Chat bots: How imperfect conversations are making AI bots smarter by Amber Nigam | Labs Stanley Kubrick’s 2001: A Space Odyssey movie’s AI, HAL 9000 in 1968 exhibited way more conversational skills than we have available to us in 2020. Why is this so? Because human language is complex, very complex…… at least for a computer. Human brains, despite their main elements and neurons, are slow compared with computer processing power. But humans can effortlessly handle mispronunciations, swapped words, contractions, colloquialisms, and other quirks, while machines are less adept at handling unpredictable inputs. It all boils down to Natural Language Understanding (NLU) and Natural Language Processing (NLP). NLU is a field of NLP that deals with the conversion of natural language into a semantic representation that a computer can interpret. These two branches of machine learning are enabling the conversion of human language to computer commands and vice versa, enabling humans to have a conversation with machines in an effective manner and augmenting human intelligence. While building a chatbot, an NLU system is expected to achieve slot-tagging and intent detection, where slot-tagging tags the keywords or the subject and object of the sentence and intent detection identifies the action to be performed or the verb. The extracted tags usually act as constraints to the kind of information the user requires. For example, for the user-query ‘Show me some colleges near Mumbai for B. Tech.’, the intent is ‘Find Colleges’ and the slots tagged are ‘Mumbai city’ and ‘B. Tech degree’. The user’s search for finding colleges is constrained by the parameters that the locality must be Mumbai and the degree must be B. Tech. The fuel for an efficient chatbot depends on the data hygiene of the dataset used, which varies from situation to situation. Unlike clean Airline Traffic Information System (ATIS3) datasets, which are continuous broadcast of recorded aeronautical information in busier terminal areas, that contain essential information, such as current weather information, active runways, available approaches, and any other information required by the pilot. Most of the sentences are syntactically incorrect, which represent the real-world scenario, especially in a non-native context where sentences might be infused with a lot of grammatical mistakes. Training an AI bot calls for a fuzzy dataset that is semantically a little noisy and contains queries from non-native speakers. This would help the greater community who are, say, non-native English speakers to use NLU-based applications like chatbots. In an attempt to find a solution to the problem of intent and slot predictions in a chatbot, builders of PeopleStrong’s chatbot, Jinie has a curated dataset that consists of real-world queries from non-native users. The Intelligent bot keeps learning and updating its dataset every time its users conversationally interacts with Jinie. One of the real-world chatbot applications is that of the bots being used as a personal assistant, professionally or casually. One that helps you stay at the top of your game through different productivity hacks One that gives you smart suggestions after intuitively understanding your schedule and work-dynamics One that reminds you of your tasks in case you forget them One that understands you (quite literally) and arranges meetings for you when you want to Provides information about workplace policies. Jinie by PeopleStrong is one such chatbot application that utilizes the complete power of NLU and NLP to deliver a cognitive conversational chatbot experience to its users. Jini also helps Provide answers to all queries in a click of a button Sync with existing HCM to simplify employees’ day-to-day tasks Appreciate fellow workers Create Polls to consider every stakeholder’s decision Connect with Amazon Alex and Google Assistance to provide a seamless experience And more. Click on the below link to read about the author’s paper on chatbot: https://ieeexplore.ieee.org/document/8665635) Submit a Comment Cancel replyYour email address will not be published. Required fields are marked *Comment Name * Email * Website Save my name, email, and website in this browser for the next time I comment.