Computational Linguistics and Chinese Language Processing
Vol. 10, No. 3, September 2005, pp. 329-346 329 ? The Association for Computational Linguistics and Chinese Language Processing
Collocational Translation Memory Extraction Based on Statistical and Linguistic Information
Thomas C. Chuang?, Jia-Yan Jian+, Yu-Chia Chang+ and
Jason S. Chang+
Abstract
In this paper, we propose a new method for extracting bilingual collocations from a
parallel corpus to provide phrasal translation memories. The method integrates
statistical and linguistic information to achieve effective extraction of bilingual
collocations. The linguistic information includes parts of speech, chunks, and
clauses. The method involves first obtaining an extended list of English
collocations from a very large monolingual corpus, then identifying the
collocations in a parallel corpus, and finally extracting translation equivalents of
the collocations based on word alignment information. Experimental results
indicate that phrasal translation memories can be effectively used for computer
assisted language learning (CALL) and computer assisted translation (CAT).
Keywords: Bilingual Collocation Extraction, Collocational Translation Memory,
Collocational Concordancer
1.Introduction
Example-based machine translation (EBMT) has been proposed as an alternative approach to automatic translation. Translations of examples range from two-word to multi-word, translations, with or without syntactic or semantic structures [Nagao 1984; Kitano 1993; Smadja 1993; Lin 1998; Andrimanankasian et al. 1999; Carl 1999; Brown 2000; Pearce 2001; Seretan et al. 2003]. In the approach, text and translations are preprocessed and stored in a translation memory, which serves as an archive of existing translations that the MT system can reuse. A number of proposed applications for machine translation and computer assisted translation systems use translation examples found in bilingual corpora; these methods include
? Department of Computer Science and Information Engineering, Vanung University, No. 1, Vanung Road, Jhongli, Taoyuan, Taiwan
E-mail: tomchuang@https://www.doczj.com/doc/e515564932.html,.tw
+ Department of Computer Science, National Tsing Hua University, 101, Kuangfu Road, Hsinchu, Taiwan
[Received January 31, 2005; Revised August 24, 2005; Accepted August 26, 2005]
330 Thomas C. Chuang et al.
[Transit 2005], [Deja–Vu 2005], [TransSearch 2005], and [TOTALrecall 2005].
Statistical methods have been proposed for automatic acquisition of bilingual collocations [Smadja et al. 1996; Gao et al. 2002; Wu and Zhou 2003] from parallel bilingual corpora [Kupiec 1993; Smadja et al. 1996; Echizen-ya et al. 2003] or from two comparable monolingual corpora [Lu and Zhou 2004]. These bilingual collocations, if acquired in quantity, can enable a machine translation system to produce more native-speaker like translations. However, parallel corpora of substantial size are harder than monolingual corpora to come by. Therefore, in small- to mid-size parallel texts, collocations may not have high enough counts for a statistical method to reliably extract them.
Consider the example of extracting verb-noun collocations for the noun “influence” from 50,000 bilingual sentences (SMEC-50000) in the Sinorama Mandarin-English Corpus (SMEC)1. Some useful bilingual collocations in SMEC have very low occurrence counts. For instance, the bilingual collocation “use influence; 發揮影響力” appears only once in SM-50000 (see Example 1). Such collocations may not be extracted by the methods proposed in the literature.
(1) These circumstances make it unlikely that APEC will be able to avoid
reform. In Lai's analysis, the implosion of the Asian economies last year
demonstrates their interconnectedness. Therefore, in order to place its own
existence on a more secure foundation, Taiwan should carefully observe
changes in APEC and use its influence to make the organization into a
vehicle driving regional consolidation.
他分析，亞洲國家近年來經濟危機持續不去，證明瞭彼此的連動性，因
此台灣應該注意觀察APEC的轉變，發揮意見的影響力，以使APEC能
夠成為區域整合的火車頭，為我國創造更大的生存利基。
A good way of extracting such bilingual collocation might be to first extract “use influence” as a collocation in a large, separate, monolingual corpus, and then identify its instances and translations in the given parallel corpus (e.g., the Sinorama Mandarin-English Corpus). At present, it is not difficult to obtain a much larger monolingual corpus (e.g., the British National Corpus) that contains enough instances of “use influence” such that extraction of such a collocation type is mostly effective. Example (2) shows one of the 60 instances of
1The Sinorama Mandarin-English Corpus was originally a database of some 6,000 bilingual articles appearing in Sinorama Magazine dated 1976 to 2001 (Copyright ?2001 Sinorama Magazine & https://www.doczj.com/doc/e515564932.html, Co.) The database is a parallel text corpus, now available from The Association for Computational Linguistics and Chinese Language Processing.
Collocational Translation Memory Extraction Based on 331
Statistical and Linguistic Information
“use influence” in BNC. Such a relatively high count makes it very easy to identify “use influence” as a collocation by using any the method proposed in the literature.
(2) I don't know who it and apparently he asked him that, are, are any of your
men gonna be there and if there are he said, I'm, I'm gonna pull out and use
all my influence to stop the march and the IRA police said no there would
not be any gunmen there so I thought yeah, fucking right, oh yeah that's
easy to say, and then if like the reporter said and, and you believe him and
you have the feeble excuse towards a small community he said, you know
what 's going on.
We will present a new method that automatically performs shallow parsing on an English corpus to identify all the statistically significant collocation types and their instances in the monolingual corpus and the English part of the given bilingual corpus. After that, the translation of each collocate of each collocation is identified based using primarily the word alignment technique. We will also present a computer assisted translation system, Tango, which accepts user queries of words, parts of speech, and types of collocation, and displays citations with bilingual collocations highlighted. An example of a Tango search for bilingual collocations for the noun “influence” is shown in Figure 1.
Figure 1. An example of a Tango search for bilingual collocations for the noun “influence”.
332 Thomas C. Chuang et al.
The rest of the article is organized as follows. We will review related works in the next section. Then we will present our method for automatically processing sentences in monolingual and parallel corpora, and extracting bilingual collocations (Section 3). As part of our evaluation, we will describe an implementation of the proposed approach using SMEC and BNC (Section 4) and discuss the results of our evaluation carried out to assess the performance of bilingual collocation extraction (Section 5).
2.Extraction of Collocational Translation Memory
It is difficult to extract bilingual collocations from parallel corpora due to their limited availability and small sizes. Methods proposed in previous works typically extract collocations based solely on co-occurrence counts and statistical association measures in bilingual corpora. Unfortunately, a substantial part of the collocations in a modest-sized parallel corpus might not have high enough frequency counts for statistical extraction methods to be effective. Many bilingual collocations useful for machine translation and computer assisted language learning may appear only once or twice in a small to medium size corpus. To extract bilingual collocations, a promising approach is to acquire collocations in from a very large monolingual corpus and obtain translations from a parallel corpus.
2.1Problem Statement
We will focus here on the first step of building a translation memory for a bilingual collocation tool this involves; extracting a set of bilingual collocations instances from a sentence-aligned parallel corpus. These collocation instances can be used for the purpose of computer assisted translation and language learning. Thus, the collocations, including those that appear only once or twice, should be identified in the part of the parallel corpus that is written in one of the two languages. At the same time, it is crucial that the translations also be identified. Therefore, our goal is to return a reasonable-size set of documents that, at the same time, must contain an answer to the question. For simplicity, we will focus on verb-noun collocations in this paper. We formally state the problem that we are addressing below.
Problem Statement: Given a parallel corpus PC of n pairs of sentences (SE i, SF i) written in the first language E and the second language F.Our goal is to identify a set of k collocations and translations (CE ij, CF ij) in (SE i, SF i). To accomplish this task, we make use of a large corpus M with m sentences ME i of texts written in E to help identify CE ij in SE i.
