Propagation-based Structured Text Retrieval

Evaluate the relevance score of text components. Approaches using propagation view the logical structure of a structured document as a tree whose nodes are components of the document and whose edges represent the relationships between the connected nodes. A document component can be either a leaf or an inner node. Leaf nodes are document components that correspond to the last elements of hierarchical relationship chains and that contain raw data (textual information). With propagation, relevance scores are first calculated for leaf components. They are then propagated upwards in the document tree structure to calculate relevance scores for the inner components.