Our bot analyzes Apple's community discussion forums to find the best solution to tech problems. We use several machine learning algorithms to analyze the collective experience of multiple people facing the same problem to find solutions that work.
The solutions they provide need to be manually added to a database by a human. This greatly limits the helpfulness of the chatbot — imagine having a conversation with someone who can only respond to you by reading off a list of a few pre-written sentences. We've taken a different approach to support chatbots by analyzing forums, which lets us give accurate solutions to millions of tech problems.
We've created a way to find accurate answers to a significant amount of tech problems. Now we're working on an even more powerful bot: one that takes a proactive approach to getting your problem solved. For years, the internet has played a passive role as a pool of information that users need to navigate through themselves, continually clicking and reading and searching. We're working on a chatbot that will proactively help throughout the whole process, from initially asking questions to obtain relevant information to answering follow-up questions and giving clarification.
We're not building a simple, frustrating robot that clears out basic and common questions. We're building a support agent with the internet's knowledge that's able to dive deep into a problem with you and get tough problems resolved. We're building a bot that's actually intelligent and useful and a bot that can give users a great support experience.
The bot starts by finding a relevant discussion forum and extracts information about each post, such as the text, what it's replying to, upvotes, and more.
It then uses a machine learning algorithm to determine the relevance of each post and sorts posts into multiple categories.
It represents the forum as a tree of replies to find the solution that worked best, and displays that solution to you.