CN112084795A - Translation system and translation service calling method and device - Google Patents
Translation system and translation service calling method and device Download PDFInfo
- Publication number
- CN112084795A CN112084795A CN201910508166.7A CN201910508166A CN112084795A CN 112084795 A CN112084795 A CN 112084795A CN 201910508166 A CN201910508166 A CN 201910508166A CN 112084795 A CN112084795 A CN 112084795A
- Authority
- CN
- China
- Prior art keywords
- translation
- original text
- intervention
- module
- service application
- 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
Images
Abstract
The application provides a translation system and a translation service calling method and device, a plurality of basic translation function modules can be created, various complex translation requirements under multiple scenes and multiple fields can be met, the native container management and scheduling capability of a cloud is combined on the basis, distributed containerized automatic deployment of the translation function modules in the translation system is realized, rapid continuous delivery and dynamic expansion of the whole translation service can be carried out on the cloud, the service deployment cost and time are reduced, and the iteration efficiency is accelerated, so that the product updating period is shortened.
Description
Technical Field
The invention relates to the technical field of cloud computing, in particular to a translation system and a translation service calling method and device.
Background
With the development of economic globalization and international trade, the rapidly growing border-crossing e-commerce is remodeling a new global trade pattern, which becomes a new driving force for trade growth concerned by various countries, and the first road barrage crossing the border-crossing e-commerce is a multilingual problem. The propagation of commodity information and the effective communication of buyers and sellers are affected by multi-language problems, and for the e-commerce scene with huge volume and higher requirement on timeliness, the low-efficiency manual translation cannot meet the actual requirement, so that the establishment of a commercially available machine translation system becomes the most appropriate solution and is widely applied.
However, the demands for machine translation in different scenarios and different fields are greatly different, especially in e-commerce scenarios. For example, the translation accuracy requirements for the model, size, specification and the like of the commodity are extremely high, and the timeliness requirements for the communication scene translation are higher. Even the most advanced neural network machine translation technology at present, only one translation decoder can not meet the actual production requirement, and a large number of upstream and downstream functional modules and engineering work are needed to be done. For example, spelling error correction is performed on an original text, forced translation intervention is performed on the original text, results of a plurality of translation models are optimized, and the like, meanwhile, the translation models need to be iterated continuously on the basis of corpus precipitation and model tuning, if deployment operation and maintenance of the whole translation system are complex, the updating iteration speed of the translation models is limited, and how to rapidly perform one-click hosting deployment on the translation system is very important.
Disclosure of Invention
In order to solve the problems, the invention provides a translation system, a translation service calling method and a translation service calling device, which can meet the requirements of various complex translation service application scenarios under multiple scenarios and multiple fields, and also combine the native container management and scheduling capabilities of the cloud on the basis to realize the distributed containerized automatic deployment of each translation function module in the system, and can perform the rapid continuous delivery and dynamic expansion of the whole translation service function on the cloud, thereby reducing the service deployment cost and time, accelerating the iteration efficiency and shortening the product updating period.
An embodiment of the present invention provides a translation system, including: deploying a controller;
the deployment controller is used for creating a plurality of translation function modules corresponding to the translation service application scenes according to the needed translation service application scenes; and respectively deploying the plurality of translation function modules in different containers.
Optionally, the system further comprises:
and a service logic layer: acquiring information of a translation function module to be created, wherein the information of the translation function module to be created comprises a translation service identifier and a translation function identifier, the translation service identifier represents the translation service application scene, and the translation function identifier represents the translation function module;
and (3) a data layer: acquiring a data file of the translation function module to be created according to the translation service identifier and the translation function identifier;
a container treatment layer: and creating an image file of the translation function module in a container according to the data file.
Alternatively,
the data layer: obtaining data files of all translation function modules corresponding to the translation service application scene by utilizing a deep learning mechanism according to the translation service application scene; and carrying out cloud storage on the data files of the translation function modules corresponding to the translation service application scenes.
Alternatively,
the service logic layer: acquiring information of a translation function module to be updated, wherein the information of the translation function module to be updated comprises a translation service identifier and a translation function identifier;
the data layer: inquiring a data file of the translation function module to be updated on the cloud storage according to the translation service identifier and the translation function identifier, and updating the data file;
the container treatment layer: and updating the mirror image file of the translation function module in the container according to the updated data file.
