Patentee Name Harmonisation - ecoom.be
Patentee Name Harmonisation - ecoom.be
Patentee Name Harmonisation - ecoom.be
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
6.3 Introducing address information (in conjunction with name<br />
similarity)<br />
When engaging in name harmonizing efforts, it seems natural to consider the inclusion<br />
of address information of patentees such as country code, city name, zip/post code and street<br />
information. Address information can <strong>be</strong> used both for the additional identification of name<br />
variations (patentees with partly different names but identical addresses) and for the<br />
identification of potential mismatches (patentees having similar names but different<br />
addresses) 14 .<br />
Indeed, when both patentees share the same address, one might examine the<br />
possibility of harmonizing them. Given sufficient levels of name similarity (to avoid mismatches<br />
when different organizations share the same premises), there would appear to <strong>be</strong> a high<br />
probability that patentees are identical in these cases. In order to assess whether such an<br />
extension would <strong>be</strong> feasible, an analysis was performed to verify whether robust and<br />
unambiguous criteria could <strong>be</strong> outlined. Consequently, we examined a sample of 5,000 EPO<br />
applications. <strong>Name</strong>s have <strong>be</strong>en cleaned as descri<strong>be</strong>d in section 4 - A content-driven name<br />
harmonization approach focusing on accuracy - and addresses have <strong>be</strong>en cleaned in a very<br />
preliminary way (removal of all non-alpha<strong>be</strong>tical characters). For those patentee names having<br />
the same address (country, city and street), the Levenshtein distance measure has <strong>be</strong>en<br />
calculated and normalized for the varying lengths of names (absolute Levenshtein distance<br />
divided by the length of the longest names). The obtained matches have <strong>be</strong>en verified in terms<br />
of correctness (is it reasonable, based on a quick verification of patentee information found, to<br />
assume that both patentees are one and the same?).<br />
Table 6 contains the 25 cleaned patentee names with the same address with the closest<br />
relative Levenshtein distance out of the sample of 5,000 names. The absolute Levenshtein<br />
distance and the result of the validation is also included: “=” means that names are variants;<br />
“≈” means that names definitely have some relationship but not clear if it is the same legal<br />
entity; and “≠” means that matched names are significantly different (but still can point to the<br />
same legal entity; name can <strong>be</strong> significantly different <strong>be</strong>cause of name changes or mergers and<br />
acquisitions).<br />
Table 6: Differences in cleaned names of patentees with matched address<br />
CLEANED NAME CLEANED NAME WITH MATCHED ABS REL VAL<br />
ADRESS<br />
DIST DIST<br />
SCHNEIDERELECTRICINDUSTRYSAS SCHNEIDERELECTRICINDUSTRY 3 0,11 =<br />
MERRELLPHARMACEUTICALS MERRELLDOWPHARMACEUTICALS 3 0,12 ≈<br />
MITSUBISHICHEMICAL MITSUBISHIGASCHEMICAL 3 0,14 ≈<br />
THGOLDSCHMIDT GOLDSCHMIDT 2 0,15 ≠<br />
COSMAINTERNATIONAL MAGNAINTERNATIONAL 4 0,22 ≠<br />
KUMIAICHEMICALINDUSTRY IHARACHEMICALINDUSTRY 5 0,23 ≠<br />
TAKEDACHEMICALINDUSTRY WAKOPURECHEMICALINDUSTRY 6 0,25 ≠<br />
SUMITOMOMETALINDUSTRY SUMITOMOELECTRICINDUSTRY 6 0,25 ≈<br />
ACCENTURELLP ACCENTURE 3 0,25 =<br />
MITSUBISHIDENKI MITSUBISHIKASEI 4 0,27 ≈<br />
SHIBANAIAKIKO SHIBANAIHIROKO 4 0,29 ≈<br />
RHONEPOULENCRORER RHONEPOULENCSANTE 5 0,29 ≈<br />
GECMARCONI THEMARCONI 3 0,30 ≈<br />
FORDGLOBALTECHNOLOGY VISTEONGLOBALTECHNOLOGY 7 0,30 ≠<br />
BOEHRINGERINGELHEIM<br />
INTERNATIONAL<br />
BOEHRINGERINGELHEIMVETMEDICA 10 0,31 ≈<br />
AUGWINKHAUS FIRMAAUGWINKHAUS 5 0,31 =<br />
SGSTHOMSONMICROELECTRONIC STMICROELECTRONIC 8 0,32 ≈<br />
MITSUBISHIGASCHEMICAL MITSUBISHIKASEI 7 0,33 ≈<br />
ASAHIKASEIKOGYO ASAHIKASEI 5 0,33 ≈<br />
JOHNSONJOHNSONCLINICAL ORTHOCLINICALDIAGNOSTICS 11 0,33 ≠<br />
14 In addition, similar addresses appearing jointly with different patentee names might trigger an assessment<br />
of ownership relationships. Such an approach can <strong>be</strong> <strong>be</strong>neficial to support legal entity harmonization efforts.<br />
However, as explained in section 2 - <strong>Patentee</strong> name harmonization and legal entity harmonization, this lies<br />
outside the scope of the methodology outlined in this paper, which is aimed at name harmonization.<br />
15