The SPECIALIST Lexicon

Antonym Candidates

The general processes of generating Lexicon antonyms are shown as bellows:

  • Generate antonym candidates from source models of LEX|SD|PD|CC|SN
  • Manually tag candidates by linguists
  • Validate and fix tagged files
  • Update tagged file to annual release tagged file
  • Generate antonym and negation detection cue words from the tagged file

This section describes the design on antonym candidates. The output file of antonym candidates has 10 fields as described in Tag section. Each line represents an antonym pair (aPair) candidate and includes all associated information. The table below shows examples of antonym candidates (before being manually tagged or further processed).

Ant-1EUI-1Ant-2EUI-2POSCanonTypeNegationDomainSource
withE0065516withoutE0065526prepCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDLEX
neverE0042565alwaysE0008403advCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDLEX
availableE0011401unavailableE0062924adjCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDPD
bacterialE0011688antibacterialE0009417adjCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDPD
carefulE0015340carelessE0015344adjCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDSD
cashableE0794454cashlessE0546479adjCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDSD
asleepE0010731awakeE0011449adjCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDCC
juniorE0036311seniorE0055206nounCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDCC
heavenE0030994hellE0031052nounCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDSN
kidE0036500parentE0045702nounCANON_TBDTYPE_TBDNEG_TBDDOMAIN_TBDSN

Antonym candidate files are generated annually and sent to linguists for tagging.

Please refer to original design documents