Ch-3: Linguistics Essentials Prepared by Qaiser Abbas (07-0906)
3.1 POS and Morphology • Important POS are noun, verb and adjectives. • Basic test for word belonging to the same category is substitution test e.g. adjectives in fig 3.1 • Many words have multiple POS e.g. candy can also be a verb as in (Too much boiling will candy the molasses). • Word Classes: divided in two. 1. open or lexical which have large number of members e.g. noun, verb and adjectives. To which new words are commonly added 2. Closed or functional such as preposition and determiners containing words like of, on, the, a which have only a few members and clear grammatical use. • POS tags: (already known in CL) • Morphological Processes like singular to plural already known in CL. However in English regular verb has 4 distinct forms and irregular verb have at most 8 forms. In Finish verb has more than 10,000 forms. • Morphology include concept of inflection (modification of root form through suffix, prefix etc. as shown in table 1.3), derivation (radical change of syntactic category and often change in meaning e.g. adverb widely from adjective wide) and Compounding (merging two words into a new word e.g. tea kettle, college degree, etc). • Antecedent: a word comes before another and influence it e.g. “He grabbled the ball and threw it in the air” here ball is antecedent of it. • The other detail is shown in table 3.2. it is pertinent to note that reflexive pronoun are also known as anaphors e.g. each other • Noun, Pronoun or Adjective when • Subject of a verb (nominative) • It is used to show possession or close connection b/t two things (Genitive) • It is indirect object of a verb or connected with the indirect object(dative) e.g. I sent her a postcard (word her is dative) • It is in the direct object or connected with the direct object (accusative) e.g. I saw him today (him is direct object)
3.1.2: Words that accompany nouns: Determiners and Adjectives • Determiner describe the particular reference of a noun. A subtype is article which indicates that we are talking about something or someone that we already know about or can uniquely determine e.g. the tree (standing in front of tree) and a tree ( not known about the tree). • Adjectives describe the properties of noun e.g. a red rose, this long journey etc. Uses such as these modifying a noun are called attributive or adnominal. • Adjective also have predicative (use after verb) use e.g. “The rose is red. The journey will be long”. • Article and Adjectives agrees with the noun because many languages mark distinction of case, number and gender on articles and adjectives as well as nouns. • Adjectives has positive (basic form), comparative, superlative, periphrastic forms (more intelligent, most intelligent) • Brown Tags: Adjective Positive form (JJ), Comparatives (JJR), superlatives (JJT), Semantically superlative adjective (JJS) e.g. chief, main and top, Numbers are subclasses of adjectives included cardinals (CD), ordinals (OD), Articles (AT), singular determiners like this, that have the tag DT, plural determiners (DTS) and others tags can be seen in the text book p88.
3.1.3: Verbs • Verbs are used to describe actions (she threw the stone), activities (she walked along the river) and states (I have $50). Have 4 morphological forms e.g. walk, walks, walking, walked • Base form with infinitive ‘to’, progressive form is ‘ing’, • Irregular verbs like drive and take. Table 3.3 summarizes grammatical features • Words that accompany verbs in a verb group are called analytically or feature indication with auxiliaries e.g. have , will and be to express aspect, mood and some tense information. • Modal auxiliaries or modal is class of verbs with special properties. They express modalities like possibility (may, can), or obligation (should) etc. • Base form of verb (VB), VBZ for 3SG(takes), VBD for past tense, VBG for gerund and present participle (taking), VBN for past participle (taken). Tag for modals is MD. See detail in table 4.6
3.1.4: Other POS (Adverbs, Prepositions, Particles) • Adverbs specify place, time, manner or degree. It modifies a verb in the same way as adjective modifies noun e.g. “She often travels to Las Vagas” or “she allegedly (fact without proof) committed perjury (crime of telling a lie in court of law)” • Some adverbs can also modify adjectives (a very unlikely event) and other adverbs (she started her career off very impressively) and do not modify verbs known as degree adverbs. Some times known as qualifiers. • Preposition: express spatial relationship e.g. in the glass, on the table, over their heads (already known in CL) • Particles: subclass of preposition that can enter into strong bond with verbs in the formation of phrasal verbs e.g. The plane took off at Sam, Don’t give in to him. Some time preposition separated from the verb e.g. He put me off. These phrasal verbs have particular meaning and quite specialized and unpredictable from the verb and particle that make them up. • We need to know the meaning of a sentence to differentiate particle and preposition e.g. She ran up a hill (up is preposition due to running on a incline plane) and She ran up a bill (up is a particle due to figurative meaning of building up a large bill) • RB (ordinary adverb e.g. simply, well etc.), RBR (comparative adverb e.g. later, better, less), RBT (superlative adverb e.g. latest, best, least) and detail can be seen in the text book • Conjunction and Complementizers ( already known in CL) however cases of subordinating conjunction like that or use of for which introduces arguments of the verb are known as Complementizers.
