English
资源说明：A toy search engine based on latent semantic indexing and singular value decomposition.
This directory contains a Python code sample implementing a toy search engine based on latent semantic indexing and singular value decomposition. The application requires Python 2.6 or greater and the numpy library. Also the Pickle file lsiData.pkl must be in the same directory as the application or the path to a similarly encoded word-document count file must be provided on the command line with the syntax -fUsage: python search_engine.py -f -n -m input-file: The path to a data file encoding page word counts, serialized using pickle. no-components: The number of components to be used in truncated svd. no-results: The number of search results to return. keywords: The keywords to search on. On execution, a list of top-ranking pages for the given search keywords is printed to stdout. All input parameters are optional and set to reasonable default values if not given on the command line. To run with defaults use: python search_engine.py or python search_engine.py for example: python search_engine.py animals python The most important parameter for determining the behavior of the search engine is no-components, which specifies the number of components to be used in truncated svd. Large values will result in more literal searches with little clustering of concepts. Small values will result in large clusters with few distinctions among keywords. As shown in the scree plot (scree.jpg) an ideal choice for the number of components for the default data set is 10. Bob Flagg bob@calcworks.net
本源码包内暂不包含可直接显示的源代码文件，请下载源码包。
