note #ai

greeting - visitor select question - article lookup / talk to people

what the goals are? the types of chatbot

rules-based chatbots - follow pre-designed rules, UI - follow a design paths - rely on a pre—determined outcome designed by a human - such as ‘yes’ or ‘no’, or keywords

AI chatbots - auto learn after an initial training period by a bot developer - AT chatbots first understand what your question is => analyze user’s request - identify intent and entities - reply - once understand the intent, an answer is delivered based on existing data - over time, by observing correct/incorrect answers, the machine gets better at understanding what the “right” answer is

 - require a training period, and more effort to start. once they understand your business well, why can be powerful

live act software -switch between chatbot / human answer questions in real-time conversations,

set goals for your chatbot - what your chatbot would do - should be intuitive and easy to use , a simple “agent like” bot will route questions - A FAQ chatbot - develop a chatbot with a great understanding of a FAQ - a frequently asked questions page

choose a channel - text-based chatbot
- offer an API so that bots can be embedded into any communication channel

 dialog flow design is important 
- decide on what the bot will say at each step

collect the data
- a collection of different ways to say the same thing - then based on rule-based to collect the data - the risk is that no answer before

- if a large collection of interactions, then machine learning to learn the most common answers to the most common questions

implement the dialogue flow
- create a classifier to map out the system response according to the incoming

Testing - a automated test, and a real-life tests with real people

deployment and revision - interactions should be closely monitored, -> show a possible space for further development