CN105069000A - Interactive prediction input method - Google Patents
Interactive prediction input method Download PDFInfo
- Publication number
- CN105069000A CN105069000A CN201510524528.3A CN201510524528A CN105069000A CN 105069000 A CN105069000 A CN 105069000A CN 201510524528 A CN201510524528 A CN 201510524528A CN 105069000 A CN105069000 A CN 105069000A
- Authority
- CN
- China
- Prior art keywords
- translation
- interpreter
- input method
- fragment
- interactive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 35
- 238000013519 translation Methods 0.000 claims abstract description 145
- 230000008569 process Effects 0.000 claims abstract description 24
- 239000012634 fragment Substances 0.000 claims description 31
- 238000002474 experimental method Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 238000012935 Averaging Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 2
- YSEXMKHXIOCEJA-FVFQAYNVSA-N Nicergoline Chemical compound C([C@@H]1C[C@]2([C@H](N(C)C1)CC=1C3=C2C=CC=C3N(C)C=1)OC)OC(=O)C1=CN=CC(Br)=C1 YSEXMKHXIOCEJA-FVFQAYNVSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Landscapes
- Machine Translation (AREA)
Abstract
The invention discloses an interactive prediction input method, relates to the technical field of language translation, and solves the technical problems of delay and low practicality of existing interactive machine translation. The technical scheme is that: the method comprises: calling a machine translation interface to divide a source language input by a user into M segments; according to segment information, translating each segment and returning N optimal candidate lists; and in a formal translation process, automatically performing retrieval in M*N candidate lists and giving out translation reminding related with an original text.
Description
Technical field
The present invention relates to Language Translation technical field, particularly a kind of interactive prediction input method.
Background technology
Current, mechanical translation is ripe is applied in teaching, scientific research and commercial field.Wherein in language service industry, mechanical translation compiles in (post-edit) technology after being applied in, and first mechanical translation translate a result original text in the art, and then interpreter modifies again according to translation.Nowadays mechanical translation is trend of the times with the combination of computer aided translation system (ComputerAidedTranslation), many manufacturers, such as GoogleToolkit, SDLTrados, MemoQ, Lingotek, Matecat, yeekit etc. are using the aid of mechanical translation as raising interpreter translation efficiency.
It is be derived from interactive machine translation technology the earliest that mechanical translation combines with translation industry.Interactive machine translation (InteractiveMachineTranslation) originates from the MIND project of Kay in 1973, develops TransType and the TransType2 project built in researcher the early 21st centuries such as Langlais.Recent years, traditional retrieval model (searchingmodel) is extended to again on various model, the SCFG model of such as Gonz á lez-Rubio, the structure prediction of Alabau and the log-linear model of Huang Guoping.Can say interactive machine translation from the beginning of last century so far, a lot of researcher proposes the quality that various method goes to improve interactive machine translation, and target improves the translation efficiency of interpreter.But, prove in practice that actual interpreter is reluctant to be intended on coarse mechanical translation translation " juggling things greatly ".Trace it to its cause is exactly that the restriction of current machine translation theory and technology causes the translation of mechanical translation to depart from practical significance even " having absolutely nothing to do with each other " very greatly.Although from the angle of scientific research, researcher proves that interactive machine translation can improve the translation efficiency of interpreter, but these class methods not adopt by the professional interpreter in reality, because be proved to be same time and effort consuming, and dynamic decoder (dynamicdecoding) process very consuming time of interactive machine translation, significantly postpone sense and allow business system be reluctant so far to receive.
Describe from formula formal a little, traditional interactive machine translation (rear technique of compiling) will consider the information that interpreter has inputted, and is called prefix (prefix), uses t
prepresent, system can generate maximally related suffix (suffix) information according to these prefixes and select for interpreter's reference, uses t
srepresent.More than describe and can be expressed as following formula:
This formula is out of shape from formula (2), wherein t
pt
s=t, in first run iteration, system will produce all possible candidate translation according to the source language message, and these candidates translation constitutes huge word graph (word-graph).Interpreter's edit-modify each time all can search for useful information from these huge word graph, and recalculates suffix information.