Alternatively,
the translation function module comprises a master control module, an original text intervention module, a translation intervention module and a decoder.
The application also provides a translation service calling method, which comprises the following steps:
preprocessing the received original text;
determining a translation model for translating the original text according to the pretreatment result;
and calling the translation model to translate the original text.
Optionally, preprocessing the received original text, including:
analyzing the original text, and performing original text intervention processing on the analyzed original text to obtain an original text intervention result;
and performing translation intervention treatment on the original text subjected to the dry prognosis of the original text according to the original text intervention result to obtain a translation intervention result.
Optionally, determining a translation model for translating the original text according to the pre-processing result includes:
determining a translation service application scene to which the original text belongs according to a translation intervention result;
determining a translation model corresponding to the translation service application scene according to the translation service application scene;
and the translation model is a translation model corresponding to the translation service application scene obtained by learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene.
Optionally, after the calling the translation model to translate the original text, the method further includes:
and post-processing the translated text, including the optimization of the translated text, case reduction, treatment of undecoded words and treatment of sensitive words.
The present application further provides a translation service invoking device, including:
the main control module is used for preprocessing the received original text; determining a translation model for translating the original text according to the pretreatment result; calling a decoder corresponding to the translation model according to the translation model;
and the decoder is used for translating the original text by utilizing the translation model according to the calling instruction of the master control module.
Optionally, the apparatus further comprises: the original text intervention module and the translation intervention module;
the master control module is used for calling the original text intervention module after analyzing the original text;
the original text intervention module is used for carrying out original text intervention processing on the analyzed original text according to the calling instruction of the master control module and returning an original text intervention result to the master control module;
the master control module is also used for calling a translation intervention module according to the original text intervention result returned by the original text intervention module;
and the translation intervention module is used for performing translation intervention processing on the original text subjected to the original text trunk prognosis according to the calling instruction of the master control module and returning an original text intervention result to the master control module.
Optionally, the total control module is further configured to determine, according to a translation intervention result returned by the translation intervention module, a translation service application scenario to which the original text belongs; determining a translation model corresponding to the translation service application scene according to the translation service application scene;
and the translation model is a translation model corresponding to the translation service application scene obtained by learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene.
Optionally, the general control module is further configured to perform post-processing on the translated text, including text optimization, case reduction, un-decoded word processing, and sensitive word processing.
The present application further provides a server, comprising: a memory, a processor, and a communication component;
the memory for storing a computer program;
the processor, coupled with the memory and the communication component, is configured to execute a computer program for performing the steps or operations of the translation system or the method.
The present application also provides a computer-readable storage medium storing a computer program, which when executed by a computer, can implement the steps or operations of the translation system or the method.
According to the translation system distributed deployment method, each translation function module can be made into a mirror image container, instantiation management is carried out on the containers through the deployment controller, the distributed deployment method is suitable for various container technologies on the cloud so as to achieve automatic deployment of the translation function modules, multiple sets of translation service systems are created, the translation service systems are isolated from one another, and the translation service identifications are used for distinguishing.
In the embodiment of the application, the translation system is deployed in a distributed manner, namely, each function of the translation system is packaged in a modularized manner, so that each function module can cover the translation service requirements of multiple scenes and multiple fields, the translation system is deployed in a distributed manner on the cloud, the whole set of translation system service can be deployed rapidly and automatically, the iterative updating of a model and the dynamic expansion of the service are supported, and the translation system has high flexibility and maintainability.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a diagram of a distributed deployment system architecture for a translation system according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a distributed deployment method of a translation system according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a distributed deployment controller of a translation system according to an embodiment of the present invention;
fig. 4 is a flowchart illustrating a translation service invoking method according to an embodiment of the present invention;
FIG. 5 is a diagram of a distributed deployment system architecture that provides a translation system according to another embodiment of the present invention;
fig. 6 is a schematic structural diagram of a translation service invoking device according to an embodiment of the present invention;
FIG. 7 is a diagram of a distributed deployment system architecture that provides a translation system according to another embodiment of the present invention;
fig. 8 is a schematic structural diagram of a server according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the present invention and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, and "a" and "an" generally include at least two, but do not exclude at least one, unless the context clearly dictates otherwise.
It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.
It is also noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a good or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such good or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a commodity or system that includes the element.
