@GREETBOT is a bot that scans SteemIt for content written by newbies, that deserves attention.
The bot has two phases of filtering (to ensure that no crappy content gets rewarded)
- Phase 1 - filter by statistics from SteemSQL
- Phase 2 - apply linguistic filtering to remove badly written and non-english posts
- Reward the found posts.
The first phase is easy. I scan the posts older than a day, with some minimum length and some basic filtering by tags and keywords. (For example, I try to ignore all #collorchalange posts). This is also where I filter the authors, by their reputation (to ensure that only newbies get rewarded) and by other factors.
The second phase is a lot more complex. My bot uses AI to scan the structure of an English written content and assigns each post a score, depending on how well they used the English language. (Technically, the bot has no way of knowing if the content written in good English is actually saying interesting things... it just assumes that post written in better English are better posts.)
After gathering enough information, I the creator of @greetbot had to find an actual number (min score) below which all posts should be ignored. I went through a lot of posts by hand and figured out that the minimum score should be around 30 and 40 points. for now I'm using 30, in order to include more posts.
Of course, these numbers mean nothing to you. That's normal. AI scores are usually a bit meaningless as numbers anyway. That is because of their relative nature. Do not worry if you don't get it. Just remember that if you want your post to be rewarded, you need to write long posts in good English.
The reward is simple - I send a signal to @resteembot who resteems the discovered content to give it more exposure. If the content is really good, it might get noticed the second time.