Optimalaccess Chatbot

close
send

Curation will help you build better chatbots, faster!

Chatbot conversations are divided into two major categories, linear dialogs, and non-linear dialogs.

In this post, we discuss a powerful method for designing non-linear dialogs efficiently.

In linear dialogs, the visitor makes a simple statement and gets a simple reply, and the intent is generally to collect information to complete an answer.

 

 

In non-linear dialogs, the answer will depend on the response of the user and may have different branches.

Our focus in this post is to explore how we can use curation to build non-linear dialogs efficiently.

Our chatbot platform of choice is Dialogflow, a Google-owned developer of human-computer interaction technologies based on natural language conversations.

Elements of a great Dialog Flow

Ben Beck, in his post on “How to Write a script for a chatbot” defines the following steps as a guideline:

  • First, you should define your goals
  • Do a quick sketch of the actual conversation flow on paper
  • Follow that up with a more polished dialog flow and some pre-testing
  • Force yourself to simplify your Bot
  • Lastly, launch a prototype version and test some more.

To demonstrate how curation can help you with this process we will use an example everyone can easily relate to, politics and advocacy.

Any politician or advocacy organization needs to appeal to a large group of people with varied interest or concerns.

For this reason, user interest is generally not bound to one set of interests, but different people have different concerns that need to be responded to.

With this in mind, let’s see how we can use curation to approach our project and build a dialog flow that satisfies the needs of our audience.

Building our Intents with Dialogflow!

In Dialogflow, the Intent module is the centerpiece of the entire platform and has two core functions.

  • Detect what users are saying
  • Answer their questions

We will show you how curation can help you with both of these objectives.

Referring back to the guidelines shared above, our first task is to set our goal.

In our example, we know that our goal is to inform voters of how the candidate will solve their concerns.

Therefore, our approach is to curate and organize candidates policies, social media posts, and any news coverage to inform our chatbot design.

Using curation to build a mind map

Our curation tools are designed and optimized for building a powerful index for our research.

Any link accessible via the Internet can be easily curated and furnished with a title, description, image, content tags, publisher and author information.

The indexing of the information is what will help us build our dialogs.

Therefore the content tags we assign to each post should be assigned with the dialog-flow in mind.

For example, each post we curate on the topic of the economy should also specify additional detail about the content like the areas of the economy addressed (jobs, minorities, urban areas, middle class, wages, etc).

From here on; the tags will basically help us design our chatbot efficiently.

To provide an example we have built an Elizabeth Warren bot using our curation page below as a guide.

Click to See Elizabeth Warren’s curated library

Curation will inform your Intents

The promise here is that once you build your curated library, building your chatbot will be a breeze.

In our example, once we built our library, describing the intents was quick.

The library helped us with both the training phrases and the flow of our bot.

“Training Phrases”, is the section in the Intent module where you train the platform to understand user questions.

With a curated library, we simply looked at phrases in the post and comment section to build our training phrases.

So for the intent “economy“, we used the following phrases

  • What is your plan on the economy
  • How will you economic plan help farmers
  • What do you mean by economic patriotism
  • How will green manufacturing affect the economy
  • Will green manufacturing produce more jobs
  • How will your plan deal with Wall Street stranglehold on the economy
  • ….

Next, we looked at our related tags to the economy and built some more specific intents like

  • rural economy
  • reform wall street
  • Green New Deal

As the campaign progresses, updating our library will automatically inform our chatbot and help keep it up to date both with the latest content and new Intents.

Using “Basic Card” to Build curated answers

There is another great advantage to using our curation solutions to build your chatbots.

Your curation can actually be the answer users see in their chatbot.

To integrate answers from our curated library with our chatbot we use a type of answer provided with “Google Assistant” called “Basic-Card”

Basic-Card lets you add an image, title and external link to your answer.

Now to embed your chatbot in your website, or create links to be shared via SMS, you need a solution we use called BotCopy.

It is actually BotCopy that opens external links inside the chatbot providing seamless integration.

To see how this works click on our example for climate change here.

Conclusion

Memory is simply put an index of what we know.

A curated library will perform a similar function as our memory does. It allows us to index and organize information for different places and present them in a Pinterest type portal to help us both remember and share our knowledge!

A chatbot is designed to help us find the best answers to our questions easily.

Both curation and chatbots have the same goal and that’s why the combination will help you build better chatbots faster.

If you like to learn more contact us.