User Transaction Log Parsing and Analysis in MetaLib QuickSearch 3.13

Abstract

Pascal Calarco, John Scofield, Univ. of Notre Dame and Rong Tang, Simmons College.

As a component of ongoing MetaLib QuickSearch user studies at the University of Notre Dame, we have developed local scripts to parse the locally-developed phrase parsing logs, server_www and search_server Metalib logs using the parent and child session keys to summarize a user's behavior during a complete search session. This has allowed us to have the raw data to be able to answer several questions that are currently not covered by the built-in Metalib statistics and reports, such as:

* How long is the average MetaLib search session?
* How many times do users, on average, reformulate their queries?
* what is the proportion of user queries that have applied quotes for "phrase-based" searches?
* What is the proportion of user queries that have utilized advanced search featuring field search options? Which search field in the advanced search option is the most popular, which are the least popular (or not at all used) fields?
* How many SFX links are usually followed in an average search session?
* What do users typically do when they encounter zero result searches?

In our presentation, we will focus on the technical mechanics of parsing the logs. The detailed content analysis of user query patterns will be conducted later this year as part of a larger MetaLib user studies research project between Dr. Rong Tang, Graduate School of Library & Information Science, Simmons College, and Pascal Calarco, Library Systems, University Libraries of Notre Dame.