The inventor discovers that in the process of implementing the invention:
in the existing machine translation, all functional modules are integrated into one module to be deployed in one machine, namely, front and back processing logics are added in a decoder module, wherein the front and back processing logics comprise modules such as clauses, participles, case conversion, original text intervention, translated text intervention, unregistered word processing, duplication removal, sensitive word processing, brand word processing, word de-participles and the like.
In order to solve the technical problem, in the embodiment of the application, the translation system is deployed in a distributed manner, that is, each function of the translation system is packaged in a modular manner, so that each function module can cover the translation service requirements of multiple scenes and multiple fields, and the distributed deployment is performed on the cloud, so that the whole set of translation system services can be rapidly and automatically deployed, and meanwhile, the iterative update of a model and the dynamic expansion of the services are supported, and the method has high flexibility and maintainability.
Interpretation of terms:
and (3) machine translation: mapping from one language to another language is implemented by a machine.
Neural network machine translation: machine translation based on deep neural networks typically consists of an encoder and a decoder.
kubernets: is an open source application for managing containerization on multiple hosts in a cloud platform.
A docker container: is an open source application container engine, so that developers can package their applications and dependency packages into a portable container and then release the package to any machine to realize virtualization.
Fig. 1 is an architecture diagram of a distributed deployment system of a translation system according to an embodiment of the present invention, as shown in fig. 1: the deployment controller is used for deploying a plurality of translation function modules corresponding to the translation service application scene according to the needed translation service application scene; and respectively deploying the plurality of translation function modules in different containers.
During specific implementation, the deployment controller creates a plurality of translation function modules in each translation service application scene by calling an Application Program Interface (API), each translation function module can be deployed in different containers of different hosts in a kubernets cluster, and each translation function module acquires a data file required by the corresponding translation function module from the cloud storage, wherein the data file is a data packet required by the corresponding translation function module.
The kubernets are an application solution of a set of open-source management containers (dockers), can be widely applied to container solutions serving as a bottom layer, and has certain standard and universality. By utilizing the container management and scheduling capability of kubernets, a plurality of sets of translation service suites with a plurality of container sets can be quickly built to provide complete translation services externally, and meanwhile, each translation service is mutually isolated and not interfered with each other, so that different translation function modules can be built according to the requirements of different translation service application scenes, and a plurality of translation service function models can be simultaneously updated in an iterative manner.
Due to the fact that requirements of different application scenes and different application fields for machine translation are greatly different, particularly in the electronic market scene, requirements for translation accuracy of models, sizes, specifications and the like of commodities are extremely high, and requirements for timeliness of translation of communicated application scenes are higher. According to the embodiment of the invention, a plurality of sets of translation function modules can be created, each set of translation function module (comprising a plurality of translation function modules) supports a corresponding translation service application scene, and each set of translation function module can be distinguished through a translation service identifier representing the translation service application scene.
Based on the architecture diagram of the distributed deployment system of the translation system shown in fig. 1, fig. 2 is a schematic flow diagram of a distributed deployment method of the translation system according to an embodiment of the present invention, and as shown in fig. 2, the method includes:
101. acquiring information of a translation function module to be created;
fig. 3 is a schematic structural diagram of a distributed deployment controller of a translation system according to an embodiment of the present invention, and as shown in fig. 3, the deployment controller includes a service logic layer, a data layer, and a container processing layer;
the business logic layer is mainly responsible for managing and organizing each translation function module deployed by the translation system in the embodiment of the invention, and supports the creation, dynamic combination and elastic expansion of each translation function module.
For example, when a set of translation service systems applied to a certain translation service application scenario needs to be created, each set of translation service system includes a plurality of translation function modules. In step 101, a service logic layer of a deployment controller needs to obtain information of a translation function module to be created according to a creation instruction of a user, where the information of the translation function module to be created includes a translation service identifier and a translation function identifier, where the translation service identifier represents a translation service application scenario, and the translation function identifier represents the translation function module.
102. Acquiring a data file of the translation function module to be created according to the translation service identifier and the translation function identifier included in the information of the translation function module;
as shown in fig. 3, the data layer of the deployment controller processes some intermediate data according to the provided information of the translation function modules to be deployed, and then generates final deployment instructions and data files by combining the deployment template files of the translation function modules.
Before step 102, a data layer learns and calculates in advance to obtain data files of each translation function module corresponding to the translation service application scene by using a deep neural network learning mechanism according to the translation service application scene; and carrying out cloud storage on the data files of the translation function modules corresponding to the translation service application scenes.
