Failed Queries: a Morpho-Syntactic Analysis Based on Transaction Log Files

The aim of the study is to elaborate on the procedure needed in order to analyze morpho-syntactically the typing-error queries submitted in Greek during the search process. In the context of our analysis a failed query is a query which returned no hits. The analysis showed that failed queries represent 36% of the submitted queries. More specifically, 19.6% of failed queries occurred due to typing errors. We discovered that for analyzing morpho-syntactically a Greek text corpus the PoS tools need to be rich in tags in order to work adequately. Open Xerox tokenizer performed well but with significant pre-processing of the queries and the analyzer seems to require additional tools to improve its performance. MS Word which was used for spelling corrections seems to perform satisfactorily. All tools were challenged in terms of named entities recognition.