CSC 350 Comp. Ling. Exam 3 Page 1 of 5 Name _____ KEY 129 pts ________ MULTIPLE CHOICE. 3 points each. Choose the SINGLE BEST answer. 1. Lexical level morphological parsing means finding the a) affixes of a word b) individual morphemes of a word B c) stem of a word d) orthographic rules of a word 2. The description of semantics that explains sentences in terms of the actions of a theoretical machine is called C a) denotational b) syntactic c) operational d) axiomatic 3. Pragmatics involves a) categorizing the structure of legal sentences. b) the use of language in context. c) finding the simplest way to say something. B d) recognition of a sentence by a computer. 4. Work in language translation has proved more difficult than at first imagined because a) massively parallel computers are necessary to attempt any translation. b) the translation of some language constructs requires world knowledge. c) it was proved that a computer could never mimic a human brain. B d) it was proved impossible for any machine to translate any language. 5. Pattern recognition includes a) recognition only of visual entities. b) recognition of visual entities only by computers. c) recognition of visual and non-visual entities. C d) recognition only of visual and audio entities. 6. Natural language processing seeks to a) embody computers with thought and feelings. b) allow computers to understand human languages. c) ascribe absolute meanings to sentences. B d) develop tools to automatically construct programs. 8. Tagsets are used to a) define the meaning of words b) define the meaning of sentences C c) show parts of speech of words d) label the antecedents of pronouns 9. Semantics is the study of a) how affixes are attached to words b) the meaning of sentences B c) the constraints in word formation d) how sentences are formed 10. Morphotactics is the study of a) how affixes are attached to words b) the meaning of sentences C c) the constraints in word formation d) how sentences are formed 11. In Interlingual Translation, the source language is translated a) into a language-independent representation before being translated into the target language. b) into a language-independent representation after being translated into the target language. c) directly into the target language. A d) directly into phonological output. CSC 350 Comp. Ling. Exam 3 Page 2 of 5 Name _____ KEY 129 pts ________ 12. Which regular expression describes the language of the given machine? B a) (v+w)v* b) (v+w)(v*(ww*v)*)* c) v + w + ww c) v + w + w(v+w) ____ ---- v,w | ---- | w ---- _______ -->| q |--------->|| q ||--------->| q | | w | 0 | || 1 || | 2 |<------ ---- | ---- |<--------- ---- ---- v | ^ |_v_| 13. Which regular expression describes the language of the given machine? D a) (v+w)v* b) (v+w)(v*(ww*v)*)* c) v + w + ww c) v + w + w(v+w) --------------------------------- | v | | V | ____ ------ ---- w | ---- | v,w | ---- | v,w ---- -->| q |--------->|| q ||-------->|| q ||----->| q | | 0 | || 1 || || 2 || | 3 | ---- | ---- | | ---- | ---- ---- ------ 14. In part-of-speech tagging, an open class a) contains words that can be used in any class. b) consists of nouns and verbs only. c) allows new words to be added to the class. C d) allows words of other classes in place of words in the class. 15. A corpus a) always contains mappings to the morphemes of its words. b) is another name for an ontology. c) is a collection of words from several languages. D d) is a collection of words from one language. 16. A unification-based grammar a) is another name for context-free grammar. b) uses unification to encode and exchange values during parsing. c) is the best way to pass attributes up and down the parse tree. B d) can not capture the notion of noun/verb number agreement. 17. What can be said about the "Chomsky Hierarchy" of 4 types of languages? a) The languages don't intersect. b) There exist theoretical machines with finite state and finite memory to recognize all these types of languages. c) There exist theoretical machines with finite state to recognize all these types of languages, but some require infinite memory. C d) Some of the languages can't be recognized by any known type of machine. 18. Bottom-up parsing a) is used by Prolog when you write a definite clause grammar. b) is the method of choice for human created parsers. c) always constructs a real parse tree as a data structure so it can be rearranged by the compiler to generate optimal code. d) uses a shift/reduce procedure to generate a tree by starting D at the leaves. CSC 350 Comp. Ling. Exam 3 Page 3 of 5 Name _____ KEY 129 pts ________ 19. Representational ambiguity a) can be resolved by using canonical forms for knowledge representation. b) can be resolved by using an unambiguous grammar for the language. c) is caused by vagueness in natural language. A d) is automatically solved by using predicate logic as the representation. 