And 102, in specific implementation, querying a data file of the translation function module to be created on the cloud storage according to the translation service identifier and the translation function identifier by the data layer.
103. And creating an image file of the translation function module in a container according to the data file.
As shown in fig. 3, the container processing layer of the deployment controller creates an image file of the translation function module by calling an Application Program Interface (API) of kubernets.
Specifically, when creating the image file of the translation function module in the container, it is necessary to determine, on the host corresponding to the translation service identifier, to create the image file of the translation function module on the container corresponding to the translation function identifier according to the translation service identifier and the translation function identifier of the translation function module.
Therefore, the deployment controller in the embodiment of the present invention can provide HTTP services to the outside, and support the example operations of adding, modifying, deleting, and querying the translation function module on the kubernets cluster. For example, taking an example of updating or iterating a certain translation function module, a conventional update or iterating translation function module generally refers to data reloading for updating a translation model in an iterative decoder.
Specifically, the deployment controller updates the translation function module as follows:
and a service logic layer: acquiring information of a translation function module to be updated, wherein the information of the translation function module to be updated comprises a translation service identifier and a translation function identifier;
and (3) a data layer: inquiring a data file of the translation function module to be updated on the cloud storage according to the translation service identifier and the translation function identifier, and updating the data file;
a container treatment layer: and updating the mirror image file of the translation function module in the container according to the updated data file.
According to the translation system distributed deployment method, each translation function module can be made into a mirror image container, instantiation management is carried out on the containers through the deployment controller, the distributed deployment method is suitable for various container technologies on the cloud so as to achieve automatic deployment of the translation function modules, multiple sets of translation service systems are created, the translation service systems are isolated from one another, and the translation service identifications are used for distinguishing.
Based on the architecture diagram of the distributed deployment system of the translation system shown in fig. 1, fig. 4 is a schematic flowchart of a translation service invocation method provided in an embodiment of the present invention, and as shown in fig. 4, the method includes:
201. preprocessing the received original text;
in practical application, when a translation request initiated by a user is received, the translation request comprises an original text to be translated. As shown in fig. 1, the deployment controller provides an HTTP service to the outside, and when a user translation request is received, the original text needs to be processed to determine an application scenario for translating the original text, so as to determine a translation service identifier of the original text, and then determine a translation system (i.e., a translation function module) corresponding to the translation service identifier according to the translation service identifier, and finally translate the original text by using a translation model in the translation function module.
Wherein, the step 201 includes, when implemented specifically:
analyzing the original text, and performing original text intervention processing on the analyzed original text to obtain an original text intervention result;
and performing translation intervention treatment on the original text subjected to the dry prognosis of the original text according to the original text intervention result to obtain a translation intervention result.
202. Determining a translation model for translating the original text according to the pretreatment result;
wherein, the step 202 includes, when implemented:
determining a translation service application scene to which the original text belongs according to a translation intervention result;
determining a translation model corresponding to the translation service application scene according to the translation service application scene;
and the translation model is a translation model corresponding to the translation service application scene obtained by learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene.
203. And calling the translation model to translate the original text.
Optionally, after the calling the translation model to translate the original text, the method further includes:
and post-processing the translated text, including the optimization of the translated text, case reduction, treatment of undecoded words and treatment of sensitive words.
Based on the distributed deployment system architecture diagram of the translation system shown in fig. 1, fig. 5 is a distributed deployment system architecture diagram of a translation system provided in another embodiment of the present invention, as shown in fig. 5, including a deployment controller, where the deployment controller calls, through an API interface, each translation function module corresponding to each set of translation service application scenarios, where each translation function module corresponding to each set of translation service application scenarios includes a master control module, an original text intervention module, a translated text intervention module, and a decoder, respectively.
In practical application, when a translation request initiated by a user is received, the translation request comprises an original text to be translated. As shown in fig. 1, the deployment controller provides an HTTP service to the outside, and when a user translation request is received, for example, in an actual application, the user selects a translation service application scenario, that is, a translation service identifier can be determined. At this time, the deployment controller may send the original text to the master controller corresponding to the translation service application scenario (translation service identifier).
Fig. 6 is a schematic structural diagram of a translation service invoking device according to an embodiment of the present invention, as shown in fig. 6, in the embodiment of the present invention, a master control module serves as a unified entry of a translation service request to the outside, and is responsible for service invoking and integration processing of an original text intervention module, a translation intervention module, and a decoder, and for pre-and post-processing of a translation result in a translation system.
