linguistics 101
TRANSCRIPT
Linguistics 101: The Conceptual Base of Natural Language Processing (NLP)
Zina Saadi
Computational Linguist
Middle Eastern Languages Specialist
Linguistics is the scientific study of human languages.
The goal of linguists is to explore the complexity of languages by creating language models that describe their minimal units and their connections.
Linguists often search for universal rules of a language.
What is Linguistics?
2
Overview of Linguistics:
Phonology
Morphology
Semantics
Syntax
Summary: A taste of Basis Technology Linguistic Solutions
[mostly Modern Standard Arabic (MSA),
the linguistic analysis at the word level]
Presentation Outline
3
Articulation
Phones
Phonemes
Prosody
Linguistic Signs
Morphemes
Affixes
Stems
Words Meaning
Roots
Lemmas
Homonyms
Words Order
Part of Speech
Linguistics: The Big Picture
4
Morphology SyntaxPhonology
Arbitrary Structure Lexical Structure Grammatical Structure
Semantics
Why is Phonology Important?
5
For a better communicationEach human language is composed of a series of sounds with functional or
distinctive characteristics that can sometimes affect the meaning of the
sentence.
اب قلب محمدتThe heart of
Mohamed repented
اب قلب محمدطThe heart of
Mohamed became kind
ال الماءسThe water leaked
ال الماءزThe water
disappeared
Emphatic
/t/ sound
vs. /t/
Voiced
sound
/s/ vs. /z/
What is Phonology?The study of the sound system of a given language in addition to the analysis and
classification of its phones and phonemes.
6
• What is a phone? sound that can be individually produced and recognized by speakers of a language.
Phonetics and Phonology
7
fb
b
bilabial
voiced
stop
oral f
labiodental
voiceless
fricative
oral
Phones: What about Arabic?
Phonetics and Phonology
8
b
bilabial
Voiced + Voiceless
stop
oralf
labiodental
Voiceless +voiced
fricative
oral
• What is a phoneme? may correspond to specific phones, or a single phoneme may correspond to any one of a set of allophones
Example: Emphatic letters distinction
NLP Uses: Text-To-Speech (TTS) applications
Phonetics and Phonology
9
ب تا taabarepented
ار س saara moved
ل ذ dhallahumiliated
Group 1
اب ط Taaba was good
ار ص Saarabecame
ل ظ Zallaremained
Group 2
What are allophones? sounds variations that do not alter the meaning of words
Phonetics and Phonology
10
[dj] as in jet
• Modern Standard Arabic
• Most Arabic countries
[g] as in game
• Egypt
• Some regions in Morocco
Prosody
Duration of sound
كتب (kataba) => to write
كاتب (kaataba) => to correspond with
Stress (emphasis of a syllable)
Present => gift or opposite of absent
Present => to offer
NLP Uses: to correct pronunciation in TTS
Phonetics and Phonology
11
Pronunciation Rules
Phonetics and Phonology
12
S (Finnish)
Lisa
[li:za]
Lisa
[li:sa]
S (French)
souvent
/s/
maison
/z/
/s/ → [voiced] / V_V
/s/ and /z/ are different phonemes/s/ → [voiced] unpredictably
/s/ and /z/ are in free variation
What is Morphology?Morphology is the systematic change that
occur at the word level to create a relationship between a given word and
other words.
13
عصفور
bird
oiseau
Vogel
The Linguistic Sign: Form and Meaning
14
• Invariant form (pronunciation)
• Invariant meaning
Why is the Linguistic Sign Arbitrary?
15
The linguistic sign is arbitrary as there is no direct connection between the
invariant form and meaning.
there are no transformational rules that connects phonological level of
words to its morphological level for a specific meaning, while there are
syntactic rules that bind words in sentences.
ل مM L
لم
lamdidn’t
لم limawhy
مل mallabored
لم lammacollect
لن
lanWon’t
Morphology: Linguistic Signs
16
Example: He writes various books
6 linguistic signs [he, write, -s, various, book, -s]
The first –s signifies the present tense
The second –s signifies plurality
The linguistic sign is the minimal meaningful unit that contributes to the meaning of a word
A Word can have:
One morpheme: ( فوق, above)
Many linear morphemes ( انكتاب ال , the two books)
Many non-linear morphemes: ( ريوط , birds), ( تبامك , desks)
Are morphemes always attached to words?
Morphemes and Words
17
Morphemes are not always linear:
MSA: تخرج ال
Algerian Dialect: شتخرج ما
French: Ne sors pas
English: do not leave
Discontinued morpheme
Morphology: Detecting Morphemes
18
Allomorph example:
Morphology: Various Morpheme Forms
19
مدد madada (extended)
ت مددأنا (I)
تما مددأنتما (you masc. dual)
تممددأنتم (you masc. plural)
مد madda (extended)
مد هو (he masc.)
