aspell-user
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Aspell-user] Can I get "edit distance" for the suggestions?


From: Kevin Atkinson
Subject: Re: [Aspell-user] Can I get "edit distance" for the suggestions?
Date: Tue, 18 Apr 2006 20:34:41 -0600 (MDT)



On Tue, 18 Apr 2006, Ethan Bradford wrote:

Here's a proposal for the API:
 We add the type AspellIntEnumeration (or maybe a better name -- lower-case
L next to cap I is ugly, but it would be nice if it was like
AspellStringEnumeration).
 We add a call to extract that from an AspellWordList:
    AspellIntEnumeration * scores = aspell_word_list_scores(suggestions);

I was more thinking of using a struct something like
  struct Suggestion {
     const char * word;
     int word_len;
     int score;
  }
  and new methods to go along with it....


Behind the scenes,  we add to AspellWordList a list of scores.  The list of
scores is set in Working::transfer when the list of words is created.

On 4/2/06, Kevin Atkinson <address@hidden> wrote:

On Fri, 31 Mar 2006, Ethan Bradford wrote:

I'm mystified as to how the definition of *AspellStringEnumeration* is
hidden from us, so I can't peek inside of that.  It seems obvious from
the
outside that edit distance (i.e. the # of changes necessary to get from
the
checked word to a proposed suggestion) is used to sort the results
presented
to the user.  I'd like to use outside information to sort the results
with
equal edit distance w.r.t. eachother.  So how do I get the edit
distance?

The "edit-distance" is not stored in AspellStringEnumeration.  It is only
stored internally inside data structures in suggest.cpp.  Also its not a
simple edit-distance function that is used to rank the suggestions, but a
combination of several factors, most of them being closely related to an
edit-distance.  In the future other factors, such as frequency, may be
used to rank the suggestions.  The more correct term to use is "score".

As far as making the score available outside of suggest.cpp, I am not
against it.  I just haven't gotten around to it.  I will gladly accept a
patch to this effect, however I will be fairly picky about how it is done.
For example I won't accept something that hacks AspellStringEnumeration to
include the score.  A separate data structure should be used.  Also,
backwards compatibility will need to be maintained.






reply via email to

[Prev in Thread] Current Thread [Next in Thread]