Words

Introduction to Linguistics for Computational Linguists

Word Structure

• Morphology - Study of Word Structure
  • Morpheme - minimal unit of meaning
  • Free vs. Bound Morphemes
  • Stem vs. Affix
    • Allomorphy
  • Inflection vs. Derivation
  • Formal Operations, Structure
  • Compounding
  • Lexicon, Size

Complex Words

• Wiedervereingungen
  • Wiedervereingung - en
    • Frau-en, Kirche-en, Schule-en, Muskel-en
  • wiedervereinig - ung
    • Regier-ung, Sitz-ung, Vorles-ung, Reif-ung, Üb-ung
  • wieder - vereinig-
    • wieder-sehen, wieder-holen, wieder-schreiben, wieder-taufen
  • ver - einig-
    • ver-dunkel(n), ver-deutlich(en), ver-grösser(n), ver-allgemeiner(n)
  • ein -ig
    • witz-ig, körn-ig, salz-ig, farb-ig, wicht-ig,
  • ein

Morphemes

• Morpheme - minimal meaningful unit
  • Wiedervereingungen
    • wieder- ver- ein -ig -ung -en
    • some meanings” very abstract

• wie - der ?
  • Both meaningful, but not parts of wieder-
  • We look for a regular process of combination, other examples of putative process

Free vs. Bound

• Free morphemes can stand alone, bound morphemes appear only as parts of words

• Test: somewhere in sentence, not necessarily as one-word answer to question
  • Morphemes in Wiedervereingungen
    • ver-, -ig, -ung, & -en are bound
    • ein is free
    • wieder- thought of as bound, not the same as independent use in ‘Morgen wieder?’

Bound Morphemes

• May be “basic”
  • -kunft in Auskunft, Herkunft, Zukunft, Einkünfte
  • No longer used independently (freely)

• Sometimes have little meaning (but always meaningful, like all morphemes)
  • “Fügungs” /-s/
    • Regierungspartei, Prüfungsangst
      • no word *Regierungs or *Prüfungs

Design Problem

• 104 -105 words in open-ended vocabulary

• Phonological systems creates forms for symbols (morphemes) to which meaning is attached.

• Symbols manipulated in syntax
  • explosion in expressiveness (later lectures)

• But why is the intermediate level (word) needed?

Duality of Patterning

• Hockett noted that all languages are “dual” in how they’re structured

• Sentences, phrases are combinations of words

• Words are combinations of sounds (phonemes)

• But there’s also a third level, morphemes, intermediate between words and sounds.

Phrases (& Words), Morphemes & Phonemes

some phonemes omitted (to simplify)

Stem vs. Affix

• Affixes are added to stems (aka roots or bases)
  • Wiedervereingungen
    • affixes: wieder- ver- -ig -ung -en; stem: ein

• Affixes always bound, stems normally free

• Affixes before stem -- prefix, after -- suffix, around -- circumfix
  • prefixes: wieder- ver-
  • suffixes: -ig -ung -en
  • circumfix: ge- plan- -t

Morpheme Variants

• German present, 3rd-person, singular <-t >

• Two allomorphs /-t/ and /-?t/
  • latter after stems ending in /d/ or /t/

Inflections

• Inflectional Morphology varies the form of a word
  • no new word, just a variant of an existing one
    • les + ɚnd-pl -t> = lest ‘[You] read’

• Derivational Morphology creates new words from old
  • more than just a variant
    • les + <-bar> = lesbar ‘readable’

“Rich” Inflectional System

• German is inflectionally rich compared to English, Dutch, Frisian, Scandinavian
  • Nouns distinguish:
    • two numbers (singular and plural)
    • (in traces) four cases (nominative, genitive, dative, accusative)
  • Adjectives, determiners have all noun distinctions plus:
    • three genders (masculine, feminine & neuter)
    • two declensions (strong and weak)
      • ein grünes Haus vs. das grüne Haus

German Inflection, cont.

• Verbs distinguish two tenses, two moods, three persons, and two numbers
  • plus three nonfinite forms (two participles, one infinitive)

• Lots of overlap in form, however
  • laufen can be 1st/3rd person plural or infinitive
  • Hand can be any singular case

Example Paradigm

Paradigm - collection of inflectionally related forms

Richer Systems

• Finnish has over twenty nominal cases

• Latin has four tenses plus inflectionally marked passive

• Possibly thousands of inflected forms per word

Inflection vs. Derivation

• Derivation creates new words, often of different syntactic category
  • A/V les - + <bar>
  • V/V <wieder> + vereinig-
  • N/V regier- + <ung>
  • A/N Mann + <lich> (männlich)
  • A/N Witz + <ig>
  • V/A <ver> + deutlich-

Derivational Results

--Natural hierarchical structure

Alternative Groupings

‘can be unlocked’

‘can’t be locked’

--notice dual role of <un->

--German parallel?

