Belfry.io - introducing the next generation moderation platform
User-generated online discussions can add real value to published content. However, it's not always easy to enforce a peaceful, constructive atmosphere. Imagine having a popular blog about science. You regularly update it with your thoughts and let people discuss the topic further or ask relevant questions. Some posts can even get over a thousand comments. When you cover a trending topic though, things can get messy. New people visit your blog, but a handful of them only do so to argue against your point in a very inappropriate manner. Others get in a heated argument on a completely unrelated topic, or simply ignore all basic rules of common courtesy in their comments.
There are not many of these users, but they eventually turn up everywhere. Initially these users are heavily outnumbered by the constructive users, but if their comments are not weeded out in time, they tend to get more replies than the quality ones. This results in long threads full of low-quality debates and that is the point when people not interested in this kind of discussion start to find it harder to ignore. Some of them reach their limit of tolerance and start ignoring the comment section all together. Without their contribution, the average quality of the comments worsens, making other constructive users think twice before even looking at the comments, fueling a vicious circle. We at belfry.io strongly believe that a similar mechanism is responsible for the observed phenomon of some comment sections often seeming to have a lower quality than you would expect based on the target audience of the content.
Different content providers fight different ways against this. Some simply close down their whole comment section, not risking the chance of scaring visitors or advertisers with an inappropriate comment feed. Others who do not want to get completely get rid of the whole comment section apply a system of pre-moderation. However this solution has the drawback of eliminating fast replies and long, meaningful discussions, which often makes the comment section much less vivid. And if there is still large traffic despite this, then the pre-moderation can take a huge amount of time. Post moderation is equally time consuming if you want to be safe and make sure that all posted comments are checked by you or a moderator before they are visible for too long. With post moderation, you can skip a fraction of the comments, reducing the workload, but the less comments checked, the more chance of unwanted comments dominating the feed. In some strong and dedicated communities, owners can mitigate these issues by involving the users in the process of moderation by implementing a flagging system or appointing moderators from the community's most trusted users.
We at belfry.io believe that comment feeds are an important thing. We want to make them better, and more constructive. We are working hard on a system that uses machine learning to identify inappropriate comments and help make the work of moderators more efficient. This will make moderation cheaper, faster, and more accurate, so more site owners can allow people to freely contribute their ideas and opinion without the risk of damaging their brand. If you want to hear more, you can sign up for our mailing list, so you can stay up to date on news about the product.
Machine learning is a powerful method for solving various natural language processing tasks, but collecting the training corpus can be a difficult job.