User generated content is a powerful way to engage users and improve journalism. The Guardian supports sharing images and other experience reports via GuardianWitness and more specific callouts. Comments remain, however, the most common way that readers share their opinions with us and one another.
The structure of comments is relatively unchanged since they were first introduced on the site;free-form text up to 5000 characters, with limited markup and no embedding of richer media. The format is liberating but freedom comes with a cost; the lack of structure limits reuse and does little to discourage or prevent abuse.
The comments are invaluable for the Guardian and can give very interesting and versatile perspectives. At their best, comments can be thoughtful, enlightening and funny. But comments can also be unhelpful or even abusive.
Earlier in the year, the Guardian conducted research into ‘the dark side of Guardian comments’. In response to that, the Guardian has recently introduced a machine-learning solution to help identify abusive comments more quickly. This has been effective in helping the moderation team tackle abuse on the site, and we expect to introduce more automation in the future. It is important to note, however, that the final decision on whether to remove a comment is, and always has been, made by one of our (human) moderators.
You can see a presentation of our recent work, with more detail about the specific approach, here:
You’ll need to create a free account with Skills Matter to view if you haven’t already got one.
And slides are available here:
https://docs.google.com/presentation/d/1vEdMSS3KrrhxUuzU5H0LjaMTl0ZsvIVRTSuOzeVztFU/edit?usp=sharing
We are continuously working to help improve the quality of conversations on the site. Stay tuned!
