Flexible semantic query expansion for process exploration

Abstract

Process exploration tools help to identify business processes that are not executed according to their documentation or lack such documentation at all. These process steps are executed by human beings, therefore process traces can often be found in unstructured documents. In order to reconstruct a process execution, these documents have to be retrieved. Natural language properties such as hyponyms, hypernyms, homonyms and synonyms make searching for a specific element a hard task. Integrating word relations in the search index is the standard solution for tackling this problem. In our process exploration scenario, however, this approach comes to its limits as ontologies defining word relations may vary from process step to process step. The problem is that the approach is rather inflexible. In order to change the relation of the words, the index needs to be rebuilt. This in turn would require running an analysis of the whole document base. Query Expansion, on the other hand, works by adding related words to a query, making it very flexible. In a classic search scenario, it still comes with a number of disadvantages such as retrieving unrelated documents. In our scenario, these disadvantages do not apply, since information from previous steps in the explored process can be used to constrain the result set.