Tuesday 5. October 2021

Being an IP manager in a big company, SME or a single consultant on your own, the same questions arise on how to manage your projects portfolio with the most efficient balanced risk/priority objectives.

•    The rapidly growing amount of patent documentation will soon no longer be manageable with conventional search methods - possibly this is already the case today.

•    For a couple of years now, the application of machine learning (ML) methods are being discussed as a potential solution for reducing human effort in searching large amounts of patent data. While some promising projects and ideas in this direction have been presented by different sources, the real breakthrough for ML as a standard and widely accepted patent search method has not happened yet.

•    The presentation looks into the challenges that still exist in this area, especially as far as practical applicability and acceptance by users is concerned, using the Intergator Smart Search project as an example.

The EXTRA classifier is a scalable solution based on recent advances in Natural Language Processing (NLP). The foundational concept of the EXTRA classifier is transfer learning, a machine learning process that enables the relatively low-cost specialization of a pre-trained language model to a specific task in a specific domain with far fewer training examples compared to standard machine learning solutions.

More specifically, the EXTRA classifier leverages BERT, a well-known pre-trained autoencoding language model that has revolutionized the NLP space in the past few years. BERT provides contextual embeddings, i.e., it provides context-aware vector representations of words that capture semantics far more efficiently than their context-free counterparts.

The EXTRA classifier contains a pre-processing module to cope with the inevitable noise in the output of standard Optical Character Recognition systems. The pre-processed plain text from a source document is then fed into a BERT-based classifier, which is built by extending pre-trained BERT with an additional linear layer trained for classification through a process commonly known as fine-tuning.

We will present preliminary results that confirm some clear benefits with respect to rule-based solutions in terms of classification performance and system scalability.

Since the 1950s, AI has been plagued by overpromise and underperformance, particularly at the interface between AI and topical Subject Matter Experts (SMEs). AI has struggled to produce results that SMEs deem effective. As the current hype around the most recent wave of AI recedes, it is time to assess if the new round of research has improved AI’s capacity to help SMEs. This presentation looks at three aspects of current AI research that might actually be useful: Composite AI, Generative AI, and Small Data. These three approaches have the capacity to reduce the distance between AI systems and SMEs by allowing experts to have more local control and input into the behavior of AI systems. This closer interaction has the potential to lead to useful, effective results for SMEs.

New tools are indeed present promising opportunities. However, the use of AI also raises new concerns regarding the transparency of the expected usability and reliability. This presentation will elaborate around some the associated questions; the impact of the training data, the black box symptoms, verification of performance and privacy considerations when using AI. But of course also some insights to best practice of AI search tool and how to take control of the process for better efficiency.

The use of machine learning in IP activities has increased exponentially over the past five years. At the same time new tools, methods and systems have begun to emerge that seek to make the analysis of patent data easier to accomplish using these techniques. Included in these new developments are a significant number of machine learning systems that have begun coming to market. As these changes continue to occur, it would be useful to review some of the tools, systems, or methods that a patent practitioner has at their disposal. Examples and perspectives on the latest advances in machine learning for IP will be provided. There will also be a tour of ML4Patents.com which is devoted to aggregating content associated with the development of this area.

What do PLI, MetOpera, ASCO, and PLOS have in common? Content management and content discovery needed major improvements. User were not getting the results they needed. The content production team including editorials, managing editorials – the whole team – could no longer cope with the volume and variety. Content quality was suffering. Brief discussions of each organization’s challenges set the stage for AI-based, human curated solutions. What worked, what didn’t, and the how and the why will be presented.