Are your patent workflows cluttered with repetitive tasks and bottlenecks? What are you spending too much time on? The second part of our AI tool evaluation blog series discusses the topic of how to recognize the need for a patent AI tool.
The AI market is expected to grow 40% annually during the 2020s, and AI continues to play an increasing role in intellectual property management. Today we launch a blog series to shed some light on why and where to consider artificial intelligence for patent related tasks and how to evaluate if a tool fits your organization.
The latest addition to the IPRally board is a tech entrepreneur, angel investor and IP attorney.
Clojure is a functional programming language, written for those who seek simplicity. The combination of a better way to build and hire top talent can help a company to move faster.
The company starts a new sprint in AI development and global operations following a Seed investment round
The round is led by Join Capital and Spintop Ventures, with the participation of previous investor Icebreaker.vc.
Defining class filters in Boolean searching is a challenging task due to the complicated classification systems. Graph-AI can mitigate the effect of human errors and biases and widen the horizon of prior art search.
There is a growing evidence that keyword search is not sufficient for finding relevant prior art. How can AI practically help with finding crucial documents that are hiding in the vast amounts of patent data?
An independent study confirms that IPRally significantly improves the quality of patent searches. Being the best performing AI platform tested, an average improvement of 29% was reported.
The amount of non-English patent documents is rapidly growing. To tackle the searchability problem, we are using machine translations to create English knowledge graphs from non-English texts.
The dynamic highlighting of the search results loads faster from now on. This is the first time we brought GPUs to the frontline, to speed up your user experience.
We used our model trained for prior art searches to classify patents into CPC classes. The performance we achieved is on par with the performance reported for state of the art classifiers in scientific articles.
Upon updating our database with millions of new documents, we expanded the filtering options with the new data sources so that you find faster the patent literature you are looking for. The database covers by now (May 2020) 75 million patents and patent applications.
Artificial intelligence (AI) opens up new opportunities to examine the patentability of inventions. The combination of AI and specialist work increases efficiency also in a telecommuting environment.
While working from home, our new UX designer Maggie connects with our customers and seeks to understand them better.
Google Cloud Pub/Sub is simple to use from Python, but what about Clojure? There is no official support and the Java interop is not straightforward. There are few libraries but none is active. We ended up building our internal mini-library jonotin, and today we have published the code.
Researchers have improved the widely used deep learning optimizer Adam, decreasing the time needed for training models and improving model generalization. We made a small contribution of our own by open sourcing our PyTorch implementation of QHAdamW.
The story of our first steps towards explainable patent search. In the end of the post you get to test our unique approach, which is one of the first few serious attempts to provide real world value from explainable NLP deep learning models.
IPRally's knowledge graph based patent search is now free to try out.
Knowledge graphs are a great fit for patents. The benefits of a graph based patent search will be increasingly hard to compete with the traditional approaches.
IPRally won the most innovative AI startup award in North Star AI conference. The event had some star speakers like Estonian president Kersti Kaljulaid and DeepMind's Samuel Ritter.
The most notable IPR innovation award of Finland, granted by the IPR University Center, was given to IPRally.
Our story about utilising neural nets with tree like data. Can a Tree-LSTM model be used for real world applications? Turns out the answer is yes, but only after the performance is improved by 7000%.
The story of IPRally in a nutshell. From initial questions to validation of ideas and a working product.