The Creation of Pubtechgator’s Facebook Chatbot
BMJ Digital Strategy team have just launched a Pubtechgator’s Facebook chatbot!
Our chatbot allows users to interact with the content published on our Pubtechgator.bmj.com aggregator website in very much in the same light as The Guardian, The Wall Street Journal and other publishers do with their Facebook bots.
The main reasons for this work is to help us to experiment with bot technologies in a safe environment without the fear of adversely affecting our users with experimental technologies. Specifically we were looking to achieve the following goals with this project:
- Understand how bots function
- What are some of the main bot engines?
- Understand how users interact with bots
- See how easy it it to integrate our content from WordPress websites and our catalog API
When learning about a new technology or learning a new skill it’s always a challenge to know where to start. I took the approach of reviewing the bot tutorials on YouTube, Udemy and Lynda to find out what technology was popular and have an understanding of what the different bots could achieve. From this process I decided to focus my attention on 2 different bot engines Dialogflow and ChatFuel.
Diaglogflow is a very good bot engine especially for free text content as its AI system is very good at predicting the meaning of free text and responding with the most appropriate action. It also had a template library of prebuilt agents that can be quickly used for different purposes such as FAQ bots process customer queries or bots to handle online shopping transactions.
On the other hand I like ChatFuel because it a very visual bot, it makes use of visual cards to help the user find appropriate content. ChatFuel also has a more visual approach to programming the bot’s behaviour. It makes use of visual building blocks that you can link together to create the desired interaction in addition to its AI core that interprets key-phrases and the corresponding bot replies. The bot understands user phrases similar to those you’ve setup and replies with some text or logical action.
Our Bot’s Functionality
Here’s a quick summary of our bot:
- Personalised welcome message
- Navigate content using cards or free text
- Free text article search of techpubgator.bmj.com
- Free text article search of BMJ’s content catalog
- Subscribe for daily content notifications
- Integrate curated content from a Google Sheet (Gator Picks)
- Respond to random questions
The Bot’s Architecture
To facilitate the above features we used the the following architecture.
What did we learn
- Creating a basic bot quite easy to do
- You can do quite a few powerful searches using a WordPress rss feed
- Images on feeds posts are vital to the look and feel of the bot
- The Facebook approval process was easier than expected
- Most users find it difficult to get started with bots
- We expect the popularity of bots to increase with systems such as Alexa & Google Assistant
- We expect to see bots being integrated into many products and more than just customer services as it’s a good mechanism to integrate multiple platforms, perform sales…
Try out our bot HERE
Dan Amos – Digital Strategy Lead at BMJ