INCREMENTAL END-USER QUERY CONSTRUCTION FOR THE SEMANTIC DESKTOP Ricardo Kawase, Enrico Minack, Wolfgang Nejdl L3S Research Center, Leibniz Universität Hannover, Appelstrasse 9a, 30167 Hannover, Germany Samur Araújo, Daniel Schwabe Departamento de Informática, Catholic University of Rio de Janeiro Rua M. de S. Vicente, 222 Rio de Janeiro RJ 224530-900, Brazil Keywords: Semantic Desktop, User interface, Visual query system, Graphical query language, Ontology, RDF, SPARQL, Personal information management. Abstract: This paper describes the design and implementation of a user interface that allows end-users to incrementally construct a query over the information in the Personal Information Management (PIM) domain. It allows semantically enriched keyword queries, implemented in the Semantic Desktop of the NEPOMUK Project. The Semantic Desktop user is able to explicitly articulate machine-processable knowledge, as described by its metadata. Therefore, searching this semantic information space can also benefit from the knowledge articulation within the query. Contrary to keyword queries, where it is not possible to provide semantic information, structured query languages as SPARQL enable exploiting this knowledge explicitly. 1 INTRODUCTION same, provided that the appropriate tools are available. Whereas traditional folder-based storage and To allow the personal information management navigation-based retrieval used to be sufficient for in the Desktop we first need to introduce an our personal information management needs, the ontology language that can be used to express personal mental models – the personal information sheer number of items on our Desktop calls for the 1 integration of Desktop search in its interfaces. The model ontology (PIMO) . The use of such an paradigm of Desktop search is similar to Web ontology gives the user a better understanding of her search, but shows some interesting differences: Desktop and her tasks, and a way to express her We typically search for items that we have knowledge about the information items. By adding created, received or stored earlier – which explicit, processable meta-data to the information means that we try to re-locate (re-find) rather items in the Desktop, it becomes a “Semantic than to find (discover) something new. Desktop”, a concept we will elaborate later. We typically have a (most likely incomplete) As it has been identified in previous works picture of how the data on our Desktop is (Teevan, 2004) (Bates, 1989) (Belkin, 1993) a organized (which means that navigation might progressive modelling search activity can provide be more apt than keyword search). better results than a single, static search. Yet, strong methods and interaction mechanisms are still From Teevan et al. (Teevan, 2004) we know that missing for searching the Desktop. users use a combination of search and navigation to In this paper we fill this gap by presenting a relocate information on the Web. For personal methodologically designed interface for the information management this is likely to be the 1 http://www.semanticdesktop.org/ontologies/2007/11/01/pimo/ 270 Kawase R., Minack E., Nejdl W., AraÞjo S. and Schwabe D. INCREMENTAL END-USER QUERY CONSTRUCTION FOR THE SEMANTIC DESKTOP. DOI: 10.5220/0001825902700275 In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page ISBN: 978-989-8111-81-4 Copyright c 2009 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved INCREMENTAL END-USER QUERY CONSTRUCTION FOR THE SEMANTIC DESKTOP (Semantic) Desktop search that combines and environment that is based on the NEPOMUK exploits current interaction mechanisms. It benefits architecture. Since NEPOMUK still requires some from what the user knows (the vocabulary and the semantic knowledge from the user, user-friendly mental model), respects what the user does not know interfaces are a crucial milestone to achieve the goal (the data structure and query languages), to finally of bringing the Semantic Desktop to the common give her what she wants. user. Walking in a two way path, first we aim at In the remaining of this paper we discuss in more designing interfaces that solve the user’s needs on details searching activities in the Semantic Desktop the Semantic Desktop, and on the other direction, we and our proposed work in Section 2. In Section 3 we design interfaces to show the user the potential of describe our implementation and the architecture. the Semantic Desktop and still hide its complexity. Finally, in Section 4 we draw our conclusions and In both cases, first we take a look at the way people the sketch future work. think and express their mental models, so that we understand how the Semantic Desktop can support this (Sauermann, 2005). 