3.2 Phrase Structure • Words are organized into phrases, grouping of words that are clumped as unit • Syntax is the study of regularities and constraints of word order and phrase structure. • Positional and Paradigmatic relationship is shown in example (3.29) and (3.30) as in fig. Noun phrases are members of one paradigm which can be replaced for each other. Two words bear a syntagmatic relationship if they can form a phrase (syntagma) like sewed clothes. Collocations are also important class of syntagmatically related words. • Overall phrase structure of English sentence is suggested in upshot (3.31) • Complex adjective phrase are less common e.g. He seemed a man who was quite certain to succeed. Verb is the head of verb phrase, verb phrase organizes all elements of the sentence that depend syntactically on the verb except most syntactic theories the verb phrase does not contain the subject noun phrase. NP is usually argument of the verb, contains optional determiner, zero or more AP, a noun head and then some post modifiers
Long distance dependency (movement theory) such as wh-extraction. which book in 3.46(b) is moved from an underlying position after the verb as in 3.46a to their surface position (the beginning of the sentence as in 3.46b. • (3.36) a. should peter buy a book? • b. which book should peter buy? • SNLP fail in non local dependency situation e.g. an n-gram model predict word after wallet in 3.45 is was, which is not true. These issues are discussed in ch-11 • Empty nodes are required in situation like eat the cake, in which subject is not present. So NP0 or e in this case. 3.2.1: Phrase Structure Grammars • Word order allows us who did what to whom as in “Mary gave peter a book” and “peter gave Mary a book” • Many languages like Latin or Russian or Urdu permit different ways of ordering without a change in meaning. However, English is very much restrictive in word order movement e.g. in general form “ The children (subject) should (auxiliary verb) eat spinach (object)” and in interrogative e.g. “should (auxiliary verb) the children (subject) eat spinach (object)?” • To produce a sentence, we start with the start symbol ‘S’ in example (3.39). Such grammars are known as CFG’s. with these rules we can derive sentence as in (3.40) and (3.41) • The more intuitive way to represent phrase structure is as a tree as shown in (3.42) and (3.43). • Another way to show constituency is via (labeled) bracketing e.g. for (3.43)[S[NP[AT the][NNS children][VP[VBD ate][NP[AT the][NN cake]]]] • Rewrite rules is recursive and this property is known as recursivity. In fig3.1the sequence of PP’s is generated by multiple application of the rewrite rule cycle “NP NP PP; PPIN NP” . Recursivity also expand sentence e.g. VP expanded to nine words as in fig 3.1 • Dependency through subject verb agreement e.g. “The women who found the wallet were given a reward”. Here women is subject and form of to be in plural.
3.2.1: Dependency: Arguments and Adjuncts • “Sue watched the man at the next table”. Here Sue and the man are arguments of the verb and dependents of watching event. Similar is the case of PP with the man. • Levels of arguments of the verb: as arguments via semantic roles contains agent, patient and instrument & goal describe other classes of semantic relationship. At syntactic level verb contains NP as sub and obj. Note: only pronoun change their forms in obj case. • Some verbs take two objects (direct & indirect) as in “she gave him the book”, him is indirect and the book is direct object. • Verbs of sending and communication such as “She sent her mother the book”. Such verbs allow an alternate expression as “She sent the book to her mother”. Language with case marking express patients in the accusative case and recipients in the dative case. • Semantic role and grammatical role is changed through voice alternation. The agent is sub and patient in obj in active is reversed in passive sentences and agent is expressed by prepositional by-phrase. • In other languages like Urdu the passive involves alternation in case marking or some morphology on the verb.
