LEX Source Model - Lexicon Records with Negations

I. Introduction

Negation plays a role to express the opposite meaning in antonyms according to the boundedness hypothesis [2006 Paradis]. Words that have negative tags in the lexicon are used as root to retrieve aPair candidates, such as “unlike”, “isn’t”, “neither” and “couldn’t” in aPairs of [like|unlike|prep], [is|isn’t|aux], [either|neither|det] and [could|couldn’t|modal], respectively.

II. Design
Lexical records with negative or broad_negative tags are used to generate antonym candidates. APair candidates are retrieved from the 7 POSs of [adv|pron|aux|modals|prep|det|conj] that have negative tags as discussed below. POSs of [noun|adj|verb|compl] do not have negative tags in Lexicon. Please see design documents for more details.

• Adverbs (adv):
  • true negative/strict negation (negative): never, no, not, nowise
  • broadly negative (broad_negative): hardly, seldom, rarely, even, either, little, scarcely, slightly, barely, seldomly.

• Pronoun (pron):
  • type=indef(neg): none, nobody, nothing, noone, neither, naught

• Auxiliary (aux) - negative:
  • variant=isn't;pres(thr_sing):negative
  • variant=aren't;pres(fst_plur,second,thr_plur):negative
  • variant=don't;pres(fst_sing,fst_plur,second,thr_plur):negative
  • variant=haven't;pres(fst_sing,fst_plur,second,thr_plur):negative
  • …

• Modal (modal) - negative:
  • variant=mayn’t;pres:negative
  • variant=mightn’t;past:negative
  • variant=mustn’t;pres:negative
  • variant=couldn’t;past:negative
  • variant=cannot;pres:negative
  • variant=can’t;pres:negative

• Preposition (prep):
  • true negative/strict negation (negative): without
  • broadly negative (broad_negative): unlikely (not used as negation cue word)

• Determiner (det):
  • true negative/strict negation (negative): no, neither, nary a, nary an

• Conjunction (conj):
  • true negative/strict negation (negative): neither, nor

III. Implementation

Java source codes are implemented in the directory of Lexicon:

• GenAntCandFromLexicon.java

Algorithm:

• Go through lexRecords and convert them into lexRecord Objects
• check if the POS has negation
• put the base form of lexRecord on the Ant-2 (because of negation)
• Use B2 or BN2 for negation field.
• put ANT_TBD, EUI_TBD, CANON_TBD, TYPE_TBD, and DOMAIN_TBD for fields of ant-1, EUI-1, canon, type, and domain.

Output: These candidates are outputted in the standard 10 field format and sent to linguists for tagging and further processing.

Notes:

• Linguists should fill all XXX_TBD fields.
• APair candidates from LEX source is rather static. It only increases when there are new lexRecord with negation tags.