PatSeer's AI Search V2 Engine Revolutionizes Patent Search with 33% Improvement in Results

PatSeer's AI Search v2 Engine Delivers 33% Better Patent Search Results

Early users of PatSeer AI search have quickly realized the power of having both Expert Search and AI search in a single platform. We've also listened closely to our users' feedback, incorporating their suggestions into the development of this new version. For instance, user directives to the AI search are given increased weightage in the engine. The result? A search tool that not only understands patent semantics better but also provides users more control over the output.

AI Searches get 33% better with the new PatSeer's AI Search v2 engine

Exactly one year ago, PatSeer launched its AI search for patents by introducing a completely new search powered by a custom-trained LLM model. Today I am excited to announce AI Search v2 which is a significant upgrade to the underlying AI search engine. The underlying LLM powering v2 has undergone extensive downstream training and fine-tuning for various technology segments resulting in a better understanding of the search context and therefore better-quality results. Early users of PatSeer AI search have quickly realized the power of having both Expert Search and AI search in a single platform. We've also listened closely to our users' feedback, incorporating their suggestions into the development of this new version. For instance, user directives to the AI search are given increased weightage in the engine. The result? A search tool that not only understands patent semantics better but also provides users more control over the output. Measuring improvements in search quality across technology domains Using sample datasets from the different WIPO technology domains, we analyzed the improvements made by the new v2 model over our previous v1 model. For this, we measure the number of X (or 102) and Y (or 103) cited records that come up in the top 100 results. Each X/102 record is given twice as much weightage as a Y/103 record in our scoring model. AI search v2 shows the highest recall improvement of 48.19% in the Instruments domain followed by Other and Chemistry domains. Importantly, all the domains have seen more than 15% improvement and an average of 33% improvement in recall. Increasing users and usage of PatSeer's AI Search The usage numbers speak for themselves. In the first month, AI searches accounted for a mere 1-2% of total searches being run on PatSeer. Today, that figure has grown to over 11%. In terms of users, one-third of the unique users on the platform are using AI searches each month. This remarkable increase demonstrates the growing popularity of our underlying engine. The Road Ahead: Continuous Improvement We continue to explore new ways to empower patent professionals with the tools they need to drive innovation. User feedback continues to be the main driving force behind our continuous enhancements. The launch of PatSeer AI Search v2 represents a significant milestone in our quest to make patent research more efficient, accurate, and accessible. About PatSeer PatSeer is an AI-driven patent research platform with integrated analytics, project workflows, and collaboration capabilities. This all-in-one patent solution is trusted by leading law firms, KPOs, technology startups, and major corporations across 38+ countries and over 8,000+ professionals globally. PatSeer includes more than 157M patents across 108+ countries with a focus on comprehensiveness and accuracy. Can be used to search global patent collections with confidence and leverage connected data points for superior research. The platform's compelling set of features and analytic tools help you generate actionable insights with ease and confidence.

GenAI Slashes the Time Needed to Draft Patents by 70%

Manish Sinha, the founder and CTO of patent research and analytics company PatSeer, recently told AIM that patent filers in India are leveraging AI and generative AI, drastically reducing the time needed to draft patents. Interestingly, this has directly contributed to an increase in the number of patents filed in India. "AI has reduced the amount of time needed to draft patents by as much as 70%," Sinha added, citing attorneys using generative AI to easily prepare a structure and all the key sections. "Earlier, they had to start from zero and go to 100, but now, they're able to start around 70 and take it to 100. Obviously, fine-tuning is still required, since AI is not going to perfectly draft it the way the inventor wants. But, it's a significant timesaver," Sinha said. And save time it does. With 'Make In India' still a core tenet fuelling the Indian ecosystem, patent filing has seen a massive surge in the last year alone. Besides, with the Indian Patent Office publishing the Patents (Amendment) Rules, 2024, allowing patent filing to become more streamlined, the number of patents filed between March 2023 and March 2024 reached over one lakh. According to Sinha, the change in the rules and researchers prioritising patenting signify India's growing ability to compete with global players. Recently, Padma Shri awardee TV Mohandas Pai implied that the reason why India was lacking in the innovation sector was a disinterest in patenting our inventions. "Many patents attributed to Western companies are the result of work conducted in India, However, these patents are registered under the names of Western companies due to the global operations," he said. Comparing India to the US, he said that American researchers prioritise patent filing for larger strategic reasons rather than for immediate utility. Meanwhile, the need for filing patents in India is considered a hindrance rather than a priority. Acknowledging the same, Sinha said that as the ecosystem in India changes towards an urgent need to compete in the global patent race, the tides seem to be turning. "This is not just the filings in India, this is Indian people also filing internationally. Whereas traditional sectors in India used to be purely pharmaceutical and biotechnology, this has extended to smaller fields of science. It is in EV research, automotives, and green tech, with a lot of startups filing patents. It's gone beyond just biopharma," Sinha told AIM. He also mentioned that a majority of patent filers use AI for their research work. Patseer helps in the research phase of patent filing as a platform that allows for patent research and analysis using AI-enabled and expert search functionalities for patent researchers and filers. Sinha describes it as "a Bloomberg for patents", drawing a comparison to Bloomberg Terminal. He said that the use of AI means that large amounts of data can be scoured to ensure that patent filings aren't repeated. Additionally, it helps cut out the tedious parts of patent research, offering insights based on the information it takes in. While Sinha said that there are no hallucinations, as it is looking for and generating very specific information, there can be issues of false positives, which cannot be entirely eradicated. "But the aim is to reach as close to 100% as it can," he said. Sinha revealed that the Indian Patent Office also made use of Patseer's services to further expedite the patent filing process using AI. "Patseer is used for examination in the Indian Patent Office. They are a client of ours. Almost all examiners use Patseer for their patent searches on a daily basis. They have also upgraded their internal software systems to speed things up," Sinha said. Apart from this, as mentioned before, the office's amendment of the Patent Rules means a more streamlined process. Included in the amended rules are a reduced timeline for examinations, an increased grace period for filing patents (giving researchers 12 months to file a patent from the time of invention), and lower fees and discounts. All of these encourage Indian researchers to prioritise patenting their inventions, removing some aspects of unnecessary bureaucracy that hinder the ability of international patenters to take advantage of indigenous research. Additionally, the overall embracing of AI, including generative AI, from both the governance and research side means that patent filing in India could continue to grow exponentially.