An organisation's custom content needs to be made available to a user by using an invocation in Google Assistant. For example, in the demo, Talk to Digital Trainer was the invocation that directed all requests from the user to the Digital Trainer content. For purposes of convenience, we call this customised content and its working as a Channel.
For purposes of ease, the hida platform groups content into Topics. Topics help a user identify the content they are interested in. Examples of a topic can be: HR Policy, or even just Leave Policy, or Product Details of ... and so on. We recommend that a topic should not be very large for better performance. A topic scope, where all the contents of a topic can get answered in about 50 questions, would be a suitable size.
Think of a topic content as a plain text file which holds all the answers to the questions related to that topic. Of course, the content creator (author) may wish to add images, or provide a link to a video, or apply some highlighting to some part of the text to make it more interesting. Check the content guidelines to know more about it.
Authors are expected to list all possible unique questions (to the extent possible) that can be answered from within the content. For every question, they need to pick that part of the content where the answer lies. The question provided by the author is an indicative one, while the actual user may ask it in a different manner. It is the job of hida to decipher that.
Using the question-answer based content of each topic, the author can choose to provide one or more of three modes to the user. They are (explained below): Ask, Explain or Test.
This is the feature where a user can ask any question related to the topic using Natural Language Query. The hida platform would interpret the query and make a best-guess of the question-answer from that topic content.
For certain topics, especially of the training course variety, the author may wish to provide a methodical explanation of the topic. In hida terms, this constitutes an ordered list of questions that the user would be taken through in steps; more like running through a set of slides.
The Test is an assessment functionality of the topic content. The platform will pick one or many assessment questions put by the author and ask the user in a random manner. Questions for test purposes may follow an exact logic or an approximate logic. The latter is used for longish answers where the user's response is compared with the provided answer and a score determined based on accuracy.
- User opens Google Assistant on phone. Makes invocation by saying, talk to ....
- Now user is in the channel. The initial response is a Greeting asking the user to choose a topic.
- User can say a topic name or say, list topics. From the list shown, the user may choose to tap on desired topic.
- Once user has selected a topic, based on the modes set by the author, the user can seek an explanation or ask questions or get tested on it.
- The user should be able to navigate easily by following the responses.
- To exit a topic, the user should say, exit topic and return to the point where the user can choose another topic or exit the channel.
- To exit the channel, the user should say goodbye. This gets the user back to the normal Google Assistant mode where general queries may be asked.
For an understanding of how all the software components of the ecosystem works, check this Technical document.
Write to: [email protected] to initiate steps to set up your own channel.