c. Negative words are relatively restricted in number. d. In some cases, the negative words are used with other negation-related words to intensify or

Size: px

Start display at page:

Download "c. Negative words are relatively restricted in number. d. In some cases, the negative words are used with other negation-related words to intensify or"

영대 구
7 years ago
Views:

1 The Improvement of Negative Sentences Translation in English-to-Korean Machine Translation Jang Chung Hyok Kim Kwang Hyok Foreign Language Faculty Computer Science College Kim Il Sung University Kim Il Sung University Pyongyang, Democratic People s Republic of Korea Abstract This paper describes the algorithm for translating English negative sentences into Korean in English-Korean Machine Translation (EKMT). The proposed algorithm is based on the comparative study of English and Korean negative sentences. The earlier translation software cannot translate English negative sentences into accurate Korean equivalents. We established a new algorithm for the negative sentence translation and evaluated it. Key Words: machine translation, negative sentence, negation Introduction Translating English negative sentences into Korean is fairly challenging since the ways of expressing negation in the two languages are not exactly the same. The generally accepted way of the negative sentence translation in EKMT was to simply negate the predicate of its corresponding affirmative sentence, which cannot give the accurate Korean equivalent for some English negative sentences, e.g. partial negatives. We did linguistic comparison between the ways of expressing negation in English and Korean and propose an algorithm for the accurate Korean equivalents to the English negative sentences in EKMT. 1. Linguistic comparison of English and Korean negation Linguistically, negation refers to a grammatical construction with negative polarity [1]. The negative sentence and its affirmative correspondent are closely related in English and so are they in Korean. This requires that we should consider the English negative sentences in close relation with their affirmative correspondents and Korean equivalents. English and Korean negative sentences have the following features in common; a. The negative sentences are declaratives, interrogatives or imperatives in both languages. b. The negative sentences have their affirmative correspondents, which can be made negative by adding negative words. 1

2 c. Negative words are relatively restricted in number. d. In some cases, the negative words are used with other negation-related words to intensify or soften the negation. e. Negation is subdivided into general negation, partial negation, double negation and intensified negation. Despite these common features, there are still differences, the most important one of which is the way of adding negative words to the affirmative sentences. The table below gives the comparison between English and Korean negative sentences. The underlined words are the negative words and the bold words are the negation-related words in English and Korean. Part of Speech Negative Words English Negative Sentences Korean Equivalents Corresponding Affirmative Sentences I did not believe at all that 나는그가온다는것을전혀 I believed that he was he was coming. 믿지않았다. coming. Not much is known about 그질병에대하여많은것이 Much is known about the not the disease. Not many people have read 알려지지는않았다. 많은사람들이그기사를 disease. Many people have read the the report. 읽은것은아니다. report. Not one of the students 학생들가운데서 한사람도 One of the students could could answer. 대답할수없었다. answer. Adverb never He has never been there. Never again would he say that. 그는거기에가본적이전혀없다. 그는다시는그런말을하지않을것이다. He has been there. He would say that again. hardly I can hardly believe it. 나는그것을거의나믿을수없다. I can believe it. little Little did I know that my life was about to change. 나는나의생활이변하리라는것을거의나알지못하였다. I knew that my life was about to change. nowhere He has nowhere to live. I am going nowhere. 그는살곳이없다. 나는아무데도가지않는다. He has somewhere to live. nothing She's nothing like her brother. 그는자기의동생과전혀비슷하지않다. She's like her brother. 2

