The SPECIALIST Lexicon

Spelling Variants Model

Introduction

From the results of non-lead-end words filters, we observed that if a term has associated spelling variant(s) exist in the same corpus (The MEDLINE n-Gram set), it is likely a valid multiword. Thus, we developed algorithms of spelling variants pattern to identify spelling variants group from n-Grams. This can be used as matchers to retrieve valid multiwords from n-Grams.

Definiton and types

By definition, spVars are terms that have same meaning, categories (POS), pronuncitaion, syntax, and different spelling. They are the different spelling from American and British English.

• Inflectional SpVars

• Genitive SpVars
• Dash-Space SpVars
• Mixed case SpVars

SpVar Pattern Algorithm

• Spelling Variants Normalization
• Phonetic algorithm tests
  • MES (Metaphone, Edit distance, and Sorted distance)
  • ES (Edit distance and Sorted distance)

  • M2CES (Metaphone, Caverphone, Edit distance, and Sorted distance)

SpVar Pattern Test

• SpVar Test on LRSPL (Lexicon SpVar)
• SpVar Test on inflVars (Lexicon)
  • GoldStd from Lexicon

  • For Amia.2016 Initial submission
  • For Amia.2016 Final submission

  • Phonetic algorithm tests

• Remove parent terms (TBD)

Applications

• SpVar Model Applications