I think the real problem behind the auto-correction issue isn't the dictionary with its supposedly useless words. The real issue is how the correction engine decides which word it suggests based on the currently typed letters.

You can see this very good on arkona's sample, where he typed "hete" but meant "here". His real mistake was that he missed the R letter and instead hit T because it's the next letter to R. If Nokia had put real thinking behind the auto-correction feature the engine would know the real mistake (because for example there'd be a dictionary with common mistyped words). Instead the auto-correction is really dumb. It seems Nokia devs thought the only usage scenario for the auto-correction would be that the user wants to save some keystrokes so they display the a word that begins exactly with the currently typed letters but is longer (the longest from all suggestions?) than the currently typed word.

In fact, this could be an usage scenario. But often (or mostly) you just mistyped a word and want it to be corrected.

Well, after taking a closer look on all suggested words when you type "hete" (in the English keyboard), you get (PR1.3): 1. "hetereosexuality", "here", "bete", "hate". So, actually, the auto-correction engine know about the typing error the problem is it doesn't prioritize this suggestion.

Another drawback of the N9 keyboard is that while typing you only get one word suggestion (you have to tap on the typed word to get all suggestions). As far as I know on Windows Phone and Android you get severeal suggestions above the keyboard and obviously this would be a much better approach for auto-correction since a software cannot always know the appropiate word suggestion.

So, there is much room for improvements and I think editing the dictionary would only treat the symptoms, not the cause.

But I don't have much hope this gets ever fixed.