CN111507113B - Method and device for machine-assisted manual translation - Google Patents
Method and device for machine-assisted manual translation Download PDFInfo
- Publication number
- CN111507113B CN111507113B CN202010193403.8A CN202010193403A CN111507113B CN 111507113 B CN111507113 B CN 111507113B CN 202010193403 A CN202010193403 A CN 202010193403A CN 111507113 B CN111507113 B CN 111507113B
- Authority
- CN
- China
- Prior art keywords
- translation
- input
- word
- determining
- source text
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/42—Data-driven translation
- G06F40/47—Machine-assisted translation, e.g. using translation memory
Abstract
The invention provides a method and a device for machine-assisted manual translation, and relates to the technical field of translation. Wherein the method comprises the following steps: acquiring a source text of a source language and acquiring an input translation of a target language; determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language; and displaying the candidate words. The candidate word is displayed to the translator, so that the prompting effect can be achieved, and the translator can conveniently understand and translate the uncommon word.
Description
Technical Field
The invention relates to the technical field of translation, in particular to a method and a device for machine-assisted manual translation.
Background
At present, along with the globalization, the communication of politics, economy, culture and the like of each country is more and more increased, and along with the communication and learning of languages are more and more frequent.
In the communication and learning of languages, manual translation of language words is often required. However, when the translation is performed manually, the problems that some obscure words are difficult to understand, a proper translation word cannot be found for a while, and the translation is jammed, and the problem of incomplete semantic expression due to careless translation are easily caused, so that the translation speed is slow, and the translation quality is poor are easily caused.
Disclosure of Invention
In view of the above, the present invention has been developed to provide a method and apparatus for machine-assisted manual translation that overcomes, or at least partially solves, the above-mentioned problems.
According to a first aspect of the present invention, there is provided a method of machine-assisted manual translation, the method comprising:
acquiring a source text of a source language, and determining a semantic space according to the source text;
acquiring an input translation of a target language;
determining a candidate word of a next input word corresponding to the input translation according to the semantic space and the input translation and based on the logic habit of the target language;
and displaying the candidate words.
Optionally, in the method, the determining, according to the semantic space and the input translation and based on the logic habit of the target language, a candidate word of a next input word corresponding to the input translation includes:
calculating first conditional probabilities of the translated words aiming at the target language in the semantic space after the input translated text according to the input translated text and the corpus;
ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word;
and taking the plurality of translated words with the ranking values smaller than or equal to a preset value as the candidate words.
Optionally, in the method, the displaying the candidate word includes:
determining a first score of each candidate word according to the first conditional probability;
and displaying each candidate word according to the first score from high to low, and correspondingly displaying the first score of each candidate word.
Optionally, the method further includes:
the method further comprises the following steps:
determining a second conditional probability of the input translation according to the source text, the input translation and the corpus;
determining a second score of the input translation according to the second conditional probability;
and displaying the second score.
Optionally, the method further includes:
when a translation completion signal is detected, determining a reference translation according to the source text and the target language, and displaying the reference translation; the reference translation is a translation for which the second conditional probability reaches a maximum value among the translations of the source text.
According to a second aspect of the present invention, there is provided an apparatus for machine-assisted manual translation, the apparatus comprising:
the acquisition module is used for acquiring a source text of a source language and acquiring an input translation of a target language;
the determining module is used for determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language;
and the display module is used for displaying the candidate words.
Optionally, in the apparatus, the determining module includes:
the first determining unit is used for determining a semantic space according to the source text;
a calculating unit, configured to calculate, according to the input translation and the corpus, a first conditional probability that each translated word in the semantic space for the target language appears after the input translation;
the ranking unit is used for ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word;
and the second determining unit is used for taking the plurality of translated words with the ranking values smaller than or equal to a preset value as the candidate words.
Optionally, in the apparatus, the display module includes:
the first scoring unit is used for determining a first score of each candidate word according to the first conditional probability;
and the first display unit is used for displaying each candidate word from high to low according to the first score and correspondingly displaying the first score of each candidate word.
