A priority-based translation enhancement method and system
By employing a priority-based translation enhancement method, translation engine resources are dynamically scheduled, addressing the challenges of insufficient device resources and the need for real-time translation, thereby achieving efficient translation output and resource utilization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- IOL WUHAN INFORMATION TECH CO LTD
- Filing Date
- 2022-10-11
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies cannot effectively schedule translation sequences when equipment resources are insufficient, resulting in resource waste and user waiting, and cannot guarantee the minimum output quality in real-time translation scenarios.
The priority-based translation enhancement method receives translation sequences and determines their priorities. It then uses the resource parameters of the translation engine array to predict future available resources, dynamically schedules translation pointers to match the priority of the translation engine model, and outputs the minimum translation result.
It optimizes translation output under limited resources, avoids resource waste and lag, and ensures that users receive appropriate translation results within a tolerable waiting time.
Smart Images
Figure CN115510877B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of translation enhancement technology, and particularly relates to a priority-based translation enhancement method and system. Background Technology
[0002] Translation enhancement, also known as augmented translation, refers to outputting the text to be translated through multimedia augmented translation when the device resources meet the requirements for enhanced output. For example, ordinary translation can output the translated sequence as text (written translation) or audio (interpretation), but augmented translation can further achieve the output of the translated sequence as text and images, video, or other forms of multimedia output. For example, after receiving a teaching sequence to be translated, it can play an augmented reality scene on a wearable augmented reality device. Obviously, augmented translation also includes ordinary translation.
[0003] While augmented translation enables multimedia output, it faces several challenges. Firstly, not all devices can allocate sufficient resources to support augmented translation at all times. Secondly, not every translation sequence requires resource allocation for augmented translation. The former raises the question of how to schedule translation sequences requiring augmented translation to devices with matching resources. The latter presents the challenge of how to achieve only regular translation when a translation sequence does not require augmented translation output, thus avoiding wasting system resources.
[0004] Furthermore, in translation scenarios that emphasize real-time or near-real-time performance, for a given sequence to be translated, even if the current translation resource does not support enhanced translation, users expect a minimum level of normal translation output; at the same time, when enhanced translation is supported, users need to receive the enhanced translation result within a tolerable waiting period, rather than experiencing frequent stuttering.
[0005] Therefore, current augmented translation still faces many technical challenges that need to be addressed. Summary of the Invention
[0006] To address some or all of the aforementioned technical problems, this invention proposes a priority-based translation enhancement method and system.
[0007] Specifically, in a first aspect of the present invention, a priority-based translation enhancement method is proposed, the method comprising the following steps:
[0008] S1: Receive multiple translation sequences and determine at least one translation priority contained in each translation sequence;
[0009] Specifically, in step S1, the translation priorities of the translation sequence include: plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority;
[0010] Step S1 determines at least one translation priority for each translation sequence based on the sequence tag or sequence keyword of each translation sequence.
[0011] S2: Determine the corresponding translation pointer for each translation priority;
[0012] Specifically, in step S2, the translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority.
[0013] S3: Obtain the resource parameters of the translation engine array;
[0014] S4: Based on the resource parameters, determine the schedulable translation pointers for each translation engine model;
[0015] S5: Each translation engine model determines at least one translation sequence as the sequence to be translated based on schedulable translation pointers;
[0016] S6: Use all arrays to be translated as input to the translation engine array, which outputs the translation enhancement result.
[0017] As a further improvement, the resource parameters of the translation engine array obtained in step S3 include:
[0018] Obtain the real-time resource parameters of the translation engine array, and predict the available resource parameters of the translation engine array within a future preset time period based on the real-time resource parameters;
[0019] The translation engine array includes multiple translation engine models, each of which can achieve translation enhancement results corresponding to different translation priorities.
[0020] Step S4, which determines the schedulable translation pointers for each translation engine model based on the resource parameters, specifically includes:
[0021] Determine the real-time resource parameters and available resource parameters for each translation engine model within a preset future time period;
[0022] When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.
