Agents and Environments with NLP Agent Environment sensors actuators 1.What do the other agents claim to believe? (NL Understanding) 2.What do the other agents actually believe or want? (Plan recognition, game theory) 3.How can I make the other agents believe X? (Planning, NL Generation) Agent Speech, Handwriting, printed text, digital text Speech, Handwriting, printed text, digital text
WHAT IS LANGUAGE? Definition with respect to form: Language is a system of speech symbols. It is realized acoustically (sound waves), visually-spatially (sign language) and in written form. Definition with respect to function: Language is the most important means of human communication. It is used to convey and exchange information (informative function) Multiplicity of languages: We know of about 7000 languages, which is estimated to be about 1% of all the languages that ever existed.
LANGUAGE AND THE BRAIN
THEORIES OF LANGUAGE Noam Chomsky claims that language is innate. B. F. Skinner claims that language is learned; it is basically a stimulus-response mechanism.
WHAT IS GRAMMAR ? When we learn a language we also learn the rules that govern how language elements, such as words, are combined to produce meaningful language. These elements and rules constitute the Grammar of a language. The Grammar is “what we know” Grammar represents our linguistic competence.
DESCRIPTIVE vs PRESCRIPTIVE GRAMMAR Prescriptive (should be) Descriptive (is)
Areas of Linguistics phonetics - the study of speech sounds phonology - the study of sound systems morphology- the rules of word formation syntax - the rules of sentence formation semantics - the study of word meanings pragmatics – the study of discourse meanings sociolinguistics - the study of language in society applied linguistics –the application of the methods and results of linguistics to such areas as language teaching, national language policies, lexicography, translation, language in politics etc.
What is the meaning of ‘meaning’? Learning a language includes learning the “agreed upon” meanings of certain strings of sounds and, Learning how to combine these meaningful units into larger units which also convey meaning.
Morphemes Morpheme is the smallest linguistic unit that has meaning. Morpheme is a grammatical unit in which there is an arbitrary union of sound and a meaning and, which cannot be further analysed (broken down into parts that have meaning).
Morphemes A morpheme may be represented by a single sound: e.g. the plural morpheme [s] in cat+s A morpheme may be represented by a syllable (monosyllabic): e.g. child+ish
Morphemes A morpheme may be represented by more than one syllable (polysyllabic): e.g. lady, water or three syllables: e.g. crocodile or four syllables: e.g. salamander
15 Words Two basic ways to form words – Inflectional (e.g. English verbs + endings other English verbs) Open + ed = opened Open + ing = opening – Derivational (e.g. adverbs from adjectives, nouns from adjectives) Happy happily Happy happiness (nouns from adjectives)
16 Syntax The study of classes of words (nouns, verbs, etc.) and the rules that govern how the words can combine to make phrases and sentences.