20. A semantic attachment specifies a) whether to use denotational, axiomatic, or operational semantics. b) how a unification grammar attaches features to on element of the parse tree. c) how to compute the meaning representation of an entity from the meanings of its constituents. C d) the set of tags attached to a corpus. 21. Lexical semantics a) relies entirely on a lexicon to ascribe meaning to a sentence. b) relies entirely on a grammar to ascribe meaning to a sentence. c) ascribes much of the meaning of a sentence to individual words. C d) specifies how to create a parse tree. 22. Rode is a homophone of C a) traveled b) bicycled c) road d) walked 23. A sentential form is any string of a) terminals. b) terminals and nonterminals. c) terminals and nonterminals derivable from the start symbol. C d) sentences. 24. Template-based approaches to speech recognition a) require visual methods to match patterns in the templates. b) compare input speech to stored units. c) analyze input speech statistically to match patterns. B d) have had no success in the translation process. 25. When viewing a Prolog program as a collection of theorems and axioms, which type of a cognitive process would you be attempting to model? B a) procedural b) logical c) data d) functional 26. What would be a reasonable translation into Prolog of "If high investment in something brought low returns then low investment in it will bring high returns." a) investment_return(high,X,high) :- investment_return(low,X,high). b) investment_return(low,X,high) :- investment_return(low,X,high). c) investment_return(high,X,low) :- investment_return(high,X,low). D d) investment_return(low,X,high) :- investment_return(high,X,low). 27. What does this Prolog code do? f([X],[]). f([X|Xs],[X|Ys]) :- f(Xs,Ys). a) returns the last element in a list. b) returns the first element in a list. c) returns the second to last element in a list. D d) returns the entire list without the last element. CSC 350 Comp. Ling. Exam 3 Page 4 of 5 Name _____ KEY 129 pts ________ 28. A referent is a) a pronoun b) an entity outside a sentence c) the first use of an entity in a sentence B d) the last use of an entity in a sentence 29. Referring to a previously introduced symbol is called A a) anaphora b) cataphora c) coreference d) situational context 30. Human dialogue is characterized by A a) turn-taking b) role-playing c) information exchange d) illocution 31. VoiceXML is meant a) for natural language processing. b) only for turn-taking dialogues. c) for Web-based dialogue control. C d) purely to control Internet telephony. 32. The order of major phases in natural language generation is a) discourse planning; sentence planning; text realizer b) sentence planning; text realizer; discourse planning c) text realizer; discourse planning; sentence planning A d) sentence planning; discourse planning; text realizer 33. A language which treats most morphemes as single words, i.e., that doesn't combine morphemes into a single surface form is called C a) agglutinative b) synthetic c) analytic d) fusional TRUE OR FALSE. 2 points each. _F_ 34. abc is in the language = {a, b} {0, 1}* {c, d} {e, f}* _T_ 35. a0c is in the language = {a, b} {0, 1}* {c, d} {e, f}* _F_ 36. a01010ef is in the language = {a, b} {0, 1}* {c, d} {e, f} _F_ 37. The set of all even length palindromes over a set A (i.e., R { w : w = uu , where u is in A* } ) is a regular set. _T_ 38. In Machine Translation, direct translation does not require elaborate analyses of the underlying structures of a sentence. _F_ 39. A corpus contains mappings to the morphemes of its words. _F_ 40. The statement "She's the sharpest woman to ever graduate from our program, presupposes that she is intelligent. _T_ 41. The statement "The insurance policy is cheaply priced, considering we drive in NYC." cancels the concept of cheaply priced. _T_ 42. Inflection, as contrasted with derivation, is the part of morphology which deals with case, number, tense, etc. _T_ 43. The goal of "information retrieval" as related to NLP can not be met by stemming. CSC 350 Comp. Ling. Exam 3 Page 5 of 5 Name _____ KEY 129 pts ________ Short Answer. 5 points each. Assume a knowledge base like the following. (The entire person/4 fact base is not necessarily shown -- this is just a sample.) % person(Who,Occupation,Gender,Height) person('Jon Stewart',comedian,male,short). person('Leonardo Dicaprio',actor,male,medium). person('Toni Braxton',singer,female,short). person('Janet Reno',cop,female,tall). 44. Write a Prolog grammar rule to recognize the following sentence, where OCCUPATION is any word at all. Who is the big-shot OCCUPATION ? sentence --> ['Who'], [is], [the], ['big-shot'], [What], ['?']. 45. Add a Prolog action to the rule you created in the previous problem to respond appropriately. Be sure to handle the case where OCCUPATION isn't in your database. your-lhs --> your-rhs, { ACTION }. Answer: sentence --> ['Who'], [is], [the], ['big-shot'], [What], ['?'], { person(Who,What,_,_), write_ln(Who); write_ln('I don''t know.') }.