L11-probability

Question 1

Structuralism

Answer 1

Language as a structural system, consisting of arbitrary relationships between concepts and their words/sounds. The connection between a word and its meaning is only determined by convention (1900s:Ferdinand de Saussure)

Question 2

Behaviorism

Answer 2

Language is learned by behaviour and reinforcement. This is how we connect objects with their sounds/words. (1960s: Skinner)

Question 3

Generativism

Answer 3

An innate, internal system of abstract language rules that governs our language competence. (Chomsky)

Question 4

Modularity

Answer 4

Language is modular and its’ computation is automatic. (Fodor)

Question 5

Transducer (modularity)

Answer 5

We capture sounds from environment through our hearing/sensing ability. The environmental signals are then transformed in a way that they can be used by the language module.

Question 6

Language module

Answer 6

It takes information from transducer and decides which sounds are relevant for language (processes language). The language module is domain specific, automatic, speedy, cannot be accessed by other systems, has a neural specific structure, same developmental stages in all humans and only its output is accessible.

Question 7

Central system

Answer 7

The central system operates on modular information. The output of the language module can be used by other systems as well.

Question 8

Probabilistic approach

Answer 8

Probability is added to existing (symbolic & innate) theories for language. The probabilistic approach is applicable to language processing, acquisition and representation.

Question 9

Representation & probability

Answer 9

Probability can be added to phonology, meaning that how a phoneme behaves in a language system and acts in it is determined by its probability in a specific language (e.g.: some phonemes are more probable than others in a language).
Probability can also be added to morphology, meaning that how a morpheme affects a language system is determined by the probability of its occurrence (e.g.: regular forms of past tense are more common than irregular forms).

Question 10

Language processing & probability

Answer 10

Language processing involves generating and interpreting probabilities. It can be used for serial and parallel processing models.

Question 11

Serial processing & probability

Answer 11

Adding probability to serial processing by using the Bayesian approach

Question 12

Bayesian approach

Answer 12

Probability is updated based on previous syntactic structures and their words. Every new piece of information updates the system.

Question 13

Bottom-up update

Answer 13

A speech signal provides you with information to extract meaning from it. –>use present information to analyze sentences

Question 14

Top-down update

Answer 14

Use a language template to extract a speech signal.

use previous knowledge to analyze received information

Question 15

Reversed probability

Answer 15

We use both top-down and down-up updating and information is shared between both updating systems.

Question 16

Syntax trees

Answer 16

Hierarchical representation of sentence structure consisting of nodes with probabilities

Question 17

Probabilistic parsing

Answer 17

Estimating likelihood of different trees given a sentence and a probabilistic model.

Question 18

Principle of minimal attachment

Answer 18

Listener prefers simplest structures prefers parse with fewest nodes. Less nodes cost less, are hence more preferred and more probable.

Question 19

Computational linguistics in language processing

Answer 19

Analyzing language processing by computational methods

Question 20

Serial parser & plausibility

Answer 20

Sometimes less probable and more specific interpretations are favoured since they can be more easily fixed. –>serial parser prefers interpretations that can be rapidly falsified.

Question 21

Lexical information & probability

Answer 21

Lexical probabilities used in parallel models. Words change parsing preference and probabilities.

Question 22

Requirements for probabilistic language comprehension

Answer 22

Model of general knowledge;
Theory of mind;
Principles of pragmatics

Question 23

Parsing preference based on:

Answer 23

Plausibility and statistics;

Proximity and relation of words (which words occur together)

Question 24

Language acquisition & probability

Answer 24

Probability added to empiricist language acquisition models by investigating linguistic corpora and to innate language acquisition model (Chomsky) by chooses between innate candidate grammars based on probability of experienced language.

Question 25

Computational linguistics & language acquisition

Answer 25

Using context-free grammar and parsed trees to study language acquisition.

Question 26

Poverty of stimulus

Answer 26

We learn language fast and can generate new sentences, by having only a limited input. This supports the existence of a faculty of language.

Question 27

Context free grammar

Answer 27

Rules that always apply in the same way and are independent of context

Question 28

Parsed trees

Answer 28

Trees are already parsed and have hence less probabilities

Question 29

Distributional methods

Answer 29

Linguistic items with similar distributions have similar meanings. Hence meaning can be inferred by relationships of words.