Heart

Menu

Stay in the loop

Never miss a beat with the latest in VET news delivered straight to your inbox.

Stay in the loop

Never miss a beat with the latest in VET news delivered straight to your inbox.

Building the Tree of Knowledge - Exploring the Use Case for Educational Chatbots

March 23, 2018

With an increasing number of tech giants getting into the business of the smart home assistant, modern life is edging closer and closer to a sci-fi movie’s depiction of future home life.

“Hey Siri, add bananas to my shopping list.”

“Alexa, buy more deodorant.”

“Ok Google, set my alarm for 7 AM.”

Beneath the dazzling home control functionalities and fancy voice interactions lies the essence of a chatbot program, which maps human utterances to intents, and from intents reaches programs to carry out the intended tasks:

Utterance → intent → task procedures → output

This is the exact model of Amazon’s Alexa, and its outputs carry out hundreds of services: ticket booking, food ordering, hotel room reservation, and more.

Smart Assistants in Education

While the use case for an education industry chatbot may be slightly different, its primary purpose is to retrieve the knowledge required by the student— be it disambiguation of concepts, clarification of contexts, or fetching resources. Like home assistants, it too is only bound by one intent: information retrieval.

The assistant may solely depend on a single or a combination of several NLP (natural language processing) algorithms to handle the query, but without the proper structuring of data, NLP algorithms alone will only generate a fixed percentage of wrong outputs.

Sensible curation of data is essential for effective information retrieval. Data is organised as an index list for a search engine, but for a domain-specific knowledge base, information is adapted into a tree structure. There is a reason we use the phrase knowledge tree — the tree branches out as knowledge goes from general to specific, from low-resolution to high-definition.

The Tree of Knowledge

Knowledge points are put together as tree nodes, and the identification of the most relevant topic is simply a traversal of the tree, each step being guided by a particular filter function.

‍The transformation of Wikipedia’s knowledge base for “Anthropology” into a tree structure. Answering a query takes log(n) time in complexity.

Organising the data in a way that’s optimal for the intended task is essential to information retrieval. A tree structure will enable a fast and accurate query, which makes it an ideal way to structure a chatbot’s knowledge base.