[0023] Based on this, step S5 includes: taking the translation sequence including the first translation priority as the sequence to be translated;
[0024] Step S6 includes: inputting the sequence to be translated into a first translation engine model, and the first translation engine model outputting a translation enhancement result corresponding to a first translation priority.
[0025] In a second aspect of the present invention, to implement the method described in the first aspect, a priority-based translation enhancement system is proposed, the translation enhancement system comprising a translation engine array, the translation engine array comprising multiple translation engine models; each translation engine model can achieve translation enhancement results corresponding to different translation priorities.
[0026] The system includes several functional units that can implement the above method steps, as follows:
[0027] A sequence receiving unit, wherein the receiving unit is used to receive and store the translated sequence;
[0028] A priority identification unit is used to identify the translation priority contained in each translation sequence, wherein the translation priority includes plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority;
[0029] A pointer determination unit determines a corresponding translation pointer based on each translation priority. The translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority.
[0030] A real-time resource parameter acquisition unit is used to acquire the real-time resource parameters of the translation engine array;
[0031] Available resource parameter prediction unit, which predicts available resource parameters of the translation engine array within a future preset time period based on real-time resource parameter prediction;
[0032] A schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters.
[0033] The sequence to be translated is determined by a unit that determines at least one translation sequence as a sequence to be translated based on a schedulable translation pointer for each translation engine model.
[0034] The translation enhancement output unit takes all the arrays to be translated as input to the translation engine array, and the translation engine array outputs the translation enhancement result.
[0035] In a specific implementation, the priority identification unit determines at least one translation priority contained in each translation sequence based on the sequence label or sequence keyword of each translation sequence.
[0036] The schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters, specifically including:
[0037] When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.
[0038] The technical solution of this invention, upon receiving a translation sequence, first determines at least one translation priority for each translation sequence; then, for each translation priority, a corresponding translation pointer is determined, and subsequently, the translation array available for that priority is invoked through the translation pointer, avoiding frequent data writing and reading; simultaneously, after obtaining the resource parameters of the translation engine array, based on these parameters, the schedulable translation pointers for each translation engine model are determined, enabling the selection of translation priorities supported by system resources, avoiding translation waiting or stuttering due to insufficient resources; furthermore, the system resources consider not only the current real-time resource parameters but also the available resource parameters within a preset future time period, ensuring that the available resource parameters within the user's tolerable waiting period are also taken into account, conforming to the objective reality of the scenario while maximizing the utilization of system resources; finally, each translation engine model determines at least one translation sequence as a sequence to be translated based on the schedulable translation pointers, and uses all sequences to be translated as input to the translation engine array, which outputs translation enhancement results. This ensures that not every sequence to be translated needs to schedule resources to achieve enhanced translation, but only when sufficient resources can be scheduled to support enhanced translation, and even when the current translation resources do not support enhanced translation, the user can still obtain a minimum amount of normal translation output.
[0039] Further embodiments and improvements of the present invention will be described in conjunction with the accompanying drawings and specific examples. Attached Figure Description
[0040] Figure 1 This is a schematic diagram illustrating the steps of a priority-based translation enhancement method according to an embodiment of the present invention;
[0041] Figure 2 yes Figure 1 A schematic diagram of a preferred embodiment of the priority-based translation enhancement method;
[0042] Figure 3 This is a schematic diagram of the architecture of a priority-based translation enhancement system according to an embodiment of the present invention;
[0043] Figure 4 This is a schematic diagram of the architecture of the translation pointers corresponding to the translation priority determination involved in various embodiments of the present invention. Detailed Implementation
[0044] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.
[0045] First, see Figure 1 . Figure 1 This is a schematic diagram illustrating the steps of a priority-based translation enhancement method according to an embodiment of the present invention.
[0046] exist Figure 1 The method includes steps S1-S6, and each step is implemented as follows:
[0047] S1: Receive multiple translation sequences and determine at least one translation priority contained in each translation sequence;
[0048] S2: Determine the corresponding translation pointer for each translation priority;
[0049] S3: Obtain the resource parameters of the translation engine array;
[0050] S4: Based on the resource parameters, determine the schedulable translation pointers for each translation engine model;
[0051] S5: Each translation engine model determines at least one translation sequence as the sequence to be translated based on schedulable translation pointers;
[0052] S6: Use all arrays to be translated as input to the translation engine array, which outputs the translation enhancement result.