The master control module is used for calling the original text intervention module after analyzing the original text; the main control module analyzes the original text, for example, html analysis, paragraph segmentation, natural sentence segmentation and the like;
the original text intervention module is used for carrying out original text intervention processing on the analyzed original text according to the calling instruction of the master control module and returning an original text intervention result to the master control module; wherein, the original text intervention module is mainly used for preprocessing the original text so that the subsequent translation model can translate the original text more accurately and smoothly, such as: in order to enable the goods to hit more search keywords, the seller can stack a plurality of vocabularies on the goods title, which is not friendly to the translation model and needs to rewrite and remove the weight of the title; spelling errors are caused by too fast input in commodity search, and the words can also influence the translated search results and need to be corrected in advance; many repeated spoken words in the communication result in too long decoding time of the translation model, and need to perform normalization in advance. In the text intervention module, therefore, functions similar to those described above for processing the text before decoding are mainly integrated.
And the master control module is also used for calling the translation intervention module according to the original text intervention result returned by the original text intervention module.
The translation intervention module is used for performing translation intervention processing on the original text subjected to the original text trunk prediction according to the calling instruction of the master control module and returning an original text intervention result to the master control module; because the training corpus of the translation model cannot cover all the words, and some words have different translation results in different application scenes, especially in the e-market scene, if wrong translation results occur in the production environment, a mechanism is needed to quickly intervene the results. Therefore, the main function of the translation intervention module is to ensure that the translation service returns the expected result by manually intervening the data input in advance; in the translation intervention module, two intervention modes of terms and sentences can be provided to adjust the translation result.
The general control module is further used for determining a translation model corresponding to the translation intervention result according to the translation intervention result returned by the translation intervention module; calling a decoder corresponding to the translation model according to the translation model;
the decoder is used for translating the original text by utilizing the translation model according to the calling instruction of the master control module; the translation model is the most central translation model in the decoder, and the translation model is used for decoding and translating the input original text into a translated text. The decoder can be loaded with translation models (such as a phrase table model and a neural network model) with different languages and different types, and the decoding is supported in a group batch mode, so that the corresponding time delay is reduced, and the translation decoding operation efficiency is accelerated.
And the master control module is also used for carrying out post-processing on the translation, including translation optimization, case reduction, un-decoded word processing and sensitive word processing.
Based on the distributed deployment system architecture diagram of the translation system shown in fig. 1, fig. 7 is a distributed deployment system architecture diagram providing a translation system according to another embodiment of the present invention, as shown in fig. 7, including a deployment controller, where the deployment controller calls each set of translation service system through an API interface to perform translation, where each set of translation service system corresponds to one translation service application scenario, and each set of translation service system shares one master control module, one original text intervention module, and one translated text intervention module, but each set of translation service system corresponds to a respective decoder, respectively, because the translation model in each decoder is a translation model corresponding to the translation service application scenario obtained through learning calculation by using a deep neural network learning mechanism according to the respective corresponding translation service application scenario.
The master control module is externally used as a uniform entrance of a translation service request, and is internally responsible for service calling and integration processing of the original text intervention module, the translation intervention module and the decoder and for the front-back processing of a translation result in the translation system.
The master control module is used for calling the original text intervention module after analyzing the original text;
the original text intervention module is used for carrying out original text intervention processing on the analyzed original text according to the calling instruction of the master control module and returning an original text intervention result to the master control module;
the master control module is also used for calling a translation intervention module according to the original text intervention result returned by the original text intervention module;
and the translation intervention module is used for performing translation intervention processing on the original text subjected to the original text trunk prognosis according to the calling instruction of the master control module and returning an original text intervention result to the master control module.
The master control module is further used for determining a translation service application scene to which the original text belongs according to a translation intervention result returned by the translation intervention module; determining a translation model corresponding to the translation service application scene according to the translation service application scene; the translation model is a translation model corresponding to the translation service application scene obtained through learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene; calling a decoder corresponding to the translation model according to the translation model;
the decoder is used for translating the original text by utilizing the translation model according to the calling instruction of the master control module;
and the master control module is also used for carrying out post-processing on the translation, including translation optimization, case reduction, un-decoded word processing and sensitive word processing.
