Day 2: Spellchecker notes

Spellcheckers (using Hunspell engine)

• paa.kwesi compiling league table of spellcheckers
• Tunde looking to work on Igbo spellchecker
• Tunde suggests creating a spellchecker how-to

• outline of manual
• explanation of how Hunspell works
• San-James asking for a checklist of what to do in order to create a spellchecker?
• work with linguist or get data from someone with linguistic

information on a language

• large amount of text as corpus
  • write tools to crawl the web as a way to collect text for corpus
    • corpus catcher from ANLoc
    • corpus crawler from Kevin Scannell
• generate word list from text
• Taha mentions there are about 400 stop-words or prepositions

which you can consider to include in wordlist for Hunspell

• make conjugator as a tool to automatically generate words for verbs in order to add to word list.

• Hunspell code complicated. includes language-specific code
• Aspell more structured, but not as advanced or as widely used

as Hunspell

• Tunde mentions spelt, which is a tool by ANLoc to help automatically recognise patterns in words
• paa.kwesi points out that for spellcheckers that already exist in

Hunspell, they could be included in applications that are still being localised