Technology’s Impact on the Four Phases of Searching

Many improvements in enhanced patent search systems relate particularly to these specific phases of the search process: 1) search strategy development, 2) strategy implementation, 3) records review, and 4) report preparation.

1. Search strategy development

I have become more productive in this phase by improving interviewing skills and developing search targets and goals as described in “Freedom-to-Operate Search Strategy: Hitting the Target” [3]. I often carry out limited initial searching in enhanced patent search systems prior to a client interview to build my knowledge of the technology and to characterize the scope and cost of a search. The process usually involves exploring patent classifications, which is facilitated by tools such as the patent Classification Explorer in PatBase. Alternatively, I may also use simple search tools such as Google Patent search or semantic search functions in enhanced patent search systems, particularly to avoid missing obvious references that clients may find on their own. Searchers never want to miss obvious references.

2. Strategy implementation

Enhanced patent search systems such as Derwent Innovation, Orbit Intelligence, and PatBase allow searchers to search and review original documents in ways that were previously impossible. New chemical structure searching in enhanced patent search systems facilitates chemical name searching and complements chemical structure searching functions in older substructure search systems, such as CAS Registry and Derwent Chemical Resource. Searchers input chemical structures, then the enhanced patent search system converts the query into text search terms in the background.

I usually supplement these structure search functions with my own text term searching, particularly for invalidity and FTO searches. This chemical search function works for documents in all of the principal languages and can be directed to claims or to full patent specifications. I would like to be able to direct structure searching to the combination of title, abstract, and claims, as I do in most text-term searching. As computing hardware and software improve, I look forward to be being able to search by chemical structure for highly posted, simple chemicals that currently exceed system limits and to use similar structure-based queries for polymers, Markush claims, and biological sequences. Developers should also add reactant, product, or other descriptive roles for these compounds and materials, which are available in some value-added databases.

Some enhanced patent search systems facilitate searching in non-English patent documents by accommodating non-Latin language search terms and providing for search in high-quality machine translations of the original documents. I anticipate systems will fully incorporate machine translations into the main search and review functions. They would replace use of separate search strings targeting machine translations or separate review screens for translated documents when search algorithms and database hardware improve adequately.

Patent citation searching and semantic or natural language searching are usually integrated in enhanced patent search systems. I conduct citation searching regularly because it is independent of terminology and patent classification. I have not found semantic or natural language searching to be particularly useful, but it is easy to use. Early search systems such as PatCafe that implemented latent semantic analysis a decade ago did not provide an appropriate means of reviewing results that we take for granted today in enhanced patent databases systems.

3. Records review

The combined benefits of advanced, color-coded, hit-term highlighting and review of machine translations of non-English documents are a big factor enabling nearly complete and efficient review of candidate patent documents. Hit-term highlighting, which may apply to full specifications and to keyword in context (KWIC) displays, enables easy screening and detailed review of documents.

Some systems have progressed beyond automated or manually designated lists of key terms for highlighting to allow for Boolean logic to designate the relationships between the terms to be highlighted. For example, in a search on coatings or treatments involving silica, the search logic might include (silica or silicon dioxide or silicic oxide or silicon iv oxide) near (coat … or treat). With this improved functionality, the search terms are highlighted only when they are in proximity rather than in every occurrence in the KWIC or full displays.

Improved access to patent legal status is another helpful aid for FTO searching. Some systems have integrated simple visual indicators for being “alive” (potentially enforceable, perhaps in green) or “dead” (no longer enforceable, perhaps in red) for individual patent family members. At this time, most systems do not allow searching these legal status attributes. When it becomes available, searchers will have to be cautious because of the potential for “dead” documents being revived, such as by the payment of late maintenance fees, litigation, or the possibility of overcoming other delays in prosecution. In theory, one could craft a search to try to eliminate from consideration older documents that may truly be stone dead. In addition, I frequently report “dead” patent documents in FTO reports to allow client attorneys to look for safe harbor references and deal with uncertainty in legal status.