Optionally, the apparatus further comprises:
and the scoring module is used for determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language and displaying the second score.
Optionally, the scoring module includes:
the second scoring unit is used for determining a second score of the input translation according to the second condition probability;
and the second display unit is used for displaying the second scores.
Optionally, the apparatus further comprises:
the reference translation module is used for determining a reference translation according to the source text and the target language and displaying the reference translation when a translation completion signal is detected; the reference translation is a translation for which the second conditional probability for the source text is maximal.
According to the method and the device for machine-assisted manual translation, when manual translation is carried out, a candidate word which is behind the input translation and matches with the logic habit of the target language is determined according to the semantic space of the source text and the input translation of the target language, and the candidate word is displayed to a translator, so that a prompt effect can be achieved, the translator can conveniently understand and translate the rarely-used word, and the problems that when manual translation is carried out, part of rarely-used words are difficult to understand, and a proper translated word cannot be found at a moment, so that translation jamming occurs are solved.
The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
FIG. 1 is a flow chart illustrating the steps of a method for machine-assisted manual translation according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a process of translating a "A" from "ABCD" by a neural network machine provided by an embodiment of the present invention;
FIG. 3 is a schematic diagram of a process of translating "B" from "ABCD" for a neural network machine provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of a process of translating "C" from "ABCD" for a neural network machine provided by an embodiment of the present invention;
FIG. 5 is a diagram illustrating a process of translating "D" from "ABCD" in a neural network machine provided by an embodiment of the present invention;
FIG. 6 is a schematic diagram of a process of translating "ABCD" by ending machine translation in a neural network according to an embodiment of the present invention;
FIG. 7 is a diagram illustrating candidate words according to an embodiment of the present invention;
FIG. 8 is a flowchart illustrating method steps for machine-assisted manual translation according to an embodiment of the present invention;
fig. 9 is a block diagram of an apparatus for machine-assisted manual translation according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Fig. 1 is a flowchart illustrating steps of a method for machine-assisted manual translation according to an embodiment of the present invention, as shown in fig. 1, the method may include steps S100 to S300:
step 100, obtaining a source text of a source language and obtaining an input translation of a target language.
In step S100, the source text is a sentence to be translated and can be directly obtained through a translation interface. The input translation is a translation text currently input by a translator, the input translation can be obtained through an input box manually translated by the translator, and the input translation can be a text directly input by a character input mode or a text generated by voice input and character conversion. In practical application, after the translator inputs a word of a target language, the current input word is acquired, so that unnecessary acquisition of input translation is reduced.
The input translation may be null, that is, when the translator does not input any translation text, the subsequent steps S200 and S300 may still be performed, that is, the steps of candidate generation and candidate display may still be performed.
And S200, determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language.
In the step S200, that is, according to the source text obtained in the step S100 and in combination with the logic habit of the target language, each possible translation text can be determined; the logic habit of the target language can be determined by accessing the corpus; meanwhile, by using the input translation acquired in step S100, a target translation text currently selected by the translator may be determined from the above possible translation texts; when the input translation is compared with the target translation text, that is, the corresponding part of the input translation can be determined in the target translation text, so that the next translation word of the corresponding part can be determined, and the next translation word is used as a candidate word.
When the input translation is empty, determining a translation text which is more in line with the logic habit of the target language in each possible translation text as the target translation text, namely, using a plurality of translation texts with higher frequency as the target translation text. Accordingly, the candidate word is the first word of the target reaction text.
And step S300, displaying the candidate words.
In the step S300, the candidate word determined in the step S200 is displayed, that is, a translation reference may be provided to the translator in the form of a candidate word for a first possible translated word of the text content that has not been translated yet, so as to prompt the translator to translate in a word-by-word or word-by-word manner. Specifically, the candidate words may be displayed in a voice broadcast mode or a text display mode. When the candidate word is displayed in a text display mode, the candidate word is displayed in an input box or after a translation is input, so that a translator can conveniently check the candidate word.
