Searching attentive tasks with document analysis evidences and Dempster-Shafer theory

Many enterprises strive toward the integration of input communication channels into their internal business processes. To help them, we propose to drive input channel document analysis (DA) by formalizing information expectations from current process instances in Attentive Task (AT) templates. This requires, however, to map incoming request documents to the related AT from a set of ATs. For this purpose, we present a search approach that prioritizes a set of ATs based on DA evidences. Our algorithm relies on the theory of Dempster-Shafer to iteratively handle DA results and further uses the string edit distance of Levenshtein to provide robustness to errors in DA results. We evaluate the search performance in terms of influence of evidences, error robustness, and ease of calibration for our approach.