Repository navigation
PHRASEDICT for joined text (enhancement) #426
Description
In one domain, words are not only joined, but may be misspelled or include clutter.
Could have a "reverse phrase" dictionary. Eg,
weathervconditions s=weather conditions
writingsendingreadingtext s=writing sending reading text
xingbridgeviaducttunnelwait s=crossing bridge viaduct tunnel wait
Could be used also for known idioms as below.
motorvehicle s=motor vehicle
licenseplate s=license plate
These phrases could be spliced into the parse tree or even replace the input text buffer.
Other ways:
1- A set of passes which deals with such issues generally.
2- Enhancement to NLP++ to allow rules such as below, which would add a phrase to the parse tree.
(license plate) <- licenseplate @@
Activity
Merge pull request #786 from VisualText/feature/dict-shape-433
ddehilster commentedon Oct 15, 2026
Done in #786 (4.2.30):
licenseplate split="license plate"replaces the token with the words. When the words spell the token, each gets the offsets of its own letters; otherwise (clutter or a misspelling, likeweathervconditions) each spans the whole token. Splitting runs before joins and phrases, so the words can matchlicense plate s=.... Docs: VisualText/visualtext-files#242.