Human-Machine Interactions for on the Fly Free Text Input Processing

Information in machine-readable form has to be entered by users of information systems in various situations. In the input process, forms with a number of individual fields are commonly used. For the creator, this input can be cumbersome and potentially counterintuitive. For this reason, free text components are becoming increasingly important. Information in the form of free text is not machine-readable as such, and extracting information is complex. The accuracy of existing approaches used to extract information from free text components depends directly on the quality of the used internal model. These models have to be created by experts or trained in advance using the appropriate machine learning methods. The main objective presented in this paper is extracting information as well as enhancing the underlying model on the fly. This implies that the information extraction algorithms apply directly to the user's input and that the user is enabled to correct possible deficiencies in the model as they arise.