19^th International CODATA Conference
Category: Infoscience

Query log analysis for information system improvement : The Arianet example

B. Delecroix (delecroix.bertrand@wanadoo.fr) and R. Eppstein (renaud.eppstein@univ-mlv.fr)
CESD/ISIS, Université de Marne La Vallée, France

For several years lots of experiments have been made which demonstrate the misfit between Search Engines and users' behaviours, needs and knowledge. It is easy to understand that the wider the users' profile, the most diverse the subjects covered by data collection, the harder it is to adapt search engine tools to the likely provision of the most appropriate documents in the initial response.

In this article, we will demonstrate that even in an Intranet with a delimited document base for "advanced level" users, this misfit still exists.

This article will provide the query log analysis of Arianet, France Telecom business, financial and technical Intranet. The log file consists in approximately 55000 entries for search requests over a period of 1 year (2003) on a document base of approximately 600000 documents concerning telecommunications and new technologies area.

First we will succinctly present the France Telecom Aria structure, Arianet Intranet and tools provided to users to complete their searches. After an outlook on analysed logs, we will more specifically analyse the use of keywords and the queries structure. We will present a first set of results which enhance our comprehension of users' behaviours and needs. We will also give an interpretation for some statistics, especially those concerning the very small number of queries having generated at least one click to view a document.

Finally these observations lead us to propose some recommendations for information system's improvement.

Keywords: query log analysis, intranet, information system, webometrics