A Distributed Retrieval System for NTCIR-5 WEB Task
نویسندگان
چکیده
We developed a distributed search system with the corresponding very large scale corpora from NTCIR5 WEB Task. And we arranged the scoring method which is based on link-structure of the Web documents to calculate lower cost. Our search system, which consists of 6 PCs could make indices for full texts size of about 1 TB. Additionally, we confirmed that our arranged scoring method made an improvement of mean average precision. Also we performed experiments with the pseudodocument vectors at every pseudo-relevance feedback. Meanwhile we made a pseudo-document vector at every relevance feedback. Therefore the results had slightly better precision than raw queries even though it had not been tuned yet.
منابع مشابه
A Distributed Retrieval System for NTCIR-5 Patent Retrieval Task
We developed a distributed search system with the corresponding very large scale corpora from NTCIR-5 Patent Retrieval Task. And we developed the method of query refining using Support Vector Machines. Our search system, which consists of 5 PCs could make indices of all claims for ten years. Additionally, we confirmed that our arranging the scoring method made an improvement of mean average pre...
متن کاملR2D2 at NTCIR-4 Web Retrieval Task
We evaluated the Relevance-based Superimposition Model at NTCIR 4 Web task A (survey retrieval) and B (target retrieval). We developed a distributed indexing / searching engine for treating the large amount of documents in a practical processing time. Some improvements of the retrieval precisions were achieved algorithmically.
متن کاملOverview of the NTCIR-4 WEB Navigational Retrieval Task 1
This paper describes an overview of the Navigational Retrieval Task 1 that was conducted from 2002 to 2004 as a subtask of the WEB Task at the Fourth NTCIR Workshop. In the Task, we attempted to assess the retrieval effectiveness of Web search systems from a viewpoint of “Known Item Search” using a common data set, and built a re-usable test collection. 100-gigabyte Web document data constructe...
متن کاملA Experiment Report about a Web Information Retrieval System for 3rd NTCIR Web Task
We joined 3rd NTCIR web task from October 2001. For this task, we constructed a small web information retrieval system. By this system, we completed “dry run” and “formal run” retrieval topics of the task. In this report we will give a brief description about our basic method for web information retrieval, our web information retrieval system and some retrieval experiment results.
متن کاملOASIS at NTCIR-5: Web Navigation Retrieval Subtask
We experienced negative results participating in this Subtask: the OASIS system, which is a distributed search system based on VSM and full text indexing, failed to retrieve relevant documents from the huge data set of Japanese Web pages when the number of relevant documents in the collection was relatively small.
متن کامل