I have just come across a data-mining tool for use in patent analysis, Spark.IP™.
In theory it is a place where people can advertise their willingness to license or sell their patents.
Much more interesting, I suggest, is the fact that this beta version, which at present is free, also offers the ability to mine patent data in US grants or published applications to create “patent landscape” information. The idea in such software is to make a visual display of interesting information about the patents, using data in them or relationships between them. I was aware that there exist companies which provide this sort of analysis for a price, but this is the first free one I’ve seen.
You could search for example for the term “wind turbine”. Numerous "cluster maps" appear, each for slightly differing wordings (and hence technologies). Within each, different wordings, shown by hexagons, are connected by lines. Larger hexagons indicate more activity, and darker colours indicate better matches.
This is what the help keys say, and I am not too sure what they mean. Apparently they rely on citations, where the search reports (as given on the front page of US patent grants) cite earlier patents as having some similarity. Presumably the bigger hexagons represent areas where there are lots of citations to the smaller ones, not the reverse. I had assumed that keywords were being used to link them, so that A has similar wording to B which has similar wording to C and so on.
“Assignees” (the owners of patents) can also be searched, where for example Google reveals cluster maps showing, not surprisingly, that their major work is in document information management and then in managing networked information. In this case I found it much clearer, if only intuitively, comprehending the data. In both cases further analysis is possible, such as the major companies involved for keywords, or a bar graph showing the numbers in each year. Titles of relevant patents can also be seen. It is possible to combine fields, so that for example an inventor in a company can be searched for.
It is not clear to me what exactly the maps are showing, and this (for me at least) limits the usefulness of the tool. I’d like to see the site offer a worked example where the precise implications of the links and the colours and size of the icons are explained in detail.
Comments