DD Lewis. Information retrieval and the statistics of large data sets. In Proc. NRC Massive Data Sets Workshop, Washington, DC, 1996.  Home/Search   Document Details and Download   Summary   Related Articles

....[51] but it provides agents with modified rewards that improve on their models of relevance and therefore on their performance. Large, distributed text collections are a typical example of massive data sets that challenge machine learning techniques due to their huge feature space dimensionality [22]. InfoSpiders deal with dimensionality reduction in a localized, situated way. Agents internalize those words that appear maximally correlated (or anticorrelated) with their objective function, in their (temporally and spatially) local context. This model of feature selection keeps the size of the ....

.... in general, we believe that the ideas incorporated into the InfoSpiders framework are a first step towards addressing some of the new challenges posed by text classification to machine learning, especially the need to extend information retrieval to deal with time varying documents and user needs [22] and with large, dynamic, and heterogeneous collections such as the Web [23] On line search makes the classification problem both simpler and harder: simpler, because there are only two classes (relevant and irrelevant with respect to the current query) and harder, because the relevant class can ....

DD Lewis. Information retrieval and the statistics of large data sets. In Proc. NRC Massive Data Sets Workshop, Washington, DC, 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC