Aggregation-Based Structured Text Retrieval

DEFINITION Text retrieval is concerned with the retrieval of documents in response to user queries. This is achieved by (i) representing documents and queries with indexing features that provide a characterisation of their information content, and (ii) defining a function that uses these representations to perform retrieval. Structured text retrieval introduces a finer-grained retrieval paradigm that supports the representation and subsequent retrieval of the individual document components defined by the document’s logical structure. Aggregation-based structured text retrieval defines (i) the representation of each document component as the aggregation of the representation of its own information content and the representations of information content of its structurally related components, and (ii) retrieval of document components based on these (aggregated) representations. The aim of aggregation-based approaches is to improve retrieval effectiveness by capturing and exploiting the interrelations among the components of semi-structured text documents. The representation of each component’s own information content is generated at indexing time. The recursive aggregation of these representations, which takes place at the level of their indexing features, leads to the generation, either at indexing or at query time, of the representations of those components that are structurally related with other components. Aggregation can be defined in numerous ways; it is typically defined so that it enables retrieval to focus on those document components more specific to the query or to each document’s best entry points, i.e., document components that contain relevant information and from which users can browse to further relevant components.