This collection of refereed papers presented at the 6th Information Retrieval Facility Conference covers a variety of topics related to multidisciplinary information retrieval. Some of the highlights are touched upon here.
In “Multilingual and Cross-Lingual News Analysis in the Europe Media Monitor (EMM)” Steinberger addresses the problem of identifying names and all of their name variants, a challenge for multilingual news analysis systems. In “Ontology Based Query Expansion with a Probabilistic Retrieval Model” Bhogal and Macfarlane show that “ontology based query expansion enhances recall, and produces bigger improvements for [pseudo-relevance feedback] compared to [relevance feedback]”. Chenlo and Losada examine the use of structural features to detect subjective expressions for opinion extraction. Audeh, Beaune and Beigbeder propose Recall Oriented Measure as a way to evaluate recall.
Pharo and Nordlie conceptualise “search transitions” as a way to measure the effort expended by searchers. Shrestha, Vulic and Moens align broadcast news data with related written news data to improve the recall and precision levels of the Stanford NER system. Martin, Schockaert, Cornelis and Naessens find that attempts to filter calls for papers (CFPs) which rely only on the content of the CFPs do not provide the best matches for scholars, whereas inclusion of scholars’ publication histories improves the results.
Andersson, Lupu and Hanbury present a Dependency Claim Graphs to facilitate comprehension of patent claim texts. They show that “performance decreases when using existing general [Natural Language Processing] tools when changing text focus from the mainstream genre text towards a specific text genre”. They further add that “a support tool for patent experts needs to be able to deal with variations in terminology and linguistic features”.
Moumtzidou, Vrochidis and Kompatsiaris use footwear as a test case to extract concepts from patent images using recursive hybrid classification. This could be useful for patent invalidation, but future work is needed to determine scalability. Loizides and Buchanan describe a document triage framework which explains manual human information retrieval of using surrogates of articles and deciding what to include in a search. Salampasis and Hanbury present their Electra framework which “builds upon this idea of integrated professional search systems have four modules”.
This collection of papers is recommended to talent scouts in the information retrieval (IR) industry as well as to graduate students working to optimise IR.