Visible from the above description, traditional rear compiling in fact search space is very large, and when 2014, researcher Koehn recognizes this problem, and last word proposing prefix to be limited to interpreter's input calculates again.The method is proved to be and can reduces the interaction process time greatly.When 2015, yellow state equality proposes a kind of log-linear model, and develops the input method of the computer-oriented auxiliary translation system of a similar Google and search dog phonetic.But both common ground are all also undertaken alternately by continuous and mechanical translation in the input of interpreter, and constantly constantly change according to the input of interpreter the content revising prompting, though the new suggestion content of real-time generation is good, but the delay sense bringing professional interpreter will reduce its practicality.
Summary of the invention
The present invention is to solve existing interactive machine translation has delay, the technical matters that practicality is low.
In order to solve the problem, the invention provides a kind of interactive prediction input method, it is characterized in that, comprising: call mechanical translation interface and the source language that user inputs is divided into M fragment; According to burst information, each fragment is carried out translating and returns N number of best candidate list; In the process of regular translation, automatically retrieve from the candidate list of M*N, and provide the translation prompting relevant to original text.
More preferably, described is phrase fragment the fragment that source language is divided in M fragment.
More preferably, it is characterized in that, each word position of each translation is indicated, get the corresponding translation relation of word of original text and translation according to automatic aligning, obtain the phrase fragment of source language.
More preferably, described according to burst information, each fragment is carried out translating and returns N number of best candidate list; Described best candidate list is deposited in the buffer.
More preferably, described input method is only relevant with the input prefix of user, again can not call mechanical translation and remove amendment suffix candidate list; Described input prefix is the information inputted.
As seen through the above technical solutions, the invention provides a kind of interactive prediction input method, have the following advantages:
(1) in the process translated, can respond fast, not postpone sense;
(2) do not need the translation input habit changing interpreter, do not need to install, translation in need is prompting just, does not have related content just not remind, does not disturb original input method, do not produce extra calculating operation, the effective input number of times reducing interpreter;
(3) effectively prevent real-time mutual with mechanical translation in a large amount of computational problems of relating to.
Accompanying drawing explanation
A kind of interactive prediction input method schematic diagram of Fig. 1;
The effect of interactive input method in Fig. 2 actual product.
Embodiment
Below in conjunction with drawings and Examples, technical scheme of the present invention is described in detail.
It should be noted that, if do not conflicted, each feature in the embodiment of the present invention and embodiment can be combined with each other, all within protection scope of the present invention.
Embodiment one, a kind of interactive prediction input method, as shown in Figure 1 and Figure 2, comprising: call mechanical translation interface and the source language that user inputs is divided into M fragment; According to burst information, each fragment is carried out translating and returns N number of best candidate list; In the process of regular translation, automatically retrieve from the candidate list of M*N, and provide the translation prompting relevant to original text.
The present invention proposes a kind of interactive prediction input method, also claims prediction input method.Wherein, following three facts are drawn according to the feedback of interpreter:
(1) first professional interpreter can scan source text information fast, and this process can continue the time of more than two seconds, just can start afterwards to translate;
(2) most interpreter a guy is accustomed to the input method commonly used, PC has installation, and is unwilling to install too much input method;
(3) wish to provide useful prompting, reduce input, as long as and in the process of translation, The faster the better to wish prompting speed.
True (1) indicates the setup time process original text sentence of at least two seconds, during this period of time can good to preparation for effective prediction; True (2) require as much as possible prediction input method to be integrated in auxiliary translation system, instead of make an independently installation kit; True (3) illustrate technically fast for interpreter provides effective prompting, will have prompting just prompting, and not relevant prompting just can be hidden, and must not disturb its original input method.
Based on this, the present invention proposes a kind of interactive prediction input method:
According to mechanical translation alignment information, source language sentence is divided into m some fragment (segment);
Each fragment is carried out calling mechanical translation, and the n of mechanical translation best candidate result is returned;
For reducing search volume, from m × n candidate result, provide interpreter's prompting.
