One of the main reasons for the Information Retrieval Facility's (IRF) existence is to find workable solutions to the challenges faced by Information Retrieval specialists operating with IP data.

The 2007 event concluded by sending all those concerned on their way with the remit to come back next year with something workable. This morning the project presentations got underway.

The first project we were shown looked at Semantic Annotation for IP. Presented by one of the project leaders Valentin Tablan, research fellow at the University of Sheffield, its main goal since the project began last February was to find a way to achieve semi-automatic semantic annotation with larges scale document collections.

I won't pretend to have fully understood some of the more complex issues about Semantic Annotation nor the complex structure of the system he and his team are developing, but from what I did see it's all about applying the annotation information (the method of adding metadata to the relevant parts of a document). Thereafter this is subjected to an ontological process which will give far more related information to data through a hierarchy of relationships in the terminology of a patent filing. Therefore this will improve the accuracy of results. However the conclusion was that for something even better, a hybrid model should be adopted that utilised both semantic annotations and knowledge management technology.

Another method on show was Text Mining, the goals of Cornelis Koster's project was the development of a deep linguistic search engine and in tandem to find an accurate parser - the analysis of linguistic structure (to put it loosely) for complex documents. By combing these to processes, Koster's aim was to create a Text Mining system that was tailored for searching through IP material.

An interesting point Koster made was that patent searchers prefer to search using older technologies like Boolean over ranked search because of the transparency that Boolean offers. Its ability to offer the searcher more control and precision with results makes it the preffered choice in many cases. This he said was what his project (PHASAR) would also offer with exact matches returned rather than ranked matches. The recepetion from delegates (who i don't think are easily won over) was warm.

The final project to dazzle us with formulas came from the University of Glasgow's Leif Azzopardi. Credited with being involved in two related world-leading projects; his task in the Findability Project has been to map how easy it is to find patents depending on the retrieval systems used. This examined the success of results by comparing the differences between Exact Match and Best Match systems as well as the analysis of how retrieval systems shape the access to patent information.

The audience seemed to like what they heard with the words 'fascinating' and 'inspired being thrown around on more than one occasion during the post-presentation Q&A session.

All in all, it would seem that the ambitions of 2007 are on their way to being fulfilled if not quite there yet.