Speaker "Antonio Valderrabanos" Details Back



NLP middleware for any multilingual chatbot and assistant –A new era of artificial data for AI


Training a conversational bot is a manual and time-consuming task. It involves feeding the bot different variations of all the potential user intents. This task is typically performed through significant manual tagging and different training iterations. We will discuss strategies to automate this process and significantly shorten training time and increase accuracy. Strategy 1. Reduce different user requests to a normalized form that captures their common meaning. Then, the bot is fed these normalized forms, linked to their respective surface forms. As a result, the complexity that your bot needs to handle is reduced drastically. Strategy 2. Given a user intent, generate all possible linguistic variations, tag them according to the intent and feed to the bot in the training phase. As a result, the bot will have a comprehensive training corpus for each intent and will be able to understand all variations during the live phase. Additionally, we will discuss other common problems in bot training: double intent, negative intent, conditional intent… We will discuss all these issues in a multilingual scenario.


Long experience on how to use Deep Linguistic Analysis to solve business problems, particularly in the areas of text analysis, now applied to chatbots. My current focus is on how to exploit linguistic knowledge to improve machine learning engines and AI in general, to make them smarter and easier to train. Chatbots are an excellent example of this trend, where deeper linguistic knowledge is the middleware needed to create the next generation of conversational interfaces. I'm also interested in new areas where computational linguistics can be revolutionary: Risk Management, Financial News Analysis, Lead Generation...