Why patent search needs to be graph based

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.

There are tens of millions English patent publications. You have an invention and you need to know whether there is a patent publication about it already. To compare your invention against all the existing patents, you clearly need computers. But how do you explain your invention for the machine?

Keyword based searches

With Google Patents and the like, you give few keywords. Generally, you get way too many irrelevant results because you cannot define the invention completely. Oh, and you forgot to define a synonym or two, and now you also missed half of the relevant hits.

Full text searches

Then there are tools where you just give the computer the description of your invention in natural text. This is more reasonable, as you can define the invention completely, and the text may be useful elsewhere as well. Current progress in natural language processing has given some amazing tools, and we are closer to the point where computers can truly understand language.

Yet we are doing patent search with graphs. Why bother, when the fully textual approach is natural for humans and quickly becoming easier for computers? For the answer we need to consider both UI and AI.

Graphs are a great fit for patents

When you define invention in a graph, you can see the same structure the computer sees. With text, it is very hard to know what is the clearest way to write for a computer. Graphs also give nice visualisation opportunities. Soon we can pinpoint the relevant parts of the result graphs and you can trust the relevance of the search result quickly, without reading 30 pages of patent text.

‍

Screenshot from our graph editor

In the end of the day, the actual results are what will make the difference. In 2015, Stanford researchers managed to get better results with graph approach compared to the best LSTM benchmarks in two NLP tasks. It could have been a new general direction for NLP research but the models were difficult to make fast enough. Nevertheless, the situation has changed. Tools have improved and last Autumn we solved the Tree-LSTM performance issues.

Graph format makes even more sense for patents than for general text. The research with graphs and text has been focused around dependency trees where all the words are kept but the computer sees them in a semantic order. With patents, we don't need all the words. We parse only the technical core of the invention to the graph, and 30 pages of text becomes a graph of 1000 items.

The new tricks that graphs give

In machine learning, data quality and quantity are usually the most important factors, and graphs give few tricks for enhancing the data. We can create new training samples. If we remove items from the graph, the invention becomes more general and if something defined similar invention before, it would still be relevant.

Graphs can also be split. If all the pieces of your invention are found from a document, the document is a search result you want to find. This changes the game for two reasons. First, now we get to solve an easier problem, as it is much easier to answer whether a patent document contains some specific part of the invention. Second, even if we could only answer the easier question as well as the other patent search providers answer the harder question, we would still win. Why? As we have split the problem, we get more data about the match. If these individual samples are merely as accurate as the competitor's one sample, our accuracy is higher after we combine the results.

Final thoughts

Because of all of this, we believe the graph based approach is simply superior to the traditional patent searches. There is, however, quite a lot of extra work. For the full benefits, we have had to convert all patent documents to graphs. Also, it's easier to make the existing machine learning solutions to work with raw text than with graphs, as text is always text but graph structures vary a lot. Eventually the graphs are nevertheless going to take over, as the benefits are too great. Already we believe our search is the state of the art for some cases.

Read more

Webinar Q&A follow-up: your questions about the IPRally MCP answered

Thank you to everyone who joined Patent answers, inside the AI you already use. Across the two sessions on 4 November, you sent us more questions than we had time to answer live. We have grouped them by theme below, merged similar questions and included the answers given during the sessions.

Your IPRally workspace, wherever you work

Your AI assistant can now pick up where your IPRally work left off. The IPRally MCP now reads your collections, tags, search cases, projects, and agent runs, so your agent starts from results you’ve already vetted instead of a blank prompt.

Webinar Q&A follow-up: your questions about IPRally Drafting answered

Thank you to everyone who joined No more drafting in the dark: prior-art-aware patent drafting with IPRally Protection. Across our EU and US sessions, you sent us more than 80 questions, from file formats and drawings to claim clarity and US disclosure duties. We have grouped the answers by theme and merged similar questions. Some were answered live or in the chat during the sessions, and we include those here too.

Webinar Q&A follow-up: your questions about IPRally Intelligence answered

Your questions on IPRally Intelligence, answered: from working with larger datasets to verifying results at paragraph level.

IPRally is SOC 2 Type II compliant

IPRally has completed its SOC 2 Type II examination. Advantage Partners, our independent auditor, tested our security controls in operation from March 4 to June 4, 2026, and issued an unqualified opinion: no exceptions noted on any control tested.

When systems start talking: IPRally's journey toward a virtual IP department

Something new is happening: people talk to systems in plain language, systems talk back, and systems have started talking to each other the same way. As AI climbs from the small decisions to the large ones, the question for IP work is no longer whether it gets automated, but how to stay in control once it is. Co-founder and CPO, Sakari Arvela, lays out where this leads – through IPRally's new API and upcoming MCP support toward a virtual IP department that's always on – and why humans end up more important in that picture, not less.

Search, Intelligence, Protection: the three pillars of IPRally’s next chapter

For nearly a decade IPRally has been obsessed with getting one task right: patent search. Having built the best patent search in the world, we’re now reorganizing around three pillars – Search, Intelligence, and Protection – the newest of which makes drafting prior-art-aware from the first sentence. Co-founder and CPO Sakari Arvela maps out the next chapter, and why Protection is the shift he's most personally invested in.

NGen selects IPRally to bring AI-native patent intelligence to Canada's advanced manufacturers

After evaluating a range of tools on functionality, usability, and cost, Canada's Global Innovation Cluster for Advanced Manufacturing chose IPRally to power its Patent Search Adoption Tool Pilot Program.

IPRally and RPX Corporation announce strategic collaboration, cross-linking AI-powered patent search with RPX Empower

IPRally today announces a strategic collaboration with RPX Corporation, the global patent risk management service provider, that further expands the capabilities of the RPX Empower platform and provides seamless access to business critical complementary information for IPRally users.

Macro meets micro: IPRally's next chapter builds intelligence on search

Excellence in patent search was always the foundation. But foundations are meant to be built on. After years of staying focused – and saying no to a lot – IPRally is now ready to bring the same rigor and precision that redefined search to the bigger, strategic questions that businesses make decisions on. CCO Andreas Cehlinder explains why now, and what comes next.