Academic writing on Artificial intelligence in Educational technology : online assessment and proctoring SaaS platform

General Outline : Introduction on how AI is laying the ground work for real Artificial Intelligence Assessment. Expand the issue with general topic categories in : 1. Natural Language Processing 2. Conversation-Based Assessment 3. Suspicious Behavior Detection and its role in proctoring Expand the last topic Suspicious Behavior Detection 1. Visual detection of unusual activity: 2. Audio analysis algorithms All of these will be a part of the next generation of Proctest software ([login to view URL]). What is Proctest? Features Its unique cutting edge in comparison to other line of similar product available in the market. the article will be 40 percent of AI and 60 percent on the current Proctest tool. source link to study: [login to view URL] [login to view URL] [login to view URL] Also read: 1. Visual detection of unusual activity: Computer vision and machine learning techniques enable the algorithm to learn throughout the video sequence, which parts of the video are repeating and therefore represent common or "usual" activity, while the parts of the video containing unique actions deviating from other activities, indicate "unusual" activity. Such unusual actions can be detected. The example of the algorithm is explained in this paper: [login to view URL]~jshi/papers/[login to view URL] With this algorithm the entire video can be processed unusual activities are detected and time stamped in the video sequence. 2. Audio analysis algorithms Signal processing and machine learning also enable audio voice analysis algorithms to inspect the audio recording and learn through the recording of certain properties of spoken voice captured in the recordings. It is possible to detect the number of different voices in the audio and also the detection of whispered speech. It must be stressed out, that these two are separate use cases with separate algorithms, but it is nevertheless possible to use some parts of the algs. in both implementations. Unsupervised learning methods are used - meaning, the algorithm performs count or detection without previous training, so it can learn and decide on the basis of single recording. Speaker count Counts the number of speaker in the room, based on the properties of human voice (e.g. pitch). Speaker count algorithm example paper: [login to view URL]~campbell/[login to view URL] Whisper Detection Detect parts of the recording representing whispered speech. Whisper detection algorithm example paper: [login to view URL] Video skimming / summation Later in the development phase (idea for the future), this technique can be used to create a short summary of the video, containing only the “interesting” parts of the video in a short summed and skimmed video sequence, for easier human preview.