The Mymory Project


Attention is the currency of the information economy, and is already the scarcest resource in many organizations.
In addition to devoting more thinking to knowledge and knowledge management, in the future all organizations will need to focus their attention on attention.

Th. H. Davenport & L. Prusak: Working Knowledge (Paperback Edition).

One of the main areas of research in Mymory is concerned with using attention evidence data collected by unobtrusive user observation in information retrieval. The general idea is that the precision of information retrieval methods can be enhanced when the current interests and task of the user are estimated with the help of user observation.

The core research questions are mainly threefold:

  • Collection of attention evidence data to get a clue about the user’s current interests
  • Utilizing attention evidence data as implicit relevance feedback for information retrieval
  • Attention-based modifications of vector-space based indexing mechanisms

Collection of Attention Evidence Data

Several attention evidence sources will be used concerning document-centered work:
  • Text mark recognition (e.g., parts of the document that have been highlighted)
  • Text work recognition (analyze the scrolling behavior in order to determine, e.g., definitely not viewed parts of a document)
  • High-fidelity eye tracking (to determine whether passages of a document have been read, skimmed or skipped)

Implicit Relevance Feedback

Manual feedback systems in information retrieval are problematic in that users find them obtrusive and time-consuming. In Mymory, it will be examined, whether such feedback information can be obtained by unobtrusive user observation methods, e.g., by applying eye tracking. This information could then be used, e.g., for attention-based query expansion techniques.

Attention-Based Indexing

Most information retrieval systems are based on a vector-space model in which a set of documents is conceptualized as a two-dimensional co-occurrence matrix, where the columns represent the documents and the rows represent the unique terms (usually words or short phrases) occurring in the documents. The value of a particular cell is based on the frequency with which the term occurs in the document. Traditionally, such transformations have been based on information internal to document itself or to the set of documents under consideration. The addition of attention-based feedback introduces a new variable for which these traditional transformations are unable to account. We will develop and test novel transformations which incorporate such feedback data.
Go to top   More info...  

This page last changed on 17-Aug-2007 20:40:14 CEST by elst.