Activity Based Metadata for Semantic Desktop Search

Abstract

Abstract. With increasing storage capacities on current PCs, searching the World Wide Web has ironically become more efficient than searching one's own personal computer. The recently introduced desktop search engines are a first step towards coping with this problem, but not yet a satisfying solution. The reason for that is that desktop search is actually quite different from its web counterpart. Documents on the desktop are not linked to each other in a way comparable to the web, which means that result ranking is poor or even inexistent, because algorithms like PageRank cannot be used for desktop search. On the other hand, desktop search could potentially profit from a lot of implicit and explicit semantic information available in emails, folder hierarchies, browser cache contexts and others. This paper investigates how to extract and store these activity based context information explicitly as RDF metadata and how to use them, as well as additional background information and ontologies, to enhance desktop search.

Semantically enhanced Information Retrieval: An ontology-based approach

“…Scope: semantic search has been applied in different environments such as the Web (Finin, Mayfield, Fink, Joshi & Cost, 2005;Fernández et al, 2008), controlled repositories (Popov, Kiryakov, Ognyanoff, Manov & Kirilov, 2004), or even the desktop (Chirita, Gavriloaie, Ghita, Nejdl & Paiu, 2005). Obtaining conceptualizations to cover the meanings involved in all Web content as well as the automatic annotation of these conceptualizations with some degree of completeness is still an open challenge.…”

“…The semantic search model detailed in Section 3, as well as other semantic approaches that have proved to work well in specific domains (Kiryakov, Popov, Terziev, Manov & Ognyanoff, 2004;Chirita, Gavriloaie, Ghita, Nejdl & Paiu, 2005), still have to undertake further steps towards an effective deployment of semantic search on a decentralized, heterogeneous, dynamic and massive repository of content such as the Web. Accomplishing this objective involves tackling several problems such as:…”

Leveraging personal metadata for Desktop search: The Beagle++ system

“…NRL was developed based on RDFS 12 , so that all RDFS schemas are legal NRL, but only RDFS schemas following NRL usage recommendations are valid NRL. Significant additions relatively to RDFS include named graphs and graph views.…”

“…Normally, a simple search for "Hannover" would be unable to retrieve the pictures in that folder. The PathAnnotator component annotates files with every token appearing in their file path, as well as additional semantic information provided by the WordNet system 17 , such as direct synonyms, hyponyms, hypernyms, meronyms and holonyms [12], for each of these tokens, but naturally excluding the trivial tokens as "/home/". This kind of additional information is particularly important since for specific types of files, which do not provide abundant textual information, the search process can rely upon the text generated for this resource, based on the effort that the user invested to organise her files in a hierarchy.…”

Beagle + + : Semantically Enhanced Searching and Ranking on the Desktop

“…Similar to the discussion above, various semantic relationships exist in other contexts such as file and web cache context. Due to space limitations, we refer the reader to [5], where we proposed several solutions to enrich the information associated to file and directory names, as well as to previously visited resources on the Web.…”