The main steps of the algorithm are outlined in Fig. When a document arrives, the system decides which evidences have to be extracted initially.Based on these evidence types, it generates the initial state, the current goal state, and a DA plan.Other approaches use established classification methods, like Naïve Bayes or Support Vector Machines. classify emails into sender intentions based on verb-noun pairs called ].Unfortunately, all these approaches do not involve a dynamic task set and are, therefore, not applicable to Attentive Tasks (ATs).We proposed the approach of process-driven document analysis (DA) .A document arriving through an input channel is mapped to the corresponding task.We then introduce our two approaches and present the results of their evaluations.
The system deals with documents coming from the main input channels in enterprises: email, mail, fax, call center, and e Docs.
They are often costly to transfer to new domains, and they do not respect the importance of search criteria or search failure.].
They assume a direct connection between heuristic and task.
Krämer recognizes the importance of tasks instances but uses manual task assignment .
However, their approach has a limited applicability for well-defined processes as they appear in enterprises.