Context-aware web retrieval

A "genre" is a classifying statement that allows us to recognize items that are similar, even in the face of great diversity. There are many different web page genres, from journal articles to personal home pages to FAQs. Genres cut across all knowledge domains. For instance, a journal article is a journal article whether in engineering or sociology, and an FAQ is an FAQ whether about how to plant a tree or how to hook up a new device to your computer. Similarly, a type of task is not domain specific; for example, shopping is a transactional task and should retrieve web pages of appropriate genres (shopping sites), whether shopping for medicine or handbags, i.e., there is a relationship between type of task and the genre(s) of web pages that should satisfy that task. This research recognizes this task-genre relationship and the research objective is leverage this relationship to provide task-based context aware web retrieval. The user task provides the context for the web search. Certain genres of web pages may be more pertinent to the task than others and, by identifying automatically the genres of web pages, we will be able to filter and/or re-order the retrieved web pages.Thus, this research will involve the automatic identification of the genre(s) of a web page and the matching of those genres to tasks. In addition, we will attempt to track the evolution of web page genres over time in order to maintain the currency of our genre definitions.The benefits to industry will include software and algorithms to improve search engines through providing a task-based context for web searching.