GLOVER, E. J., FLAKE, G. W., LAWRENCE, S., BIRMINGHAM, W., KRUGER, A., GILES, C. L. and PENNOCK, D. 2001. Improving category specific Web search by learning query modifications. In Proceedings of IEEE Symposium on application and the Internet (SAINT). 23-31.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....or personal homepage , the refining process could be eliminated by applying a query of XML upon the classes of resume or personal homepage. Researchers have realized these problems, and proposed the classifications of user interesting classes such as calls for paper , personal homepage [9]. This involves binary classification techniques that distinguish Web pages of a userinteresting class from all others. This binary classifier is an essential component for Web mining because identifying Web pages of a particular class from the Internet is the first step of mining interesting data ....

E. J. Glover, G. W. Flake, and S. Lawrence. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet (SAINT 2001.

....In contrast, our method finds a query modification e#ective for various queries in a specific domain in advance, and instantly returns relevant documents to the user without any interaction. Improving domain specific (or category specific) web search by query modifications is also described in [17]. and they extracted query modification rules for finding personal homepages and call for papers. They formed the training set by combining the positive examples collected by human from may resources and negative examples from logs of a search engine. To solve the problem of the mismatch between ....

Eric Glover, Gary Flake, Steve Lawrence, William P. Birmingham, Andries Kruger, C. Lee Giles, and David Pennock, "Improving category specific web search by learning query modifications," in Symposium on Applications and the Internet, SAINT.

....engines, especially those with very large index like Google, o#ers a very promising source for question answering. Agichtein et al. 12] presented a technique on how to learn search engine specific query transformations for question answering. A similar transformation technique also appeared in [13]. The idea is that the current query interfaces of most generic search engines, such as Google, etc. does not provide enough capability for direct question answering in natural language mode. By transforming the initial natural questions into a certain format which include more domain specific ....

Eric J. Glover, Gary W. Flake, Steve Lawrence, William P. Birmingham, Andries Kruger, C. Lee Giles, and David M. Pennock. Improving category specific web search by learning query modifications. In The Proceedings of Symposium on Applications and the Internet, SAINT

....leads to word sense disambiguation and can allow a richer search interface functionality. The type of specialized search engine discussed here does not restrict its searches to a subset of the Web pages available but rather restricts the types of search that can be done. For example Glover et al. [6] used learning techniques to automatically determine successful query modifications to find Web pages in a number of categories such as personal homepages, calls for papers, product reviews and guide or FAQ documents. They used Support Vector Machines to do text classification. This involves ....

....One way to do this would be for the search engine to search for the products name along with such words as competitor or rival and then intelligently extract the names of competitors. The process of determining which of these query modifications work best can be automated as has been done in [6]. Alternatively link analysis can be used as is done by GoogleScout (http: www.google.com) Once the competitors are known articles that mention a number of the competitors and the original product along with possible query modifications such as versus or compared can be searched for. Once ....

E. Glover, G. Flake, S. Lawrence, W. Birmingham, A. Kruger, C. Lee Giles, D. Pennock. Improving Category Specific Web Search by Learning Query Modifications. In Symposium on Applications and the Internet, SAINT 2001.

....accessing the Web for users, automated systems for learning from the Web have primarily been installed in crawlers, or spiders. A new generation of algorithms is seeking to augment the set of search capabilities by combining other kinds of topic or target directed searches. Glover and colleagues [9] use machine learning to automatically augment user queries for specific documents with terms designed to find document genres, such as home pages and calls for papers. Rennie and McCallum [15] use reinforcement learning to help a crawler discover the right kinds of hyperlinks to follow to find ....

E. Glover, G. Flake, S. Lawrence, W. P. Birmingham, A. Kruger, C. L. Giles, and D. Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12 2001.

....accessing the Web for users, automated systems for learning from the Web have primarily been installed in crawlers, or spiders. A new generation of algorithms is seeking to augment the set of search capabilities by combining other kinds of topic or target directed searches. Glover and colleagues [9] use machine learning to automatically augment user queries for specific documents with terms designed to find document genres, such as home pages and calls for papers. Rennie and McCallum [16] use reinforcement learning to help a crawler discover the right kinds of hyperlinks to follow to find ....

E. Glover, G. Flake, S. Lawrence, W. P. Birmingham, A. Kruger, C. L. Giles, and D. Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12 2001.

....a single document or a set of keywords in the target concept, our methods can learn to generate queries that can acquire a reasonable number of documents in Slovenian from the Web and that our approach also generalizes to other languages that are also minority languages on the Web. Glover et al. [7] use machine learning to automatically augment user queries for specific documents with terms designed to find document genres, such as home pages and calls for papers. Rennie et al. 9] use reinforcement learning to help a crawler discover the kinds of hyper links to follow to find research ....

Glover, E., Flake, G., Lawrence, S., Birmingham, W. P., Kruger, A., Giles, C. L., & Pennock, D. (2001). Improving category specific web search by learning query modifications. Symposium on Applications and the Internet. San Diego, CA.

No context found.

E. Glover, G. Flake, S. Lawrence, W. P. Birmingham, A. Kruger, C. L. Giles, and D. Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12, 2001.

No context found.

E. Glover, G. Flake, S. Lawrence, W. Birmingham, A. Kruger, C. Giles, and D. Pennock. Improving Category Specific Web Search by Learning Query Modifications. In Proceedings of the Symposium on Applications and the Internet, SAINT 2001, San Diego, CA, January 2001,pp. 23-31.