Method of the present invention is exactly allow interpreter not only provide input prompt but also consider the source language message when translation by mechanical translation.Interactive prediction input method no longer blindly provide a lot of useless prompting, but provide the text relevant to original text.
In order to more practical service specialty interpreter, the present invention proposes the interactive predicted method of following three steps, wherein " interactive mode " is embodied a concentrated reflection of two aspects: (1) interpreter takes the process that original text carries out thinking deeply, this process is the process that original text follows mechanical translation mutual; (2) interpreter thinks deeply complete, the process of regular translation, and this process is that the candidate list that interpreter follows mechanical translation to produce carries out mutual process, and this prompting is all relevant to the source language message.
A kind of interactive prediction input method, is specially:
(1) first call mechanical translation interface and source language is divided into M fragment.
Wherein mechanical translation is the hybrid machine translation system of specific research and development, contains alignment information in mechanical translation result, and this alignment information can indicate which part of each fragment from the source language message of the text of target translation.As shown in Figure 1, " interactive input method can accelerate the translation speed of interpreter to given source language.", source language can be divided into 6 phrase fragments according to mechanical translation result: [interactive, input method, can accelerate, interpreter's, translation, speed.] this point-score is different from traditional participle, participle traditionally, source language will be divided into 9 similar result fragments: [interactive, input method, passable, accelerate, interpreter, translation, speed.]。
Wherein, the process that above-mentioned 6 phrase fragments produce is as follows: first system indicates each word position of each translation, and program adopts src-start, src-end, tgt-start and tgt-end as identifier.Then the corresponding translation relation of word of original text and translation is got by the MGIZA automatic aligning of Gao Qin.Finally obtain the fragment of original text.Wherein the meaning of identifier is as follows:
I.src-start: the beginning word position after source language participle, from 0
Ii.src-end: the position that source language phrase terminates, comprises this word
Iii.tgt-start: in the target language after traditional participle, the position that phrase starts, from 0
Iv.tgt-end: the phrase end position in the target language after traditional participle, comprises this word
As shown in Fig. 1, wherein the mechanical translation result of source language is: " Interactiveinputmethodcanspeeduptheinterpreter'stranslat ionspeed. " is according to this translation and alignment information, original text can be split into 6 fragments, wherein each fragment can provide alignment information, in the description of table 1.For fragment 3, alignment information indicates, from the word position 2 of source language to word position 3, snap to the word position 3 of target translation to word position 6, namely source language word " can " " quickening " snap to word " can " " speed " " up " " the ".Finally obtain 6 fragments of source language by that analogy: [interactive, input method, can accelerate, interpreter's, translation, speed.]。
Table 1: the alignment information that mechanical translation provides
(2) according to the burst information of previous step, again call mechanical translation, each fragment is carried out translating and returns N number of best candidate list.
Consider the quality of candidate list in reality, during system realizes, general employing is no more than the method for five candidate lists.When programming realization, the translation that candidate list obtains, deposit in the buffer, object, when treating that next step interpreter translates, can meet with a response fast.
(3) in the process of interpreter's regular translation, utilize the method for similar Autocomplete to retrieve from the candidate list of M*N, and provide most suitable and relevant to original text translation and remind, reduce the input of interpreter.
In practice, in auxiliary translation system now, much source language is carried out punctuate operation, namely a chapter is split into sentence and translate for interpreter.And the word limited amount of each sentence, general language is no more than 200 words (Chinese calculates with single character).Compared with traditional interactive approach, last interpreter goes retrieval from limited prompting translation, and method of the present invention is only relevant with the input prefix of user, again can not call mechanical translation and remove amendment suffix candidate list.Interpreter can not be caused to have in the process of input and to postpone sense.
Fig. 1 gives the example of above-mentioned three steps, wherein in step 1, system call mechanical translation, draws the burst text of source language according to the information of translation, in second step, system can call mechanical translation again each block of burst information, draws the mechanical translation candidate list of every a slice, and is stored in the buffer memory of client, so far with the terminating alternately of mechanical translation, last interpreter reads candidate list from buffer memory, therefrom provides the prompting of interpreter's the best, actual product as shown in Figure 2.
