Article Preview
TopIntroduction
Information retrieval (IR) methods play a crucial role when it comes to organizing and finding relevant information in large collections. Standard information retrieval methods aim at finding material of an unstructured nature from within large collections (Mќuller, Schimkat, & Mќuller, 2002) that highly resembles/matches the input query. In general, a query does not uniquely identify a single object in the collection. Instead, several objects may match the query, perhaps with different degrees of relevancy in the form of numeric scores, and rank the objects according to these values. The top ranking objects are then shown to the inquirer.
When searching knowledge bases by query concepts, contextual information can be useful to clarify ambiguous situations (Tulving & Thomson, 1973; Keßssler, 2007) and thereby play a crucial role when measuring the similarity of two concepts (Keßssler, 2007). Contextual information or context can be defined as a set of constraints that influence the behavior of a system (a user or a computer) embedded in a given task (Bazire & Brézillon, 2005). It belongs to an individual and is an integral part of the representation that this individual is building of the situation where he is involved. Context cannot be separated from the knowledge it organizes and the way this knowledge is organized provides clues that serve the retrieval process (Tulving & Thomson, 1973). When searching knowledge bases by query concepts, results often depend on whether and how contextual information is taken into account independent of the applied information retrieval method.
In order to develop a useful context mode for information retrieval, we must have a good understanding of the knowledge representation mechanism used, what information needs to be considered, and what is out of scope for the current task. Recently, it has been noted that the human ability to interpret social schema queries, depending on the current context, has been mostly neglected in inference-based approaches to information retrieval from semantically annotated information (Keßssler, Raubal, & Wosniok, 2009; Li, Lee-Urban, Appling, & Riedl, 2012). One example is Sapper (Veale & O'Donoghue, 2000) that uses retrieval cues in its retrieval mechanism but lacks a model of temporal progression (Veale & O Donoghue, 2000). Sapper uses semantic networks for knowledge representation to represent narratives, such as 'Star Wars' and 'The Dambusters'. Hence, temporal progression seems to be ‘neglected’ in the existing structural retrieval approaches despite its importance in providing deep structural knowledge.
One way to capture temporal progression in societal information is to use cognitive scripts (Shank and Ableson, 1977). A cognitive script, which is a type of schema, is a predetermined, stereotyped sequence of actions that defines a well-known situation. It formally represents a powerful feature of our thinking processes: the fact that our understanding of new situations is often driven by stereotypes, which can be applied in a rough-and-ready way while avoiding the need for extensive inferential processes in order to build up an understanding from scratch (Ritchie, 2004). Cognitive scripts are comprised of slots and the connections between these slots. For each of these slots, there are default values that reflect a well known stereotyped event, such as going to a restaurant. Cognitive scripts are characterized in two aspects. First, the slots in cognitive scripts specify actions in a sequence; second, the connections between actions are temporal and causal (Chen, 2004). For these reasons, cognitive scripts act as a suitable representation for social schemas. The element of time and the sequence of actions are tightly bound by the structure of a cognitive script (Ip, 2011) and this structure is what provides the context.