Close report

Language Processing : Humans and Computer

Exercises

1. Speech errors

a. For each speech error, state what kind of linguistic unit or rule is involved, that is, phonological, morphological, syntactic, lexical, or semantic.

b. Stats, to the best of your ability, the nature of the error, or the mechanisms which produced it.

(1) brake fluid → blake fruid

a. phonological vowel segment        

b. reversal or exchange of segments

(2) drink is the curse of the working classes → work is the curse of the drinking classes (Spooner)

a. semantic, syntactics change      

b. reversal or exchange of words however there are continuous grammatical plural morphemes

(3) we have many ministers in our church → …many churches in our minister

a. syntactics change                

b. reversal or exchange of words however there are continuous grammatical plural morphemes

(4) untactful → distactful

a. eclipsis morphology                

b. incorrect application of negative rule to exceptional example.

(5) an eating marathon → a meeting arathon

a. phonological vowel segment        

b. reversal or exchange of segments and intonation contour

(6) executive committee → executor committee

a. semantic word change

b. making the meaning to explain relative two words.

(7) lady with the dachshund → lady with the Volkswagen

a. semantic word change

b. two word are a common denominator; Germany. Not speaker's intend, however  change the word. 

(8) stick in the mud → smuck in the tid

a. phonological vowel segment

b. reversal or exchange of segments

(9) he broke the crystal on my watch → he broke the whistle on my crotch

a. phonological vowel segment and semantic word change

b. reversal or exchange of segments and intonation contour for making comic effect. It is meant to be almost certainly done because the second one is funny.

(10) a phonological rule → a phonological fool

a. semantic word change

b. making the meaning to express back word; rule and fool.

(11) pitch and stress → piss and stretch

a. phonological vowel segment        

b. reversal or exchange of segments

(12) big and fat → pig and vat

a. phonological vowel segment

b. single feature are switched; [+voiced] becomes [- voiced] in big and [- voiced] becomes [+ voiced] in fat.

(13) speech production → preach seduction

a. phonological vowel segment        

b. reversal or exchange of segments and intonation contour

(14) he's a New Yorker → he's a New Yorkan

a. inflectional morphology                      

b. incorrect application of an over-extension of plural rule. As like people from Texas are Texan, Korea are Korean.

(15) I'd forgotten about that → I'd forgot abouten that.

a. syntactic rule and inflectional morphology

b. Frist, failure to move the past participle to form a past sentence. Second, incorrect application of past-tense. Just add the exceptional rule in target word.

2. The use of spectrogram for speaker identification, Can you think of any reasons for such differences?

Spectrogram은 단시간 퓨리에 변환(short-time Fourier transform 이하 STFT)을 통해 인간의 발화를 시간 축에 따라 형태를 특정화 하는 스펙트럼 형식을 의미한다. 즉, 천만 분의 일초로 시간을 나누어 x 좌표로 정하고, 그 시간에 따른 음 진동수의 강약을 y 좌표로 나타내는 것을 의미한다.

지문과 홍채의 모양 그리고 DNA처럼 인간 개체의 특질을 정의하는 수단 중에 하나로 쓰이곤 하는 Spectrogram은 우리에게 Voice-Print라는 이름으로 쉬이 알려져 있다. 어째서 그러한가? 왜 인간의 발화가 그 인간 고유의 특질로 남을 수 있는가?

첫 번째로, 인간의 구강구조는 아이 때부터 발달한다. 사용하는 언어, 먹는 음식, 치아의 상태와 모양 그리고 후천적 구강구조의 변화 등 여러 가지 이유로 인간의 구강구조가 형성된다. 이러한 변화는 설령 흉내 내려 하여도 쉽게 되지 않을뿐더러 사소한 차이만으로도 결과는 다양하게 변화한다.

두 번째로, 인간의 후두는 쉽게 변화한다. 첫 번째 이유와 마찬가지로 인간의 후두 또한 어떻게 사용하느냐에 따라 그 형태가 쉽게 바뀐다. 고음, 장기간의 담배에서 사용하는 단어에 따른 발성법의 차이까지 변화의 요인은 다양하다.