In summary, according to the method for machine-assisted manual translation provided by the embodiment of the present invention, when performing manual translation, according to the input translation of the source text and the target language, the candidate word which is after the input translation and matches the logic habit of the target language is determined, and is displayed to the translator, so that a prompt effect can be achieved, and the translator can understand and translate the obscure word conveniently, thereby solving the problems that some obscure words are difficult to understand, and a proper translated word cannot be found at a moment and a translation jam occurs during manual translation.
Optionally, in an embodiment, the step S200 includes steps S201 to S204:
step S201, determining a semantic space according to the source text.
In step S201, the semantic space extracts information of the source text, and replaces an original sentence with an abstract semantic. In practical application, the semantic space of the source text can be determined by calling neural network machine translation and the like, namely, when the source text is provided for manual translation, the source text is converted into a vectorized semantic space by using the neural network machine translation.
Neural network machine translation is a neural network-based translation model that employs an encoding-decoding framework, where the encoding process and the decoding process each include one or more layers of neural networks. The coding process is to extract the information of the original sentence and use an abstract semantic meaning to replace the original sentence; the decoding process is to convert the abstract semantics into the sentences of the target language, so that the generated sentences can perfectly express the meaning of the original sentences and accord with the logic habit of the target language.
Step S202, calculating a first conditional probability of each translated word aiming at the target language in the semantic space after the input translated text according to the input translated text and the corpus.
In step S202, the corpus includes a bilingual corpus of the source language and the target language. The method comprises the steps of firstly determining each translation word which corresponds to a source text and is in the target language by combining a semantic space with a corpus, and meanwhile, counting the probability of each translation word determined after the input translation by using the corpus, namely the first conditional probability, wherein the first conditional probability reflects the frequency degree of each translation word combined after the input translation in the use logic habit of the target language, namely the coincidence degree of the translation word and the logic habit of the target language.
Step S203, ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word.
In step S203, the possible translated words of the target language corresponding to the source text are ranked according to the frequency of the possible translated words appearing immediately after the input translated text, so as to determine the possibility that the possible translated words appear immediately after the input translated text.
And S204, taking the plurality of translated words with the ranking values smaller than or equal to a preset value as the candidate words.
In the step S204, the more posterior translation word is ranked, the lower the probability that the translation word appears after the translated text is input is, that is, the less the semantic logic and the usage habit of the target language are met, and the more unlikely the translation word is the target translation word; if all the translated words are used as candidate words, not only too many display pictures are occupied, but also the candidate words are not convenient for the translator to observe, so that the translated words with the ranking values smaller than a preset value are set as the candidate words, that is, the translated words with the preset value number which appears after the input translated text and has high frequency are set as the candidate words, and the preset value can be set and adjusted according to the habit or the need of the translator, and can be 3, for example.
In practical application, considering that each operation is to translate a next word or phrase based on an existing translation when the neural network machine translates a source language character into a target language character, the neural network machine translation may be invoked to perform the steps S201 to S204, that is, the neural network machine translation is used to determine a candidate word corresponding to a next input word of the input translation, specifically, the input translation is used as an input of the neural network machine translation, and the candidate word is obtained through the neural network machine translation, thereby achieving an effect of assisting manual translation. Therefore, the method for machine-assisted manual translation provided by the embodiment of the invention is embedded into a machine translation system for practical application through software development.
