Two Articles from the Proceedings of the 8th International Conference on Scientometrics & Informetrics


Abstract:
The first two of what will hopefully be many article reviews that may be of interest to people working in the field of patinformatics. The first article covers some tips for cross-country comparisons using patent information and the second discusses the use of artificial neural networks for mapping science and technology.

Enjoy!

Proceedings of the 8th International Conference on Scientometrics & Informetrics, Sydney, Australia July 16th-20th 2001:

Methods for Using Patents in Cross-Country Comparisons, Eric Archambault, pg. 41
Suggests only using data from granted patents as opposed to pending applications (makes the comparison between published papers versus papers submitted for publication) since overestimation rates (the percentage of applied for versus granted patents) varies from country to country. Distinguishing between contributions from individuals versus organizations (Which begs the question of how this can be done when some documents are unassigned and upon follow-up they are clearly part of an industrial research project. US pre-grant publications, even though from above, applications should not be used, do not even require that an assignee be named.). There is also some discussion of using inventor address information to establish patterns of collaboration (a very useful method which can also be used to map location to things like IPC code for getting an idea of what research is being done in what locations) and a word of caution about simply relying on the address of the first inventor as being representative of the entire document.

Using Artificial Neural Networks for Mapping of Science and Technology: Application to Patent Analysis, Jean-Charles Lamirel, et al, pg. 339
Authors give a nice, general overview of Self-Organizing Map (SOM, a form of Neural Network) technology and describe the use of MultiSOM for this particular application. The authors take advantage of use and advantage fields (it appears they used Derwent data for their study even though this is not explicitly mentioned) to organize documents based on different viewpoints (documents with a similar use are organized together in one view, similar advantages are organized in another, et...). The other viewpoints used were terms from the title and the assignee field (which they claim they normalized but I am wondering if they only looked at Derwent PACOs for this). Having multiple viewpoints allows an object in one view to be highlighted in another. This can be very powerful and may even have been a little ahead of it’s time. Tools from ClearForest and especially OmniViz can be used these days to see how a data point would be distributed across different views.

Posted: Wed - March 12, 2003 at 10:00 PM   Patinformatics   Interesting Reference Articles   Email Comments

© Anthony Trippe