Knowledge workers spend a significant portion of their time looking for information in documents – and are unsuccessful up to 50% of the time. Automatic Text Summarisation tools are a type of technology that could potentially aid such users find the information they are looking for. However, a summary is only particularly useful if it supports the underlying task of the user, which can be difficult to capture and represent.
To make summaries more responsive to user needs, a system must collect information about that need. One of the key observations in this work is that the content currently displayed in document browser (the reading context) is an approximate snapshot of the user's current interest and is easily obtainable. This reading context can be exploited to generate a summary of a linked document. Based on this observation, we have developed IBES, a contextually-aware text summarisation tool built into an internet browser (Firefox). It can be characterised as follows:
User Need: Tell me more about the sentence that I have just read with content from the Linked document.
Possible User Tasks: Justify/Dismiss the sentence just read; Learn more about proposition in the Linking sentence; Decide if the Linked document is worth reading.
Interaction: The user moves the mouse over the hyperlink. This sets the Linking sentence as the user focus. This user focus is passed to the summariser.
System Output: IBES pops up a window containing the summary of the linked document. This summary is created dynamically taking the user focus into account.
We are experimenting with a number of ways to use the reading context in order to find the appropriate sentences in the linked document.

Authors: Stephen Wan and Cécile Paris

Event: SF08: Speed Papers

← View all submissions for this event.