Referring to fig. 2 to 6, taking the translation text "ethyl propyl butyl" of the source text "ABCD" into the target language as an example, the process of neural network machine translation is shown:
in fig. 2, firstly, the ABCD is processed by an N-Layer neural network (Layer) in an encoding model (ENCODER) to obtain a semantic space, and then the content of the semantic space is input into a decoding model (DECODER); meanwhile, the DECODER takes a translation start identifier "< SOS >" as an initial input, combines the semantic space and carries out the operation of an N-layer neural network in the DECODER, and the output is 'A'
In fig. 3, the obtained "a" and "< SOS >" in fig. 2 are spliced together to obtain "< SOS > a" as an input, and the latest output "b" is obtained by combining the semantic space and performing DECODER operation;
in fig. 4, "b" and "< SOS > a" obtained in fig. 3 are spliced together to obtain "< SOS > a" and "b" as input, and a latest output "c" is obtained by combining the semantic space and performing DECODER operation;
in fig. 5, the "c" and the "< SOS > ethyl-propylene" obtained in fig. 4 are spliced together to obtain "< SOS > ethyl-propylene" as an input, and a latest output "d" is obtained by combining the semantic space and performing DECODER operation;
in fig. 6, "d" obtained in fig. 5 and "< SOS > ethyl-propylene" are spliced together to obtain "< SOS > ethyl-propylene-d" as an input, and the latest translation end identifier output "< EOS >" is obtained by combining the semantic space and performing DECODER operation, so that a final translation result "< SOS > ethyl-propylene-d < EOS >" is obtained, that is, the translation of "ABCD" into "ethyl-propylene-d".
According to the translation process of the neural network machine translation, each calculation of the decoding model is based on the translated text content, and the next word or word of the translated text content is determined by combining the semantic space of the source text. And determining the next word or word with the highest probability after the existing translated text content in each translated word corresponding to the semantic space by calling the semantic library, outputting the translated word decoding model with the highest probability, splicing the output with the original translated text content, taking the spliced new translated text content as a new input, and continuing the decoding operation to determine the next word or word corresponding to the new translated text content.
The embodiment of the invention just needs to obtain the candidate word of the next input word of the input translation, so the input translation can be used as the input of the neural network machine translation, and the probability of each translation word corresponding to the source text appearing after the input translation is calculated through the neural network machine translation, thereby obtaining a plurality of candidate words with higher probability. For example, when "ABCD" is translated manually, assuming that "ab" has been translated, neural network machine translation may be invoked, and a candidate word is obtained through machine translation using "< SOS > ab" as a machine translation input.
Optionally, in an embodiment, the step S300 includes: and if the input translation is not updated within the preset time, displaying the candidate words.
In this embodiment, if it is detected that the input translation is not updated within the preset time, that is, the input translation is not changed within the preset time, it indicates that the translator is likely to encounter difficulty in translation, and at this time, the translator can be prompted with more pertinence for displaying the candidate words; if the input translation is updated in bathroom time, the translator has no difficulty, so that translation prompt is not needed, and the candidate words can be set not to be displayed.
Optionally, in an embodiment, the step S300 includes steps S301 to S302:
step S301, determining a first score of each candidate word according to the first conditional probability.
In step S301, the larger the first conditional probability is, the more the combination of the corresponding translated word and the input translated text conforms to the language logic and usage habit of the target language, so that each candidate word can be scored according to the first conditional probability to obtain a corresponding first score. And setting the candidate words with higher first conditional probability, wherein the corresponding first scores are also higher.
In step S301, the first conditional probability of each translated word may also be obtained by invoking neural network machine translation, and each translated word is scored according to the first conditional probability to determine a corresponding first score.
And S302, displaying each candidate word according to the first score from high to low, and correspondingly displaying the first score of each candidate word.
In the step S302, the candidate words determined in the step S200 are displayed to the translator according to the first score from high to low, and the corresponding first scores are displayed, so that the translator knows which translated words can be used as the next input word, and the logical habit fit degree between each translated word and the currently input translated text can be visually compared, so that the translator can obtain the best candidate word.
For example, when "ABCD" is translated manually, assuming that "ab" has been translated, then neural network machine translation may be invoked, and "< SOS > ab" is used as a machine translation input, candidate words are obtained through machine translation, each candidate word is scored, then sorted from high to low according to the score, and displayed to the translator for reference, as shown in fig. 7, a better candidate word prompt effect may be achieved.
Optionally, in an implementation manner, the method for machine-assisted manual translation provided by the embodiment of the present invention further includes step S400:
step S400: and determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language, and displaying the second score.
In the step S400, the input translation translated from the source text is scored according to the logic habit of the target language, so as to reflect the accuracy and reasonableness of the input translation translated by the translator.
Optionally, the step S400 may include steps S401 to S403:
step S401, determining a second conditional probability of the input translation according to the source text, the input translation and the corpus.
