Context, Data and Queries Annika Hinze1 , George Buchanan2 , Andrea Schweer1 , and Matt Jones2 1 University of Waikato, New Zealand University of Swansea, United Kingdom
[email protected],
[email protected] [email protected],
[email protected] 2
Abstract. Context is typically viewed at as meta-data that supplements the ‘real’ data of an application. In this paper, we contrast and compare the concepts of context in four different applications. We show that our notion of context needs to be extended and that wider and more flexible notions need to be supported. In particular, a key issue is how to provide systems that support computation in regard to any context.
1
Introduction
Research into context-aware systems has received increasing attention, especially with advanced technology fostering mobile systems and location-aware systems. However, the term “context” and the use of context continues to develop; the concepts used mirror the different application focus and background of the researchers. This paper compares and contrasts the concept of context in four different applications. For simplicity we concentrate on the context of location. The paper is structured as follows: Section 2 introduces the four context-aware systems. Section 3 compares their notion of context. Section 4 suggests a taxonomy to capture the different characteristics of the systems and their context models. We conclude the paper in Section 5 by highlighting and summarizing our observations.
2
Four Context-Aware Systems
Tourist Information Provider (TIP). The TIP system delivers location-based information to mobile users [2]. The information delivered is selected and its presentation configured using a user’s context, such as their current location, their interest in semantic groups of sights and topics, and their travel history. Examples for sight groups are public art, buildings, or beaches; topics may be history or architecture. The system implements a number of additional services such as location-based recommendations. Greenstone Digital Library. Greenstone is an open source digital library toolkit [7]. Digital libraries that have been created using Greenstone include collections of historic newspapers, books on humanitarian aid, first editions of works by Chopin, scientific repositories to personal collections. The TIP/Greenstone bridge [3] provides location-based access to documents in the digital library. R. Meersman, Z. Tari, P. Herrero et al. (Eds.): OTM 2007 Ws, Part I, LNCS 4805, pp. 222–225, 2007. c Springer-Verlag Berlin Heidelberg 2007
Context, Data and Queries
document
user
item
query
223
user
query
A B
location
location
location
(a) TIP
(b)
location
Greenstone
location
location
(c) QnA
location
location
(d)
Dg.Parrot
Fig. 1. Agents and location context
Questions-not-Answers (QnA). The QnA project investigates the potential utility of queries made by mobile users as an information context [5]. Specifically, search engine queries made on the move can be coupled with details of the place at which the search occurred. The aggregation of searches made at a place may give other mobile users an insight into the function or nature of a location. For example: many searches made regarding “Claude Monet” “Impressionism” and “Renoir” may suggest a relationship of the location in question to art. Digital Parrot. This project is developing a context-aware augmented memory system – a personalised software that runs on a mobile device [6]. The digital parrot supports people while they are attending conferences or during travels; it helps them remember information they already knew but forgot, by recording incidental information. Examples are names of people, topics of conversations, and under which circumstances the user met a person before. The system allows automatic information capture as well as manual information input. Stored data can be accessed based on whichever piece of information the user remembers to help in recalling the user’s experiences.
3
The Concept of Context
To make comparisons clearer, we focus here particularly on the context of location. Traditionally location-based systems query for data that is related to a given location. That means, the query carries location information about the agent and the data have searchable location meta-data attached. The TIP system uses this traditional approach: The user queries for data items by specifying their location context; the query retrieves items whose location is in close proximity to the location given in the query. TIP uses explicit context as meta-data for user and items (see Fig. 1(a)). Greenstone traditionally supports both full–text search and meta-data–based access. For locationbased access via the TIP/GS bridge, location information (place names) is used to retrieve relevant documents. The location markup was added to the documents in a preprocessing step. Within the documents, further location information is used as hyper-link points to other documents supporting location-based browsing (see Fig. 1(b)). The digital parrot has the most open approach to context: every concept covered in a user’s data model may turn into contextual
224
A. Hinze et al. Table 1. Taxonomy for example systems
Task
TIP Retrieval of items based on proximity to user location
Greenstone search and browsing for documents based on location
QnA Digital Parrot retrieval of query access of items terms based on lo- based on linked cation network
location
User infl. direct History immediate, rect history Collab. indirect Reflect. indirect
key
rd wo
item
item
item
document
R. space
location
er y qu
location
direct none indi- immediate, no his- historical only tory none essential none direct
ta da
location
direct and indirect immediate, history possible none limited
information for a query. The user’s recorded items are stored in an interlinked network; the direction of querying the network depends on the user’s way of recalling past experiences. Location may be the context of either query or answer, or it may be the requested data point. Fig. 1(d) illustrates location as context of a conversation. The QnA approach to context and location is contrary to the traditional one described earlier: The user is presented with a map of query terms and their location context (which does not appear explicitly in the query) – see Fig. 1(c). Reflection on context is here used as trigger for further user actions.
