One of the best parts of my job is that I get to play with cool stuff. Sometimes I try and explain these very cool things to my family, however, most of the time they fail to see the “cool factor” or understand why I’m so excited by it. Just the other night during dinner, I decided to raise the topic of what I had been working on that day. I started with the phrase, “Hey, this was pretty cool. I was using K2, UIPath, and Azure Cognitive Services to perform sentiment analysis on current news stories.”
At which point my 5-year-old decided to return to watching an episode of Paw Patrol that he has seen at least 300 times, my 2-year-old daughter immediately threw most of her dinner on the floor, and my wife looked me dead in the eye and said, “ I don’t understand one single word you just said, but as long as you are happy, “Go you!” As such, writing this blog is cathartic, as I may have the opportunity for you to read past the first paragraph, and you may by the end get something out of it.
Working towards intelligent automation with K2 and RPA
For those who don’t know, K2 is a powerful low-code app development platform that enables organizations to automate and build end-to-end applications. These apps can integrate with multiple line-of-business systems and data sources including some clever technology that sits perfectly with digital process management such as robotic process automation (RPA) and strands of artificial intelligence.
So, over the last week I thought – wouldn’t it be cool to utilize some great work from my colleague who hooked up K2 to the new Azure Text Analysis API. This API allows you to pass in text and respond with key data about the text you passed in like the language it detected, key phrases, linked entities (URLs to articles about the entities it found), and in my case sentiment. Now for those who kind of know what sentiment is but would like a further description, here you go:
Sentiment score is a percentage given between 0% and 100% that describes how negative or positive the words are in the analyzed text. 0% being very negative and 100% being very positive. This can be used to score people’s responses in certain contexts such as analysis of the general trends of people’s comments on social media pages of your organization. Are people being positive or negative about your company or product?
In my case, I didn’t particularly care about targeting a specific company or product. All I wanted was to pull the tech together and hook it up. So where could I get real-time data or text that would allow me to gauge a mood using sentiment analysis? “Enter Google Search. Enter the Google News tab. Great!” I have a data source but how can I get unstructured data into K2.
We’re very model-based at K2, and although I could probably find a way to do this with a Search API to Google, why bother. What I really need is a tool that would allow me to scrape all the news stories automatically without me needing to lift a finger, welcome to the stage UiPath – a leader in RPA.
With UiPath, I created a robot that would open a web browser, navigate to Google Search, type in the search term I gave it, navigate to the news tab on the search results page then scrape all the lead paragraphs from that page. The robot then tells K2 it’s done and passes the data back into the K2 app.
Finally, I created a simple K2 SmartForm that allows you to enter a search term that would be fed into the robot and show a list of results when done. A simple workflow is used to coordinate the results coming back from the robot and the sentiment analysis is performed on the text extracted.
From concept to delivery in hours
This took a few hours to build from scratch. From concept to delivery. This was all out-of-the-box plug and play K2 functionality with no code integration into UiPath and Azure Cognitive Services.
I realize this example is pretty narrow in intention and a bit rough in terms of proven capability however if we shift our eyes towards the horizon, even slightly, this technology can be placed in the hands of organizations who would otherwise not have access to it. If we think about the day-to-day interactions with our customers – whether it be Twitter, email, Facebook, or even scanned in “snail mail” could we screen and redirect all incoming correspondence, regardless of the platform to the correct people based on the content, language, keywords or sentiment?
By creating this level of intelligent automation we as customer success managers can see where the pain points are, deal with the most urgent requests without having to triage the interactions manually. You can then take that a step further by creating heat maps with the most common questions that are asked of your organization. In fact, maybe we could screen outgoing correspondence to make keep our “tone of voice” in check.
To summarize, within a few hours I connected three very powerful tools in a simple example that demonstrates the unlimited technological potential organizations can grasp when using K2, with UiPath and a smattering of cognitive services from Microsoft Azure.
I also now know I must change my “tone of voice” if I will have any chance of my family understanding or getting excited by what I do every day when I dance down my hallway into my office every morning.