Semantic Searching – The Evolution of Keyword Patent Searching

Artificial intelligence (AI) is transforming the art of patent searching. What has traditionally been an expensive and time-consuming exercise is now becoming more cost-effective and accessible. However, is AI-based searching reliable and can it really replace a human?

One of the ways AI is having a major impact on patent searching is through semantic searching on text. Traditional patent searching has usually involved some form of keyword searching using human input, i.e. searching for exact terms. However, the process has several limitations and can be very time consuming. First, a strategy must be constructed by the patent searcher which involves brain storming all relevant terms and their synonyms and using Boolean operators (e.g. AND, NOT, OR) to connect these terms. Not searching for a particular term could mean missing a relevant patent. On the other hand, searching for terms that are too broad could result in an unmanageably high number of results. Broad terms and homonyms (words that are spelled the same but have different meaning) can also result in irrelevant results which must then be manually filtered out.

What is Semantic Searching?

Semantic searching allows you to enter a block of descriptive text, e.g. an article, invention disclosure or abstract into the search engine. The text is then analysed for context and meaning and a result set of the most relevant patents is produced. The results can be further refined by date, status, jurisdiction, keywords and patent classification codes to produce a result set that is more specific to your needs.

At Watermark, we use several searching platforms including Innography™ to provide a low-cost searching alternative for our clients. Innography™ is a patent searching and analysis software and one of the leaders in semantic patent searching.


It must be stated that the semantic search is not a comprehensive search. Instead, it is more suitable for providing an overview of the patent landscape. A comprehensive search will use several methods such as keyword, patent classification and citation searching in order to minimise the limitations carried by any one searching method. Furthermore, the semantic search will not analyse figures or drawings, so when these are used to describe significant aspects of the invention, the semantic search will be less effective.

As a semantic search is not comprehensive, depending on the reason for the search, human expertise may be required. In instances where the consequences of missing a relevant patent could be extremely costly, such as a Freedom-to-Operate (FTO) search, it is recommended that further, more comprehensive patent searching be performed. Semantic searching can be used to complement the search in these instances. The semantic search can also be used as an initial screening search. Once the results of the initial search are assessed, a decision can be made as to whether additional searching using other methods is required.

As search engines continue to mature, semantic searching will become more and more accurate. For now, it is a useful tool to gain an overview of a technology area, or as an initial or complementary search. Being aware of its limitations will help you decide whether a semantic search will suit your needs, or whether a more comprehensive search is required.


Whether you require a patent search to assess the novelty of your invention, or to stay on top of the latest developments in your field, consider a semantic search as a fast and cost-effective option, or as a first step towards your patent searching needs.


Author: Annie Nguyen