And second, publishers love to hate comments sections. Comments can indicate how readers feel about a publisher’s content, but if not moderated properly, discussions can spin out of control, which alienates users, the newsroom or both.

While these factors may seem unrelated, tackling the challenges of the comments section can materially help publishers fulfill their need for first-party data. User-generated content, like that in comments sections, can provide publishers with valuable audience insight to continue to engage with their readers meaningfully.

What’s necessary to make comments sections rich sources of data?

Gathering useful insights from comments sections depends on at least a couple of conditions. Firstly, comments need to be moderated by AI. And secondly, only people who are registered with the site can be enabled to comment. The minute a reader registers, the process of capturing data can begin.

When a publisher follows a registered user’s movements on their site, there’s quite a lot to glean from their interactions with the comments section alone.

There are declarative data, which includes information that a user actively gives, such as when they post a comment. And there’s also inferred data, which tracks a user’s interest, tone or sentiment. For instance, inferred data could come from a reader reacting to a comment with a “like” instead of commenting.

Then there’s augmented or enriched data, which is acquired when the AI learns qualitative information about commenters to gauge individual users’ behaviors, preferences and desires and how people feel about specific content. This type of data can get very granular because it eventually allows publishers to identify purchase intent and brand affinity.

For instance, imagine a controversial author publishing a high-performing opinion piece on a publisher’s site, and they want to get a sense of whether or not readers think it’s too inflammatory. Most publishers probably don’t have the resources to sift through, say, several thousand comments — even if the AI doesn’t publish the uncivil ones in the first place. In this example, the machine-learning tool can instead process the language in the comments to determine the positive percentage (in agreement), negative (in disagreement) or neutral.

The potential applications go deeper as well. If, in another example, one user always comments on restaurant reviews, but only for fancy establishments and never for a snack shack or greasy-spoon diner, the AI tool would learn that this user is a foodie. This individual is likely to respond positively to beautiful images of food and elevated restaurant experiences. Beyond just learning some information about the user, this data also presents an opportunity to tailor more content …read more

Source:: Digiday

Author: Aaron