About CSpell

CSpell is a generic spelling detection and correction tool developed in Java. It is distributed by NLM via an Open Source License agreement. It was originally developed for Consumer Health Question Answering project and thus the consumer health corpus are used for word frequency and context scores (word vectors). Easy configurable options are provided for customizing different data files. The correction features include:

• Non-dictionary based correction

  Type	Input Text	Output Correction
  Xml/Html handler	"germs"	"germs"
  Informal expression handler	pls	please
  Leading digit splitter	1.5years	1.5 years
  Ending digit splitter	from2007	from 2007
  Leading punctuation splitter	volunteers(healthy)	volunteers (healthy)
  Ending digit splitter	cancer?if so	cancer? if so

• Dictionary based correction
  • non-word

    Type	Input Text	Output Correction
    Spelling	dianosed	diagnosed
    split	knowabout	know about
    merge	stiff n ess	stiffness

  • real-word

    Type	Input Text	Output Correction
    Spelling	bowl movement	bowel movement
    split	for along time	for a long time
    merge	early on set	early onset

• Combination of above:
  • Example-1:

    Input Text	He was dianosed early on set deminita 3years ago.
    Output Correction	He was diagnosed early onset dementia 3 years ago.

    Input text	Output Correction
    dianosed	diagnosed	non-word, spelling
    on set	onset	real-word, merge
    deminita	dementia	non-word, spelling
    3years	3 years	non-dictionary, split

  • Example-2:

    Input Text	No bowl movement for along time.
    Output Correction	No bowel movement for a long time.

    Input text	Output Correction
    bowl	bowel	real-word, spelling
    along	a long	real-word, split