I want to share with you an overview of the steps a user takes to build a project in Watson Analytics for Social Media and why each step makes social data more relevant, trustworthy, and valuable.
Bring Watson Analytics for Social Media the questions that make you scared, excited, confused, or optimistic. Use it to relieve or confirm your fears about disruption in your industry. Use it to understand reputational risk, loyalty among your own customers, and those of your competitors. Use it to find answers to questions you didn’t even know to ask.
Social media is vast – there are thousands of thousands of posts added to conversations across social media sites each day.
Some of it is relevant to you, and some of it is not. What’s relevant really depends on the question you need to answer. To trim social media conversations down to just the relevant ones, create topics. Each topic (you need one, but you could create dozens) is a list of keywords that must appear in a post in social media for that post to be added to your dataset. Basically, you’re creating a custom subset of social data that contains only relevant content.
You should create a list of include terms for each topic because there are often many ways to refer to the same topic – tennis shoes and sneakers and trainers, for example. You aren’t on your own to figure this out, you have Watson’s cognitive power to help you. The feature is called Topic Suggestions, and it pulls a real sample of social data based on what you’ve added to your topic so far. If your Topic Suggestions are relevant, you can be sure that your final dataset will include only relevant data.
Themes are the ways in which your custom subset of social data will be classified – this is where text analytics comes in. If all we showed you was how many times people mentioned your topics, you might be interested, but you’d probably be no closer to answering the question that brought you into Watson Analytics for Social
Media in the first place. You need themes.
When I’m making a Watson Analytics for Social Media project I often use themes like “expensive” “I want to buy this” and “quality,” but you should use themes that help you answer your question. You need to build out your themes the way that you build out topics. Sure, some people will say, “these sneakers are expensive” or “these tennis shoes are awesome. I want to buy them” but many people won’t use those exact words. So I usually add “pricey, big bucks, an arm and a leg, too much money, out of budget” etc. to capture a variety of ways that people could express the feeling that something is too expensive. Similarly for “I want to buy this” I usually add, “going to get” “gonna get” “wanna get” “planning to buy” “will buy” etc.
There are a few other decisions to make when you’re setting up a Watson Analytics for Social Media project – dates, languages, and sources. If your project depends on seasonal availability, make your date range a year. If you’re wondering about a week-long conference, limit it to the duration of the event. If your brand name is an irrelevant word in another language, unselect that language to keep those mentions out of your project. If you only want to see product reviews, unselect all other sources.
Once you finish this process you’ll click “create dataset” and Watson Analytics for Social Media will do just that: quickly grab all of the documents that include your topics, and read and classify them by theme. When this process is finished, you’ll be presented with prebuilt visualizations that you can filter and manipulate until your question is answered.
For more information about Watson Analytics for Social Media and the free trial, visit this page.