Intellectual Assets Management: From Patent to Knowledge Paper Appears in World Patent Information


Abstract
The June 2003 issue of World Patent Information contains a paper from an Italian group which has put together a patent collection and analysis tool that they are calling ArchIPat. The software was developed by an academic spin-off within the AREA Science Park of Trieste. A user can conduct a search using the system and select which patent provider they would like to have the data downloaded from. Once retrieved ArchIPat will create a relational database from the data (in either Oracle or Access format) and allow the user to analyze the data using various text and data mining methods.

I have been saying for some time that Oracle would make a good database system for building a very large relational patent database. Most of the commercial efforts are built on SQL and within recent enhancements have
learned to scale quite well for the amount of data that is typically available (many millions of full-text patents and images) but originally had issues. I would like to see what a really large database of relational patent data looks like in Oracle.

In some ways this system looks a little like MicroPatent's Aureka on a budget. The authors claim that the goal of this project is to provide an inexpensive, non-proprietary desktop patent analysis system. This is a laudable goal and may in fact help bring patent analysis to more companies and people who are interested in the topic. The idea of also only downloading and working with the patents of interest instead of building and maintaining a huge database is attractive. It will be interesting to see if it becomes frustrating to users if they have to wait for an extended period of time while their documents download and are processed before they can start/continue doing their work. Most people will probably decide to set up a download and let it run overnight so they can work with the results in the morning. Some features of the system include: Information Retrieval: once documents are downloaded the user can see the full-text and image, see the cited references, select the patent text and open a notepad for keeping references, send or save patents in a desired file format or share the patent or analysis within the company over the LAN (local area network). Data and Text Mining: a co-occurrence matrix can be browsed and expanded upon and Hierarchical and Binary Relational Clustering can be conducted on the patent text. Work with Non-Patent Information: a sister system for working with other data sources (including PubMed) called ArcHirudO can be connected with ArchIPat. The paper was a little light on images of the system and since I have not actually used it I can not say much more on well it performs but in principle ArchIPat looks like something to look for in the future.

Posted: Wed - April 30, 2003 at 10:04 PM   Patinformatics   Interesting Reference Articles   Email Comments


© Anthony Trippe