Because the types of searches a user sometimes has to perform (Patentability/Novelty, Clearance/Freedom-to-Operate, Validity, State-of-the-Art and others) differs, a patent research tool—especially the content contained within it—needs to be “fit for purpose.” The ability to research the claims, to be able to distinguish between dependent and independent claims, to have access to Granted patents, even the means of searching terms embedded in drawings can all be critical.
1. Look for databases that offer the broadest and deepest full-text coverage. At the very minimum, most searchers need access to data from the USPTO, EPO, WIPO, China, Japan and S. Korea. But, depending upon where filing or opposition is taking place, consider major European authorities like Great Britain, France and Germany, as well as English language publishing countries like India, Australia and Canada. Coverage for these and many other authorities is quite comprehensive today, so try to ensure that the back files are complete and that front files are timely.
2. Ensure that full-text always encompasses the Titles, Abstracts, Descriptions, Claims and Citations. Some, “so called” full-text collections do not include the Description or only contain the first Claim. Ideally, look for collections where sub-elements of the Description, such as Prior Art, Embodiment, Examples, Experimental and Drawing Captions have been parsed and indexed and where Claim 1 or independent only Claims can be searched.
3. Look for databases where the text has been published directly from the XML or where the standard of OCR is high. Most of the major patent issuing authorities publish their full-text data as XML (eXtensible Markup Language). This is always to be preferred over databases created from OCR which can introduce errors into the data leading to false hits. Use of OCR sometimes fails to record ends of sentences or paragraphs accurately which can result in misleading hits where proximity searching is important and special characters such as Greek symbols and diacritics are not always picked up. Using OCR does, though, allow users to search text elements within tables, charts, drawings, diagram and figures.
4. Seek out databases where the search engine supports sophisticated searching in English and other languages. Full-text searches can generate “noise” but worse is not retrieving critical documents. Look for databases that allow you to search both in the language of publication, including non-Latin scripts, as well as in the language of choice. Insist on the ability to create complex searches in all languages.
5. Ensure that you are able to search within the PDF. Given that many full-text patent databases are derived from OCR document scanning of the PDF, look for solutions where the PDFs are also searchable. This has the added benefit in that portions of the PDF can be copied and pasted into reports, thus minimizing post process reporting errors.