The increasing amount of textual information on the Internet is a good example of information explosion. Previous research on information retrieval mostly focused on improving the retrieval effectiveness of the system. As the amount of online information and the number of users accessing these information resources continue to grow at an exponential rate, system efficiency becomes an important concern of information retrieval systems since users demand better results to be found in a shorter period of time....[ Read more ]

The increasing amount of textual information on the Internet is a good example of information explosion. Previous research on information retrieval mostly focused on improving the retrieval effectiveness of the system. As the amount of online information and the number of users accessing these information resources continue to grow at an exponential rate, system efficiency becomes an important concern of information retrieval systems since users demand better results to be found in a shorter period of time.

Our research focuses on improving the efficiency of information retrieval systems. Several aspects concerning the efficiency of an information retrieval system are first discussed. A new index structure, called the Weight-Partitioned Inverted File, is proposed. Experiments are carried out on a subset of the TREC collection to obtain the tradeoff between retrieval effectiveness and efficiency. It is found that the Weight-Partitioned Inverted File is able to improve efficiency noticeably while maintaining the effectiveness at about the same level as a regular inverted file implementation.