3 none She seems none the worse for her experience. 그는자기경험에비하여한심해보이지는않는다. She seems the worse for her experience. no one No one could hear me. 그누구도나의말을듣지못하였다. Someone could hear me. nobody I knocked on the door but nobody answered. 내가문을두드렸지만그누구도대답하지않았다. I knocked on the door and somebody answered. Pronoun nothing We said nothing about her. 우리는그에대하여아무것도말하지않았다. We said something about her. none None of my friends phones me any more. 내 친구들중 그 누구도 나에게 더 이상 전화하지 않는다. Some of my friends phone me. little Little is known about the causes of the problem. 그문제의원인에대해서는거의알려지지않았다. Something is known about the causes of the problem. no No food is left in the fridge. 랭동기에는음식이남은것이하나도없다. Some food is left in the fridge Determiner few Few people knew he was ill. 그가앓는다는것을아는사람은거의없다. Some people knew he was ill. little I paid little attention to what the others were saying. 나는다른사람들이하는말에거의관심을돌리지않았다. I paid attention to what the others were saying. Table 1. Comparison between English and Korean negative sentences As the table shows the negative words in English include adverbs, pronouns and determiners, therefore the subject, predicate, object, adverbial modifiers or attributes can be negated, while in Korean the negation is expressed by adding agglutinative suffixes (the underlined words in Korean sentences) to the predicates and by adding other related words (the bold words in Korean sentences) to the sentences in some cases. This significant difference poses a main problem for the translation of the English negative sentences into Korean. 2. Classification of English negative sentences Some English sentences with negative words are not negative sentences, so they are not translated into Korean negative sentences. He just stopped working for no reason. ( 그는아무리유도없이일을그만두었다.) The victory was nothing less than a miracle. ( 그승리는그야말로기적이였다.) 3

4 Also, the following examples show that there is a big difference in word order between the negative sentence and its corresponding affirmative sentence in English, while their Korean equivalents have less difference except the negative words and the negation-related words. Little did I know that my life was about to change. ( 나는나의생활이변하리라는것을거의나알지못하였다.) I knew that my life was about to change. ( 나는나의생활이변하리라는것을알고있었다.) Considering the features described above, we concluded that it is very important to decide whether the sentence with a negative word is a negative sentence and what sort of negative sentence it is. a. If the subject, predicate or the object of the sentence is negated or a negative word itself, it is a negative sentence. No one could hear me. (negative subject) We said nothing about her. (negative object) He has never been there. (negative predicate) However, if the negative word is not the subject, object or predicate of a sentence but a part of an idiomatic expression (e.g. for no reason, be nothing less than, etc), the Korean equivalent might not be a negative sentence. b. We subdivided the English negative sentences into single negative (one negative word in the sentence) and double negative sentences (two or more negative words in the sentence) according to the number of negative words in the sentence, and also into general negative, partial negative and intensified negative sentences according to the part of sentence which is negated. The Korean equivalent of the general negative sentence can be made from its corresponding affirmative sentence by adding agglutinative suffix, while those of the partial negative and intensified negative sentences require other negation-related words as well. 3. Translation of English negative sentences 3.1 The determination of English negative sentence To translate English negative sentence, the parser should decide; a. whether the sentence with a negative word is a negative sentence. According to which part of the sentence is negated, we subdivided the negative sentences into six categories; Negative Subject Affirmative Predicate Affirmative Object (NS-AP-AO) Negative Subject Negative Predicate Affirmative Object (NS-NP-AO) 4

