Submission Type: Long Presentation
Email search is an under-explored problem. In contrast to web search, which
successfully exploits lots of information about the structure of the web
(e.g., hyperlinks, page titles, URIs), the state of the art in email search
is quite primitive, and often does little more more than look for exact
strings in the body of a message.
Can we take what we know about web search and dialogue and make use of both
the internal and conversational structure of email messages to improve email
search? Within an email message, can we give higher weight to text in a
subject line? Assuming we can reliably identify different zones in an email
message, should we disregard text quoted from earlier messages? Should we
also ignore legal disclaimers and advertising content? How should we treat
the content of email signatures?
Moving beyond individual messages, how can we make use of email click data?
What are its characteristics and how does this compare with web click data?
How do we compensate for the lack of citation-style external evidence that
is successfully applied in web search? How else can we exploit the latent
social graph, conversation structure and other characteristics of email to
improve the utility and effectiveness of email search?
Abstract: Better Email Search slides
Authors: Andrew Lampert
Event: Fourth HCSNet Next-Generation Search Technology Workshop (NGS09)
← View all submissions for this event.