Collocational Translation Memory Extraction Based on 333
Statistical and Linguistic Information
We attempt to identity collocations in a parallel corpus by leveraging another larger monolingual corpus in which collocations appears with higher occurrence counts. Our method is shown in Figure 2.
(1)Annotate the English Sentences with parts of speech, chunk, and clause information
(Section 2.2)
(2)Extract English collocation types in ME i (Section 2.3)
(3)Extract English collocation instances in CE ij in SE i (Section 2.3)
(4)Identify translation equivalents CF ij to collocation CE ij in SF i (Section 2.4)
Figure 2. Outline of the process used to extract bilingual collocations
2.2Taggers for parts of speech, chunks, and clauses
Using an annotated corpus with texts written in E, we can develop a tagger based on three Hidden Markov Models: one each for parts of speech, chunks (i.e. basis phrases), and clauses. The training corpus consists of sentences with three levels of annotation for each word: parts of speech, chunk, and clause. Figure 3 shows three levels of annotation for Example (3): w i t i u i v i This DT B–NP S is VBZ
X
B–VP
X the DT B–NP
11th JJ I–NP X consecutive JJ I–NP X
quarter NN I–NP X
S in IN B–PP
which WDT B–NP X
S the DT B–NP
company NN I–NP X
X has VBZ B–VP
paid VBN I–VP X shareholders NNS B–NP X
an DT B–NP
X extra JJ I–NP X dividend NN I–NP X
X of IN B–PP
five CD B–NP X cents NNS I–NP X . . O X Figure 3. Examples of three levels of tagging performed on the sentence “This is the 11th consecutive quarter in which the company has paid shareholders
an extra dividend of five cents.”
334 Thomas C. Chuang et al.
(3) This is the 11th consecutive quarter in which the company has paid
shareholders an extra dividend of five cents.
As shown in Figure 3, each word is tagged with a parts of speech tag (e.g., DT for determiner and VBZ for third person singular verb), a chunk tag indicating the basis phrase type (e.g., noun phrase, NP, verb phrase VP, etc.) Plus the position of the word (e.g., “B” for words beginning a chunk and “I” for all other words in the chunk), and a clause tag (e.g., “S” for a clause-beginning word and “X” otherwise). The chunk annotation shown in Figure 3 indicates that “This,” “the 11th consecutive quarter,” “the company,” “shareholder,” “an extra dividend,” and “five cents” are nouns on noun phrases, while “is” and “has paid” are verb phrases. Similarly, the clause annotation results indicate that “This is the 11th consecutive quarter” and “in which the company has paid shareholders an extra dividend of five cents” are the only two clauses in the sentence. With a substantial amount of annotated sentences like the above we can develop three taggers for each level of analysis.
For the parts of speech tagger, the HMM operates on a set of states represented by all possible POS tage, goes through a sequence of state t i, and produces words w i , i = 1 to n . An example of transition probability function ) | P(1-i i u u for the chunk tagger is shown in Figure 4. A first order HMM is characterized by the emission probability, P(w i | t i ), and state transition probability, P(t i | t i-1). An example of the emission probability function ) | P(i i u t for the chunk tagger is shown in Figure 5.
Figure 4. Example of transition probability function ) | P(1-i i u u for the chunk
tagger
) | i i u for the
chunk tagger
P(I-NP | I-NP) = 0.1 P(I-NP | B-NP) = 0.6 P(I-NP | I-NP) = 0.4
00.40.44.5
Collocational Translation Memory Extraction Based on 335
Statistical and Linguistic Information
Specifically, we have
P(w ) = P(w 1 w 2 … w n ) = 111i i-1i i 2,P( ) P( | )P( | ) P( | ) i n
t w t t t w t =∏. (1)
Therefore, we can to derive the parts of speech t = (t 1 t 2 … t n ) for the sentence w = (w 1 w 2 … w n ) by calculating ( t 1 t 2 … t n ) that maximizes P( t | w ). Thus, we have
( t 1 t 2 … t n ) = argmax t P(t | w ) = argmax t P(w | t ) P(t )/P(w ) = argmax t P(w | t ) P(t )
= 111i i-1i i 2,argmax P( ) P( | )P( | ) P( | ) t i n
t w t t t w t =∏. (2)
Similarly, we can derive the chunk tags ( u 1 u 2 … u n ) and clause tags ( v 1 v 2 … v n ) for the given sentence (w 1 w 2 … w n ) using Equations (3) and (4):
(u 1 u 2 … u n ) = 111i i-1i i 2,argmax P( ) P( | )P( | ) P( | ) u i n
u t u u u t u =∏, (3)
(v 1 v 2 … v n ) = 111i i-1i i 2,argmax P( ) P( | )P( | ) P( | ) v i n
v u v v v u v =∏. (4)
The optimal values of parts of speech tags ( t 1 t 2 … t n ), chunk tags ( u 1 u 2 … u n ), and clause tags ( v 1 v 2 … v n ) can be derived by using a dynamic procedure [Manning and Shutze 1999]. The tagging process is carried out for the n source sentences SE i in a given parallel corpus PC and in the m sentences ME i in a large corpus M . We will describe in Section 3 how the training data provided the results from the common task CoNLL-2000 and CoNLL-2001 shared tasks (CoNLL, 2000) can be used to estimate the probabilistic functions involved, including P(t 1 ), P(t i | t i-1), P(w i | t i ), P(u 1 ), P(u i | u i-1), P(t i | u i ), P(v i ), P(v i | v i-1) and P(u i | v i ).
2.3 Extraction of Collocation Types in M
With the chunk and clause tags for the sentences in the monolingual corpus M , we can proceed to extract a set of verb-noun collocation types from M . To that end, we can consider the heads of phrases in three prevalent verb-noun collocation structures in the corpus: VP+NP, VP+PP+NP, and VP+NP+PP. To extract verbs and nouns that appear in a predicate-object relation, we need to have a full parse of the sentences. However, a state-of-the-art parser can not produce a full parse of unrestricted texts with a very high precision rate. Therefore, we simply assume that a noun phrase following a verb phrase is in a predicate-object relationship unless they belong to two different clauses.
For instance, consider the sentence shown in Example (4):
336 Thomas C. Chuang et al.
(4)Confidence in the pound is widely expected to take another sharp dive if
trade figures for September due for release tomorrow fail to show a
substantial improvement from July and August’s near-record deficits.
The taggers described in Section 2.2 will produce the parts of speech tags, chunk tags, and clause tags shown in Examples (5)-(7):
(5)Confidence/NN in/IN the/DT pound/NN is/VBZ widely/RB expected/VBN
to/TO take/VB another/DT sharp/JJ dive/NN if/IN trade/NN figures/NNS
for/IN September/NNP ,/, due/JJ for/IN release/NN tomorrow/NN ,/,
fail/VB to/TO show/VB a/DT substantial/JJ improvement/NN from/IN
July/NNP and/CC August/NNP 's/POS near-record/JJ deficits/NNS ./.
(6)Confidence /B-NP in/B-PP the/B-NP pound/I-NP is/B-VP widely/I-VP
expected/I-VP
to/I-VP take/I-VP another/B-NP sharp/I-NP dive/I-NP if/B-SBAR
trade/B-NP figures/I-NP
for/B-PP September/B-NP due/ADJP for/B-PP release/B-NP
tomorrow/I-NP ,/O
fail/O to/O show/O a/B-NP substantial substantial/I-NP improvement/I-NP
from/B-PP
July/B-NP and/O August/B-NP 's/B-NP near-record/I-NP deficits/I-NP ./O
(7)Confidence /S in/X the/X pound/X is/X widely/X expected/X
to/S take/X another/X sharp/X dive/X if/S trade/X figures/X
for/X September/X due/X for/X release/X tomorrow/X ,/ *
fail/S to/X show/X a/X substantial/X improvement/X from/X
July/X and/X August/X 's/X near-record/X deficits/X ./X
The words in the same chunk can be further grouped together (as shown in Figure 6) to make it easy to examine the phrase types of two adjacent chunks and extract the head word of each phrase. For instance, we can extract a VN pair (e.g., “take” and “dive”) from an annotated sentence by taking the last words of two adjacent VP and NP chunks.
