Frequently Asked Questions
Where do you get your opinions?
There are 30 bizillion (ok, that’s a made up word) pages of data out there on the Internet and much of it contains people’s opinions. Our ultimate goal is to provide a comprehensive collection of all opinions wherever they may exist but we had to start somewhere so we will initially be focusing on Twitter and then expanding out to blogs, Facebook public and Google+. We respect your privacy so we’re only interested in posts from these sources in the public domain. Which means if you have a private blog, don’t worry, we can’t see it, and if we could we’d do everything we could to not use it. We’ll continue to look into new sources of opinions and hopefully update our data from time to time to bring new info into the mix.
Do you collect every opinion out there?
We’d love to and hope to do so one day but in the short run our goal is to analyze way more than enough to get a statistical representation of public opinion on just about all of the topics we cover. Does this mean we’ll have solid info on most people, products, entertainment and issues out there? We certainly think so. Does it mean we’ll have good coverage on your 6th grade science diorama? Probably not. When the level of discussion is low, we’ll try to let you know not to bet the farm on the results as they may not be terribly representative. And please remember, since this is an alpha release we are only processing a 1% Twitter feed for the time being. We will be adding content over time.
How current is your data?
Pretty current. We are currently updating our indexes every 4 hours but will bring that down to within a few minutes of someone hitting the publish button over time. Sometimes we come in early and sometimes we’re a little late but we are talking minutes and not days here so when you hit the “Tune In” or “Today” button you’re getting very current opinions.
What do the percentages mean?
The percentages are the percent positive opinions and negative opinions based on the total posts we’ve been able to determine had a sentiment we could analyze. What on earth does that mean? Well, we can’t determine sentiment on every post. Some posts are truly neutral (i.e. “Standing in line at the post office”) and some are just too tricky for us to take a definitive stand on (i.e. “If my senator had voted for the farm bill I might have voted for him again.”) But one great thing about writing on the Internet is most of us don’t pull any punches about our opinion on something. This means we’re likely to analyze just about all of the opinions we see. So, back to the question. Our percentages are really simple math – % Good equals the number of posts on a subject we’ve determined to be positive divided by the sum of the total positives and negatives determined for the same topic. You don’t have to go too far out on a limb to figure out how we do the % Poor.
Are there other adjectives or sentient indicators used by people out there for my topic or are the ones you provided the whole enchilada?
In many cases there are additional adjectives, emoticons, etc. that people have used to describe the topic of interest to you, however, we had to draw the line somewhere and we use the ones with the highest confidence to be positive or negative in the context of the language we analyzed. One interesting side note about adjectives. We tend not to include too much obscenity in our sentiment analysis. No not because we’re opposed to it (you should see our engineering meetings) but because it often doesn’t give you much to go on to determine sentiment. For example, something is likely to be as F bomb great as F bomb bad.
How accurate is your sentiment?
Pretty accurate. We’ll try not to put you to sleep or fell trees unnecessarily with academic papers but the reality is humans have a hard enough time agreeing about the sentiment of something between themselves, let alone when we bring computers into the mix. And, (at least for the time being) humans run computers so humans have control over the knobs and levers which can be turned and pulled to adjust the accuracy of the sentiment. Sometimes algorithms are tuned to be as accurate as possible (for example, if we only score sentiment for “I love / hate cheeseburgers”) but they don’t score many posts so the volumes are very low and the results may not be reliable. Other times they open up the spigot too much and go after volume at the expense of accuracy. We try to find the right balance to give you best mix of enough volume with the highest accuracy to provide results you can trust. Does that mean we’ll be as accurate as humans? Sometime yes and sometimes no. But we believe our results will provide the right guidance about where pubic opinion is, how it is changing and how it compares to other topics of interest.
One other thought about sentiment. Not all topics are ideal for determining sentiment. News for example is a category which can be a challenge. If a natural disaster befalls a city people may express their negativity towards the event and their support for the heroic efforts of the disaster relief agencies, but either way these sentiments are not typically directed at the city itself but represent positive and negative discussion involving the city. In these cases it is often best to drill in a little deeper into the co-occurring topics to uncover where the opinions are directed. On the other hand a politician who is caught on the wrong side of a campaign finance indictment may face a tsunami of social media vitriol aimed directly at him or her. In this case the assessed sentiment may be spot on.
When using the app please keep in the back of your mind the topic you’re exploring and if sentiment is really relevant to the circumstances. Even when the sentiment stars don’t align there is often great value in exploring the co-topics and sentiment wording around your query to discover some very interesting information.
