How Can You Analyze the Relevance and Sentiment of Online Data?
Salience and polarity.
These words sound like complex chemistry terms. In reality, salience and polarity help our team at BrightPlanet determine which pieces of content from the Internet are most relevant to your business. We leverage our technology partner Rosoka to help access salience and polarity in the data we harvest.
When we harvest data from open sources, we’re looking at millions of entries, from Tweets and news stories to bankruptcy databases and medical journals. Combing through the vast amount of information online, we use salience and polarity to select the most important documents, which you can use to protect your brand and educate your team.
What is Salience?
When Rosoka’s technology is looking at a document for salience, it is assessing the importance of an entity within a document we’ve harvested. This isn’t just a count of how many times a person or thing is mentioned in the content, but the context of those mentions.
Salience is reported on a scale of 0 to 100, with higher number indicating more salient — or more important — entities.
But how does that help businesses using services like BrightPlanet?
Political campaigns try to track relevant mentions of candidates within the race and use data to better analyze polls. Oftentimes they’re limited to simply keyword mentions within news articles giving them every single mention regardless of relevance.
Unless you use open source intelligence (OSINT) with salience.
Combining collection of OSINT, in this case news media, with salience detection, political managers could filter in on only mentions of candidates that have substance and importance to help further filter through the noise. Filtering in on higher salient news articles would ensure the news articles are indeed about the candidate.
How Does Polarity Fit in?
Rosoka can take this data to the next level by adding in polarity. It tells us whether an entity is mentioned in a positive, negative or neutral context.
Polarity is reported on a scale from negative three to positive three, with zero being neutral.
Returning to our example of the political campaign manager, this would allow them to quickly sort between any news articles collected that are flagged as very positive or very negative. Keep in mind polarity is calculated for each individual entity within a webpage – an entity being a name of a person in this example.
This means that there are many positive or negative scores within one webpage. We calculate based on entity as opposed to webpage because it gives you more accurate data. It’s also impractical to take a webpage and perform analysis to see how positive the webpage is as a whole compared to detecting how positive snippets within a webpage are.
How it Works for You
BrightPlanet harvests data from any source — regardless of style or language. Even though things like Twitter are short and include abbreviations and slang, we can turn that into relevant information your business can put into use.
It’s time for companies of all shapes and sizes — across all industries — to take advantage of the wealth of information available online. If you want to learn how to put OSINT technologies into action at your business, reach out to us about a free demo.
//