17 Basic classes of words Classes of words aka parts of speech (POS) – Nouns – Verbs – Adjectives – Adverbs The above classes of word belong to the type open class words We also have closed class words, or function words – Articles, pronouns, prepositions, particles, quantifiers, conjunctions
18 Basic phrases A word from an open class can be used to form the basis of a phrase The basis of a phrase is called the head
19 Examples of phrases Noun phrases – The manager of the institute – Her worry to pass the exams – Several students from the English Department Adjective phrases – easy to understand – mad as a dog – glad that he passed the exam
20 Examples of phrases Adverb phrases – fast like the wind – outside the building Verb phrases – ate her sandwich – went to the doctor – believed what I told him
21 Grammars and parsing syntactic parsing: Determining the syntactic structure of a sentence Basic steps – Identify sentence boundaries – Identify what part of speech is each word – Identify pairs of words that form phrases – Identify pairs of phrases that form larger phrases …
Context Free Grammar S -> NP VP NP -> det (adj) N NP -> Proper N NP -> N VP -> V, VP -> V PP VP -> V NP VP -> V NP PP, PP -> Prep NP VP -> V NP NP 22
23 Parses V PP VP S NP the mat satcat on NP Prep The cat sat on the mat Det N N
24 Parses V PP VP S NP time an arrow flies like NP Prep Time flies like an arrow. N DetN
25 Parses VNP VP S NP flies like an N Det Time flies like an arrow. N time arrow N
26 Semantics and Pragmatics Semantics: the study of meaning that can be determined from a sentence, phrase or word. Pragmatics: the study of meaning, as it depends on context (speaker, situation)
27 Language to Logic John went to a book store. s. bookstore(s) ^ go(John, s) Every boy loves a girl. ∀ b. boy(b) ∃ g. girl(g) ^ loves(b, g) Who broke the vase? λx. broke(x, vase17)
28 Headlines Police Begin Campaign To Run Down Jaywalkers Iraqi Head Seeks Arms Teacher Strikes Idle Kids Miners Refuse To Work After Death Juvenile Court To Try Shooting Defendant
31 Issues in NLP Ambiguity Lack of Knowledge – it’s needed for understanding, but computers don’t have it
Ambiguity Computational linguists are obsessed with ambiguity Ambiguity is a fundamental problem of computational linguistics Resolving ambiguity is a crucial goal
Ambiguity Find at least 5 meanings of this sentence: – I made her duck
Ambiguity Find at least 5 meanings of this sentence: – I made her duck I cooked waterfowl for her benefit (to eat) I cooked waterfowl belonging to her I created the (plaster?) duck she owns I caused her to quickly lower her head or body I waved my magic wand and turned her into undifferentiated waterfowl At least one other meaning that’s inappropriate for gentle company.
Ambiguity is Pervasive I caused her to quickly lower her head or body – Lexical category: “duck” can be a N or V I cooked waterfowl belonging to her. – Lexical category: “her” can be a possessive (“of her”) or dative (“for her”) pronoun I made the (plaster) duck statue she owns – Lexical Semantics: “make” can mean “create” or “cook”
Ambiguity is Pervasive Grammar: Make can be: – Transitive: (verb has a noun direct object) I cooked [waterfowl belonging to her] – Ditransitive: (verb has 2 noun objects) I made [her] (into) [undifferentiated waterfowl] – Action-transitive (verb has a direct object and another verb) – I caused [her] [to move her body]
Ambiguity is Pervasive Phonetics! – I mate or duck – I’m eight or duck – Eye maid; her duck – Aye mate, her duck – I maid her duck – I’m aid her duck – I mate her duck – I’m ate her duck – I’m ate or duck – I mate or duck
Kinds of knowledge needed? Consider the following interaction with HAL the computer from 2001: A Space Odyssey Dave: Open the pod bay doors, Hal. HAL: I’m sorry Dave, I’m afraid I can’t do that.
Knowledge needed to build HAL? Speech recognition and synthesis – Dictionaries (how words are pronounced) – Phonetics (how to recognize/produce each sound of English) Natural language understanding – Knowledge of the English words involved What they mean How they combine (what is a `pod bay door’?) – Knowledge of syntactic structure I’m I do, Sorry that afraid Dave I’m can’t
What’s needed? Dialog and pragmatic knowledge – “open the door” is a REQUEST (as opposed to a STATEMENT or information-question) – It is polite to respond, even if you’re planning to kill someone. – It is polite to pretend to want to be cooperative (I’m afraid I can’t…) – What is `that’ in `I can’t do that’? Even a system to book airline flights needs much of this kind of knowledge
Computational models of how natural languages work These are sometimes called Language Models or sometimes Grammars Three main types (among many others): 1.Document models, or “topic” models 2.Sequence models: Markov models, HMMs, others 3.Context-free grammar models
Computational models of how natural languages work Most of the models I will show you are -Probabilistic models -Graphical models -Generative models In other words, they are essentially Bayes Nets. In addition, many (but not all) are -Latent variable models This means that some variables in the model are not observed in data, and must be inferred. (Like the hidden states in an HMM.)