have an intelligent AI do the groundwork for you? In short, it allows us to provide much more holistic and valuable solutions to our clients, in areas where they are many times using outdated or inefficient methods.
Sakari: The amount of patent data has become impossible to manage by traditional ways. And the business environment is changing rapidly as well, with new areas emerging and new players appearing in the field. How long are traditional competitor monitoring profiles maintainable? AI classifiers can be used to help identify relevant technologies with less effort, but also categorize them on-the-fly. Also, people change workplaces. Would it be useful to harness the work done by previous employees to benefit the next ones? A well-trained classifier is like a team member having superpowers and being constantly present.
How will it be made available?
Andreas: The classifier module will be available in IPRally early September, and some clients that have been involved in the process of building this will have access even before this. Anyone that is using IPRally today will be able to implement the classifier and start building their own classifiers immediately after the launch, if they decide to add the new module to their IPRally accounts. Anyone else will be able to get access to IPRally with the classifier if they decide to acquire licenses to IPRally. We will have a promotional fee for the classifier module throughout the rest of 2022, and if you’re interested to learn more – just reach out to us!
.webp)
.webp)
