@grammarnazi: Introduction + Recent Updates

Created at the end of March, the Hive account@grammarnazi bot has been active for more than a month now, so a proper introduction post shedding some light on its behavior seems like a good idea.

The idea:

You know that feeling when you read through one of your old posts and stumble upon an embarrassing error, but Steemit doesn't allow you to edit the post anymore, and you really wish someone had told you about that one mistake before it's stored on the blockchain forever? That's how the idea for Hive account@grammarnazi was born.

The creation:

The actual implementation of the bot was done as an exercise to get familiar with the Steem.js library.

The philosophy:

For Steemit to succeed in the long run there needs to be a healthy balance between content creators and content consumers. YouTube is an example of a platform which has achieved this, providing advertisers with a valuable group of consumers. STEEM more or less directly translates into influence on Steemit, but for this influence to be worth something, there ought to be consumers who get exposed to the content which people have inested in STEEM to promote. For Steemit to become a go-to media aggregator like Medium, the quality of the content on Steemit needs to be elevated. This is where Hive account@grammarnazi comes in. In its current implementation, the bot scans (almost) every new post for some of the most common typos, spelling mistakes and grammatical mistakes in the English language and notifies the author about the first mistake it detects. While this does not serve to make the whole text spotless, it increases the author's chances of never making that exact mistake again. In that sense, Hive account@grammarnazi improves the quality of Steemit posts one mistake at a time.

How the bot initially worked:

Every five seconds, the bot would load the most recent Steemit post, unless no new posts had been made during the last five seconds. Working its way through a list of the most common mistakes in the English language, it would check whether the text contained any of them. If the text contained an error, the bot would post a comment claiming that it had found a grammatical error and quote the sentence containing the error. The final days of March served as an initial test period and the data collected made for some significant upgrades in April.

Upgrades made in April:

• Initially, the bot did not distinguish between grammatical mistakes, misspellings and typos. It simply claimed every mistake was a grammatical mistake, which was simply wrong. Now the bot is updated to specify which kind of mistake the author has made.
• Not everyone likes to be corrected, and understandably, some authors prefer to be left alone with their mistakes. To avoid getting constantly flagged, the bot has been updated with a constantly updating list of authors who have previously flagged it, so it will never bother them again.
• Some authors were happy about the corrections pointed out by Hive account@grammarnazi, but wondered why it didn't upvote their post. To avoid unhappiness, Hive account@grammarnazi has now been programmed to always give a 50% upvote to the posts it comments on, unless its voting power is too low.
• The bot sometimes detected misspelled words in image links, which are not always visible when viewing the post. Now, the bot has been updated to ignore errors found in links.
• Sometimes, a misspelled word is simply a part of a longer word which is not misspelled. Initially, the bot did not detect whether the mistake it was correcting was a standalone word or phrase. It simply had a lot of hard-coded exceptions. To make sure the bot could run without making mistakes, the bot now makes sure that the mistakes it's correcting are not part of something bigger. In many cases, the previous implementation would correct multiple conjugations of a word. Now each conjugation has to be corrected separately, so the bot's dictionary still needs to expand to be as broad as before.
• Some of the feedback from the first few weeks of the bot's existence, claimed that it was a bit rude and negative. To compensate, it now has a happy profile picture which spreads joy and happiness and encourages correct writing.
• Sometimes, the bot would correct words which are wrong in English, but correct in another language. To avoid this, the bot now has a clever detection system which pretty accurately determines whether the text is written in English or not. If it's not written in English, the text is simply ignored.

How can you help?

• While Hive account@grammarnazi is appreciated by most authors whose posts it comments on, and receives more upvotes than downvotes overall, its reputation is still low thanks to a few heavy flaggers. If you want to help increase Hive account@grammarnazi's reputation, make sure to upvote this post and perhaps some of the bot's comments on other people's posts.
• A key factor of Hive account@grammarnazi's future popularity is its ability to upvote the posts it corrects. Currently, its Steem Power does not allow it to upvote every post it comments on, so again, upvotes, and maybe even some delegated Steem Power would be of massive help.

Let's hope this post does not contain any mistakes, because that would be embarrassing.