4
Taxonomy and Comparison
We introduce a taxonomy to compare the four systems: The Task describes the main interaction of the user with the system and the systems activity. The Retrieval space captures the available dimensions of data, meta-data and retrievable items. User influence captures the ability of the user to influence the system strategy and the data used for context taken into account. History captures the time period the data retrieved in a user interaction will be available for use. Collaboration between people may be required or facilitated when using the system. Reflection is an HCI interaction paradigm that describes a support for user (inter)action triggered by the outcome of previous interactions. Table 1 shows an overview of the systems using the introduced taxonomy. Note especially the different dimensions of the retrieval space and the possible query directions (indicated by the arrows).
Context, Data and Queries
5
225
Conclusion and Observations
A key concept in hyper-text is the viewing of any hyper-text link as a query, pre-computed to arrive at a specific hyper-text location [1]. The actual link is a concrete, fixed reification of a more abstract logical connection between two documents. This concept has already been powerfully exploited in the creation of the PageRank algorithm for information retrieval. Similarly, the location links between Greenstone documents form a network based on concepts of location and place. The QnA project exploits one natural corollary of this concept: that queries made in a place reveal something of the nature of the location, just as the links between documents reveal conceptual links. Furthermore, information science researchers who study the behaviour of users engaged in information seeking and retrieval have specifically identified the significant issue of the user’s own conceptual context – i.e. what is “in their head” [4]. A query, thus, implies a hidden context not directly encoded in the query itself (The digital parrot endeavours to tease out and use these implicit connections). In short, links and queries both provide a context upon which computation can be performed, and imply contexts such as the user’s interests or physical location. Specific context can also be applied to a query or link – e.g. its document of origin, or place of execution. The different dimensions in the retrieval space for the projects may be likened to data cubes in data warehouses: but there only certain dimensions are available for drill-down retrieval. This rich combination of queries, links, and context provides a powerful framework for further research. Existing systems limit which features (queries, links, metadata) can be configured by or used as context. A universal solution that permits any feature to play any role is a profound computational challenge. The projects discussed in this paper provide examples of this potential, but much more research is required to abstract, systematize and organise this area of study.
References 1. Golovchinsky, G.: What the query told the link: The integration of hypertext and information retrieval. In: Procs. Hypertext 1997, Manchester, UK, pp. 67–74 (1997) 2. Hinze, A., Buchanan, G.: The challenge of creating cooperating mobile services: Experiences and lessons learned. In: Proceedings of the ACSC 2006, Hobart, Australia, pp. 207–215 (January 2006) 3. Hinze, A., Gao, X., Bainbridge, D.: The TIP/Greenstone Bridge: A Service for Mobile Location-Based Access to Digital Libraries. In: Gonzalo, J., Thanos, C., Verdejo, M.F., Carrasco, R.C. (eds.) ECDL 2006. LNCS, vol. 4172, pp. 99–110. Springer, Heidelberg (2006) 4. Ingwersen, P., Jaervelin, K.: The Turn: Integration of Information Seeking and Retrieval in Context. Springer, Heidelberg (2005) 5. Jones, M., Buchanan, G., Harper, R., Xech, P.-L.: Questions not answers: a novel mobile search technique. In: Proc. SIGCHI 2007, pp. 155–158 (2007) 6. Schweer, A., Hinze, A.: Challenges in interface and interaction design for contextaware augmented memory systems. In: Proc. of the SIGCHI-NZ (July 2007) 7. Witten, I.H., Bainbridge, D.: How to Build a Digital Library. Elsevier, Amsterdam (2002)