Fig. 2 is explained as follows:
Interpreter opens web site url: http://www.ourime.cn/
Text to be translated is inputted: interactive input method can accelerate the translation speed of interpreter in input frame.
Interpreter translates in output box.If there is prompting to provide prompting, selected by the key up and down of keyboard, carriage return can choose good prompting.
In order to verify the effect of the input method model that the present invention proposes, carried out the experiment test in following two language directions, translator of Chinese becomes English (zh-en) to become English (fr-en) with French Translator.Wherein Chinese source language is from first 500 (can reference: http://www.ai-ia.ac.cn/cwmt2015/evaluation.html) in the news testing material 1000 of CWMT2015, and French is from first 500 (can download in this: http://statmt.org/wmt14/translation-task.html) in the news test set in WMT2014 3000.Wherein this text message of 500 is as described in Table 2:
| Language to be translated | Sentence number | Mean sentence length | Total words |
| Chinese | 500 | 38.81 | 19,405 |
| French | 500 | 17.99 | 8,999 |
Table 2: text message to be translated
This time test, adopts particular task translate duration to be unit of account, i.e. 500 average translate durations of text interpreter.Find out 12 translation interpreters with more than 3 years from translation company and carry out this time test, wherein 6 interpreters carry out Chinese-English translation, and 6 interpreters carry out the translation of French to English.In order to ensure the translation efficiency of professional interpreter, the translate duration limiting each interpreter is two hours every days, in wherein one hour morning, in one hour afternoon, within each hour, is split into half an hour again, interval 20 minutes.
Can does mechanical translation adopt the Moses system of increasing income in the world (in http://www.statmt.org/moses/index.php? n=Main.HomePage downloads), wherein Sino-British training corpus is from the training set of CWMT2015, and the training set of method English is from the Europarl200 Wan Faying data in WMT.Training data does not comprise text to be translated.
Experiment will be divided into two parts, wherein tests in (1), do not comprise text to be translated, and interpreter is unfamiliar with text to be translated in the training result of mechanical translation.In experiment (2), after all interpreter's cypher texts, the sentence that translated text selection has unanimously been thought is carried out mechanical translation re-training, and namely mechanical translation comprises text sentence to be translated.The object of experiment (1) translates when being all unfamiliar with original text, and even mechanical translation can not provide the effect that useful information carries out testing prediction input method; The object of experiment (2) is when known machine translation can carry out effective prompting to translation, tests oneself input completely and has the effect pointing out input.
There is the system A (system B interface is similar, just have disabled prediction prompting) of prompt facility
For this reason, the basis of http://www.ourime.cn/ is developed the system of two webpage versions, and one of them has the input prompt carried, i.e. system A, and one containing prompt facility, i.e. system B.500 texts to be translated are placed on server end, and object ensures that interpreter cannot see other sentences and operate translation in advance by the time advance beyond translation.And automatically providing time countdown, the time is defined as 30 minutes, and the time one arrives, and namely stops translation.System will record the sentence number of interpreter's translation in every 30 minutes automatically, and can record the time of translation every words.Each language centering, three interpreters operate in A system, and other three interpreters operate translation in B system, and six interpreters independently translate 500 all texts.For experiment (1), through the times of 22 days (French Translator 13 days), all translate ends, draw the result of table 3 by system measuring and calculating:
| Interpreter | System A translate duration (minute) | System B translate duration (minute) |
| Interpreter 1 | 2688 | - |
| Interpreter 2 | 2442 | - |
| Interpreter 3 | 2611 | - |
| Interpreter 4 | - | 2734 |
| Interpreter 5 | - | 2797 |
| Interpreter 6 | - | 2683 |
| Averaging time | 2580 | 2738 |
Table 3: experiment (1) Chinese 500 translate duration contrasts
| Interpreter | System A translate duration (minute) | System B translate duration (minute) |
| Interpreter 7 | 1207 | - |
| Interpreter 8 | 1223 | - |
| Interpreter 9 | 1169 | - |
| Interpreter 10 | - | 1401 |
| Interpreter 11 | - | 1233 |
| Interpreter 12 | - | 1295 |
| Averaging time | 1199 | 1309 |
Table 4: experiment (1) French 500 translate duration contrasts
As can be seen from Table 3, in the system A providing automatic-prompting, the translation of 500 has been carried out in interpreter's mean consumption for 2580 minutes, and in the translation process not providing automatic-prompting, employ the time of 2738 minutes.Improve the translation efficiency of 6.12%.Observed by table 4, the efficiency of 9.11% can be improved to usage forecastings input method in the translation of French.
