For the points with a i the class memberships are

This preview shows page 31 - 34 out of 80 pages.

, then the patient is put under group 2 where Y=1. For the points with a i = 0 , the class memberships are intermediate. This method aims to obtain an optimal w s j where the resulting classifications will have a maximum agreement with the outcomes, meaning there is the least misclassification error. Subliminally, the best weights need to allow two factors. First, the sign of a i to is similar to Y i , so that the classification is correct. Secondly, the magnitude of a i must be far away from 0 to minimize the vagueness of classification. To achieve these, w j s that cause minimum loss of quadratic function is selected. Besides, if the assumption is made that the new patients are from the same population, the resulting w j s can use these patients' traits in classifying them (Figure 9). 31
Figure 9 : SVM illustration 2.5.3 Natural Language Processing (NLP) NLP extract vital information from a narrative text to aid clinical decision-making. There are two main components in the NLP pipeline, text processing, and classification (Jurafsky and Martin, 2000, 10, Kruschwitz, 1999, 35). According to historical databases, NLP uses text processing to detect disease- relevant keywords from the narrative text. After that, a subset of the keywords is used, and their effect on the classification of abnormal and normal cases is examined. These validated keywords enter and buffer the structured data that ropes decision-making by the physician (Wilensky, 2000, 79). The NLP pipeline assists clinical decision making by monitoring adverse events, generating alerts on treatment arrangements, and others. 32
2.6 Chatbots and Autism The widespread use of personal computerized systems and the need for conversational agents have likewise increased where people wish to converse with the machine they would have with real people. The concept is called Human-Computer Interaction (HCI). It allows humans to express themselves in either textual or auditory form. These qualities have existed in chatbots, which were created for the sole purpose of HCI (Abu Shawar and Atwell, 2005a, 23, Atwell, 1996, 45). Chatbots incorporate computational algorithms and integrate a language model in mimicking human conversations. Building chatbots for this subset of the population assists them in communication and reduces the disparities existing between them and the non-autistic adults. In the subsequent paragraphs of this section, we discuss some of the existent chatbots employed in the current study. 2.6.1 EDDI The Enhanced Dialogue Driven Intelligence (EDDI) is an opensource platform used for developing chatbot solutions. Some of the features the platform offers are: - · An in-memory Natural Language Processing principle. The feature increases the efficiency of the chatbots built on this platform, since the processing does not need to be fetched from storage, thus reducing the time constraints.

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture