[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [aspell-devel] suggestion algorithm
From: |
Karl Chen |
Subject: |
Re: [aspell-devel] suggestion algorithm |
Date: |
Wed, 26 Oct 2005 22:50:15 -0700 |
User-agent: |
Gnus/5.110004 (No Gnus v0.4) Emacs/22.0.50 (gnu/linux) |
decipehr soundslike TSPR
decipher soundslike TSF
corretcing soundslike KRTSN
correcting soundslike KRKTN
Summary of my understanding:
- Metaphone is a transformation from text to a "sounds-like"
representation, very similar to Soundex.
- Ispell suggests words with small edit-distance in the text plane
- This is good at finding typos
- Aspell suggests words with small edit-distance in the Metaphone
plane
- This is good at finding the correct spelling for a word a
user doesn't know how to the spelling
- This is poor at finding typos, a single transposition or
substitution in the text plane can yield very different
Metaphone strings
- E.g. transposition in text plane can lead to 2-character
change in the metaphone plane
I suggest:
- Populate the suggestions list with the regular edit-distance
(ispell) algorithm. The scoring algorithm needn't change.
What do you think?
--
=========== begin suggest decipehr ===========
TRYING SCAN 1
SCORING LIST
Threshold is: 295
try_for: 250
Size of scored: 11
Size of ! scored: 82
Size of scored: 49
Size of ! scored: 42
Try Harder: 0
decipehr TSPR
Desiree 235 TSR
desire 242 TSR
Cipro 252 SPR
despair 255 TSP
disciple 260 TSPL
drapery 260 TRPR
despise 262 TSPS
despite 262 TSPT
Desiri 265 TSR
dispel 267 TSPL
Desirae 270 TSR
sprier 277 SPR
despot 280 TSPT
tempera 280 TMPR
despoiler 287 TSPL
dapperer 290 TPR
despairs 292 TSPS
Diaspora 295 TSPR
diaspora 295 TSPR
disappear 302 TSP
^^^^^^^^^^^ end suggest decipehr ^^^^^^^^^^^
& decipehr 20 0: Desiree, desire, Cipro, despair, disciple, drapery, despise,
despite, Desiri, dispel, Desirae, sprier, despot, tempera, despoiler, dapperer,
despairs, Diaspora, diaspora, disappear
deciphre
=========== begin suggest deciphre ===========
TRYING SCAN 1
SCORING LIST
Threshold is: 192
try_for: 150
Size of scored: 4
Size of ! scored: 62
Size of scored: 5
Size of ! scored: 59
Try Harder: 0
deciphre TSFR
decipher 77 TSF
deciphers 115 TSFS
deciphered 117 TSFRT
desire 187 TSR
^^^^^^^^^^^ end suggest deciphre ^^^^^^^^^^^
& deciphre 4 0: decipher, deciphers, deciphered, desire
--
Karl 2005-10-26 22:29