Page 27 -
P. 27

(ω agg   = 4.3±0.1(γ) + 41.1±2.5, p < 0.1), which results from the prior art fields in the
                   aggregator websites and provides little new information on the value of the patent even
                   though the signal is amplified.

























                     Figure 8. Web hits versus future prior art citations from patents in 1995 (A) and 2005 (B)
                                                    solar energy patents

                   However, signal websites (ω sig  = ω −ω agg ) can provide unique information that diverges
                   from the predominantly linear correlation observed in aggregator Web hits because their
                   Web presence is the result of human intervention and is a measure of market importance
                   or commercial interest.  Indeed, the signal Web hits contain significantly more variability

                   and do not correlate (ω = 1.1±0.2(γ) + 17.4±2.8, p > 0.5).  Interesting outliers can be
                                          sig
                   identified when the patents have a significantly higher ratio of signal Web hits to
                   citations, such as those patents that are circled by the dotted lines in Figure 8.  These data
                   points highlight Web information, which may be unique compared to prior art

                   information. Qualitative inspection of the highlighted patents shows that the Web hits
                   result from an active description similar to the aforementioned Wikipedia example.  Only
                   one of seven highlighted patents has a high signal count that comes from a true error
                   wherein the search terms occur in websites; these websites that use the same terms as
                   designations for an unrelated company.  These data indicate that while the measure is
                   nuanced and requires some manual validation, the signal contained in these Web hits is
                   different from the citation data and worth further investigation.

                   To extend this initial investigation, we also compared two populations of patents, those
                   that were licensed (n = 81) and those that were not (n = 304).  An assumption is that
                   licensed patents are of more commercial or market importance than the unlicensed
                   patents [76].  We expect that they may have a different signature of Web hits as a
                   measure of that commercial impact.  Each patent received a mixture of aggregator (B)
                   and signal (A) Web hits (Figure 9), which appeared to have different signatures for
                   licensed versus unlicensed patents.




                                                             20
   22   23   24   25   26   27   28   29   30   31   32