Sub categorization • Different verbs differ in the number of entities (persons, animals, things) that they relate e.g. “She brought a bottle of whiskey” in which of whiskey is the object of brought (Transitive case). Moreover, the verb bring can not be used without object. • However in “She walked (along the river)” there is no object except the activity in parentheses (Intransitive Case). • Dependents of verb: The sub, obj and direct obj are the arguments of the verb. Arguments centrally involved in the activity of the verb. Arguments are expressed as NP’s, PP’s, VP’s or as clauses (that clause after verb). • Adjuncts have a less tight link to the verb and are optional, easy to move around in the sentence. Where as complements are obligatory e.g. the object of bring in obligatory in the above sentence. Some example of adjuncts tell us the time, place, or manner of action or state that verb describe e.g. a. She saw a Woody Allen movie yesterday (time). b. She saw a Woody Allen movie in Paris (place).c. She saw a Woody Allen movie with great interest(manner of action). • Subordinate Clauses (Sentences within a sentence): can also be adjuncts or subcategorized arguments. Sometime it is difficult to distinguish adjuncts and complements e.g. “she put the book on the table” here put forces the PP’s as complement and “he gave his presentation on the stage” here adjunct is optional. Some examples of subcategorized arguments are given in fig as. • Similarly subcategorization frames is particular set of argument that a verb can appear with. It tells us the syntactic regularities about complements and also semantic regularities e.g. bark prefers dogs as subject and verb eat prefers edible things as objects. Some subcategorization frames are listed in fig as
3.2.3: X Theory • Due to limitation of phrase structure w.r.t. systematicity and regularities to construct phrase and appearance of dependents in clauses. An idea of head of phrase is floated in modern syntax. • A head forms small constituent with its complements which can be modified by adjuncts to form a bigger constituent and finally combine with specifier( a sub or sth else determiner) to form a maximal phrase as shown in example 3.64 N’ nodes are intermediate constituents. It is two bar level theory. recursive adjunction of modifiers are allowed at the N’ leval. This general pattern of constituency is repeated across phrase types. This is Known as X’ theory. Where X represents variable across lexical categories.
VP makes statement about the instrument spoon Children used a spoon while eating the cake 3.2.4 Phrase Structure/Syntactic Ambiguity: • Different parses for a particular sentence is known as phrase structure ambiguity (already discussed a lot in previous lecture and CL). • Attachment Ambiguity & Semantics: phrases generated from two different nodes. Different attachment have different meaning as shown in fig 3.2 • Garden Pathing: A garden pathing leads you along a path that suddenly turns out not work e.g. “The horse raced past the barn fell”. • First Meaning of above example is “the horse ran past the barn (area of storage of wheat)” and second is “the horse fell (close, bandhna) after it had been raced past the barn • Garden pathing is not problem in spoken language due to speaker’s communicative maxims (sensible behavior) and intonational pattern (rise and fall in spoken) and others like that. • What are the reasons if sentence will have no parse at all? NP makes statement which cake was eaten i.e. cake with a spoon and not icing Garden pathing is the phenomenon of first being tricked into adopting a spurious parse and then having a backtrack to try to construct the right parse 1.The rule does not exists 2. Ungrammatical sentence
3.3: Semantic and Pragmatics • Semantic: The study of meaning of words, construction and utterances. Divided into lexical semantic (individual word) and larger unit semantic (combination of words). • We can organize words into lexical hierarchy as in WordNet for lexical semantics such as antonym, synonym etc. • Hypernym or hyperonym is word with more general sense e.g. animal is a hypernym of cat. • Hyponym: word with more specialized meaning e.g. cat is hyponym of animal. • Meronym: part-whole relationship of word e.g. tire is meronym of car, leaf is meronym of tree. And holonym is opposite of meronym contains whole corresponding to a part. • Others can be seen in book. Also covered in CL • Natural language does not hold principle of compositionality, so assembling the meaning of whole sentence from the individual sentence is hard problem e.g. white refers to very different color in white paper, white hair, white skin, white wine. Here is meaning include meaning of the part and some additional semantic component that can not be predicted from the part. • Idiom: meaning of word and meaning of phrase is completely opaque (difficult to understand) e.g. to kick the bucket (mean dying) and carriage return.
Scoping problem with quantifiers and operators e.g. “everyone did not go to the movie” when quantifier having scope over not then it means that not one person went to movie and in contrast at least one person did not go to the movie. • After word and sentences, the next larger unit is the discourse. Narrative Discourse (sentence is examined to be example, elaboration, restatement) and Conversation Discourse (model relationship between turns and kinds speech acts (question, statements, requests, acknowledgement etc.) • Problem of anaphoric relation (relation b/t NP refers to the same thing or person) in discourse analysis e.g. “Mary helped peter get out of the cab. He thanked her”. • Anaphoric relations are important for information extraction as clear in example 3.72 • Pragmatics: the study of how knowledge about the world and language conventions interact with literal meaning. Anaphoric relations are pragmatic phenomenon e.g. to resolve the relation in discourse 3.72 , it is necessary to know the hurricanes are disasters. SNLP did not give much attention to pragmatics due to its complexity and lack of training data. Two areas of pragmatics in which work started is resolution of anaphoric relations and modeling of speech acts.
3.4: Other Areas • Linguistic is subdivided into phonetics (physical sound of language), phonology (structure of sound system), sociolinguistics (interaction of social organization and language), historical linguistics (change of languages over time), linguistic typology (languages make different use of linguistic devices and how they are classified as per their use), language acquisition (how children learn language), psycholinguistics (focus on real time production issues and perception of language and representation of language in brain), Mathematical linguistics (non quantitative mathematical methods). 3.5: Further Readings • References from the book