Inflection vs. Derivation

• New variants of existing words

• Required by syntax, e.g., 3rd-sg subject

• No change in part of speech

• Regular in meaning

• Results in new words

• No interaction with syntax

• May change part of speech, e.g., A/N

• Possibly irregular

New Variants

• Inflectional variants “count as the same” in grammatical rules
  • Ich schreibe an Gabi, und Du schreibst an Karin
  • Ich schreibe an Gabi, und Du ? an Karin

• Derivational results are always new words
  • Es ist nicht deutlich, aber man kann ’s verdeutlichen
  • *Es ist nicht deutlich, aber man kann ’s (ver- ? )

Productivity

• Some inflectional variants are missing, but few in comparison to derivation
  • No perfect participle for scheinen used as auxiliary
    • Sie schien zu laufen
    • Sie hat zu laufen geschienen/scheinen

• Lots of “gaps” in derivation
  • <ver> + Adj (verdeutlichen, etc.) does not apply to most adjectives (hübsch, viereckig, krank, mutig, …)

Regularity

• Inflection can change meaning, but regularly
  • past tense, plural change meanings

• Derivational meanings often “shift”
  • <wieder-> + V ‘do V again’, but wiederholen?
  • V + <-ung> ‘event, result of V-ing’ but Sitzung?

• Inflection/Derivation still controversial

Formal Operations

• Not limited to attachment of affixes
  • Umlaut in plural, derivation
    • Umlaut changes [back]? [front], retaining rounding
      • u/ü [u/y], o/ö [o/?] a/ä [a/e]
      • [back,closed,round] (u) ? [front, closed,round] (y)
      • Buch/Bücher; Loch/Löcher; Angst/ängstlich
      • Historically vowel harmony with following [i]
    • Ablaut vowel alternation
      • laufen, liefen; helfen, half, geholfen; singen, sangen, gesungen

Other Formal Means

• ktb ‘write’ (Arabic)

• katab -perf. active

• kutib -perf. passive

• aktub -imperf. active

• uktab -imperf. pass.’

• kitaab ‘book’

• kutayyib ‘booklet’

• kaatib ‘scribe’

• kuttab ‘Koran school’

• katiib-a ‘document’

• ma-ktub ‘office’

• isti-ktaab ‘dictation’

--Semitic consonantal roots are filled variously in morphology

Borderline Cases

• Inflected Prepositions
  • German zum/zur/beim/ins/fürs/ans
  • French du/des/au/aux -- de le, de les, a le, a les

• If analyzed as one word, they would be in an otherwise non-existing category
  • (zu (der Schule)) (bei (dem Feld))

• Aka “portmanteau words” -- two-in-one

Separable Prefixes

• Genuine prefix or two words?
  • Mir kam einer entgegen / ist entgegengekommen

• Occasionally irregular semantics, missing productivity
  • vor-schreiben, nach-tragen, wieder-holen, vor-lesen
  • mitkriegen ? mitbekommen, kriegen = bekommen

Clitics

• Affixes or words?
  • Affixes are parts of words, cannot stand alone (bound morphemes), words are free

• Pronouns are words:
  • Ich kann ihm und will ihm helfen
  • Ich kann und will ihm helfen

• -st is an affix
  • Du kannst und willst Tom helfen
  • *Du kann- und willst Tom helfen

Clitics

• English auxiliaries simple clitics
  • Mary and I’ll leave soon
  • Tom and his wife’ve left already

• French clitics affix-like
  • Marie l’ai lu et l’ai compris

Clitics

• Affixes or words?
  • Ich kann’s und will’s nicht begreifen
  • Ich kann und will’s nicht begreifen

• Infinitival zu
  • Er versprach, zu helfen und zu arbeiten
  • *Er versprach, zu helfen und arbeiten

• Proclitic ’s
  • ’s Geschäft

English Possessive

• No normal affix
  • The king of England’s hat
  • Seems to apply to phrases, not words

• “Special” forms (e.g., ?) after plural, etc.
  • The boys’ hats [??.b?Iz.h?ts] *[??.b?Iz.?z.h?ts]
  • His kid’s toy is here, hers’ is over there.
    • [h?Rz.Iz] *[h?Rz.?z.Iz]

Compounding

• Combining two free words to make a third
  • Donaudampfschifffahrtsgesellschaft

• Increases number of words enormously

• Only unpredictable compounds in dictionaries

• Introduces more question of associativity:

• Softwarevirusprüfung Softwarevirusprüfung

How Many Words are there?

• Size of dictionaries (Miller, p.135)

Computational Morphology

• Lexicography (collecting words) -- entirely computational

• Models of affixation as finite automata

• Lexical structure as inheritance networks
  • lexical properties inherited (as in O-O design)
    • morphological, syntactic properties
    • semantic (hyponym, synonym): WordNet

Word Structure

• Morphology - Study of Word Structure
  • Morpheme - minimal unit of meaning
  • Free vs. Bound Morphemes
  • Stem vs. Affix
    • Prefix/Suffix/Circumfix; Allomorphy
  • Inflection vs. Derivation: word forms vs. novelty
  • Formal operations varied
  • Compounding creates words
  • Lexicons