ا مد هما (they masc. dual)
وامد هم (they masc. plural)
Analytic Relationship (each word is a morpheme)
Example: Chinese & Japanese
Agglutinative Relationship (words with linear morphemes)
Example: Turkish and Korean
Morphological Relationship (words with non-linear morphemes)
Example: Arabic and Hebrew
Morphology and Languages
20
Agglutinative vs. Morphological languages
Morphology and Languages
21
French
maison
ma maison
maisons
mes maisons
English
house
my house
houses
my houses
Turkish
ev
evi
evier
evieri
Arabic
منزل
يمنزل
زلامن
يزلامن
د ر س
درسlesson
ةدرس م school
رساد student
سم درteacher
سيدر ت teaching
Why Arabic is Morphological?
22
Arabic words are derived from a root adhering to a list of patterns
Arabic words with the same pattern and different root share something in meanings (Semantics => word classes)
Why Arabic is Morphological?
23
مفعلةmifaalah
مطرقة mitraqahhammer
مكنسةmiknasah
broom
مسطرةmistarah
ruler
ملعقة milaaqah
spoon
مسبحةmisbahah rosary
All these
words refer to
instruments
(إسم آلة)
خرج • to exit
• to leave
ج خر • to remove
أخرج
• to take out
• to emit
• to extricate
إستخرج
• to extract
• to derive
• to conclude
ج تخر • to graduate
Lexical Stucture in Arabic
24
Root
Lemmas as verbs
خرجم (an exit)رجاخ م (exits)
ج تخر (graduation)
Lemmas as nouns
يج خر (graduate)يجــون (graduates masc. pl)خريجـات خر (graduates fem. pl)
Stems
The study of meaning
Word-to-words relationship
What is Semantics?
25
Word Relationships
Pronunciation Spelling Meaning Examples NLP Uses
Homonyms same same different bow /boʊ/ -> a musical instrument of the violinbow /boʊ/ -> a tied ribbon
Translation
Heteronyms different same different bow -> /baʊ/ to bendbow -> /boʊ/ a tied ribbon
TTS
Homophones same different different (lessen, lesson), (maid, made)
SpellCorrection
Synonyms different different same (to begin, to start) InformationRetrieval
The study of the principles and rules for constructing sentences in natural languages.
Word order
English: SVO (subject-verb-object)
Persian: SOV (subject-object-verb)
Arabic (most often): VSO (verb-subject-object)
NLP Uses: Spell Checking
What is Syntax?
26
SummaryA Taste of Basis Technology Linguistic
Solutions
27
Articulation
Rosette Name
Translation (RNT)
Rosette Name
Indexer (RNI)
Linguistic Signs
Rosette Base
Linguistics (RBL)
Rosette Entity
Extractor (REX)
Words Meaning
Rosette Base
Linguistics (RBL)
Rosette Entity
Extractor (REX)
Words Order
Rosette Base
Linguistics (RBL)
Rosette Entity
Extractor (REX)
Linguistics: The Big Picture
28
Morphology SyntaxPhonology
Arbitrary Structure Lexical Structure Grammatical Structure
Semantics
AyadAllawi
إياد عالوي
EyadAllawi
اياد عالوى
Defines arbitrariness
Basis Technology
29
Enriches lexicology
Basis Technology
30
.الحفالتتخرج جامعة هارفرد في العام الحالي من أهم حفلةكانت
يجـونحيث كان عدد يجاتو الخر آالف يتجاوز ال الخريج .خر
The graduation ceremony at Harvard University this year was among the most important
ceremonies, where the number of female and male graduates exceeded 10 thousands graduate.
Affixes
(Morphology)
ةحفل
تحفالال
يجـال ونخر
يجوال اتخر
Pronunciation & Prosody (Phonology)
The final letter in ةحفل is silent (instead of /t/), medial alef in يجاو ت الخر has longer
duration than the initial alef in the same word
Triage grammatical complexity
Basis Technology
31
طه حسين قصص مؤثرة طبعت على هيئة كتب.كتب
Taha Hussein wroteimpressive stories published
in the form of books.
Homonyms (Semantic):
same spelling and
pronunciation but different
meaning
Word Order
(Syntax)
Arabic VSO
Part-of-Speech
(Syntax)
Verb <- كتب
كتب -> Noun
Manner of articulation (Phonology)
kataba كتب is different from قطب qaTaba, even though many English speakers
can't differentiate between them
Thank [email protected]