Artificial Intelligence (AI) is attracting much attention and it will be a major driver of technology in the coming years.
It will bring a big transformation to many industries, such as transportation, manufacturing, healthcare, communications,
financial services, and more. This is possible because of the big data availability, the advances in hardware capabilities and
the inventions of new models, methods, algorithms capable to offer new solutions for long-standing research problems.
Question Answering (QA) is a complex task that requires the ability to understand the natural language (NLU) and to reason over relevant contexts. Almost all Natural Language Processing (NLP) tasks can be seen as QA problem (e.g. entity extraction, sentiment analysis, machine translation).
Recently, QA by using novel AI techniques has seen scientific and commercial popularity that attracted media attention, but effective QA is a challenging task for machines that try to simulate the human behaviour.
Some solutions are based on Information Retrieval (IR) techniques, other on Information Extraction (IE) processes that enable to create Knowledge Bases (KBs), so logic-based query languages are used to infer answers from KBs. KB-based solutions can be satisfactory for closed-domain problems, but they are unlikely to scale up to answer questions on any topic. Novel approaches for QA over documents are based on Deep Neural Networks that encode the documents and the questions to determine the answers. A lot of research has focused on learning from fixed training sets of labeled data, but other try to learn through online interaction (dialogue) with humans or other agents. This is the case of conversational agents (or conversational interfaces/bots/chatbots) that adapt their model based on teacher's feedbacks (Reinforcement Learning) and change beliefs in response to new information.