And lastly, rest assured as long as the lights are on at 30dB, we are working on improving our sentiment as well as all of the other moving parts which contribute to our results.
How many opinions does a query need to have results provided?
We were very concerned with this one when we began designing our system. Seems to us that if you get 2 posts on a topic it’s probably not terribly representative of what the Internet population thinks on the subject. It also opens you up to getting gamed – people supporting or slamming something not based on their true opinion but because they have a vested interest in the topic or the competitor. We know this is hard to imagine, especially in the anonymous world of the Internet. However, after giving it some thought we decided to show you what people are saying and let you be the judge. Still, our math guys couldn’t sit quite still in their chairs so we’ve provided a little warning flag whenever the guys with the pocket protectors feel the count is just too low. In this case it may be best to increase the date range of your query to see if you can get more opinions.
What does “Low Volume” mean”?
This is one way we want to let you know not to bet the farm just yet on the results. The threshold is currently set at 10, which is still a ways from statistically representative, however, we had to draw the line somewhere and this is where things sit today. And even if you get the Low Volume warning there is often great value in the discovery elements of the app. You can still click through to posts on topics or sentiment terms which catch your eye.
Can I compare two things?
Absolutely. This may be one of our coolest features. It’s easy to do. Just hit the Compare button and drop in the topic(s) you want to compare to your original topic. Done. You’ll see the positive sentiment for the two topics either side by side (in the towers view) or trended against each other (in the trend view). You’ll also see the co-occurring topics which they share in common.
Do I have to select co-occurring topics that exist in the widget or can I add in my own?
Both. To drill further into a topic you can either select a co-occurring topic from the topic slider below the towers or the trend lines, or, if you don’t see what you’re interested in, you can add anything to the query line. Just use a comma to separate the terms of interest. For example, “South Park, Kenny” will give you results for posts expressing opinions on the the character Kenny from South Park. If you add in a topic which is not in our index or one which is not used in connection with your query (i.e. Mt. Vesuvius, Socks) you’ll come up empty. If this happens, no problem. Adjust the query in the query box or click the first topic in the breadcrumbs in the widget to back the results up.
How do I publish your widget into my blog, website, etc?
This one’s a lay up. Hit the “Embed” button on the results widget. Chose whether you want to use live data (the widget will constantly update) or locked data (the data will stay the same as the day you published it until someone tinkers with the widget). The embed code is already highlighted so just use the copy function on your computer and and paste it wherever you’d like to see it online. See also Get an Infographic.
Why doesn’t the embed code work?
We know that some blogging and other publishing platforms do not readily allow the embedding of widgets. For example, unless you host your own WordPress blog they might shut you out. We’re working to make it as easy for everyone to embed our widgets as it is to embed YouTube widgets but it may take a bit of time. In the interim, you can always embed a link via the Share feature or embed an image of the results. Better yet, please let your blogging host know you’d like to embed our widgets.
Can I share the results on my social media network?
Please do. Just hit the Share button at the bottom of the results widget and you can share an image or a link on your favorite network. Facebook, Twitter, Google+, Pintrest, Tumblr, Gmail and others are all up for grabs. We’ll keep adding more social networks over time, but 30dB finally gives you the goods to spark a conversation or just tell the world how right you were – full stop.
So, this is just like an online poll right?
Close but not exactly. Polls are all about asking a specific question to a targeted group of people. We’re all about listening not asking, so we report back on issues and opinions expressed by millions of people writing in social media online. Think of us as a game of Jeopardy – we have all the answers sitting in our database from people expressing their opinions online and you’re the contestant coming up with the question. Unlike polls, we never stop collecting new opinions expressed online and our topics are limited only by what people write about so you can get online opinions about anything on your mind updated as frequently as you’d like to see it. Now if we could only get Alex Trebeck to do a vanity search on himself.
Can I see all of the posts on my topic on the samples pages?
Nope. It’s just too much data for us to keep around. Some day maybe but for now we are shooting for a few pages of samples on your query.
How are the topics in the Best Ofs determined?
The topics in our Best Of categories are determined primarily through a combination of internal and external resources. We use externally published information from sources online such as Billboard, Nielsen, Box Office Mojo, etc. along with our internal data on Tweet volumes with sentiment and query volumes in the app. To that we add a bit of editorial license to make sure some old standards make the cut.
We know some of your favs may not make it on to the list but you are always welcome to run a query in the app and compare your choice to any on the list or others.
Lastly, there is currently no magic to the order of the topics in a Best Of. We’ll keep an eye on how they are used and tinker with some display options down the road.