....Third, we describe our method for combining the results to improve accuracy. Fourth, we describe how to name a cluster using the features selected from the virtual documents. 2. 1 Full Text Classifier In our earlier works, we described our algorithm for full text classification of web pages [10, 11]. The basic algorithm is to generate a feature histogram from training documents, select the important features and then to train an SVM classifier. Figure 2 summarizes the high level procedure. 2.1.1 Training Sets and Virtual Documents To train a binary classifier it is essential to have sets ....

....there can be hundreds of thousands of unique features, most that are not useful, occurring in just hundreds of documents. To improve performance and generalizability, we perform dimensionality reduction using a two step process. This processes is identical to that described in our earlier works [10, 11]. First, we perform thresholding, by removing all features that do not occur in a specified percentage of documents as rare words are less likely to be useful for a classifier. A feature f is removed if it occurs in less than the required percentage (threshold) of both the positive and negative ....

[Article contains additional citation context not shown here]

Eric Glover, Gary Flake, Steve Lawrence, William P. Birmingham, Andries Kruger, C. Lee Giles, and David Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12 2001.

....of feature selection for information retrieval [8] Expected entropy loss is computed separately for each feature. It ranks the features lower that are common in both the positive set and the negative set but ranks the features higher that are effective discriminators for a class. Glover et al. [9] used this method for feature selection before training a binary classifier. We use the same technique. Feature selection increases both effectiveness and efficiency since it removes non informative terms according to corpus statistics [19] A brief description of the theory [1] is as follows. ....

Glover, E. J., Flake, G. W., Lawrence, S., Birmingham, W. P., Kruger, A., Giles, L. C., and Pennock, D. M. "Improving category specific web search by learning query modification." IEEE Symposium on Applications and the Internet (SAINT 2001.

....uncertain, and separating them out demonstrated a substantial improvement in accuracy. The second method is to combine results from the extended an chortext based classifier with the less accurate full text classifier. Our observations indicated that the negative class accuracy was approaching 100 for the extended anchortext classifier, and that many false negatives were classified as positive by the full text classifter. As a result, our combination function only considered the full text classifier when a document was classified as negative, but uncertain, by the extended anchortext ....

....we chose several sub categories to add documents. Table 2 lists the results for each of the classifiers from Table 1. In addition to the Yahoo categories, we tried applying SVM classification to the WebKB categories of courses and faculty. For training of courses, we used 144 positive and 1000 negative (from the otheW category) and for training of the faculty category we used 84 positive and the same 1000 negative. For the category courses there were 1000 negative test documents, and 70 positive test examples, for an accuracy of 96.8 negative and 67 for the positive. For the ....

[Article contains additional citation context not shown here]

E. Glover, G. Flake, S. Lawrence, W. P. Birmingham, A. Kruger, C. L. Giles, and D. Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8 12, 2001.

....search engines in the form of a metasearch engine. Moreover, by using SVMs to guide the rule search, our extracted rules are predisposed to have many of the same generalization qualities that the originating SVM possesses. This approach differs from our other work on learning query modifications [6] by producing a set of query modifications that work together to improve recall, as opposed to a set of individually effective modifications, that may or may not work well together. This paper is divided into five sections. In Section 2, we discuss WWW search engines and metasearch engines, with ....

E. Glover, G. Flake, S. Lawrence, W. P. Birmingham, A. Kruger, C. L. Giles, and D. Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12 2001.

....to the original query, in case the transformations are not successful. Inquirus 2 has proven to be highly effective at improving the precision of search results within given categories. Recent research related to Inquirus 2 includes learning methods that automatically learn query modifications [18, 28]. 2.2 Automatically inferring context information Inquirus 2 can greatly improve search precision, but requires the user to explicitly enter context information. What if search context could be automatically inferred This is the goal of the Watson project [11, 12, 13] Watson attempts to model ....

Eric Glover, Gary Flake, Steve Lawrence, William P. Birmingham, Andries Kruger, C. Lee Giles, and David Pennock. Improving category specific web search by learning query modifications. In Symposium on Applications and the Internet, SAINT, San Diego, CA, January 8--12 2001.

No context found.

GLOVER, E. J., FLAKE, G. W., LAWRENCE, S., BIRMINGHAM, W., KRUGER, A., GILES, C. L. and PENNOCK, D. 2001. Improving category specific Web search by learning query modifications. In Proceedings of IEEE Symposium on application and the Internet (SAINT). 23-31.

No context found.

E. Glover, G. Flake, S. Lawrence, W.P. Birmingham, A. Kruger, C.L. Giles, and D. Pennock, "Improving Category Specific Web Search by Learning Query Modifications," Proc. 2001.

No context found.

Glover, E.J., Flake, G., Lawrence S., Birmingham W.P., Kruger A., Giles C.L., and Pennock D., "Improving category specific Web search by learning query modifications," Symp. on Applications and the Internet, 23-31, 2001. Available at http://citeseer.nj.nec.com/golver01improving.html

No context found.

E.J. Glover, G.W. Flake, and S. Lawrence, "Improving Category Specific Web Search by Learning Query Modifications," Proc. 2001.

No context found.

E. Glover, G. Flake, S. Lawrence, W. Birmingham, A. Kruger, C. Giles, and D. Pennock. Improving Category Specific Web Search by Learning Query Modifications. In In Symposium on Applications and the Internet, SAINT, pages 8--12, San Diego, CA, January 2001.

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