Collocational Translation Memory Extraction Based on 337
Statistical and Linguistic Information Phrase Type
Confidence NP
In PP
the pound NP
is widely expected to take VP
another sharp dive NP
If SBAR trade figures NP
For PP
September NP
Figure 6. Example of grouping words in a chunk together record by record
Care should be taken to avoid extracting verbs and nouns from two clauses in a sentence.
For instance, in Example (8), in some cases, considering only chunk information is not sufficient. For example, the VN pair (“think,” “people”) taken from the two separate clauses “why do you think” and “people cannot see the top of the building on some days” should be excluded from consideration.
(8) W hy do you think people cannot see the top of the building on some days?
Some VN pairs extracted at this stage are free combinations, while some recur more frequently than is likely according to chance and should be considered collocations. After obtaining a list of instances of candidate collocations, we proceed to find distinct collocation types and calculate their word counts and phrase counts in order to verify whether each of them is a valid collocation. After that, we calculate the strength of association between each verb-noun pair in the collocations by using Logarithmic Likelihood Ratio (LLR) statistics. Equation (5) is the formula that computes the LLR.
11122
1112221122(1)(1)(;)2log (1)(1)k n k n k k n k k n k p p p LLR x y p p p p ??????=???; (5)
k 1 :
count of sentences that contain x and y simultaneously; k 2 :
count of sentences that contain x but do not contain y; n 1 :
count of sentences that contain y; n 2 : count of sentences that do not contain y;
p 1 = k 1 / n 1;
p 2 = k 2 / n 2;
p = (k 1+k 2) / (n 1+n 2 ).
338 Thomas C. Chuang et al.
We have described a method for handling VN collocations. This method can be easily extended to handle VPN and VNP collocations as well. The idea is quite simple. After identifying a VN collocation type where the verb and noun are separated by a preposition, we go on to consider the preposition that comes between the verb and the noun or that follows the verb and noun. The VPN and VNP collocations are validated again by calculating the LLR between each VN pair and the preposition.
2.4Extraction of Collocation Instances in PC
We subsequently identify collocation instances in the n sentence SE i of the give parallel corpus PC. First, each sentence SE i is subjected to the same POS, chunk, and clause analyses as is applied to the corpus M. The collocation instances of the forms VN, VPN, and VNP are extracted in a similar way to that described in Section 2.2. There are two cases in which a collocation instance will be considered as a valid collocation:
1.if it passes the LLR threshold calculated based on the counts of words and co-occurrences in
PC;
2.if it is in the list of valid collocations found in M.
The quantity and quality of collocations in a very large monolingual corpus surely will facilitate collocation identification in a smaller bilingual corpus with better statistical measures.
2.5Extracting Collocation Translation Equivalents in a Bilingual Corpus After instances that are most likely valid collocations are obtained from a bilingual corpus, we go on to work on the second part of the parallel corpus PC. We exploit statistical word-alignment techniques [Melamed 1997] and dictionaries to find translation candidates for each of the words in a given collocation. Using Melamed’s approach, we can establish a word translation based on corpora to supplement English-Chinese dictionaries, which generally suffer due to insufficient information. We first locate the translation of the noun. Subsequently, we locate the verb nearest to the noun translation to find the translation of the verb. Figure 7 shows some examples.
English sentence Chinese sentence
If in this time no one shows concern for them, and directs them to correct thinking, and teaches them how to express and release emotions, this could very easily leave them with a terrible personality complex 如果這時沒有人關心他們，引導他們正確思考，教他們表達、渲洩情緒，極易在人格成長上留下一個打不開的死結。
Collocational Translation Memory Extraction Based on 339
Statistical and Linguistic Information
they can never resolve.
Occasionally some kungfu movies may appeal to foreign audiences, but these too are exceptions to the rule. 偶爾有一些武打片對某些外國觀眾有吸引力，但也是個案。
Figure 7. Examples of identifying translations of nouns (in bold) and verbs (shaded) of VN collocation instances in bilingual sentence pairs
3.Implementation and evaluation
We have implemented a program for extracting bilingual collocations based on the proposed method and experimented with 50,000 bilingual sentences (SMEC-50000) from the Sinorama Mandarin-English Corpus (SMEC). We wanted to assess the performance of the program and
verify whether useful bilingual collocations in SMEC with very low occurrence counts (e.g.,
“use influence; 發揮影響力”) could be extracted. Such collocations are beyond the reach of methods previously proposed in the literature.
We used the Brown corpus to develop a parts-of-speech tagger and the CoNLL-2000 benchmark database to build a chunk tagger and clause tagger. The chunk tagger relied on the transition and output probabilities of chunks. Figures 8 and 9 show examples of these two processes. The average precision rate of the chunk tagger was about 93.7%, based on CoNLL testing data.
Chunk tag u i Chunk
tag
u i+1adj. count(u i u i+1) count(u i) P(u i | u i+1)
B-NP I-NP 46327.3675030.686300 B-NP B-VP 8762.3675030.129806 B-NP O 5418.3675030.080268 B-NP B-PP 3878.3675030.057454 B-NP B-NP 1974.3675030.029248 B-NP B-ADVP 645.3675030.009560 B-VP I-VP 9830.3261250.345313 B-VP B-NP 9021.3261250.097619 B-VP B-PP 2550.3261250.065811 B-VP O 1719.3261250.039782 B-VP B-ADJP 1039.3261250.031169 B-VP B-ADVP 814.3261250.025428 B-VP B-SBAR 664.3261250.011265 Figure 8. Example data of transition probabilities of chunks
340 Thomas C. Chuang et al.
Chunk tag u i POS tag t i adj. count(u i u i+1)count(u i)P(t i | u i) B-NP at 19307.3675030.286021 B-NP nn 7555.3675030.111925 B-NP np 7541.3675030.111718 B-NP jj 5587.3675030.082771 B-NP nns 4473.3675030.066268 B-NP cd 2836.3675030.042017 B-NP pp$ 2261.3675030.033499 B-NP pps 2080.3675030.030818 B-NP ppss 1620.3675030.024003 I-NP nn 34671.3776830.446318 I-NP nns 13143.3776830.169191 I-NP jj 8247.3776830.106166 I-NP np 7250.3776830.093332 I-NP cd 4727.3776830.060854 I-NP cc 1727.3776830.022235 I-NP vbg 1147.3776830.014769 I-NP vbn 940.3776830.012104 I-NP ap 862.3776830.011100 Figure 9. Example data of emission probabilities of chunks
Using the chunk and clause information, w e proceeded to extract a list of collocation
types from the monolingual British National Corpus. We mainly used this list to identify collocation instances in SMEC. Finally, we applied the Competitive Linking Algorithm to
SMEC to obtain word alignment results. We then applied the results of word alignment to
extract the matching translations of the noun and verb collocates. The collocation extraction
program produced a much larger set of collocation candidates than could be obtained from
BNC. The corpus consists of over 100 million words in about 5 million sentences. After
filtering out incomplete sentences, we obtained around 4 million sentences for use in extracting valid English collocations. After implementing our proposed method as described
in Sections 2.2 and 2.3, we obtained over half a million collocation types of the forms VN,
VPN, and VNP. We were able to identify over 30,000 collocation instances in SMEC. Figures
10 and 11 show some examples in BNC.
