Negation on Derivation


The meaning of derivational variants usually change (related but different) from the original term. This characteristic allows wider searching coverage (better recall rate) through derivations. On the other hand, some NLP projects are sensitive on negation and most of the times derivations with opposite (reversed) meaning should be excluded for better precision. The options of negative derivations are now available in Lexical Tools. By default, the negative derivations are excluded. Two negation types are defined in Lexical Tools.

OOtherwise (not negation)
BBoth N and O.
The class-B only exist at prefix level not the prefixD level. In other words, all prefixD pairs should be either N or O.


  • Processes:
    • All prefixes are tagged with N or O with associated sense
    • Apply automatically tagging algorithm for class-N and class-O prefixes with consistent negation tag
    • Class-B prefixD (both N and O) requires manual negation tagging during derivation tagging (only for valid prefixD pair for class-B prefixes)
      • Some prefixes might have different sense with different classes and they are tagged as class-B.
      • Only N or O is allowed for the negation tag on the prefixD pairs level.
      • Tag a prefixD pair as O when it could be N and O because Lexical tools preserves the recall rate (not to increase the precision) for using derivational variants.
      • This means a prefixD pair is N has to be strictly negative.
      • Example:

        => getting rid of methylate
        => removing (atoms of) methylase, a description of what they contain more than a negation.

    • Apply negation tag from linguists (prefixD.tag.txt) on class-B prefixes (a-, an-, de-, dys-, in, under-)

  • Results:

    B (N/O)6a-, an-, de-, dys-, in-, under-
    N10anti-, contra-, counter-, dis-, il-, im-, ir-, mis-, non-, un-
    • anti- could be class O it has spVar of ante-


    • dis- could be class O it has spVar of di-

    • im- could be class O it has spVar of em-


    O129 abs-, af-, Afro-, ambi-, amphi-, ana-, Anglo-, ante-, apo-, arch-, astro-, at-, auto-, be-, bi-, bio-, bis-, circum-, cis-, col-, com-, con-, co-, cor-, cryo-, crypto-, demi-, demo-, dif-, di-, down-, du-, duo-, eco-, electro-, em-, en-, epi-, Euro-, ex-, extra-, fore-, Franco-, geo-, gyro-, haemo-, hemi-, hemo-, hetero-, homo-, hydro-, hyper-, hypo-, ideo-, idio-, Indo-, infra-, inter-, intra-, iso-, macro-, mal-, maxi-, megalo-, mega-, meta-, micro-, midi-, mid-, mini-, mono-, multi-, neo-, ob-, oc-, of-, omni-, op-, ortho-, out-, over-, paed-, paedo-, paid-, paido-, paleo-, pan-, para-, ped-, pedo-, peri-, per-, photo-, pod-, poly-, post-, pre-, preter-, pro-, pros-, proto-, pseudo-, pyr-, pyro-, quasi-, re-, retro-, self-, semi-, socio-, step-, sub-, super-, sup-, supra-, sur-, syl-, sym-, syn-, sy-, tele-, trans-, tri-, twi-, ultra-, uni-, up-, vice-, with-

  • Reference:
    • A Grammar of Contemporary English, by Quirk, Greenbaum, Leech and Svartvik (the Quirk et al. I have been referring to), has a classification of English prefixes (pp.981-990).
    • Neutral negation: a-, an-, dis-, in-, il-, im-, ir-, non-, un-
      => a-, an-, dis-, in-, un-: B (N/O)
      => il-, im-, ir-, non-: N
    • Negative attitude or position: anti-, contra-, counter-, co-, pro-
      => excludes co-, pro- and tag them as O for our reqiurements. They are part of a larger class of attitude or position, which includes both negative and non-negative prefixes.
      anti-, contra-, counter-: N
    • Reversatives (for verbs)/privatives (for nouns): de-, dis-, dys-, un-
      => group these with neutrally negative to fit our requirements for dis- and un-
      => de-, dys-: B (N/O), because it is also pejoratives
    • Pejoratives: dys-, mal-, mis-, pseudo-
      => dys-: B (N/O)
      => mis-: N
      => mal-, pseudo-: O

suffix of: -less, -aster are negative. However, -aster is not common and thus it is not included in Lexical Tools. So, the only negative SD-Rules (or facts) are:

  • $|noun|less$|adj
  • $|verb|less$|adj