Fraunhofer FOKUS is receiving a US patent
Intelligent search overcomes language barriers and knowledge gaps
The US Patent and Trademark Office has granted a patent for a technology developed by Fraunhofer FOKUS. The invention forms the basis for a novel semantic search function that guides users to the relevant information even when they do not use the same terms as the authors of the content.
Companies and various specialist fields often develop their own terminology. Those searching for information are often unfamiliar with the technical language used. Conventional search engines quickly reach their limits here, as they primarily search for exact terms. This is changing thanks to a technology developed by Fraunhofer FOKUS, for which a US patent (US-12775651-B2) has been granted under the title ‘Method and Device for Pre-Selecting and Determining Similar Documents’. Instead of searching for exact terms, the technology focuses on the meaning of words and their use in the respective context. This means that similar or semantically related terms can be indentified and utilized in the search.
Technology reduces maintenance effort
Until now, knowledge gaps between search queries and documents were often bridged using synonym lists, thesauri or ontologies. However, creating and maintaining these is time-consuming and labor-intensive. The method developed and patented by Fraunhofer FOKUS uses modern clustering techniques to automatically identify similar terms from documents, terminologies or search queries. In this way, search results can be improved without the need to manually build and maintain additional terminologies.
Patented solution improves search accuracy
Other search methods in vector databases also use similarity or distance measures and must limit their results to a fixed number. However, whilst these previously lacked a criterion for limiting results based on content, the new patent has found a way to restrict the search to the most important and most similar alternative terms. This automatically limits the document search. At the same time, the alternative terms used make the search results easier to understand.
The technology is particularly suitable for organizations with large information repositories, complex specialist fields or a high volume of search queries that wish to avoid time-consuming modelling. The solution can be combined with tried-and-tested open-source search engine technology to guarantee fast response times.
Last modified: