Revising Dicionário Olímpico : a log-files analysis of a Brazilian frame-based online dictionary

Although still underused in Digital Lexicography, log-files analysis has great potential for empirical dictionary-use research through free implicit feedback and unobtrusive monitoring. In the context of revising and improving Dicionário Olímpico, a log-files analysis using AWStats was designed to investigate user behaviour wrt patterns of navigation and, more specifically, page views. Dicionário Olímpico is an online dictionary based on the theoretical foundations of Frame Semantics that describes the lexicon of the 40 Summer Olympic sports. The present work reports on the main outcomes of a pilot analysis and points to future perspectives for broadening the research scope to include all the available data. For the pilot analysis we had, amongst others, only access to the top 1,000 pages visited per month. Also, the usage data collected for this experiment corresponds to just three months (August, September, and November), the months with the highest average unique visitor numbers, for the years 2016-2019. Bearing in mind Dicionário Olímpico’s lexicographic structure (i.e., the fact that the tool presents three main levels with information on sports, scenarios, and words), the results of this investigation are divided into three categories. Regarding the SPORTS level, the analysis revealed that not all 40 sports pages appear in the top 1,000 viewed pages’ list. At the same time, the total number of viewed sports decreases throughout the four-year period, while the number of views per sport that is consulted increases. Exceptionally popular is rhythmic gymnastics, whose page-view numbers are 2.5 times higher than the second sport in the list, being volleyball. Results concerning the SCENARIOS level indicate that only around half of the total number of 780 scenarios appear in the top 1,000 viewed pages; but also that the total number of viewed scenarios remains stable over time. For the WORDS level, the logs indicate that users only explore a small part of them: on average 508 per month, out of a total of 3,930 words. In general, the results from this initial exploration point to the following major finding: most dictionary data is not (frequently) seen by users. This begs the question: Do users not see that data because they do not need it, or do they not see it because they fail to find it? Lexicographers secretly hope that all their work may be of use at some point in time, so we shall assume for now that users simply failed to navigate to ‘all’ the dict