Figure 10. The results for collocation types extracted from the BNC and SMEC
Collocational Translation Memory Extraction Based on 341
Statistical and Linguistic Information
VN type Example
Exert influence That means they would already be exerting their influence by the time
the microwave background was born.
Exercise influence The Davies brothers, Adrian (who scored 14 points) and Graham
(four), exercised an important creative influence on Cambridge
fortunes while their flankers Holmes and Pool-Jones were full of fire
and tenacity in the loose.
Wield influence Fortunately, George V had worked well with his father and knew the
nature of the current political trends, but he did not wield the same
influence internationally as his esteemed father.
Extend influence The CAB extended its influence into the non-government sector,
funding research by the Cathedral Advisory Commission and the Royal
Society for the Protection of Birds.
Diminish influence To break up the Union now would diminish our influence for good in
the world, just at the time when it is most needed.
Gain influence In general, women have not benefited much in the job market from
capitalist industrialization nor have they gained much influence in
society outside the family through political channels.
Counteract influence To try and counteract the influence of the extremists, the moderate
wing of the party launched a Labour Solidarity Campaign in 1981. Reduce influence Whether the curbs on police investigation will reduce police influence
on the outcome of the criminal process is not easy to determine.
Figure 11. Examples of collocation instances extracted from SMEC
With the collocation types and instances extracted from the corpus, we built an on-line collocation reference tool called TANGO to support searching for collocations and translations of a given word.
Figure 12. TANGO, a web-based bilingual collocation tool
342 Thomas C. Chuang et al.
TANGO accepts a query word in English and a collocation type, and returns a list of collocation types and examples. Figure 12 shows a screen returned for a query for VN collocations of “influence.” One instance for each collocation type is shown first. All instances can be shown on demand. Besides showing bilingual collocation extractions, TANGO also color codes the translation counterparts of the collocation instances. This informative, bilingual reference tool has been used in language learning classes and by professional translators. Initial responses have been quite positive, indicating that this new tool is very useful for EFL learners and translators.
To assess the quality of the extracted bilingual collocations, we randomly selected 100 sentences with extracted bilingual collocations from SMEC for manual evaluation. Many of these sentence had more than one collocation, 50 we evaluate each collocation individually. Students majoring in English assessed each bilingual collocation in the context of the corresponding pair of sentences. The evaluation process involved judging the validity of translation of the collocation. There were three levels of validity: satisfactory translation, approximate translation (partial matching), and unacceptable translation. Figure 13 shows examples for each level of validation. For the purpose of this research, satisfactory translations and approximate translations were considered useful. Therefore, we determined the percentages of bilingual collocations that fell into these two categories. As indicated in Figure 14, the average precision rate for the extracted bilingual collocations was about 90% for satisfactory translations and approximate translations.
Level of quality English sentences Chinese sentences
satisfactory translation Thus when Chinpao Shan put out its
advertisement last year, looking for new
people to develop its related enterprises,
the notice frankly stated "Southern
Taiwanese preferred.
去年，金寶山在發展關係企業徵
招新人的廣告上，就坦白指明
「本省籍南部人優先」。
approximant translation Ah-ying relates that "Teacher Chang"
friendly and easy-going, is always there
to answer her questions. She even goes to
him for answers when her friends have
legal questions.
阿英表示，「張老師」親切隨和，
只要有不懂的事，都去問老師，
就連朋友有法律上的問題，也去
請教他。
unacceptable translation Said one observer, "If I can speak bluntly,
the mainlanders are robbing graves of
their treasures and smuggling them away,
and the situation is bad. In reality, though,
it is Taiwan that is behind it all
committing the crime.
「說得不好聽，大陸近年來盜
墓、文物走私情形嚴重，台灣其
實是背後的劊子手！」有人這樣
認為。
Figure 13. Three levels of quality of the extracted translation memory
Collocational Translation Memory Extraction Based on 343 Statistical and Linguistic Information
Type % of
satisfactory
translations % of satisfactory and approximate translations*
VN 73 90
VPN 66 89
VNP 78 89
Figure 14. Evaluation of bilingual collocations extracted from SMEC
4.Discussion and limitation
Collocation is an important part of translation task yet it has long been neglected. Traditional machine translation tends to translate input texts word by word, which easily leads to literal translations. Therefore, even with abundant vocabulary, dictionary and grammar rule-based model systems fail to generate fluent translations in a target language. For example, due to its lack of collocational knowledge, a machine translation system may recognize “take” as “na”
(i.e., take away) and “medicine” as “yao” (i.e., medicine) in Chinese, respectively. Thus, systems are inclined to literally translate “take medicine” as “na yao” (i.e., "take away the medicine" in Chinese), resulting in an odd translation or mistranslation. We suggest that machine translation systems should take collocational translation memory into consideration to improve the translation quality.
Due to the limitations of the word-alignment technique, our method may incorrectly recognize some matching translations. We need better word-alignment to align translations more correctly. Moreover, expansion of the bilingual corpora would also increase the precision achieved in retrieving collocational translation memory. This would enable us to obtain high enough counts for each collocate (i.e., verbs and nouns in VN collocations) in the target language so as to increase the confidence level of the LLR statistics, which in turn would eliminate the anomalous collocational translation memory.
5.Conclusion
In the field of machine translation, Example-Based Machine Translation (EBMT) exploits existing translations in the hope of producing better quality translations. However, collocational translation has always been neglected and is hard to deal with. We have proposed the use of collocational translation memory to develop a better translation method that can solve some problems resulting from literal translation. Encouraged by the satisfactory precision rates in collocation and translation extraction obtained in this study, we hope that collocational translation memory can be further applied in machine translation, cross language information retrieval, computer assisted language learning, and other NLP applications.
344 Thomas C. Chuang et al.
Acknowledgements
This work was part of the “CANDLE” project funded by the National Science Council of Taiwan (NSC-2213-E-238-015 and NSC92-2524-S007-002). Further information about CANDLE is available at https://www.doczj.com/doc/e515564932.html,.tw/.
References
Andriamanankasina, T., K. Araki, and T. Tochinai, “Example-Based Machine Translation of Parts-Of-Speech Tagged Sentences by Recursive Division,” In Proceedings of MT SUMMIT VII, Singapore, 1999.
Brown, R. D., “Automated Generalization of Translation Examples,” In Proceedings of the Eighteenth International Conference on Computational Linguistics (COLING-2000), Saarbrücken, Germany, August 2000, pp. 125-131.
Carl, M., “Inducing Translation Templates for Example-Based Machine Translation,” In Proceedings of MT Summit VII, 1999.
CoNLL yearly meeting of the SIGNLL, the Special Interest Group on Natural Language Learning of the Association for Computational Linguistics. The shared task of text chunking in CoNLL-2000 is available at http://cnts.uia.ac.be/conll2000/.
Echizen-ya, H., K. Araki, Y. Momouchi, and K. Tochinai, “Effectiveness of Automatic Extraction of Bilingual Collocations Using Recursive Chain-Link-Type Learning,” The 9th Machine Translation Summit, 2003, pp.1 02-109.
Gao, J., J. Nie, H. He, W. Chen, and M. Zhou,“Resolving Query Translation Ambiguity Using
a Decaying Co-occurrence Model and Syntactic Dependence Relations,” The 25th
Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 2000, pp.183 -190.
Kitano, H., “A Comprehensive and PracticalModel of Memory-Based Machine Translation,”
In Proceedings of IJCAI-93, 1993, pp. 1276-1282.
Kupiec, J., “An algorithm for Finding Noun Phrase Correspondences in Bilingual Corpora,” In Proceedings of the 31th Annual Meeting of Association for Computational Linguistics, 1993, pp. 17-22.
