Where Is The Value?
With a lot of the services we used daily such as Google incorporating Natural Language Processing into their applications it’s easy for us to take for granted the power of simply answering people’s questions as opposed to giving them a huge suite of tools and letting them figure out where the value is. However, a lot of apps are still made with the sole purpose of lumbering the users with a never-ending supply of tools with which is perceived as great value.
To be fair, at VIX we were guilty of this too until very recently. Our software tools began to suffer feature bloat, support emails increased and when we actually measured it, only a few of the features in the tools actually got a substantial amount of use. The main reason behind all of this was that it was simply too confusing and/or time consuming for people to do what they needed to do with the product.
Enter The Ask Feature
We don’t have a name for it yet, but codenamed ASK our newest suite-wide feature is essentially a search box where you can type whatever it is you’re looking for, or a question you need the answer to.
When we first came up with the idea to include this in our products, we didn’t give much thought to exactly how it would work. We just knew that we needed it and would figure out the hard part later. No gain without pain and all that jazz.
We decided to start simple, by getting everyone to type in questions they would like to get quick answers to and we recorded these questions and occurrences in a database. With this data, we could see the most common questions and identify key pain points within the tools. Before long we realised how badly this feature was needed - things we thought were easy or convenient for people to do already turned out to be some of the most commonly asked questions.
Coming Up With The Answers
From the bunch of questions we gathered in the first test we realised two things;
- A few of the questions were asked multiple times by different users and had straightforward answers
- There will always be questions that surprise you
This information made us realise we could tackle answering questions with a two pronged approach. There were questions that had solid answers that were simple, and the questions themselves were very straightforward with little variation between occurrences - this meant that we could manually write answers to the most frequently asked questions pretty easily.
However, what I meant by “there will always be questions that surprise you” is that some questions are actually quite specific and might be things we’ve never even considered people would want to know or see. Certain comparisons of different data on one chart for example - it would take us a long time to manually go through and create all of these answers for every eventuality.
Playing The Brain Game
Some of the areas that we at VIX are really delving into at the moment are Machine Learning and Neural Networks. We currently use a variety of different Machine Learning algorithms to spot trends and make smart predictions based on data we collect.
It seemed obvious to us that we could extend and apply this experience to Natural Language Processing and produce an AI ‘brain’ that could generate and suggest its own answers to the asker and also produce charts, and perform calculations requested by the user on a whim.
While this approach has a lot of upfront work involved, it could allows us to build a system that we can plug into all of our applications in the future to allows anyone of any IT literacy to open the tool of their choice, and simply tell the brain what they want or what they need to do.
Look out for more in-depth blog posts on our machine learning algorithms and neural networks in the near future!