Mining Consumer Opinion in Comments

An interview with Daniel Ha and Steve Roy from Disqus on mining opinion in comments. 

Commonly known as a comment system, Disqus facilitates comments from over 2.5 million sites. The team at Disqus, Daniel Ha and Steve Roy, like to think of themselves as a community of other communities. But how do they distinguish themselves?

 Communities and Identity

Any discussion that happens on Disqus, by its nature is its own community. Disqus found that the majority of users’ time was spent below the fold, in the comments. Part of what fuels this is the ability to act under a pseudonym. Disqus maintains that by embracing a pseudonym, people can act as their “real” self. They find that people who embrace a pseudonym reveal a more passionate interest than they normally would. It gives people a voice they wouldn’t typically be able to use, enabling a user to pursue things that mainstream media may not be covering, or to be part of a community they couldn’t otherwise.

Brands

Brands can tap into Disqus in a couple ways:

  1. On their properties utilizing Disqus: Brands like HP have launched destination websites with Disqus to participate in the conversation naturally happening.
  2. Disqus’ ad product: Brands can pay to have a presence in other websites (like a Tumblr blog) and place their content above the comment feed. The response to this placement of content is higher as well because it’s located where the audience is more engaged.
  3. Learned Insights: Brands can use pattern detection to learn stories about their brands. A great example of this is when there needs to be a product recall, because a lot of this type of discussion takes place in these stories.

Data Learnings

Disqus recently achieved a major milestone, reaching 1 billion monthly unique visitors. Often considered US focused, the majority of their growth in recent months is international. Disqus supports 40+ languages worldwide. Through its many users, Disqus has been able to understand the behavior patterns on their networks and noted 3 things in particular:

  •  Comment Length: The amount of characters can tell a lot about the level of interest in users. Steve says 57% of all comments are essentially the lengths of Tweets (under 140 characters) and not using links.
  • Time of Day: The worldwide pattern for commenting shows a peak in volume at 10 am in every time zone. Not only does this mean more people comment at this time of day, they also engage with other comments and read comments then too.
  • Categories: Disqus buckets their sites into about 45 different types. Each category has various statistics associated with their category as well. For instance, gamer sites average about 10 characters per comment. Religious sites, on the other hand, average closer to 600 characters per comment. As a brand, this is valuable data that can help shape how they engage with users.

Disqus is proud of the use cases of their data too. Several examples were mentioned, like Gooqus, a search engine utilizing both Google custom search and Disqus.This allows a user to not only see the top Google results, but also add a layer of richness, allowing for more sentiment to be derived from the data.

Big Boulder is the world’s first social data conference. Follow along at #BigBoulder, on the blog under Big BoulderBig Boulder on Storify and on Gnip’s Facebook page.