Building a Chatbot Part 2: Progress & Platforms
Nina Cialone shares her journey as she builds a chatbot for Lehigh University's student-run publication, The Brown and White - with no coding background! Follow along as she documents the process.
Photo Source: Story Xpress
We all make mistakes, right? Progress is never linear.
It’s been a little while since my first piece in the Building a Chatbot series, and that’s because we’ve been finding our footing and figuring out what direction we want to take this thing. Along the way, we’ve made a few errors, which I won’t shy away from discussing, so that you guys can learn from my experience.
In that first piece, I mentioned I would be taking the chatbot building class by IBM through CognitiveClass.ai. The course opened with the following statement.
“We have created this course so that anyone with basic computer skills would be able to learn to create a chatbot. The only prerequisite for this course is your desire to learn.”
I have to say, after completing this course, they were absolutely right! Unfortunately, I also realized along the way that it wasn’t exactly the course I was looking for. Right now, I could build a simple chatbot, such as one of those little customer service bots that many websites have adopted, and deploy it to a WordPress website.
The course teaches you how to leverage IBM Watson Assistant and does a great job explaining each of the features you are instructed to use. So, if you’re looking to build a simple chatbot that answers a predetermined set of questions, or performs a few simple tasks, I would highly recommend this course. It took me about seven hours to complete, and each module was broken down into very simple steps with pictures to help you follow along. The resulting chatbot is extremely responsive and scalable, and it could probably reduce the workload of a human customer care team by half.
However, that is not the type of chatbot I was looking to build. The key feature of our chatbot will be its ability to scour a database of previous publications and find the correct answer to a user’s query.
After some digging, I found that IBM does have a tutorial on connecting your chatbot to a database, but it involves a bit of coding using Python. The level of programming doesn’t look unreasonable, and the tutorial includes most of the code for you, so if you know a bit of Python, this might be incredibly useful for you!
However, in order to demonstrate how accessible generative artificial intelligence has become to anyone interested in using it to their advantage, we aim to build the chatbot with no coding at all.
With all of that being said, we decided to stick with our initial goal of a zero-code chatbot and pivot away from IBM. OpenAI has really been at the forefront of this generative AI explosion, so we’re looking into using ChatGPT as our natural language processing (NLP) tool.
What I’ve gathered from some further research is that regardless of which company you choose, you can find various how-to articles online with instructions on how to build a generative AI-powered chatbot using Python. Most of these articles even provide all of the code that you need. Here is an example of one such article that uses ChatGPT.
Although these instructions do make the process significantly more accessible than if you were to develop the code yourself, what we’re looking for is the ultimate level of accessibility, so I kept searching.
After a while, I began to feel frustrated. There are a number of courses out there that can teach you how to build a chatbot without coding, as long as you only need it to answer a specific set of questions for which you can directly input the answers. You can also connect your chatbot to a generative AI service using products such as Zapier, but accessing a database is where it gets tricky and seems to require some level of programming.
Finally, I came across three services that I want to explore further.
Zapier’s ChatGPT integration: This service allows you to create a Slackbot that connects to ChatGPT. I’d like to see if figuring out how to upload files to ChatGPT would solve our database issue.
Udemy’s free chatbot building course: This quick course claims no coding is required. Maybe it will answer some of the questions I had after IBM’s course and demonstrate how to use other models, such as ChatGPT.
Botsonic: A no-code chatbot builder that emphasizes its ability to train on your own data – just what we’ve been looking for!
I promise to write an update sooner than last time, and hopefully one of these three paths (or something that I’ve yet to discover!) will be the strategy that we’ve been looking for. Also on the agenda is a meeting with the Advisor and Editor-In-Chief of The Brown and White to nail down the areas we need this chatbot to service. I’m really enjoying this journey so far, even despite any missteps, and I look forward to writing another update as soon as one is available!
Best,
Nina for the Don’t Count Us Out Yet Team
Related Articles:
 | A guest post by |