5 Negative Subject Affirmative Predicate Negative Object (NS-AP-NO) Affirmative Subject Affirmative Predicate Negative Object (AS-AP-NO) Affirmative Subject Negative Predicate Affirmative Object (AS-NP-AO) Affirmative Subject Negative Predicate Negative Object (AS-NP-NO) If the sentence matches one of the above sentence structures, the parser decides that it is a negative sentence. b. what sort of negative sentence it is. According to the number of negative words and the negated part of sentence we conclude that; Single negative sentence is general negative, partial negative or intensified negative. The negative sentence with negative subject, negative object or negative predicate which is modified by the words like every, all, always, entirely, necessarily, etc is a partial negative sentence. Single negative sentence with negative predicate which is not a partial negative sentence is a general negative sentence. Single negative sentence with negative subject or negative object which is not a partial negative sentence is an intensified negative sentence. Double negative sentence is an intensified negative sentence. c. which negative agglutinative suffixes and related words should be added to the corresponding Korean affirmative sentence. Korean language has various negative agglutinative suffixes to the predicate such as ~ 이아니다, ~ 지못하다, ~ 지않다 and ~ 지말다. It is not easy to decide which one makes a natural expression in Korean. In 3.2 and 3.3, we describe the principles to choose correct suffixes and related words to the corresponding Korean affirmative sentences. 3.2 Negative agglutinative suffixes to Korean predicate The negative agglutinative suffixes to Korean predicate is decided by; a. The kind of predicate verb in English sentence and its tense and aspect If the predicate is a copular verb + noun, add ~ 이아니다 or ~ 이없다 to the predicate of the Korean affirmative sentence, and if the predicate is a lexical verb and the auxiliary verb do is used, add ~ 지않다 or ~ 지못하다. She is not my younger sister. 그녀자는나의녀동생이아니다. He doesn t get up early in the morning. 5

6 그는아침일찍일어나지않는다. If the predicate is a lexical verb in perfect aspect, add ~ ㄴ적이없다. He has never been there. 그는거기에가본적이없다. b. The kind of English negative sentence If the negative sentence is a general negative, the above rule is applied to choose a suffix to Korean predicate. If it is an intensified negative, the Korean negative suffix is added as general negative sentence. No trains will be affected by this accident. 렬차들은이사고에의하여전혀영향을받지않을것이다. He has made promises, but kept none of them. 그는약속을하였지만하나도지키지않았다. If it is a partial negative, the agglutinative suffix ~ ㄴ것은아니다 is added. It is not always true that combining two substances into one releases energy. 두물질을하나로결합하면에네르기가방출된다는것이언제나진실인것은아니다. If an auxiliary verb is negated in a partial negative sentence, the suffixes ~ 지는않다 or ~ 수는없다 is added to the Korean equivalent of the lexical verb and change it according to the tense of the English predicate. The particles will not necessarily move the same way under identical conditions. 립자들이동일한조건에서반드시같은방식으로움직이지는않을것이다. 립자들이동일한조언에서반드시같은방식으로움직일것이지는않다. c. The collocation of the Korean predicate In Korean, the terminative predicates of sentences are verbs or adjectives. According to the characteristics of Korean verb or adjective collocation, the suffix ~ 지않다 or ~ 지못하다 is added. Their effort accomplished little or nothing. 그들의노력은거의아무것도달성하지못하였다. 그들의노력은거의아무것도달성하지않았다. The Korean equivalents of some English negative predicates are completely different from their Korean affirmatives, thus making them impossible to be negated by adding agglutinative suffixes. In this case, the Korean predicate is replaced by appropriate word. I don t know what these symbols mean. 6

7 나는이기호들이무엇을의미하는지모른다. 3.3 Negation-related words Some negatives sentences (e.g. intensified negatives and double negatives) are accurately translated in Korean not only by adding negative suffix to the Korean predicate but also by adding adverb and changing the particle to the subject or the object. a. If it is a single negative (one negative word in the sentence), the negation-related words are added according to the English negative word. - If the predicate verb is negated by adverb not, no related words are added. The rooms don t have air-conditioners. 방들에는공기조화기가없다. - If the negative word in the sentence is never, little or few, the related adverbs are added to the Korean equivalent. She never looked like her photograph. 그녀자는자기사진과전혀비슷해보이지않는다. b. If it is a partial negative, the particle added to the subject of a Korean equivalent changes from 는 ( 은 ) to 가 ( 이 ). Not all my friends were at the wedding. 나의모든친구들이결혼식에참가한것은아니였다. c. If it is an intensified negative in which the subject or the object is negated, the particle to the Korean subject or object noun should be changed or other related words added. - If the subject or object is a negative pronoun or is negated by the determiner no, the particle to the Korean subject or object noun should be changed. Not one person remembered my birthday. 한사람도나의생일을기억하지못하였다. I have seen none of them before. 나는그들중누구도이전에본적이없다. - If the subject or object is negated by the determiner little or few to make a quasi-negative, negation-related words are added. Few students could answer the question. 거의모든학생들이그질문에대답할수없었다. He showed little interest in my suggestion. 그는나의제안에거의흥미를보이지않았다. d. If it is a double negative, the particles to the subject or the object should be changed and add negation-related words as the intensified negative. 7