위에 제시한 두 가지 이유를 비롯해서 인간의 발음 구조 내의 사소한 변화가 STFT에서 변화를 일으킨다. 그러므로 누구라도 다른 사람과 같은 발화 특징을 가질 수 없다.

3. Using a bilingual dictionary of a language you don not know, attempt to translate he following English sentences by looking up each word.

a.     The children will eat the fish.

→    Les enfants manger le poisson.

b.    Send the professor a letter from your new school.

→    Envoyer le professeur la lettre de votre nouveau école.

c.     The fish will be eaten by the children.

→     Le poisson mangé chez les enfants.

d.     Who is the person that is hugging that dog?

→     Qui est la personne lequel est caresse ce chien?

e.     The spirit is willing, but the flesh is weak.

→     Le esprit est disposé, majs la chair est faible.

A. Using your own knowledge, give a grammatically correct translation of each sentence.

a. Future-tense

b. Past-tense

c. an article

d. application of word

e. word place

B. Have a person who knows the target language translate the grammatical translation back into English.

a. The children eat the fish. (infintive form = to eat)

Grammatically correct it attach be : Les enfants mengent le poisson.

b. Sending the professor the letter of/from your new school. (infintive form = Sending)

c. The fish eats at the children's. (at the children's place.)

d. Who is the person who is stoked by the dog? or Who is the person who was stroking the dog? (It is a bit incorrect.)

e. The mind is willing but the flesh is weak. or The mind is willing but the body weak.

4. Suppose you were given a manuscript of a play, Describe how you would use the computer to help determine the true authorship of the mysterious play.

글을 분석할 때, ‘작가의 성격을 상상하면서 읽는다.’ 라고 말하곤 한다. 글에는 작가 고유의 것, 문체, 단어, 음성적 배열 등을 고려하여 Marlowe와 Shakespeare의 이전 작품과의 유사점을 비교한다면, 충분히 진실을 가려낼 수 있으리라고 생각한다.

즉, STFT의 방법과 유사하다. 단어의 유사성을 분간할 때와 비슷한 방법으로 작품 간의 유사성을 분간하게 한다는 뜻이다. 물론 여기에서 어느 정도의 오차가 생기는 것은 감수해야 하며, 이는 각 작품이 쓰인 시간의 간격과 이전 작품들 간의 간격에서 생긴 작가의 변화의 정도를 참고로 하여 오차의 범위를 가설로 세워두면 문제가 되지 않는다.

5. Speech synthesis is useful because it allows computers to convey information without requiring the user to be sighted.  Think of five other uses for speech synthesis in our society.

음성합성의 사용은 두 가지 분류로 나눌 수 있다. 반복적인 작업에서 인간을 벗어나게 해주는 것과 완전한 음성을 사용할 수 없거나 혹은 음성이 없을 경우에 부족함을 보강하기 위함이다.

첫 번째 분류의 경우 간단히는 대중교통 수단의 방송을 대체하는 것에서부터 단어, 문장, 책 등을 반복적으로 읽어주어 학습의 용도까지 폭 넓게 사용할 수 있다. 또는 감정이 섞이면 곤란한 직업에서 사용할 수 있다.

두 번째 분류의 경우 첫 번째 분류보다 더 활용 범위가 넓다. 말을 잘 하지 못하는 장애인에게 일반인과 비슷할 정도로 발음하게 도와주거나, 특정한 목적이 있는 컴퓨터가 그 목적을 위한 도구로 음성을 사용한다거나, 기계의 상태를 기계가 직접 제어자에게 말하는 것도 가능하다. 최근의 예로는 컴퓨터 가수를 들 수 있다.

7. Consider the following ambiguous sentences. Explain the ambiguity, give the most likely interpretation, and state what a computer would have to have in its knowledge base to achieve that interpretation.

(1) For those of you who have children and don't know it, we have a nursery downstairs. (Sign in a church)

a. "It" can refer to the children or you.

b. "It" refers to you.

c. Children don't interesting about the nursery.

(2) The police were asked to stop drinking in public places.

a. "ask" can refer to request or inquire.

b. "ask" refers to request.

c. Police is an executor for the law.