By exchanging with interpreter, learn in the translation of Chinese, in given 500, the professional term in a lot of sentence, mechanical translation does not provide effective prompting, but the effect pointed out in French is with regard to showed increased.Finally know by analyzing, although Sino-British training corpus is also from News Field, but have a lot different from given translation of the sentence, namely unregistered word is many, that such as " souvenir middle school, middle mountain " mechanical translation provides is " ZhongShanMemorialMiddleSchool ", and in fact interpreter want adopt be " SunYat-senMemorialMiddleSchool ".Like this according to algorithm set forth above, interpreter oneself can only input " SunYat-sen ", result in the increase of translate duration.So will the fact of larger corpus be had for the mechanical translation in actual conditions, 500 texts of interpreter's translation are trained as training corpus again.Then, 12 interpreters translate again in new Machine Translation Model.Same, interpreter still translates on system A, B.Table 5 and table 6 provide the time contrast of this translation respectively.
| Interpreter | System A translate duration (minute) | System B translate duration (minute) |
| Interpreter 1 | 1942 | - |
| Interpreter 2 | 2033 | - |
| Interpreter 3 | 1844 | - |
| Interpreter 4 | - | 2167 |
| Interpreter 5 | - | 2366 |
| Interpreter 6 | - | 2254 |
| Averaging time | 1939 | 2262 |
Table 5: experiment (2) Chinese 500 translate duration contrasts
| Interpreter | System A translate duration (minute) | System B translate duration (minute) |
| Interpreter 7 | 1083 | - |
| Interpreter 8 | 986 | - |
| Interpreter 9 | 1092 | - |
| Interpreter 10 | - | 1285 |
| Interpreter 11 | - | 1109 |
| Interpreter 12 | - | 1137 |
| Averaging time | 1053 | 1177 |
Table 6: experiment (2) French 500 translate duration contrasts
By merge translation of the sentence in translation system, change clearly can be seen, in same translator of Chinese, the average translation efficiency of interpreter with prediction input method improves 16.65% (being reduced to 1939 minutes from 2262 minutes), and in the translation of French, translation efficiency improves 11.71% (being reduced to 1053 minutes from 1177 minutes).
Propose a kind of interactive mode prediction input method of Corpus--based Method mechanical translation herein, the method takes into full account the source language message, when interpreter translates input, can provide the maximally related prompting with original text.The method is divided into three steps: (1), by mechanical translation alignment information, carries out a point fragment source language; (2) again call mechanical translation, each fragment is carried out mechanical translation, and return some candidate lists; (3) interpreter is after to source language analysis, in the process of translation, will search for candidate result, and provide maximally related prompting, reduces and beats keyboard number of times.During at 500, translation of the sentence is tested on a small scale, by interactive mode prediction input method, translator of Chinese becomes English can improve maximum 16.65%, and French Translator becomes in English, can improve 11.71% at most.
Prove in true Practice of Translation, the method model also proves the translation efficiency that can improve professional interpreter.The method is applied in the practice of Liang Jia translation company, from the feedback of interpreter, this input method produces a desired effect: in the process that (1) translates, can respond fast, does not postpone sense; (2) do not need the translation input habit changing interpreter, do not need to install, translation in need is prompting just, does not have related content just not remind, does not disturb original input method.
From the angle of scientific research, a kind of information making full use of mechanical translation is proposed to promote the method for interpreter's translation efficiency herein, this method solve the matter of time of " interactive mode ", researcher proposes before effectively prevent, also want in the process of interpreter's input real-time mutual with mechanical translation in a large amount of computational problems of relating to.From the angle of practice, " interactive input method " is dissolved in the translation process of interpreter, only provides effective prompting function, does not produce extra calculating operation, the effective input number of times reducing interpreter.