In step S401, the size of the possibility of translating the source text according to the input translation is statistically determined through the corpus, and the size of the possibility determines the reasonable degree of translating the source text according to the input translation.
In practical applications, the second conditional probability can be calculated according to the formula p (y | x), where x represents the source text and y represents the input text.
And S402, determining a second score of the input translation according to the second conditional probability.
In step S402, the larger the second conditional probability, the more the language logic and usage habit of the target language are satisfied as the source text is translated according to the input translation, so that the input translation can be scored according to the probability to obtain a corresponding second score. Wherein, the input translation with higher probability of setting the second condition is also higher corresponding to the second score
And S403, displaying the second score.
In step S403, the translation result is reported to the translator in the form of score, so that the translator knows the reasonable situation of the translated sentence obtained by translation.
In practical application, the second conditional probability of the input translation for the source text can be calculated by calling neural network machine translation, and the second score can be obtained by outputting the second conditional probability in a scoring form as a result.
In the embodiment, by scoring the input translation, the translator can reflect the accuracy and the reasonable degree of the input translation translated by the translator.
Optionally, in an implementation manner, the method for assisting manual translation provided by the embodiment of the present invention further includes step S500:
step S500, when a translation completion signal is detected, determining a reference translation according to the source text and the target language, and displaying the reference translation; the reference translation is a translation for which the second conditional probability for the source text is maximal.
In the step S500, after the translator has translated a sentence of source text or a piece of source text, the translator is presented with the reference translation to provide the translator with the reference for error correction. The reference translation is a translation which has the maximum second conditional probability for the source text, that is, a translation which best meets the logic habit of the target language, and thus the reference translation is used as a reasonable translation result and is shown to the translator in the form of the reference translation.
In practical applications, the above-mentioned reference translation may also be obtained by invoking a neural network machine translation to obtain y that maximizes p (y | x), where x represents the source text and y represents the translation of the target language. That is to say, the
Y of (a) as the above reference translation, providing an error correction reference.
Referring to fig. 8, a flowchart of a method for machine-assisted manual translation according to another embodiment of the present invention is shown. The method for machine-assisted manual translation according to another embodiment of the present invention, as shown in fig. 8, includes steps S801 to S805:
step 801, acquiring a source text of a source language and acquiring an input translation of a target language.
The above step S801 can refer to the detailed description of step S100, and is not repeated here.
And S802, determining a semantic space according to the source text.
The above step S802 can refer to the detailed description of step S201, and is not repeated here.
Step S803, according to the input translation and the corpus, calculating a first conditional probability that each translated word in the semantic space for the target language appears after the input translation.
The above step S803 can refer to the detailed description of step S202, and is not repeated here.
Step S804, ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word.
The above step S804 can refer to the detailed description of step S203, which is not repeated herein.
Step S805, using the translation words with the ranking values smaller than or equal to a preset value as the candidate words.
The above step S805 can refer to the detailed description of step S204, and is not described herein again.
Step S806, determining a first score of each candidate word according to the first conditional probability.
The above step S806 can refer to the detailed description of step S301, and is not repeated here.
And step S807, displaying each candidate word according to the first score from high to low, and correspondingly displaying the first score of each candidate word.
The above step S807 can refer to the detailed description of step S302, and is not repeated herein.
Step S808, determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language, and displaying the second score.
The above step S808 can refer to the detailed description of step S400, and is not repeated herein.
Step S809, when a translation completion signal is detected, determining a reference translation according to the source text and the target language, and displaying the reference translation; the reference translation is a translation for which the second conditional probability for the source text is maximal.
The step S809 refers to the detailed description of the step S500, and is not repeated here.