(3) Our bikinis are exciting; they are simply the tops. (Bathing suit ad in newspaper) 

a. "tops" can refer to shape of the bikini or express the first company in this kinds.

b. "tops" refers to express the first company in this kinds.

c. This is ad in newspaper. They want to effect of ad.

(4) It's time we made smoking history. (Antismoking campaign slogan)

a. "time" can refer to present action or interrupt.

b. "time" refers to interrupt.

c. Antismoking means interrupt the smoking.

(5) Do you know the time?(Hint: This is a pragmatic ambiguity.)

a. "time" can refer to matching the period or real time.

b. "time" refers to matching the period.

c. "know" helps making meaning of time.

(6) Concerned with spreading violence, the president called a press conference.

a. "violence" can refer to brute force or misinterpretation.

b. "violence" refers to misinterpretation.

c. "press conference" directions "violence" meaning.

(7) The ladies of the church have cast off clothing of every kind and they may be seen in the church basement Friday. (Announcement in a church bulletin)

a. "they" can refer to the ladies or clothing.

b. "they" refers to clothing.

c. The ladies are not usually open the basement Friday and announcement in a church. Not target, making the ceremony.

The following three items are newspaper headlines;

(8) Red Tape Holds Up New Bridge

a. "Holds Up" can refer to support or put on.

b. "Holds Up" refers to support.

c. Bridge can't put on with Red Tape. Maybe, It can help relation between person to person.

(9) Kids Make Nutritious Snacks

a. "Make" can refer produce or get.

b. "Make" refers to get.

c. Kids are usually consumer.

(10) Sex Education Delayed, Teacher Request Training

a. "Sex" can refer physical activity or two kinds of distinction.

b. "Sex" refers to two kinds of distinction.

c. Education are not usually how do such activity.

8. Here is transition network for the Noun Phrase(NP) rule given in chapter 4:

NP → (Det) (AP)N(PP)

Using this as a model, draw a transition network for the Verb Phrase Rules:

VP → V(NP)(PP)

(Hint: Recall from chapter 4 that the above rule abbreviates four rules.)

9. A. Based on the examples in the text, give a semantic network representation for each example.

B. Here are five more sentences and a semantic network representation for each. Give a representation of each of them using the predicate logic notation.

(6) Seals swim swiftly. SWIM(SEALS, SWIFTLY)

(7) The student doesn't understand the question. NOT (UNDERSTAND [THE STUDENT, THE QUESTION])

(8) The pen is on the table. ON(THE PEN, THE TABLE)

(9) My dog eats bones. EAT(BONES, [POSSESSIVE (I, DOG)])

(10) Emily gives mony to charity. GIVE (EMILY, MONEY, CHARITY)

10. "Torment the computer." Think of ten more words that are likely to lead to false analyses.

       Killarneyshot : Kill + Arney + Shot / Killarney Shot

       cattraingo : Cat + Train + Go / Cat Train + Go

       penaidsbroad : Penaid + s + Broad / Penaids + Broad

       gamebagsgot : Gamebag + s + got or Game + baguets / Gamebags + got

       foodaholicrostrum : Foodaholic + Cross + Trum / Foodaholic + Rostrum

       Drinksbeer : Drink + s + Beer / Drinks + Beer

       Paysback : Pay + s + Back / Pays + Back

       Trashygirl : Trash + Y + Girl / Trashy + Girl

       Traysplay : Tray + S + Play / Trays + Play

       Zadkielmorrison : Zadki + Elmorrison / Zadikiel + Morrison

11. Do this exercise ten times to see the average number of surnames you can pronounce with confidence before becoming unsure. This gives some measure of the vast difficulty facing computers that have to read names from phone directories.

For maximum benefit, you may wish to read aloud the phonetic pronunciations in the telephone directory. This same technique can be helpful when you encounter an unfamiliar surname: say it aloud, several times, changing the accented syllable or the vowel sounds. Listen for the sound of a name which is familiar to you. That may suggest an alternative spelling to include in your research.¹⁾

So on, You must believe you're a sixth sense; intuition. Pronounce example is add end of this paper.