The translation system deployed in a distributed manner can be used for creating a plurality of basic translation function modules, can meet various complex translation requirements under multiple scenes and multiple fields, combines the native container management and scheduling capabilities of the cloud on the basis, realizes the distributed containerization automatic deployment of the translation function modules in the translation system, can perform the rapid continuous delivery and dynamic expansion of the whole translation service on the cloud, reduces the service deployment cost and time, and accelerates the iteration efficiency so as to shorten the product updating period.
Fig. 8 is a schematic structural diagram of a server according to an embodiment of the present invention, as shown in fig. 8, including:
a memory, a processor, and a communication component;
the memory for storing a computer program;
the processor, coupled with the memory and the communication component, is configured to execute a computer program for performing the steps or operations of the method described in the method embodiments of fig. 2 or fig. 4.
Further, as shown in fig. 8, the server further includes: display, power components, audio components, and the like. Only some of the components are schematically shown in fig. 8, and the server is not meant to include only the components shown in fig. 8.
The server shown in this embodiment may execute the method embodiments shown in fig. 2 or fig. 4, and the implementation principle and technical effect thereof are not described again.
Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a computer, the steps or operations related to the server in the method embodiments shown in fig. 2 or fig. 4 can be implemented, which are not described herein again.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (15)
1. A translation system, comprising: deploying a controller;
the deployment controller is used for creating a plurality of translation function modules corresponding to the translation service application scenes according to the needed translation service application scenes; and respectively deploying the plurality of translation function modules in different containers.
2. The system of claim 1, further comprising:
and a service logic layer: acquiring information of a translation function module to be created, wherein the information of the translation function module to be created comprises a translation service identifier and a translation function identifier, the translation service identifier represents the translation service application scene, and the translation function identifier represents the translation function module;
and (3) a data layer: acquiring a data file of the translation function module to be created according to the translation service identifier and the translation function identifier;
a container treatment layer: and creating an image file of the translation function module in a container according to the data file.
3. The system of claim 2, wherein:
the data layer: obtaining data files of all translation function modules corresponding to the translation service application scene by utilizing a deep learning mechanism according to the translation service application scene; and carrying out cloud storage on the data files of the translation function modules corresponding to the translation service application scenes.
4. The system of claim 3, further comprising:
the service logic layer: acquiring information of a translation function module to be updated, wherein the information of the translation function module to be updated comprises a translation service identifier and a translation function identifier;
the data layer: inquiring a data file of the translation function module to be updated on the cloud storage according to the translation service identifier and the translation function identifier, and updating the data file;
the container treatment layer: and updating the mirror image file of the translation function module in the container according to the updated data file.
5. The system of any one of claims 1-4, further comprising:
the translation function module comprises a master control module, an original text intervention module, a translation intervention module and a decoder.
6. A translation service invocation method, comprising:
preprocessing the received original text;
determining a translation model for translating the original text according to the pretreatment result;
and calling the translation model to translate the original text.
7. The method of claim 6, wherein preprocessing the received text comprises:
analyzing the original text, and performing original text intervention processing on the analyzed original text to obtain an original text intervention result;
and performing translation intervention treatment on the original text subjected to the dry prognosis of the original text according to the original text intervention result to obtain a translation intervention result.
8. The method of claim 7, wherein determining a translation model for translating the original text based on the pre-processing result comprises:
determining a translation service application scene to which the original text belongs according to a translation intervention result;
determining a translation model corresponding to the translation service application scene according to the translation service application scene;
and the translation model is a translation model corresponding to the translation service application scene obtained by learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene.
9. The method of claim 6, wherein invoking the translation model to translate the original text further comprises:
and post-processing the translated text, including the optimization of the translated text, case reduction, treatment of undecoded words and treatment of sensitive words.
10. A translation service invocation apparatus, comprising:
the main control module is used for preprocessing the received original text; determining a translation model for translating the original text according to the pretreatment result; calling a decoder corresponding to the translation model according to the translation model;
and the decoder is used for translating the original text by utilizing the translation model according to the calling instruction of the master control module.
11. The apparatus of claim 10, further comprising: the original text intervention module and the translation intervention module;
the master control module is used for calling the original text intervention module after analyzing the original text;
the original text intervention module is used for carrying out original text intervention processing on the analyzed original text according to the calling instruction of the master control module and returning an original text intervention result to the master control module;
the master control module is also used for calling a translation intervention module according to the original text intervention result returned by the original text intervention module;
and the translation intervention module is used for performing translation intervention processing on the original text subjected to the original text trunk prognosis according to the calling instruction of the master control module and returning an original text intervention result to the master control module.