Certainly; the present invention also can have other various embodiments; when not deviating from the present invention's spirit and essence thereof; those of ordinary skill in the art are when making various corresponding change and distortion according to the present invention, but these change accordingly and are out of shape the protection domain that all should belong to claim of the present invention.
Claims (5)
1. an interactive prediction input method, is characterized in that, comprising: call mechanical translation interface and the source language that user inputs is divided into M fragment; According to burst information, each fragment is carried out translating and returns N number of best candidate list; In the process of regular translation, automatically retrieve from the candidate list of M*N, and provide the translation prompting relevant to original text.
2. input method as claimed in claim 1, it is characterized in that, described is phrase fragment the fragment that source language is divided in M fragment.
3. input method as claimed in claim 2, is characterized in that, each word position of each translation is indicated, and gets the corresponding translation relation of word of original text and translation, obtain the phrase fragment of source language according to automatic aligning.
4. input method as claimed in claim 1, is characterized in that, described according to burst information, each fragment is carried out translating and returns N number of best candidate list; Described best candidate list is deposited in the buffer.
5. input method as claimed in claim 1, it is characterized in that, described input method is only relevant with the input prefix of user, again can not call mechanical translation and remove amendment suffix candidate list; Described input prefix is the information inputted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510524528.3A CN105069000A (en) | 2015-08-24 | 2015-08-24 | Interactive prediction input method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510524528.3A CN105069000A (en) | 2015-08-24 | 2015-08-24 | Interactive prediction input method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105069000A true CN105069000A (en) | 2015-11-18 |
Family
ID=54498374
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510524528.3A Pending CN105069000A (en) | 2015-08-24 | 2015-08-24 | Interactive prediction input method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105069000A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401000A (en) * | 2020-04-03 | 2020-07-10 | 上海一者信息科技有限公司 | Translation real-time preview method for online auxiliary translation |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0798652A3 (en) * | 1996-03-27 | 2002-03-27 | Hitachi Business International, Ltd. | Document composition supporting method and system, and electronic dictionary for terminology |
| JP2003016066A (en) * | 2001-06-29 | 2003-01-17 | Fujitsu Ltd | Translation support device, translation support program, and recording medium |
| JP2005284912A (en) * | 2004-03-30 | 2005-10-13 | Nec Corp | Text and its translation sentence alignment editor system, method and program |
| CN101034407A (en) * | 2007-03-30 | 2007-09-12 | 传神联合（北京）信息技术有限公司 | Quick input method based on scene lexicon in process of translation |
| JP2008027027A (en) * | 2006-07-19 | 2008-02-07 | Toshiba Corp | Machine translation apparatus and machine translation program |
| US7346488B2 (en) * | 2000-07-10 | 2008-03-18 | Fujitsu Limited | Automatic translator and computer-readable storage medium having automatic translation program recorded thereon |
| CN102662490A (en) * | 2012-04-08 | 2012-09-12 | 北京语智云帆科技有限公司 | Inputting method having assistant translation function |
| CN104462072A (en) * | 2014-11-21 | 2015-03-25 | 中国科学院自动化研究所 | Input method and device oriented at computer-assisting translation |
| JP2015082204A (en) * | 2013-10-22 | 2015-04-27 | 富士通株式会社 | Machine translation device, machine translation program, and machine translation method |
-
2015
- 2015-08-24 CN CN201510524528.3A patent/CN105069000A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0798652A3 (en) * | 1996-03-27 | 2002-03-27 | Hitachi Business International, Ltd. | Document composition supporting method and system, and electronic dictionary for terminology |
| US7346488B2 (en) * | 2000-07-10 | 2008-03-18 | Fujitsu Limited | Automatic translator and computer-readable storage medium having automatic translation program recorded thereon |