Lin, D., “Extracting Collocation from Text Corpora,” First Workshop on Computational Terminology, 1998, pp. 57-63.
Lü, Y., and M. Zhou, “Collocation Translation Acquisition Using Monolingual Corpora,”
Association for Computational Linguistics 2004, pp. 167-174.
Melamed, I. D., “A Word-to-Word Model of Translational Equivalence,” In Proceedings of the Association for Computational Linguistics 1997, Madrid Spain, 1997, pp. 490-497. Nagao, M., “A Framework of a Mechanical Translation between Japanese and English by Analogy Principle,” Artificial and Human Intelligence, A. Elithorn and R. Banerji (eds.) North-Holland, 1984, pp. 173-180.
Collocational Translation Memory Extraction Based on 345
Statistical and Linguistic Information
Pearce, D., “Synonymy in Collocation Extraction,” In Proceedings of the NAACL 2001 Workshop on WordNet and Other Lexical Resources: Applications, Extensions and Customizations, CMU, 2001.
Seretan, V., L. Nerima, and E. Wehrli, “Extraction of Multi-Word Collocations Using Syntactic Bigram Composition,” International Conference on Recent Advances in NLP, 2003, pp. 424-431.
Manning, C.D., and H. Schutze, “Foundations of Statistical Natural Language Processing” MIT Press, Cambridge, Mass, 1999.
Smadja, F., “Retrieving Collocations from Text: Xtract,” Computational Linguistics, 1993, 19(1), pp.143-177.
Smadja, F., K. R. Mckeown, and V. Hatzivassiloglou, “Translation Collocations for Bilingual Lexicons: a Statistical Approach,” Computational Linguistics, 22, 1996, pp.1-38.
Wu, H., and M. Zhou, “Synonymous Collocation Extraction Using Translation Information,”
The 4Jth annual conference of the Association for Computational Linguistics, 2003, pp.
120-127
Related web pages:
Deja–Vu (https://www.doczj.com/doc/e515564932.html,/).
TOTALrecall (https://www.doczj.com/doc/e515564932.html,.tw/Counter/Counter.asp?funcID=1).
Transit (https://www.doczj.com/doc/e515564932.html,/eng/software/sprachtech/transit.html).
TransSearch (https://www.doczj.com/doc/e515564932.html,/).
346 Thomas C. Chuang et al.
Chapter 1 Passage 1 Human Body In this passage you will learn: 1. Classification of organ systems 2. Structure and function of each organ system 3. Associated medical terms To understand the human body it is necessary to understand how its parts are put together and how they function. The study of the body's structure is called anatomy; the study of the body's function is known as physiology. Other studies of human body include biology, cytology, embryology, histology, endocrinology, hematology, immunology, psychology etc. 了解人体各部分的组成及其功能，对于认识人体是必需的。研究人体结构的科学叫解剖学；研究人体功能的科学叫生理学。其他研究人体的科学包括生物学、细胞学、胚胎学、组织学、内分泌学、血液学、遗传学、免疫学、心理学等等。 Anatomists find it useful to divide the human body into ten systems, that is, the skeletal system, the muscular system, the circulatory system, the respiratory system, the digestive system, the urinary system, the endocrine system, the nervous system, the reproductive system and the skin. The principal parts of each of these systems are described in this article. 解剖学家发现把整个人体分成骨骼、肌肉、循环、呼吸、消化、泌尿、内分泌、神经、生殖系统以及感觉器官的做法是很有帮助的。本文描绘并阐述了各系统的主要部分。 The skeletal system is made of bones, joints between bones, and cartilage. Its function is to provide support and protection for the soft tissues and the organs of the body and to provide points of attachment for the muscles that move the body. There are 206 bones in the human skeleton. They have various shapes - long, short, cube - shaped, flat, and irregular. Many of the long bones have an interior space that is filled with bone marrow, where blood cells are made. 骨骼系统由骨、关节以及软骨组成。它对软组织及人体器官起到支持和保护作用，并牵动骨胳肌，引起各种运动。人体有206根骨头。骨形态不一，有长的、短、立方的、扁的及不规则的。许多长骨里有一个内层间隙，里面充填着骨髓，这即是血细胞的制造场所。 A joint is where bones are joined together. The connection can be so close that no movement is possible, as is the case in the skull. Other kinds of joints permit movement: either back and forth in one plane - as with the hinge joint of the elbow - or movement around a single axis - as with the pivot joint that permits the head to rotate. A wide range of movement is possible when the ball - shaped end of one bone fits into a socket at the end of another bone, as they do in the shoulder and hip joints. 关节把骨与骨连接起来。颅骨不能运动，是由于骨与骨之间的连接太紧密。但其它的关节可允许活动，如一个平面上的前后屈伸运动，如肘关节；或是绕轴心旋转运动，如枢轴点允许头部转动。如果一根骨的球形末端插入另一根骨的臼槽里，大辐度的运动（如肩关节、髋关节）即成为可能。 Cartilage is a more flexible material than bone. It serves as a protective, cushioning layer where bones come together. It also connects the ribs to the breastbone and provides a structural base for the nose and the external ear. An infant's skeleton is made of cartilage that is gradually replaced by bone as the infant grows into an adult. 软骨是一种比一般骨更具韧性的物质。它是骨连结的保护、缓冲层。它把肋骨与胸骨连结起来，也是鼻腔与内耳的结构基础。一个婴儿的骨骼就是由软骨组成，然后不断生长、
