A premise of our patent system lies in striking a bargain with the public. A monopoly of limited duration is granted in return for an invention’s disclosure. The hope is that patent publication will “promote the progress of science and the useful arts” by publicizing new and innovative ideas.
Underlying the bargain is an assumption that issued patents and published applications are available to the public, and the public upon conducting a search can find relevant information. However, patents and legal documents in general are unique compared to other knowledge. In the realm of patents, relevant search results need to identify whether a patent contains relevant or related knowledge as opposed to an exact match of search terms. While this is the same type of problem that plagues searches for legal material in general, challenges facing patent searchers are increasingly acute. Patents are increasingly global in scope and require searching across diverse classification schemes and languages.
The problem of accurate and precise patent searches has the potential to undermine the goals of the patent system by impeding progress and wasting investment dollars. For example, in May 2010, Rolls-Royce sued United Technologies Corp.’s Pratt & Whitney unit claiming infringement of a method of manufacturing quieter and more efficient jet engines. Pratt & Whitney invested more than $1 billion to independently develop a more-efficient design. (See Rolls-Royce PLC v. United Technologies Corp., 10cv457, 2009 U.S. Dist. LEXIS 127214 (Alexandria)); see also interesting counter-suit as reported by Bloomberg.com).
You can be sure that a sophisticated party such as Pratt & Whitney performed a freedom to operate search and analysis. Yet such searches are inherently difficult and time consuming. I would like to highlight a foreign approach that has been around since 2006, but has recently received an important improvement that will improve patent searching. The article is forthcoming in a technical journal, but may pass below the radar of mainstream legal publications.
In 2006, at the first international conference on Semantics and Digital Media Technology in Athens, researchers presented an article titled, PATExpert: Semantic Processing of Patent Documentation outlining the technical details of their algorithm. PATExpert is an EU funded consortium, and includes the European Patent Office as a partner. (For a competing and different approach to patent searching used in the Korean Patent Office, see Segev, A.; Kantola, J. Patent Search Decision Support Service, 2010 Seventh International Conference on Information Technology: New Generations (ITNG), Digital Object Identifier: 10.1109/ITNG.2010.99.)
In distilled form, the PATExpert software relies on an ontology to create relationships between data common to different patents. In addition to traditional linguistic matching, the software attempts to return results based on how a particular patent is related to others. In its current state PATExpert uses the following search methodologies: full text search, metadata search, image similarity search, semantic search, and document similarity search. Few or non-existent connections indicates that there is no relationship between two patents. However, a large number of connections between two patents signifies a strong relationship, and despite linguistic differences may indicate two patents have a lot of substance in common.
For example, images such as photographs, diagrams, flow charts, and drawings are important factors in the representation of a patent’s disclosure and are modeled by an ontology. Additional data such as the patent holder, inventor affiliation, and assignee are also captured in an ontology. An additional ontology captures linguistic similarities. By looking at each ontology independently or as a combination, a representation of related patents may be obtained. The hope is that more relevant results, independent of linguistics, can be returned in performing a patent search.
The forthcoming article Iterative Integration of Visual Insights during Scalable Patent Search and Analysis by Koch, S; Bosch, H; Giereth, M; Ertl, T (IEEE Transactions on Visualization and Computer Graphics, Volume: PP , Issue: 99 Digital Object Identifier: 10.1109/TVCG.2010.85) outlines development of a visual interface for the PATExpert algorithm. The interface called “PatViz” makes PATExpert user friendly and potentially more powerful. Using PatViz, users can now graphically construct patent queries, similar to constructing a flow chart using Microsoft Visio. Furthermore, and perhaps more important than query construction, each step or node in a query can be enlarged to show the patents that comprise the results returned at the particular node.
Take a simple example as described in the article. In searching for patent that contains the word “disk recorder” and not the terms “CD” or “DVD,” a user can enlarge each of the nodes containing intermediate results without the term “CD” or “DVD” or enlarge the node containing the term “disk recorder” without any other restrictions.
Data at a particular node, or at a final node, may be represented in 11 different visualizations. For example, one can view a graph of how results relate to different subject matter classification schemes or which results are U.S. patents. Another helpful representation is a “Term Cloud.” A page of text is displayed containing all search terms used. PatViz highlights the most frequently found terms by enlarging the font of the term, the largeness of the font corresponding to the frequency of the term in search results. This gives much more of a “feel” for the content and scope of a patent search.
The article is full of technical details, but important for patent lawyers is the possibility of easier and less time consuming freedom to operate analysis. How these more efficient patent searchers will affect patent prosecution and litigation remains to be seen.