Focused and aggregated search: A perspective from natural language generation

Users are often faced with complex information needs that are not easily represented as a single query. With current technology, the burden of issuing these individual queries, analysing retrieved documents for relevance, as well as aggregating results falls upon the time-poor and informationally overloaded user. Aggregated search techniques represent the new generation of search applications that endeavour to help users perform these complex tasks. However, the way in which different data types are combined in current aggregated search applications is often performed using static hard-coded structures. We suggest that a useful alternative is to marry techniques from natural language generation, such as text planning and summarisation, in order to dynamically determine the best organisation of retrieved information. These organisations can be motivated by linguistic theories that consider issues such as the role that the information plays to facilitate a task, and the relationships between different pieces of information. With reference to a discourse strategy, it is possible to draw on several data sources automatically to generate a useful, focused, and coherent answer. We focus on exploring the parallels between aggregated search and natural language generation in the hope that the fields can be mutually informed, leading to further advances in the way search technologies can better serve the user. These issues are discussed and presented with examples of existing systems across different domains.