Better keyword creation algorithm

[jannemann]

The current keyword creation algorithm creates a lot of keywords that don't fit at all because they are "everyday" verbs with no relation to the article at all. I understand that it is not easy to enhance this.
At least for languages that use capitalization there is a not-so-difficult enhancement method: extract only the words that are capitalized = substantives. There is a much higher chance that they represent the article as for the lower-case words.
Another method might be to present all the generated keywords (of all articles or on a per article basis) in a different view and have the user either pick the keywords from it or block words from that list. As many verbs will be used again and again a block list of just a few verbs would block a lot of verbs in general. To cater for verb conjugation you could remove the last two or so letters from each blocked verb and match on the remaining part of the word. There are probably other methods to enhance this algorithm.