User:ClueBot NG/FalsePositives: Difference between revisions
Third parties could also notice false positives. |
Reported "false negatives" -- edits that should have been reverted by the bot but weren't |
||
| Line 25: | Line 25: | ||
Once the entry is reported it will be reviewed by an admin, then added to the [http://tools.wmflabs.org/cluebot/?page=Main review interface] if valid and used to train the bot. Status of the report and comments can be seen from within the report panel. |
Once the entry is reported it will be reviewed by an admin, then added to the [http://tools.wmflabs.org/cluebot/?page=Main review interface] if valid and used to train the bot. Status of the report and comments can be seen from within the report panel. |
||
==False negative== |
|||
First, a thank you to everyone involved in maintaining the bot and the dataset. Today, I'm reporting a false negative -- that is, edits that were not reverted but should have been. The edits go back several months and were completed by two IP users (possibly the same user) who added false edit summaries to their removal of content. The user(s) add edit summaries like "removed bias" and "removed anti-muslim content" when the actual changes are the removal of sourced content from various articles. I'm hoping that the bot can learn from these changes so as to reduce this type of vandalism in the future. One of the IPs is [[Special:Contributions/173.174.203.95|173.174.203.95]] (22 edits since June 2016) and the other is the now-blocked [[Special:Contributions/69.112.166.30|69.112.166.30]] (33 bad edits, 1 good edit, since June 2016). The IPs geolocate to different US states, but either or both may be proxies (or just wrong), and other IPs may have been used that I haven't noticed yet. The point is not the specific editor but the style of the edit which has a definite signature that (I'm hoping) the bot can learn to recognize. Thank you for your attention to this. <span style="text-shadow: 4px 4px 12px #ceff00, -4px -4px 12px #ceff00;">[[User:Etamni|Etamni]] | [[User_talk:Etamni|✉ ]]| [[Special:Contributions/Etamni|✓]] </span> 16:27, 15 August 2016 (UTC) |
|||
Revision as of 16:27, 15 August 2016
 | Due to bot spam on the report interface, reporters without accounts (or those who do not log in) will be required to enter a CAPTCHA. We apologize for this; however, it was the only way to combat the spam until Cobi re-appears. |
| To report a false positive: | ||||
You need to get ClueBot NG's revert ID. There are two ways to do this:
| ||||
| Now, click here and put the revert ID in the box when prompted. |
Once the entry is reported it will be reviewed by an admin, then added to the review interface if valid and used to train the bot. Status of the report and comments can be seen from within the report panel.
False negative
First, a thank you to everyone involved in maintaining the bot and the dataset. Today, I'm reporting a false negative -- that is, edits that were not reverted but should have been. The edits go back several months and were completed by two IP users (possibly the same user) who added false edit summaries to their removal of content. The user(s) add edit summaries like "removed bias" and "removed anti-muslim content" when the actual changes are the removal of sourced content from various articles. I'm hoping that the bot can learn from these changes so as to reduce this type of vandalism in the future. One of the IPs is 173.174.203.95 (22 edits since June 2016) and the other is the now-blocked 69.112.166.30 (33 bad edits, 1 good edit, since June 2016). The IPs geolocate to different US states, but either or both may be proxies (or just wrong), and other IPs may have been used that I haven't noticed yet. The point is not the specific editor but the style of the edit which has a definite signature that (I'm hoping) the bot can learn to recognize. Thank you for your attention to this. Etamni | ✉ | ✓ 16:27, 15 August 2016 (UTC)