12. The apparatus of claim 11, wherein:
the master control module is further used for determining a translation service application scene to which the original text belongs according to a translation intervention result returned by the translation intervention module; determining a translation model corresponding to the translation service application scene according to the translation service application scene;
and the translation model is a translation model corresponding to the translation service application scene obtained by learning calculation by utilizing a deep neural network learning mechanism according to the translation service application scene.
13. The apparatus of claim 12, wherein:
and the master control module is also used for carrying out post-processing on the translation, including translation optimization, case reduction, un-decoded word processing and sensitive word processing.
14. A server, comprising: a memory, a processor, and a communication component;
the memory for storing a computer program;
the processor, coupled with the memory and the communication component, to execute a computer program for performing the steps or operations of the system of claims 1-5 or the method of any of claims 6-9.
15. A computer-readable storage medium storing a computer program, wherein the computer program is capable of implementing the steps or operations of the system according to claims 1-5 or the method according to any one of claims 6-9 when executed by a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910508166.7A CN112084795A (en) | 2019-06-12 | 2019-06-12 | Translation system and translation service calling method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910508166.7A CN112084795A (en) | 2019-06-12 | 2019-06-12 | Translation system and translation service calling method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112084795A true CN112084795A (en) | 2020-12-15 |
Family
ID=73734611
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910508166.7A Pending CN112084795A (en) | 2019-06-12 | 2019-06-12 | Translation system and translation service calling method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112084795A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022135309A1 (en) * | 2020-12-22 | 2022-06-30 | 张龙哺 | Cloud translation calling method, apparatus and system |
| CN114721719A (en) * | 2022-04-20 | 2022-07-08 | 上海道客网络科技有限公司 | Method and system for containerized deployment of heterogeneous applications in cluster |
-
2019
- 2019-06-12 CN CN201910508166.7A patent/CN112084795A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022135309A1 (en) * | 2020-12-22 | 2022-06-30 | 张龙哺 | Cloud translation calling method, apparatus and system |
| CN114721719A (en) * | 2022-04-20 | 2022-07-08 | 上海道客网络科技有限公司 | Method and system for containerized deployment of heterogeneous applications in cluster |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220058343A1 (en) | Written-modality prosody subsystem in a natural language understanding (nlu) framework | |
| US11093707B2 (en) | Adversarial training data augmentation data for text classifiers | |
| US11373049B2 (en) | Cross-lingual classification using multilingual neural machine translation | |
| US10339916B2 (en) | Generation and application of universal hypothesis ranking model | |
| US10019437B2 (en) | Facilitating information extraction via semantic abstraction | |
| US11520992B2 (en) | Hybrid learning system for natural language understanding | |
| US10547747B1 (en) | Configurable natural language contact flow | |
| US10601740B1 (en) | Chatbot artificial intelligence | |
| KR20200108775A (en) | Training corpus generating method, apparatus, device and storage medium | |
| US11189269B2 (en) | Adversarial training data augmentation for generating related responses | |
| US20210342547A1 (en) | System for focused conversation context management in a reasoning agent/behavior engine of an agent automation system | |
| CN111221842A (en) | Big data processing system and method | |
| US20220327012A1 (en) | Software validation framework | |
| US20230141398A1 (en) | Data augmentation for intent classification | |
| CN112084795A (en) | Translation system and translation service calling method and device | |
| CN111062521B (en) | Online prediction method, system and server | |
| US10891114B2 (en) | Interpreter for interpreting a data model algorithm and creating a data schema | |
| US11907678B2 (en) | Context-aware machine language identification | |
| US20230317058A1 (en) | Spoken language processing method and apparatus, and storage medium | |
| WO2022078506A1 (en) | Transliteration based data augmentation for training multilingual asr acoustic models in low resource settings | |
| CN112799658B (en) | Model training method, model training platform, electronic device, and storage medium | |
| US11645095B2 (en) | Generating and utilizing a digital knowledge graph to provide contextual recommendations in digital content editing applications | |
| CN115858556A (en) | Data processing method and device, storage medium and electronic equipment | |
| US20230016962A1 (en) | Multilingual natural language understanding model platform | |
| JP7029351B2 (en) | How to generate OOS text and the device that does it |
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 |