全国2013年7月高等教育自学考试 英语翻译试题 课程代码：00087 一、单项选择题（本大题共15小题，每小题2分，共30分） （一）本部分共有10小题，在每小题列出的四个备选项中只有一个在意义和表达方面与原文最接近，选择最佳译文。错选、多选或未选均无分。（20分） 1. Several times on his trips to China, which he made as a guest of the Chinese Government, Bill’s birthday occurred while he was in Beijing. A. 有好几次在他作为中国政府客人的中国之旅中，比尔的生日都是在北京过的。 B. 比尔有好几次是在作为中国政府客人的中国之旅中在北京过的生日。 C. 比尔作为中国政府的客人访问中国，好几次都赶上在北京过生日。 D. 在比尔的好几次中国之旅中，他作为中国政府的客人，生日都是在北京过的。 2. Brindley had begun on his own account, out of interest, to survey the waterways that he traveled as he went about his engineering projects for mills and mines. A. 布林德雷在为他的磨坊和矿井工程到处奔走的时候，出于自愿和兴趣，对沿途经过的河道进行了勘察。 B. 布林德雷由于个人的原因，为磨坊和矿井工程奔走时在兴趣之外已经开始勘察他走过的河道。 C. 按布林德雷自己的解释，他在为磨坊和矿井工程奔走的过程中，对于所走过的河道的勘察完全是出于兴趣。 D. 在为他的磨坊和矿井工程到处奔走的时候，布林德雷受利益驱动，用自己帐户里的钱勘察了沿途经过的河道。 3. And as rain fell in the Ethiopian highlands and the snows melted in the Mountains of the Moon, the Nile was everlastingly renewed. A. 随着埃塞俄比亚高原上雨水的降临和月亮山上积雪的融化，尼罗河被不断地更新着。 B. 埃塞俄比亚高原上降临的雨水和月亮山上融化的积雪，使尼罗河的水长年不断地更新着。 C. 尼罗河的水一直都来源于埃塞俄比亚高原的降水和月亮山的积雪融化。 D. 埃塞俄比亚高原上的雨水和月亮山上融化的积雪，为尼罗河提供了无穷无尽的水源。 4. Since 1978, when China began opening its economy to increased foreign investment and trade, aggregate output has more than doubled. A. 自从1978年中国经济向日益增加的外国投资和贸易敞开大门以来，总产量增加了两倍以上。 B. 自从1978年以来，中国经济为日益增加的外国投资和贸易敞开大门，总产量增加了一倍以上。
2019年10月自考《00087英语翻译》复习资料 Unit 1 stories 1. He was thirty-six, his youth had passed like a screaming eagle, leaving him old and disillusioned. 他已三十六岁,青春像一路鸣叫的鹰,早已一闪而逝, 留给他的是衰老和幻灭. 2. average height 普通高度 3. gleaming eyes 两眼闪着光辉 4. in his middle twenties 大概是二十五六岁的年龄 5. ignoring the chair offered him, Chu The stood squarely before this youth more than ten years his junior and in a level voice told him who he was, what he had done in the past, how he had fled from Yunnan, talked with Sun Yat-sen, been repulsed by Chen Tu-hsiu in Shanghai, and had come to Europe to find a new way of life for himself and a new revolutionary road for China. 朱德顾不得拉过来的椅子,端端正正地站在这个比他年轻十岁的青年面前,用平稳的语调, 说明自己的身份和经历: 他怎样逃出云南, 怎样会见孙中山, 怎样在上海被陈独秀拒绝, 怎样为了寻求自己的新的生活方式和中国的新的革命道路而来到欧洲. 6. When both visitors had told their stories, Chou smiled a little, said he would help them find rooms, and arrange for them to join the Berlin Communist group as candidates until their application had been sent to China and an answer received. 两位来客把经历说完后,周恩来微笑着说,他可以帮他们找到住的地方,替他们办理加入党在柏林的支部的手续, 在入党申请书寄往中国而尚未批准之前,暂时作候补党员. 7. Chinese Communist Party 中国共产党 8. 两条要求, 忠实------内容, 通顺-------语言 9. Several times on his trips to China, which he made as a guest of the Chinese Government, Bill’s birthday occurred while he was in Beijing. 以中国官方客人的身份, 比尔来访中国已多次了, 而且在北京停留期间恰适他生日也有好几次了. 10. ‘This is for you,’Bill Morrow heard on many occasions he would never forget----such as when he was taken a boat down the Grand Canal and every boat that passed sounded its siren in salutation. Or when he shown over the great Nanjing bridge, built where the ferries used to carry trains across the Changjiang River. He was given a chair and asked to wait a little as darkness came on, then suddenly the whole bridge was outlined in lights. “这是为你安排的.”这句话比尔.莫诺听到过好几次, 每一次都令他难以忘怀. 有一次, 他沿大运河乘船顺水而下,途经的每艘船都鸣笛致敬. 还有一次, 他参观雄伟的南京长江大桥------以前没桥
医学专业英语翻译 医学专业英语翻译如下： portable electric dental engine 轻便电动钻牙机，轻便牙钻portable hearing aid 袖珍助听器 portable microtome 手提式切片机 portable monitor 手提式监护仪 portable obstetric table 轻便产床 portable operating table 轻便手术台 portable photoelectric colorimeter 便携式光电比色计 portable suction unit 便携式吸引器 portable testing set 便携式测试仪器 portable typewriter 手提式打字机 portable X-ray machine 手提式X 光机 portacid 移酸滴管，滴酸器 portal 门，入门 portal venography 门静脉造影术 port B/L 港口提单 portcaustic 腐蚀药把持器 porte 柄 porte-acid 移酸滴管，滴酸器
porte-aiguille 持针器 porte-caustique 腐蚀药把持器 porte-ligature 深部结扎器，缚线把持器porte-meche 填塞条器 porte-noeud 瘤蒂结扎器 porte-polisher 握柄磨光器 porterage 搬运费 portial impression trays 局部牙托portion 部分，段，份 portligature 深部结扎器，缚线把线器port of arrival 到达港 port of delivery 交货港 port of departure 出发港 port of destination 到达港目的港 port of discharge 卸货港 portogram 门静脉造影片 portoraphy 门静脉造影术portovenogram 门静脉造影片 posion 阴离子，阳向离子 position 位置，状态 positioner 定位器（牙），位置控制器
全国高等教育自学考试(2018年4月) 00087英语翻译(试题及答案) I. Word and Phrase Translation (20 points, 1 point each) A. Translate the following words and phrases into Chinese. Write your translation on the Answer Sheet. 1.martial arts 2.ticket check https://www.doczj.com/doc/e515564932.html,itary forces 4.cultural taboos 5.agricultural waste 6.property law 7.humanitarian aid 8.World Economic Forum 9.public-service advertisements 10.a research university B. Translate the following words and phrases into English. Write your translation on the Answer Sheet. 11.红包12.交通民警 13.面部肌肉14.民族认同 15.虚拟世界16.植物病害 17.消费市场18.人才外流 19.主办城市20.外汇储备 II. Translation Revision (20 points, 2 points each) A. Correct or improve the translation of the following English sentences. Write your revised versions on the Answer Sheet. Example: 原文：Adelaide enjoys a Mediterranean climate. 译文：阿德莱德享有地中海型气候。 改译：阿德莱德属地中海型气候。 21.原文：The organization begins to feel a need to systematize its management of human assets. 译文：该组织开始意识到将人力资源管理系统化的必要。 改译： 22.原文：Of the forces shaping higher education, none is more sweeping than the movement across borders. 译文：在影响高等教育的众多力量中，任何其他力量都比跨国界迁移更加重要。 改译：
英语自考本科翻译重点词汇
