This sounds like a great idea, but I think if there was to be an actual system it should be a mix of 2 things:

• Automatic recommendation. This should be for all the common cases you described. But this will leave quite a number of false positives/false negatives. Not to mention a lot of people will probably not even bother reading the warning message.
• Flag that could be at the same level as other flags, where users would be able to tell that this looks suspiciously like a solved answer, and a mod could take action.

The reasons for the mix of these 2 is that, after browsing a bit in the search, there seems to be a great diversity and creativity in the way people mark something as solved, and an efficient system would need to find a good compromise between false negatives and false positives. If we just add a warning, I get the feeling that a lot of people will keep doing it, there needs to be a way to take action if we really want to go that way.

False negatives: There is still a certain number of posts who would be non-detected by a script but easily by humans (assuming we want to avoid too many false positives). Some examples of posts who would be hard to catch:

• SOLVED at the end, no delimiter
• solved but still puzzled
• SOLVED at the beginning, no delimiter
• Actual sentence could make sense if you don't look at the upper case
• solved myself
• commas can be used for anything
• SOVLED and not SOLVED
• What about a C++ class called Solved?

False positives:

• Typo and a separator, but actually just a class name
• solved after a comma and before a question mark
• not solved !

In summary, I think the automatic system should be very basic otherwise this will end of flagging too much stuff, you'd need to do some kind of AI to have a reasonable algorithm and I don't think it's worth the effort. People will do a much better job at flagging those. But overall that's a pretty minor issue if you look at the number of "solved" posts, and a lot of headache for not that much impact.

This sounds like a great idea, but I think if there was to be an actual system it should be a mix of 2 things:

• Automatic recommendation. This should be for all the common cases you described. But this will leave quite a number of false positives/false negatives. Not to mention a lot of people will probably not even bother reading the warning message.
• Flag that could be at the same level as other flags, where users would be able to tell that this looks suspiciously like a solved answer, and a mod could take action.

The reasons for the mix of these 2 is that, after browsing a bit in the search, there seems to be a great diversity and creativity in the way people mark something as solved, and an efficient system would need to find a good compromise between false negatives and false positives. If we just add a warning, I get the feeling that a lot of people will keep doing it, there needs to be a way to take action if we really want to go that way.

False negatives: There is still a certain number of posts who would be non-detected by a script but easily by humans (assuming we want to avoid too many false positives). Some examples of posts who would be hard to catch:

• SOLVED at the end, no delimiter
• solved but still puzzled
• SOLVED at the beginning, no delimiter
• Actual sentence could make sense if you don't look at the upper case
• solved myself
• commas can be used for anything
• SOVLED and not SOLVED
• What about a C++ class called Solved?

False positives:

• Typo and a separator, but actually just a class name
• solved after a comma and before a question mark
• not solved !

In summary, I think the automatic system should be very basic otherwise this will end of flagging too much stuff, you'd need to do some kind of AI to have a reasonable algorithm and I don't think it's worth the effort. People will do a much better job at flagging those. But overall that's a pretty minor issue if you look at the number of "solved" posts, and a lot of headache for not that much impact.