If you have visited one of any number of online stores these days, you would have noticed something – an overly helpful, sometimes annoying, pop-up chat window offering help (“may I help you?”) or advice (“chat with agent”). Despite the picture of a smiling human agent (often with a headset) adorning that window, you will likely not be chatting with a real person, but with a piece of software called a ‘bot’ (short for ‘robot’). Welcome to a world increasingly populated by such chatbots or virtual assistants. More and more companies are deploying such chatbots either on their main websites or even on their mobile apps to interact with potential customers efficiently and in a cost-effective manner. When they work well, the customer’s queries are answered immediately (without them having to make a time-consuming telephone call) and the company saves a significant amount of money in human operator costs. To be clear, chatbots do not eliminate human agents altogether. For more complex customer queries or situations that simply cannot be handled by a bot, there exist well-defined mechanisms for a chatbot to hand-off the conversation to a human agent, or for the customer to just pick up the phone and call someone. But for a majority of basic, boilerplate queries, chatbots are an increasingly cost-effective and efficient alternative.

​

The bots that have the potential to really improve our lives are the “smart” or AI bots. These bots typically incorporate a natural language processing (NLP) engine that uses machine learning to train a language model. Once trained, the model can determine the intent of the user even if the user’s input does not exactly match the sample utterances provided to the model during training. Depending on the device configuration (e.g., presence of microphone on the device), such bots can accept speech input (using automatic speech recognition (ASR) technology) and talk back to the user (using text-to-speech (TTS) technology), providing a truly hands-free, natural user interface.