TermsAndConventions

+ Table of Contents

Source:

• WordNetDocumentation
• ParserEvaluation
• Communications Forum: dictionary of speech and language terms
• The Hyperlearning Center, i.e Statistics Refresher

Words

Distinguish:

• word form: the lexical appearance of a word word meaning: the semantic concept expressable by a set of word forms (see synset).

Word meanings might be definable constructive or differential:

• constructive word meaning definition: The representation should contain sufficient information to support an accurate construction of the concept.
• differential word meaning definition: Meanings can be represented by any distinguishable symbol marking its existance.

Synonymy

• = Gleichbedeutung
• Leibniz: Two expressions are synonymous if the substitution of one for the other never changes the truth value of a sentence in which the substitution is made. Problem: Are there any synonyms at all according to this definition.
• Weak synonymy: Two expressions are synonymous in a context if the substitution of one for the other never changes the truth value of a sentence in which the substitution is made.
• Consequence: only words of the same class (noun, verb, adjective, adverb) are synonymous.
• Synsets: set of words which are synonyms of each other
• properties: transitive, symetric

Antonymy

• = Gegensatz
• e.g. [poor,rhich], [rise,fall], [ascend,descend]
• semantic relation between word forms, not between word meanings: [rise,descend] , [{rise,ascend},{fall,descend}]
• properties: transitive, symetric

Hyponomy / Hypernomy

• A concept represented by the synset {x1, x2,...} is said to be y hyponym of the concept represented by the synset {y1, y2, ...} if a native speaker of english accepts the sentence "An x is a (kind of) y ".
• = ISA relation
• = Abstraktion / Spezialisierung
• = subordination / superordination
• {maple} is a hyponym of {tree} , {tree} is a hyperonym of {maple}
• hyponym = subordinate, hyperonym = superordinate
• main organization principle for nouns (in WordNet)
• semantic relation between word meanings
• properties: transitive, asymetric
• inheritance system: a hyponym inherits all features of the superordinate and adds at least one distinguishing feature

Meronymy / Holonymy

• part-whole-relation
• = HASA relation
• A concept represented by the synset {x1, x2,...} is a meronym of the concept represented by the synset {y1, y2, ...} if a native speaker of english accepts the sentence "An x is a part of =y=".
• A concept represented by the synset {y1, y2,...} is a holonym of the concept represented by the synset {x1, x2, ...} if a native speaker of english accepts the sentence "A y has an x as its part".
• properties: transitive, asymetric

Pertonymy

• lexical relation btw. word form and stemma

Evaluation Measures

• phrase based a sentence is correctly parsed if precision = recall = 100%
  precision = |A - K| / A : percentage of answers found in the key
  recall = |A - K| / K : percentage of keys found in the answer
  where
  
  • A : set of answers
  • K : set of keys
  • A - K : set of correct answers
• sentence based a sentence is correctly parsed if the number of crossing-brackets is zero
  crossing-brackets : two pharses [i,j] and [i',j'] are a pair of crossing brackets if i < i' <= j < j'
• dependency based a sentence is correctly parsed if all dependency relations are correct. a dependency relation is
  structural correct if the modifier is attached to the correct modifiee
  labelled correct if it is structural correct and the dependency relation is labeled correct
  lexical correct if it is labelled correct and the modifier word form is using the correct lexical reading
  strict correct if it is lexical correct and the modifiee is using the correct lexical reading (caution: wrong lexical readings might count multiply - as a modifier and as modifiees in other dependency relations)
  dependency relations might be restricted only to those describing main levels, e.g. SYN, SEM, not OBL1,OBL2 which are auxilliary levels and motivated out of technical reasons.

Example

1. [They [[came] yesterday]] (key) phrase boundaries: [0,2],[1,1],[1,2]
2. [[They [came]] [yesterday]] (answer) pharse boundaries: [0,2],[1,1],[0,1],[2,2]
   precision = 2 / 4 = 50%, recall = 2 / 3 = 66.7%, crossing-brackets = (your exercise )
3. [They came yesterday] (answer) phrase boundaries: [0,2]
   precision = 100%, recall = 1 / 3 = 33.3%

Dysfluency

A disorder which affects the fluency of speech production. Problems with fluency may include disruptions in the flow of speech, e.g. hesitations, blocking or repetitions of speech. These may be accompanied by associated non-verbal mannerisms and / or psychological and emotional difficulties.

Linear Regression

Fitting two dependent variables by a line

see also:

• The Statistics Refresher
• http://www.okstate.edu/artsci/botany/ordinate/STATS.htm

Multiple Regression

testing whether a dependent variable (y) is related to more than one independent variable (e.g. x1, x2, x3).

see also:

• The Statistics Refresher
• http://www.okstate.edu/artsci/botany/ordinate/MULTIPLE.htm