In summary, in the method for machine-assisted manual translation provided by the embodiment of the present invention, when performing manual translation, according to the input translations of the source text and the target language, the candidate word which is after the input translation and matches the logic habit of the target language is determined, and is displayed to the translator, so that a prompt effect can be achieved, and the translator can understand and translate the rarely-used word conveniently, thereby solving the problems that when performing manual translation, some rarely-used words are difficult to understand, and a proper translated word cannot be found at a moment, so that translation jamming occurs; meanwhile, the method for machine-assisted manual translation provided by the embodiment of the invention also scores the input translation through machine translation and provides a reference translation, so that references are provided for the translator to carry out gap and omission checking, error correction and the like on the input translation.
Fig. 9 is a block diagram of an apparatus for machine-assisted manual translation according to an embodiment of the present invention, where as shown in fig. 9, the apparatus may include:
the acquiring module 10 is used for acquiring a source text of a source language and acquiring an input translation of a target language;
a determining module 20, configured to determine, according to the source text and the input translation, and based on a logic habit of the target language, a candidate word of a next input word corresponding to the input translation;
and the display module 30 is used for displaying the candidate words.
In the device for machine-assisted manual translation provided by the embodiment of the invention, when manual translation is carried out, the determination module 20 determines the candidate word which is matched with the logic habit of the target language after the translation is input according to the semantic space of the source text and the input translation of the target language, and the display module 30 displays the candidate word to the translator, so that the prompting effect can be achieved, the translator can conveniently understand and translate the obscure word, and the problem that when manual translation is carried out, part of obscure words are difficult to understand, and a proper translation word cannot be found at a moment, so that translation jamming occurs is solved
Optionally, in the apparatus, the determining module 20 includes:
the first determining unit is used for determining a semantic space according to the source text;
a calculating unit, configured to calculate, according to the input translation and the corpus, a first conditional probability that each translated word in the semantic space for the target language appears after the input translation;
the ranking unit is used for ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word;
and the second determining unit is used for taking the plurality of translated words with the ranking values smaller than or equal to a preset value as the candidate words.
Optionally, in the apparatus, the display module 30 includes:
the first scoring unit is used for determining a first score of each candidate word according to the first conditional probability;
and the first display unit is used for displaying each candidate word from high to low according to the first score and correspondingly displaying the first score of each candidate word.
Optionally, the apparatus further comprises:
and the scoring module is used for determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language and displaying the second score.
Optionally, the scoring module includes:
the second scoring unit is used for determining a second score of the input translation according to the second condition probability;
and the second display unit is used for displaying the second scores.
Optionally, the apparatus further comprises:
the reference translation module is used for determining a reference translation according to the source text and the target language and displaying the reference translation when a translation completion signal is detected; the reference translation is a translation for which the second conditional probability for the source text is maximal.
Compared with the prior art, the machine-assisted manual translation device and the machine-assisted manual translation method have the same advantages, and are not described in detail herein.
In summary, the method and the device for machine-assisted manual translation provided by the embodiments of the present invention determine, with the aid of machine translation, a candidate word that matches the logic habit of a target language after the input of the translation according to the input translations of a source text and the target language, and display the candidate word to a translator, so as to achieve a prompt effect, and facilitate the translator to understand and translate rare words, thereby solving the problem that some rare words are difficult to understand and a proper translated word cannot be found in the manual translation, so that a translation jam occurs; meanwhile, the method for machine-assisted manual translation provided by the embodiment of the invention also scores the input translation through machine translation and provides a reference translation, so that references are provided for the translator to carry out gap and omission checking, error correction and the like on the input translation.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
As is readily imaginable to the person skilled in the art: any combination of the above embodiments is possible, and thus any combination between the above embodiments is an embodiment of the present invention, but the present disclosure is not necessarily detailed herein for reasons of space.