fellow countrymen 同胞militarist军阀 Chinese Communist Party 党组织 the Grand Canal （京杭）大运河 Industrial Revolution 工业革命 dawn to dark从早到晚 especially English 具有英国特征的 cottage industry 农村手工业 Indian summer 最后的余辉 stock-raising 畜牧业 Homestead Act 宅地法 wild west 西部荒原 the gold rush 淘金热 basic occupation 基础产业 Department of Agriculture 农业部门 the Nile Delta 尼罗河三角洲 coastal erosion 水土流失 Mediterranean climate 地中海型气候 surveyor general 测量总监 sovereign nation 主权国家 aggregate output 总产量 coastal area 沿海地区 telephone connections 电话用户 economic well-being 生活水平 two way trade 双向贸易 lucrative export market 出口市场 foreign exchange 外汇 sum total 总数 per capita 人均 GNP 国民生产总值 joint ventures 合资企业 net income 净收入 world investment system世界投资体系 the stock of foreign investment 外资存量
cardiovascular diseases; 脑垂体的功能the function of pituitary; 泌尿道urinary tract; 分子molecule; 动脉artery; 内分泌学endocrinology; 呼吸困难dyspnea; 唾液saliva; 组织学histology; 血液循环blood circulation; 血液学hematology; 生理学physiology; 解剖学anatomy; 女性生殖系统femal reproductive system; 神经细胞nerve cell; 免疫学immunology; 消化不良dyspepsia; 随意肌voluntary muscle; 胚胎学embryology; 心理学psychology; 细胞学cytology; 原生质protoplasm; 细胞膜cell membrane; 细胞核nucleus; 细胞质(浆)cytoplasm; 脱氧核糖核酸deoxyribonucleic acid; 能半渗透的semipermeable; 分子生物学molecular biology; 碳水化合物carbohydrate; 有区别性的differentially; 使…完整intact; 根据according to; 遗传特性hereditary trait; 渗滤diffusion; 转换transaction; 蓝图blueprint; 染色体chromosome; 色素pigment; 排出废液excrete waste fluid; 散开disperse; 脉冲信号impulse; 核糖核酸ribonucleic acid; 损害正常功能impair the normal function; 污染环境pollute environment; 功能失调malfunction; 致病因子causative agents; 易受侵害的人群vulnerable groups; 局部化的感染localized infection; 花柳病venereal disease; 抗原与抗体antigen&antibody; 肌电图electromyogram; 多发性硬化multiple sclerosis; 心电图electrocardiograph; 疾病的后遗症sequelea of disease; 光纤技术fiber optic technology; 造血系统hematopoietic system; 致命的疾病fatal disease; 体液body fluid; 无副作用的治疗hazard-free treatment; 无侵犯的实验检查non-invasive laboratory test; 核磁共振nuclear magnetic resonance; 葡萄糖耐糖实验the glucose-tolerance test; 乐观的预后optimistic prognosis; 超声波检测法ultrasonography; 病史medical history; 随访活动follow-up visit; 营养不良nutritional deficiency; 使细节显著highlight detail; 脑电图electroencephalogram; 缺血的组织blood-starved tissue; 肌纤维muscle fiber; 随意肌voluntary muscle; 消化道alimentary canal; 肌腹fleshy belly of muscle; 横纹肌striated muscle; 肌肉痉挛cramps of muscle; 肌肉收缩muscle contraction; 肌肉附着点attachment of the muscle; 肌肉放松relaxation of muscle; 动脉出血arterial hemorrhage; 止端insertion;起端origion;供血blood supply; 屈肌flexor; 蛋白分子protein molecule; 纤维结缔组织fibrous connective tissue; 伸肌extensor; 意志力willpower; 横切面transverse section; 起搏器pacemaker; 肌萎缩muscle atrophy; 重症肌无力myasthenia gravis; 弥散性局部缺血diffuse ischemia; 常染色体隐性autosomal recessive; 全身性感染systemic infection; 受累的肌肉muscle involved; 显著相关性significant correlation; 神经末梢nerve terminal; 自体免疫反应autoimmune reaction; 神经支配innervation; 肌营养不良muscular dystrophy; 慢性营养不良chronic mulnutrition; 先天性肌病congenital myopathy; 预期寿命life expectancy; 免疫紊乱immunologic derangemant; 发病高峰年龄the peak age of onset; 胸腺肿瘤thymoma; 呼吸肌受累the involvement of respiratory muscle; 感染性肌炎inflammatory myositic; 去神经支配denervation; 矿物质吸收mineral absorption; 机械应力mechanical stress; 骨基质有机部分the organic parts of bone matrix; 青春期早熟premature puberty; 蛋白溶解酶protein-digesting enzyme; 破骨细胞osteoclast; 松质骨spongy bone; 骨折fracture; 不规则骨irregular bone; 骨骼系统skeletal system; 维生素吸收vitamin absorption; 骨钙丧失the loss of calcium from bone; 生长激素growth hormone;
自考,英语翻译重点 Translation Conclusion Unit 1 stories his youth had passed like a screaming eagle～ 1. He was thirty-six～leaving him old and disillusioned. 他已三十六岁～青春像一路鸣叫的鹰～早已一闪而逝～留给他 的是衰老和幻灭。 2. average height 普通高度 3. gleaming eyes 两眼闪着光辉 4. in his middle twenties 大概是二十五六岁的年龄 5. ignoring the chair offered him～ Chu The stood squarely before this youth more than ten years his junior and in a level voice told him who he was～ what he had done in the past～ how he had fled from Yunnan～ talked with Sun Yat-sen～ been repulsed by Chen Tu-hsiu in Shanghai～ and had come to Europe to find a new way of life for himself and a new revolutionary road for China. 朱德顾不得拉过来的椅子～端端正正地站在这个比他年轻十岁的 青年面前～用平稳的语调～说明自己的身份和经历: 他怎样逃出云 南～怎样会见孙中山～怎样在上海被陈独秀拒绝～怎样为了寻求 自己的新的生活方式和中国的新的革命道路而来到欧洲。 6. When both visitors had told their stories～ Chou smiled a little ～
这里汇聚了中西医学行业的大部分英语词汇和详细解说，如果要查询相关词汇，你可以点此word 文档工具栏的“编辑”，找到“查找”，然后点开输入你要查询的词汇就可以查询了。 西医篇： 1、医院部门及科室名称 2、医务人员名称 3、诊断和治疗常用词汇 4、常见疾病名称 5、常见手术名称 6、常用药物名称 7、常用护理术语 8、常用临床医学术语 9、医疗器材 10、医学英语快速记忆－后缀 11、主要人体系统名称 12、医院类型名称 13、医学词汇 14、医学常用字首与字根 1.医院部门及科室名称 out-patient department 门诊部 In-patient department 住院部 Nursing department 护理部 Admission office 住院处 Discharge office 出院处 Registration office 挂号处 Reception room, waiting room 侯诊室 Consultation room 诊察室 Isolation room 隔离室 Delivery room 分娩室 Emergency room 急诊室 Ward 病房室 Department of internal medicine 内科 Department of surgery 外科 Department of pediatrics 儿科 Department of obstetrics and gynecology 妇科 Department of neurology 神经科 Department of ophtalmology 眼科 E.N.T.department 耳鼻喉科 Department of stomatology 口腔科 Department of urology 泌尿科 Department of orthopedic 骨科 Department of traumatology 创伤科 Department of endocrinology 内分泌科
