Bigram on personal dictionaries

[vironite-23]

Bigram (two-word sequence) are very useful for predicting the next word you'll typed in. But currently the only way to add them is by import the dictionary file, which is only limited to three bigrams per word. And sometimes it's not the word you'd used the most.

So maybe on dictionary editor, between shortcut and weight, there's a text box labeled "Bigram" which is the list of the words that'll be suggested after typing that word (separated by comma).

For example, for the word "I", the bigram list is "need,love,want,have,etc)

This is how it looked like on dictionary file before it got converted into binary (example, but not accurate), I need, I love, I want (only three words is listed).

word=I,f=254
bigram=need,f=1
bigram=love,f=2
bigram=want,f=3

[cinnabar777]

I've thought about this as well. My custom dictionary has no bigrams in it because I've found that they aren't that useful for me and often override the internally built up next word prediction.

I personally think that the internal dictionary, prediction, next word, weight, etc should all be synced to the user dictionary.

This would not be that large because most people don't use that many words for writing.

This would allow the user to edit things, adjust word frequency for words they no longer use that much or increase it. And would become a very precise dictionary for the user, and the developer could leverage this over time to decrease emphasis on the dictionary and increase this focused dictionary, essentially that's what the algorithm is supposed to be doing anyway, but in cases of new install or clearing data the user doesn't loose any ground, as there are times you cannot restore a backup without some problems.

Ultimately this would be the user's typing dictionary while the main dictionary would be spell checking.