In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the invention and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features included in other embodiments, rather than other features, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
Claims (4)
1. A method of machine-assisted manual translation, the method comprising:
acquiring a source text of a source language and acquiring an input translation of a target language;
determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language;
displaying the candidate words;
determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language, including:
determining a semantic space according to the source text;
calculating a first conditional probability of each translation word aiming at the target language in the semantic space after the input translation according to the input translation and the corpus;
ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word;
taking a plurality of the translated words with the ranking values smaller than or equal to a preset value as the candidate words;
the determining a semantic space according to the source text comprises:
determining a semantic space of the source text using neural network machine translation;
after the semantic space is determined according to the source text, the method further comprises the following steps:
determining the next word or word of the input translation on the basis of the input translation in combination with the semantic space of the source text; the next word or word is a translated word with the highest probability after the input translated text in the translated words corresponding to the semantic space is determined by calling a semantic library, the translated word with the highest probability is used as the output of a decoding model, the output is spliced with the input translated text, the input translated text obtained after splicing is used as the input of a new decoding model, and then the decoding operation is continued to determine the next word or word corresponding to the new input translated text;
the method further comprises the following steps:
determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language, and displaying the second score; the second score is determined according to a second conditional probability;
the method further comprises the following steps:
when a translation completion signal is detected, determining a reference translation according to the source text and the target language, and displaying the reference translation; the reference translation is a translation for which the second conditional probability for the source text reaches a maximum value; after a translator finishes translating a sentence of source text or a section of source text, displaying a reference translation for the translator to provide error correction reference for the translator; the second conditional probability is calculated according to the formula p (y | x), where x represents the source text and y represents the input text.
2. The method of claim 1, wherein the presenting the candidate word comprises:
determining a first score of each candidate word according to the first conditional probability;
and displaying each candidate word according to the first score from high to low, and correspondingly displaying the first score of each candidate word.
3. The method of claim 1, wherein determining a second score for the input translation based on the logical habits of the target language from the source text and the input translation and presenting the second score comprises:
determining a second conditional probability of the input translation for the source text according to the source text, the input translation and the corpus;
determining a second score of the input translation according to the second conditional probability;
and displaying the second score.
4. An apparatus for machine-assisted manual translation, the apparatus comprising:
the acquisition module is used for acquiring a source text of a source language and acquiring an input translation of a target language;
the determining module is used for determining a candidate word of a next input word corresponding to the input translation according to the source text and the input translation and based on the logic habit of the target language;
the display module is used for displaying the candidate words;
the determining module includes:
the first determining unit is used for determining a semantic space according to the source text;
a calculating unit, configured to calculate, according to the input translation and the corpus, a first conditional probability that each translated word in the semantic space for the target language appears after the input translation;
the ranking unit is used for ranking each translation word according to the first conditional probability from big to small, and determining the ranking value of each translation word;
the second determining unit is used for taking the plurality of translated words with the ranking values smaller than or equal to a preset value as the candidate words;
the device further comprises:
the scoring module is used for determining a second score of the input translation according to the source text and the input translation and based on the logic habit of the target language and displaying the second score;
the device further comprises:
the reference translation module is used for determining a reference translation according to the source text and the target language and displaying the reference translation when a translation completion signal is detected; the reference translation is a translation for which the second conditional probability for the source text reaches a maximum value; after a translator finishes translating a sentence of source text or a section of source text, displaying a reference translation for the translator to provide error correction reference for the translator; the second conditional probability is calculated according to the formula p (y | x), where x represents the source text and y represents the input text.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010193403.8A CN111507113B (en) | 2020-03-18 | 2020-03-18 | Method and device for machine-assisted manual translation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010193403.8A CN111507113B (en) | 2020-03-18 | 2020-03-18 | Method and device for machine-assisted manual translation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111507113A CN111507113A (en) | 2020-08-07 |
| CN111507113B true CN111507113B (en) | 2021-03-02 |
Family
ID=71874096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010193403.8A Active CN111507113B (en) | 2020-03-18 | 2020-03-18 | Method and device for machine-assisted manual translation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111507113B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPWO2022130940A1 (en) * | 2020-12-15 | 2022-06-23 | ||
| CN114139560B (en) * | 2021-12-03 | 2022-12-09 | 山东诗语信息科技有限公司 | Translation system based on artificial intelligence |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510194A (en) * | 2009-03-15 | 2009-08-19 | 刘树根 | Statement component apparatus and multilingual professional translation method based on statement component |
| CN108460027A (en) * | 2018-02-14 | 2018-08-28 | 广东外语外贸大学 | A kind of spoken language instant translation method and system |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100555270C (en) * | 2004-01-13 | 2009-10-28 | 中国科学院计算技术研究所 | A kind of machine automatic testing method and system thereof |
| CN102789451B (en) * | 2011-05-16 | 2015-06-03 | 北京百度网讯科技有限公司 | Individualized machine translation system, method and translation model training method |
| CN102193914A (en) * | 2011-05-26 | 2011-09-21 | 中国科学院计算技术研究所 | Computer aided translation method and system |
| CN104714943A (en) * | 2015-03-26 | 2015-06-17 | 百度在线网络技术（北京）有限公司 | Translation method and system |
| US9836457B2 (en) * | 2015-05-25 | 2017-12-05 | Panasonic Intellectual Property Corporation Of America | Machine translation method for performing translation between languages |
| CN106484682B (en) * | 2015-08-25 | 2019-06-25 | 阿里巴巴集团控股有限公司 | Machine translation method, device and electronic equipment based on statistics |
| CN106649288B (en) * | 2016-12-12 | 2020-06-23 | 北京百度网讯科技有限公司 | Artificial intelligence based translation method and device |
| CN108710616A (en) * | 2018-05-23 | 2018-10-26 | 科大讯飞股份有限公司 | A kind of voice translation method and device |
| CN108874785B (en) * | 2018-06-01 | 2020-11-03 | 清华大学 | Translation processing method and system |
| CN109446534B (en) * | 2018-09-21 | 2020-07-31 | 清华大学 | Machine translation method and device |
-
2020
- 2020-03-18 CN CN202010193403.8A patent/CN111507113B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510194A (en) * | 2009-03-15 | 2009-08-19 | 刘树根 | Statement component apparatus and multilingual professional translation method based on statement component |
| CN108460027A (en) * | 2018-02-14 | 2018-08-28 | 广东外语外贸大学 | A kind of spoken language instant translation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111507113A (en) | 2020-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10073843B1 (en) | Method and apparatus for cross-lingual communication | |
| US8423346B2 (en) | Device and method for interactive machine translation | |
| CN105917327B (en) | System and method for entering text into an electronic device | |
| US9805718B2 (en) | Clarifying natural language input using targeted questions | |
| US20120047172A1 (en) | Parallel document mining | |
| US8219381B2 (en) | Dictionary registration apparatus, dictionary registration method, and computer product | |
| KR20090084818A (en) | Web-based collocation error proofing | |
| JP2006252382A (en) | Question answering system, data retrieval method and computer program | |
| JPWO2003065245A1 (en) | Translation method, translation output method, storage medium, program, and computer apparatus | |
| US20130297284A1 (en) | Apparatus and method for generating polite expressions for automatic translation | |
| CN111507113B (en) | Method and device for machine-assisted manual translation | |
| WO2010020087A1 (en) | Automatic word translation during text input | |
| JP2008225963A (en) | Machine translation device, replacement dictionary creating device, machine translation method, replacement dictionary creating method, and program | |
| KR101709693B1 (en) | Method for Web toon Language Automatic Translating Using Crowd Sourcing | |
| US8738353B2 (en) | Relational database method and systems for alphabet based language representation | |
| US10650195B2 (en) | Translated-clause generating method, translated-clause generating apparatus, and recording medium | |
| JP2007149109A (en) | Translation support device | |
| KR100958340B1 (en) | Device and Method for Real-time Interactive Machine Translation | |
| Gerlach | Improving statistical machine translation of informal language: a rule-based pre-editing approach for French forums | |
| US10984191B2 (en) | Experiential parser | |
| JP2005078318A (en) | Method for evaluating mechanical translation and device for evaluating mechanical translation | |
| KR20150043065A (en) | Apparatus and method for correcting multilanguage morphological error based on co-occurrence information | |
| KR101501459B1 (en) | Translation apparatus and method for providing various style of translatability | |
| JP5185343B2 (en) | Machine translation apparatus and machine translation program | |
| Davis | Tajik-Farsi Persian Transliteration Using Statistical Machine Translation. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |