What is Language Is it simply communication conveying

  • Slides: 31
Download presentation
What is Language? • Is it simply communication (conveying meaning)? Animal calls have meaning

What is Language? • Is it simply communication (conveying meaning)? Animal calls have meaning Cries = predator Mating sounds = Boom chicka wow-wow Sounds often related to meaning (e. g. shrill warning sound) Human language Sound usually aren’t correlated with events We can lie and say “not” Word meanings are usually arbitrary Human language is generative (creative) We can say a virtually unlimited number of things with limited words. “My blue elephant plays the violin at midnight”

What is Language? Languages must have both Syntax (grammar) Semantics (meaning of words and

What is Language? Languages must have both Syntax (grammar) Semantics (meaning of words and interactions in language) Syntax without semantics Colorless green ideas sleep furiously. Semantics without syntax Jumped truck car over cow head the at.

Valid sentences The horse raced past the barn fell. The bird flew. The bird

Valid sentences The horse raced past the barn fell. The bird flew. The bird the cat chased flew. The bird the cat the dog attacked chased flew. The dog the dog chased the cat. Buffalo buffalo.

What is Language? Beyond syntax and semantics…. Laws of conversation (e. g. relevance): Pragmatics

What is Language? Beyond syntax and semantics…. Laws of conversation (e. g. relevance): Pragmatics Intonation and inflection Contrastive stress “government of the people, by the people, for the people” Sarcasm “yeah, right” What do words refer to? (Objects in the world, or concepts? ) All of these are extremely complex problems. Imagine trying to get a computer to understand language, considering all of these factors.

The Structure of Language Levels of analysis: sentences phrases words morphemes phonemes letters (e.

The Structure of Language Levels of analysis: sentences phrases words morphemes phonemes letters (e. g. “The linguist was a total idiot. ”) idea, event (e. g. “the linguist”, “a total idiot”)complex concept (e. g. “the”, “linguist”, “total”, “idiot”) simple concept (e. g. “the”, “lingu-”, “-ist”, “tot-”) meaning unit (e. g. /th/, /l/, /u/, /t/ ) sound unit (e. g. “t”, “h”, “e”, “l” ) written unit

Smallest unit of meaning? -ag lag, flag, sag, bag, drag -amble, ramble, scramble -ash

Smallest unit of meaning? -ag lag, flag, sag, bag, drag -amble, ramble, scramble -ash gnash, crash, trash, bash -ee jamboree, jubilee spree, whoopee, glee, free -eazy sleazy, greasy, queasy -ush lush, slush, gush, flush `droopy' and `flabby’ `locomotion' `breaking’/‘fragmants’ `absence of restraint’ / `carefree abandon' `the sleaze factor' `moist' and `oozy' crflglslvi- `bent' ‘movement' ‘light or visual phenomenon’ `horizontal movement’ crawl, cringe, creep, crumple flitter, flow, flicker, flurry glitter, glow, glare, gleam slip, slide, sled, sluice vituperative, vitriolic, vindictive, vicious `intense ill-temper'

Language Learning The computational problem: INPUT: OUTPUT: The language! • The lexicon: words and

Language Learning The computational problem: INPUT: OUTPUT: The language! • The lexicon: words and their meanings • The grammar: rules of syntax for combining words and grammatical morphemes should be able to correctly identify meanings of words should be able to correctly identify whether a sentence is grammatical or not

Language Learning The computational problem: INPUT: R o y r e a d the

Language Learning The computational problem: INPUT: R o y r e a d the w i l l

Language Learning The computational problem: INPUT: formant transitions formants R o y r e

Language Learning The computational problem: INPUT: formant transitions formants R o y r e a d the w i l l

Language Learning The computational problem: INPUT: Schematic representation of formants and transitions

Language Learning The computational problem: INPUT: Schematic representation of formants and transitions

Language Learning The computational problem: INPUT: 3000 2000 1000 EE (beet) i (bit) e

Language Learning The computational problem: INPUT: 3000 2000 1000 EE (beet) i (bit) e (bet) a (bat) o (sod) aw oo (jaw) (spoon) Vowel sounds are distinguished by relative frequencies of the first two formants

Language Learning The computational problem: INPUT: 3000 2000 ba da formant transition 1000 Consonants

Language Learning The computational problem: INPUT: 3000 2000 ba da formant transition 1000 Consonants are determined by the slopes of fast transitions (less than 100 ms). pa

Language Learning The computational problem: INPUT: 3000 2000 bet ba wet da formant transition

Language Learning The computational problem: INPUT: 3000 2000 bet ba wet da formant transition 1000 Vowel combinations are determined by slow formant transitions: dipthongs duet pa

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate relative position of two formants (vowels) • Calculate slopes of formant transitions (consonants) This allows you to learn phonemes in a language

Language Learning

Language Learning

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate relative position of two formants (vowels) • Calculate slopes of formant transitions (consonants) This allows you to learn phonemes in a language This could be learned in conjunction with visual cues from watching adults, and muscular cues from their own babbling phase, about phonemes. • Once learned, phonemes can be grouped into words.

Language Learning Not that simple. . .

Language Learning Not that simple. . .

Categorical Perception same percentage of /b/ responses Not that simple. . . different

Categorical Perception same percentage of /b/ responses Not that simple. . . different

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate

Language Learning The computational problem: INPUT: • Identify the first two formants • Calculate relative position of two formants (vowels) • Calculate slopes of formant transitions (consonants) • Once learned, phonemes can be grouped into words.

Grouping into Words

Grouping into Words

Language Learning FIRST STEP in the computational problem: INPUT: OUTPUT: WORDS

Language Learning FIRST STEP in the computational problem: INPUT: OUTPUT: WORDS

Language Learning SECOND in the computational problem: INPUT: Words in different orders / combinations

Language Learning SECOND in the computational problem: INPUT: Words in different orders / combinations OUTPUT: Grammatical rules syllable length, loudness, pitch,

How do we learn grammar? Prosody Syllable length Loudness Pitch Motherese Sing-songy way parents

How do we learn grammar? Prosody Syllable length Loudness Pitch Motherese Sing-songy way parents talk to their children Gets infant attention but does not seem to help them learn language Negative evidence Negative feedback about incorrect grammar Does not seem to be big factor Universal Grammar a set of restrictions of applied to how people learn languages. (E. g. Looking for particular types of phrases universal across languages and seems to be an innate bias. )

The Mind’s Purpose Why do we have a mind? To process info so we

The Mind’s Purpose Why do we have a mind? To process info so we can make good decisions for the physical implementation of the mind (e. g. , a person)

Bad Decisions Our physical implementation can lead to bad decisions when the system is

Bad Decisions Our physical implementation can lead to bad decisions when the system is messed up … Buy a house while you are high on drugs Decide whether or not you and your partner want to have a baby while you are drunk Decide to accept a job while you are on heavy amounts of prescription pain killers Decide to marry someone while under the influence of chemicals linked to infatuation /love

The Mind’s Purpose Why do we have a mind? To process info so we

The Mind’s Purpose Why do we have a mind? To process info so we can survive and procreate – a big time constraint of the physical implementation How do we make decisions about mates that help us procreate and help our procreations survive?

Factors in Selecting a Mate Rate 1 - 10 How important each of these

Factors in Selecting a Mate Rate 1 - 10 How important each of these are Hotness Patience Flexibility Wealth

Love: Nature’s Trick of Biased Sampling Early Dating “Love” chemicals biased sample Your behavior

Love: Nature’s Trick of Biased Sampling Early Dating “Love” chemicals biased sample Your behavior is biased sample of you Your partner’s behavior is biased sample Pheromones How much of long term relationship will be spent in middle-old age?

Factors Importance Over Time Qualities that increase in importance Patience Wealth Flexibility Qualities that

Factors Importance Over Time Qualities that increase in importance Patience Wealth Flexibility Qualities that decrease in importance Hotness

Biased Samples Early Dating Playing with other people's kids Fighting with someone you love

Biased Samples Early Dating Playing with other people's kids Fighting with someone you love Job interviews Inspirational speeches (Steve Jobs) Politicians People talking about politics Reviews of Professors (“Mc. Awful!”) Evaluating your own behavior

Final Piece of Advice: Listen to My Mom Date someone for at least two

Final Piece of Advice: Listen to My Mom Date someone for at least two years before marrying someone Wait until you are 30 before you get married (that's a joke), (no it isn't)