| JP2003016066A (en) * | 2001-06-29 | 2003-01-17 | Fujitsu Ltd | Translation support device, translation support program, and recording medium |
| JP2005284912A (en) * | 2004-03-30 | 2005-10-13 | Nec Corp | Text and its translation sentence alignment editor system, method and program |
| JP2008027027A (en) * | 2006-07-19 | 2008-02-07 | Toshiba Corp | Machine translation apparatus and machine translation program |
| CN101034407A (en) * | 2007-03-30 | 2007-09-12 | 传神联合（北京）信息技术有限公司 | Quick input method based on scene lexicon in process of translation |
| CN102662490A (en) * | 2012-04-08 | 2012-09-12 | 北京语智云帆科技有限公司 | Inputting method having assistant translation function |
| JP2015082204A (en) * | 2013-10-22 | 2015-04-27 | 富士通株式会社 | Machine translation device, machine translation program, and machine translation method |
| CN104462072A (en) * | 2014-11-21 | 2015-03-25 | 中国科学院自动化研究所 | Input method and device oriented at computer-assisting translation |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401000A (en) * | 2020-04-03 | 2020-07-10 | 上海一者信息科技有限公司 | Translation real-time preview method for online auxiliary translation |
| CN111401000B (en) * | 2020-04-03 | 2023-06-20 | 上海一者信息科技有限公司 | Real-time translation previewing method for online auxiliary translation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9767095B2 (en) | Apparatus, system, and method for computer aided translation | |
| CN104461484B (en) | The implementation method and device of front-end template | |
| WO2018176758A1 (en) | Method and device for generating article | |
| CN110046261A (en) | A kind of construction method of the multi-modal bilingual teaching mode of architectural engineering | |
| Way et al. | On the Role of Translations in State‐of‐the‐Art Statistical Machine Translation | |
| Yan et al. | Studies on translator and interpreter training: A data-driven review of journal articles 2000–12 | |
| CN104615589A (en) | Named-entity recognition model training method and named-entity recognition method and device | |
| CN103246732A (en) | Online Web news content extracting method and system | |
| CN112836525A (en) | Human-computer interaction based machine translation system and automatic optimization method thereof | |
| Chang et al. | Dart: A lightweight quality-suggestive data-to-text annotation tool | |
| CN105740370B (en) | A kind of online Web news contents extraction system | |
| Sharifpour et al. | An investigation of retranslation hypothesis in English translations of Rūmī’s poetry | |
| CN105069000A (en) | Interactive prediction input method | |
| Valdeón | From translatology to studies in translation theory and practice | |
| Wang | Translation and the relocation of global cultures: mainly a Chinese perspective | |
| US9311302B2 (en) | Method, system and medium for character conversion between different regional versions of a language especially between simplified chinese and traditional chinese | |
| Guo et al. | Character-level dependency model for joint word segmentation, POS tagging, and dependency parsing in Chinese | |
| Zhou et al. | Domain adaptation for SMT using sentence weight | |
| CN104317505A (en) | Pinyin outputting system and method | |
| Salunkhe et al. | A research work on English to Marathi hybrid translation system | |
| CN114154092A (en) | Method for translating web pages and related product | |
| CN108959555B (en) | Query type expansion method and device, computer equipment and storage medium | |
| Han | Statistics in Corpus Linguistics: A Practical Guide: by Valclav Brezina, Cambridge, Cambridge University Press, 2018, xix+ 296 pp., ISBN: 978-1107565241 | |
| Tanasescu | Using computers in the translation of literary style. Challenges and opportunities: by Roy Youdale, London and New York, Routledge, 2020, 242 pp.,£ 115.00/£ 36.89 (hardback/e-book), ISBN 9780367131233/9780429030345 | |
| Bellandi et al. | When translation requires interpretation: Collaborative computer–assisted translation of ancient texts |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 100040 Shijingshan Road, Shijingshan District, Beijing, No. 20, 16 layer 1601 Applicant after: Chinese translation language through Polytron Technologies Inc Address before: 100040 Shijingshan Road, Shijingshan District, Beijing, No. 20, 16 layer 1601 Applicant before: Mandarin Technology (Beijing) Co., Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20151118 |
|
| RJ01 | Rejection of invention patent application after publication |