8 I never said nothing about the plan. 나는그계획에대하여아무것도말하지않았다. 4. The improved algorithm for the English negative sentences On the basis of the linguistic study described above, we established an improved algorithm for the English negative sentence translation in EKMT. The following diagram shows it. act The workflow of Translating English Negative Sentences into Korean Initial :Negation Count Negation Count English Negative Sentence Decide the type of English negative sentence :Negation Type 1 Single Negation 2 Double Negation Negation Type Decide the negative word and its insert position :Insert Position of Negative Words Korean Negative Suffix to the Predicate General Negation Partial Negation Intensified Negation Insert Position of Negative Words After Predicate After Auxiliary Decide the additional negation-related words Korean Negation-Related Adverb Korean Particle to Subject and Object {'~ 이아니다,' '~ 지 않다,' '~ 한적이없다,' '~ 인것은아니다,' '~ 하지는않는다 '} {' 전혀,' ' 거의나,' ' 결코,'} Korean Negative Sentence Place the Korean negative word and its related words {' 도 '} Final Fig.1. Algorithm for the English negative sentence translation 5. Evaluation We applied the improved algorithm to English-to-Korean Translation System Ryongnamsan to improve its translation accuracy. The sample text was about the computer operating system. The sentences with negative words are 518 or 8.2% of the whole 6377 sentences. The following table shows the number of negative sentences of each kind. 8

9 Single Negative Double Negative Others NS-AP-AO AS-NP-AO AS-AP-NO NS-NP-AO NS-AP-NO AS-NP-NO Table 2. Number of negative sentences of each kind There were 5 partial negative sentences among 505 single negatives. The following examples show the translation results of the partial negative and intensified negative by the proposed algorithm, in comparison with the earlier one. When a process is executed, it is not always executed at the site in which it is initiated. 프로쎄스가실행될때그것은언제나호출이시작된싸이트에서실행되지않는다. (by the earlier algorithm) 프로쎄스가실행될때그것이언제나호출이시작된싸이트에서실행되는것은아니다. (by the proposed algorithm) Servers cannot do so, since they have no knowledge of the purpose of the client's requests. 그것들이의뢰기의요청들에대한지식을가지지않기때문에봉사기들은그렇게할수없다. (by the earlier algorithm) 그것들이의뢰기의요청들에대한그어떤지식도가지지않기때문에봉사기들은그렇게할수없다. (by the proposed algorithm) As the examples show, the proposed algorithm improved the translation of negative sentences, making it accurate and similar to the human translation. 6. Conclusion Negative sentences don t account for large proportion in English text, but their translation is important. We did a comparative study on the negation in English and Korean to establish a new translation algorithm in EKMT and introduced it to English-to-Korean Machine Translation System Ryongnamsan and conclude that it is effective. 7. References [1] Ronald Carter and Michael McCarthy (2006). Cambridge Grammar of English, Cambridge University Press 9

Page 2 of 5 아니다 means to not be, and is therefore the opposite of 이다. While English simply turns words like to be or to exist negative by adding not,

Page 2 of 5 아니다 means to not be, and is therefore the opposite of 이다. While English simply turns words like to be or to exist negative by adding not, Page 1 of 5 Learn Korean Ep. 4: To be and To exist Of course to be and to exist are different verbs, but they re often confused by beginning students when learning Korean. In English we sometimes use the