4. Report preparation

Clients appreciate getting selected claims and excerpts of specifications that support the selection of a patent document for further consideration. This used to be done by multiple copy-and-paste steps during the review process. Now some search systems have added convenient, time-saving functions that allow searchers to create notes associated with patent records or patent documents. These notes may then be exported with each selected patent record and processed for inclusion in the final report. I provide links in spreadsheets to full patent records and to legal status summaries when provided by patent search systems. I would like to be able to provide links directly to particular sections of patent text, such as claims or examples, and to have clients see records with highlighted terms.

The Future for Patent Searching

My ability to carry out higher-quality searches and provide better search reports, as well as my overall increased productivity, correlates well with increased reliance on enhanced patent search systems relative to value-added databases. Enhanced patent search system developers deserve tremendous credit for providing professional patent searchers with tools that make our jobs economically competitive and enjoyable, particularly in the face of ever-increasing numbers and languages of patent documents. They have done this even as they have designed their search systems to accommodate casual searchers as well. In particular, they have developed very critical interfaces that have significantly decreased the amount of time needed to analyze individual patent documents—claims, specifications, or both—for invalidity and FTO searching. For example, the ability to focus on search terms with advanced hit-term highlighting and the availability of more and better machine translations mean that searchers are less likely to miss important content and can make better decisions about accepting or rejecting patent documents.

Further improvements will undoubtedly come with advances in computing resources. I look forward to complete integration of machine translations in search and review processes. Machine translations would be treated exactly as original language documents and would be used by searchers as an equal information source. Searchers would still have to recognize that machine translations are not perfect and do not have the legal status of certified expert translations. Patent searchers should always be conservative in their judgments in order to provide all appropriate information to their clients.

Enhanced patent search systems will continue to create new alternatives to standard text term searching. Chemical structure searching should be followed by similar functionality, enabling searching for Markush claims, for addition, condensation and natural polymers, and for biological sequences. Information must be efficiently retrievable from tables; the relationship between elements in table rows is valuable and should be search able with appropriate proximity operators applicable within individual or nearby rows and columns. These systems should provide access to textual and non-textual information in figures and images. Searchers would like to search for specific visual information contained in utility patents just as they might in design patents. While inventors are expected to describe figures within the specification, there is always other information that could be learned from the images themselves.

These systems should add numerical property searching by indexing numbers and their units within specifications, thereby providing full numerical searching. Such searching is currently available in some value-added databases and could be implemented in enhanced patent search systems much in the way chemical structure searching was. For example, this functionality would enable searching for a boiling point range in either Fahrenheit or Celsius, or particle size range in microns or nanometers. In each case, the property should be searchable in proximity to a target substance or material with the required numeric property.

Predicting Revolutionary Changes

Up to this point, I have been describing improvements that evolve in the course of product development. Revolutionary changes are more difficult to predict. I recall first hearing about natural language processing a couple of decades ago and auto mated chemical structure assignment in patent documents a few years ago. The latter quickly became implemented by commercial patent search systems, whereas the former has yet to make its mark.

The next extension of natural language processing seems to be the application of artificial intelligence (AI). IBM calls this augmented intelligence, because such enhancements aim to work along with searchers during the course of their work. This was a big topic at the PIUG 2018 Annual Conference and elsewhere. My concern is that AI seems to offer yet another way to find candidate references. That could be useful, particularly to casual searchers and some patent examiners, but is not likely to make a significant difference in how professional patent searchers do their work.

I understand that there is interest in using AI processes to improve the patent review process. Systems are being developed that would learn from the searcher’s acceptance and rejection of candidate records, presumably by updating the ranking of candidate record sets. I leave it to others to program how an AI system could determine the rationale that a searcher used for rejecting a patent, particularly in FTO or invalidity cases which require detailed evaluations of claims or specifications. In addition, I would be concerned about using AI systems that were not integrated with patent search systems from strategy development to report generation. Developers of new systems will make inroads with professional patent searchers only if they recognize that we rely on such well-integrated systems.