自考英语翻译技巧一 定语从句：定语从句是由一些关系代词或者关系副词引导的从句组成，用来修饰名词中心词。定语的翻译要特别注意位子上的变化，即：如果是单词修饰名词就放在中心词的前面，这和汉语的词序是一样的；如果定语太长，定语就要放在中心词的后面，这时就要注意断句，重复先行词，例如： 1.I have the same problem as you concerning the learning of English 在学英语的问题上，我和你有同样的问题。 2.She has a perfect figure. 她身材不错。 3.This is indeed a most pressing problem. 这的确是一个非常棘手的问题。 4.The only other people who knew the secret were his father and mother. 别的知道这个秘密的人就是他的父母。 5.The effective work of maintaining discipline is usually performed by students who advise the academic authorities 有效的维持纪律通常是由一些学生来做的，而这些学生则负责给学校的领导提建议。 6.The nomads who made these markings lived by hunting a d fishing during the last Ice Age which began about 35，00 B.C. and ended about 10，000 B.C.作出这些刻记的游牧人，依靠狩猎和捕鱼生活在冰河时代的晚期。这段时期大约始于公元前35，000 年而结束于公元前10，000 年 7. For several hundred years， cases have been reported of children who have been reared in the wild by animals kept isolated from all social contact.在过去的几百年中，已经有记载说，确实有被动物养大的与世隔绝的狼孩的存在。 另外，翻译成汉语时还要注意未省略关系代词和关系副词的定语从句，翻译时所遵守的原则和上述的一样：关系代词引导的定语从句： 8.He is the very person （whom） I‘m looking for.他正是我要找的人 9.This is the flexible policy that our government has pursued consistently. 这就是我们政府一直追求的灵活政策。 10.Those whose families cannot afford board and expenses can be allowed to receive scholarship.那些家里提供不起食宿的学生可以得到奖学金。11.There are forms that occasionally occur which should not be taught to students.有些形式是不能告诉给学生的。
医学英语课后翻译答案完 整版 Revised by Jack on December 14,2020
Translation： Unit one 1 医院，在现代意义上，就是由专业人员和专业的设备为病人提供医疗服务的机构，通常但也不总是提供长期的住院治疗。它的历史意义，直至最近，是给退伍战士提供服务的地方。 2 在现代社会，从广义上说，医院或者由它所处的国家的政府提供资助，或者在私营部门通过竞争在经济上生存下来。目前，为医院提供赞助的通常有公共部门，医疗机构，赢利或者非赢利的，医疗保险公司或者教会，包括直接由教会提供捐赠。在历史上，医院通常由宗教机构或者教会的成员或领导来建立和资助。与此不同的是，现代医院主要由专业的内科医生、外科医生和护士组成，在过去这些工作主要由成立医院的宗教机构或者志愿者来完成。 3 一些病人去医院只是去接受诊断和治疗，然后离开，不在医院过夜（门诊病人）；也有一些病人被收入院，在医院过夜，甚至长达几个星期或几个月（住院病人）。医院的等级划分通常由它们所提供的医疗设施的类型和为住院病人提供的医疗服务，以及其他的在门诊所提供的医疗服务水平所决定。（医院与其他医疗机构的区别在于它具有将病人收住院并为住院病人提供医疗服务的能力，而其他一类通常称为诊所。） Unit two 1它们平均是每两位医生为一千位病人提供预防性和初级医疗服务。下一级别的为城镇医疗中心，这些中心通常主要为一万到三万位病人
提供门诊医疗服务。每一个中心有十到三十个床位，中心的专业医生为医士(助理医师）。这两个较低级别的医疗机构构成了乡村合作医疗系统，为大多数的乡村病人提供医疗服务。 2强调公共卫生和预防性治疗为主的医疗政策始于50年代早期，在那个时候，共产党开始有大批的人口流动，（动员群众）参与大规模的“爱国卫生运动”，目的在于改进较差的环境卫生，预防某些疾病。 3这一转变对农村医疗卫生产生了一系列重要影响，合作医疗缺乏资金来源，致使赤脚医生数量减少，这就意味着健康教育、初级保健和家庭保健都受挫，在一些村庄，卫生状况和饮用水(水源检测这个大家自己看哈）很少得到监控。 Unit three 1金刚烷阻碍(阻断）了由m2蛋白质形成的离子通道，这一通道在vRNPs释放进入细胞质的过程中是至关重要的。对金刚烷的排斥（耐药）迅速形成，大多数的人类传染性H1N1和H3N2病毒，一些H5N1病毒，大多数的欧洲猪H1N1、H1N2和H3N2病毒都能抵制（耐受）金刚烷。 2（然而），万一（一旦）发生流行性疾病（大流行），鸡蛋供应就会短缺，相反的是，细胞培养是可高度控制的系统，可以促使疫苗的大批量生产，包括这些对抗高致病性H5N1的禽流感病毒（可以促使包括那些针对高致病性H5N1病毒在内的疫苗大批量生产），通过以狗肾传代细胞或者非洲绿猴子肾细胞生产的流感疫苗的纯度和免疫能力
2020年10月自考00087英语翻译真题答案已公布 一、词汇翻译1. travel companion 【答案】旅伴 2. e-bank 【答案】电子银行，网上银行 3. government purchasing 【答案】政府采购 4. flower shop 【答案】花店 5. 弱势群体 【答案】disadvantaged group 二、单句翻译1. 我应集中精力，读我所未读过的书。 【翻译】I should have concentrated my energies on reading whatever books I have not yet read. 2. 古人云：“不识庐山真面目，只缘身在此山中。” 【翻译】As a line of classical poem goes, “You fail to see the true features of Lushan just because you are among the mountains yourself.” 3. A group of Italian youths waited for 45 minutes before they realized they would have to fetch their own. 【翻译】一伙意大利年轻人等了45分钟才明白他们得自己去买酒。 4. 网友提出的问题大多聚焦在民生热点问题上，***一一坦诚作答。 【翻译】Most of the questions raised by netizens focused on hot issues of livelihood, and Wen answered them frankly one by one. 【解析】①原文中有几个词语翻译较难，如“聚焦”应译为focused on，“热点问
00015英语（二）》新题型考试技巧 前言 在你打算阅读以下前，先问心自问自己有多少词汇量，没有词汇量您最好回去找本词汇书啃啃·再好的技巧，也要点词汇量，英语二这个词汇量基本上等于四级，只少不多， 但绝对高于高中的3000，而且新题型基本上是通篇阅读，什么意思呢··就是要不懂语法压根就没关系，主语 谓语那是个什么东西压根不重要，重要的是十几个单词拼在一起组成的个句子，你能翻译过来，理解他的大 概意思就OK, 关于如何背单词，还有在文中猜单词意思就先不在这块说了·因为每个人背单词的方法各不相同，感兴趣的可以私聊交流下，今天主要是把自己整理总结的一些应试技巧分享给大家， 愿这些方法能够帮助大家榜上有分··· 第一部分——阅读判断题 考试内容是根据给出的一篇短文提供的信息，对10个陈述句做出判断：Right(对)， Wrong(错)， Not Mentioned(未提及)。也就是在A、对，B、错，C、未提及，三个选项中选一个正确的答案。由于不类题目本身是不完善、不严密，存在着一定的缺陷，所以，做起来很容易模棱两可，较难把握。最麻烦的是，判断对、 错及不提及，仅就原文所给的信息做出，与生活常识、平时知识理论没有任何关系，所以容易误导我们做出错 误判断。在没有把握读懂原文的情况下，只有抓住大多数选项的规律，并掌握做题技巧，才能争取多选择正确 的选项和多拿分。由于这类题是是非题，只要确定了一个选项就行了，所以，我们只需要根据原文的信息对给 出的陈述句做出是非判断就行了。 (注意：本题有个重要规章就是，原文为大！假如有题说pigs in trees 猪会上树还会飞，千万别以常识选错，如果原文没有提到这件事应该选的是Not Mentioned(未提及)) 1、做题步骤技巧： 做题总思路：先看各选项给出的陈述句内容并找出其关键词，接着浏览原文找出与陈述句关键词相关的句 子或陈述句的出处位置，最后根据原文文意和其规律进行对比并做出是非判断。 (1)第一步：仔细考究和确定各陈述句内容的关键词。对不认识的关键词可直接查字典。 关键词的确定可根据： 一是名词(含专有名词)或名词性短语; 二是形容词、副词与其短语及其比较级; 三是数字和年代; 四是在文章中出现较少的前三项单词; 五是一目了然、容易找并有利于确定答案的单词;
医学英语翻译常用专业词汇 学英语教学担负着促使医学生从外语学习阶段向使用阶段过渡的任务,下面为大家分享医学英语翻译常用专业词汇，希望对大家有用。 医学英语翻译常用专业词汇如下： M (①meter ②molar) ①米②容积克分子的（溶液） MA (milliampere) 毫安（培） maceration 浸渍，浸软 maceration tank 浸渍槽 macerator 浸渍器 Mach 马赫（音速单位） machcardiogram (abbr. MCG) 心机械图（即对心电、心音、颈动脉波，心尖搏动图进行同步 记录，可对各图型和时相及其相互关系进行研究） machine(abbr. mach.) 机，机器，装置 machine parts 机器零件 machinery 机器，机械装置 macro- 巨，大，长，宏观 macroanalysis 常量分析 macrochemistry 常量化学
macroroentgenogram X 射线放大照片macroroentgenography X 射线放大工业照像术 made-up 人工的，制成的，预制的 madisterion 拔毛镊 madisterium 拔毛镊 madistor 晶体磁控管 magamp 磁性放大器 magazine ①杂志，期刊②箱，盒，办片匣，暗盒③软片盒magic 魔术，不可思议的 magic eye ①电眼②光调谐指示管 magic lantern 幻灯 magmeter 直读式频率计 magnalium 镁铝合金 magnesia anastomosis tube 镁氧吻合管 magnesium (abbr. Mg) 镁 magnet 吸铁器，磁铁 magnet for eye 眼用吸铁器 magnetic 磁（性）的 magnetic amplifier 磁性放大器，磁增强器 magnetic attraction 磁吸引 magnetic belt 磁性治病带 magnetic core memory 磁心存储器