2 SEARCHING THE As we mentioned before, we use the Personal Information Model Ontology (PIMO) to describe SEMANTIC DESKTOP and work within the PIM in the Semantic Desktop. The PIMO forms the basis for all custom, user- In (Sauermann, 2005), the authors argued, that the created types and relations. It defines basic types typical user uses the (information on the) Desktop to such as Document, a Person, a Location, etc. and complete a certain task. For this, documents that are relations such as creator, hasLocation, etc. and is relevant to the user’s current task are retrieved, intended to be extended by the user in any way he or processed and stored. Such documents contain she likes. Note that we use the terms “type” and relevant information that is processed by the user, “relation” throughout the paper as user-friendly allowing the user to generate knowledge. This terms for RDF Class and RDF Property. We also use knowledge is implicitly stored in the documents. these less technology-based terms consistently in our Making this implicit knowledge explicitly user interface and the implementation. expressible and machine-processable, is one of the The user can extend the PIMO ontology and use goals of the Semantic Desktop. Allowing the user to it to articulate arbitrary knowledge in an explicit exploit knowledge for retrieval at the Semantic way. For the task of re-finding information on the 2 Desktop, as implemented in the NEPOMUK Semantic Desktop, NEPOMUK provides two project, is the goal of our proposed user interface. different mechanisms. First, a type and instance The definition given by Sauermann et al. browser that is analogous to most operating system (Sauermann, 2005) depicts, that the Semantic file browsers where users type hierarchy and the Desktop paradigm brings the ideas of the Semantic instances of each type. Alternatively, there’s a full- Web paradigm to the user’s personal Desktop where text keyword search that analyses the extracted the conceptualization of the personal mental model metadata from the instances returning a relevance is described in formal ontologies. The standard data sorted list classified using complex ranking format for a common representation is RDF algorithms. (Resource Description Format). Finally, the different It happens that in both cases the potential Desktop applications are integrated using the same combination of user knowledge and system concept of the Semantic Web, for exchanging data functionality is not fully merged. The browser does and accessing resources. not allow the user to input her knowledge about the The NEPOMUK project integrates research, instances and its relations. Conversely, the keyword industrial and open source community efforts to search does not permit the user to use her knowledge develop a new technical and methodological of her PIMO. platform: the Social Semantic Desktop. This is an Consider the case when the user is looking for an extension of the personal Desktop that aims at email (or a presentation document) that was sent (or collaboration and personal information management. created) by a certain person, containing a certain 3 The NEPOMUK framework PSEW (P2P information (e.g., a telephone number or a quote). In Semantic Eclipse Workbench) is an integrated a pure keyword-based interface, the user should input a query such as “email sent by person 2 http://nepomuk.semanticdesktop.org/ telephone number”. However, this is very unlikely, 3 http://nepomuk-eclipse.semanticdesktop.org/xwiki/bin/view/ since most purely keyword-based interfaces assume Main/PSEW 271 WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies the search will be made exclusively in the contents, the solution. The results are given based on the and it would not be expected to understand the records that match the pattern of this table. semantics of “sent by”, or the type of item (email). Considering the graph properties of the RDF Being able to leverage such semantic information model (Pérez, 2006), our solution was based on greatly reduces the search space, since, for instance, allowing the user to create a graph visually and only a small fraction of items in the Desktop would incrementally, which then is translated to a graph- have a relation “sent by”. matching query language. Analogous to what was A similar argument can be made when the query done by Zloof, we are giving to the user a visual must span several items, e.g., “email sent by the model where she can provides examples of the author of a given presentation”. In such cases, reality it wants to get. The incremental feature of our keyword-based searches would, at best, retrieve solution is that each new node added to the graph, partial answers, that must be combined by the user the user is able to view the intermediate result of her by filling in the semantic information that allows the query. binding between the partial answers returned. Below is an example of how the user can make a The user would now need to know the exact query using a graph structure. Suppose you want to semantic terms that are used in her semantic get all emails sent by John. This query can be made repository.