[0053] Figure 2 yes Figure 1 A schematic diagram of a preferred embodiment of the priority-based translation enhancement method is provided.
[0054] Next, in Figure 1 Based on the combination Figure 2 This paper specifically describes the implementation of each step of a priority-based translation enhancement method according to an embodiment of the present invention.
[0055] S1: Receive multiple translation sequences and determine at least one translation priority for each translation sequence.
[0056] In this step, the translation sequence can be generated in various ways, such as by inputting the text sequence to be translated through a text input device or by receiving the speech sequence to be translated through a voice input device.
[0057] The method can be executed by a translation terminal, which includes multiple translation engine models, and the multiple translation engines constitute a translation array.
[0058] An example of a translation engine model could be a machine translation engine that performs translation from language A to language B;
[0059] As an example, translation engine F1 can be a Chinese-to-English translation engine, and translation engine F2 can be an English-to-Chinese translation engine.
[0060] Different translation engines have different schedulable system resources, and even the same translation engine has different schedulable system resources at different times. This results in different forms of translation results that can be supported by the translation engines. Therefore, different times / different translation engines can only meet different translation priorities of the translation sequence.
[0061] Specifically, the translation priorities in the translation sequence include: plain text translation priority, plain audio translation priority, image and text translation priority, and video translation priority.
[0062] Step S1 determines at least one translation priority for each translation sequence based on the sequence tag or sequence keyword of each translation sequence.
[0063] In one embodiment, at least one translation priority contained in each translation sequence can be determined based on the sequence label of each translation sequence.
[0064] Specifically, when a text input device inputs a sequence of text to be translated, or when a voice input device receives a sequence of voice to be translated, a priority label can be added at the beginning of the sequence of text to be translated or the beginning of the sequence of voice to be translated to indicate its priority.
[0065] For example, assuming the sequence to be translated is XXXXXX, the received translation sequence can be "text output #XXXXXX", "voice output #XXXXXX", "voice output #text output #XXXXXX", etc. Priority labels can be identified through the special marker #.
[0066] In one embodiment, at least one translation priority can be determined for each translation sequence based on the sequence keywords of each translation sequence.
[0067] For example, assuming the sequence to be translated is "Next, show the chemical formula of hydrogen burning in oxygen" in the first language, then based on the pre-established sequence keyword database, the sequence keywords can be identified as "hydrogen, oxygen, chemical formula, show", thus determining the translation priority of the sequence to be translated, including at least the plain text translation priority and the graphic translation priority.
[0068] That is, the expected enhanced translation result of the sequence to be translated is the chemical formula image corresponding to the chemical formula "2H2O+O2=2H2O" in the second language. For example, the chemical formula image includes the molecular structure diagram of "H2O, O2, H2O".
[0069] For example, assuming the sequence to be translated is "playing a video of hydrogen burning in oxygen" in a third language, then based on the pre-established sequence keyword database, the sequence keywords can be identified as "hydrogen, oxygen, play, video", thus determining that the translation priority of the sequence to be translated includes at least the plain text translation priority and the video translation priority.
[0070] The expected enhanced translation result for this sequence to be translated is video translation, meaning the translation result should display the combustion scene corresponding to the chemical formula "2H2O+O2=2H2O" in the fourth language.
[0071] It is understandable that any sequence to be translated can include plain text translation priorities, while other translation priorities are optional.
[0072] In other words, any translation engine model can output plain text translation results for the current sequence to be translated at any time, because outputting plain text translation requires the least amount of resources to be scheduled, and the system can almost always meet this requirement.
[0073] However, if other forms of enhanced translation output are required, the system's resource availability needs to be considered.
[0074] Therefore, we proceed to step S2 of the present invention: determine the corresponding translation pointer for each translation priority.
[0075] To avoid repeatedly reading sequence data when performing enhanced translation outputs with different priorities for the same sequence to be translated, this embodiment creatively determines a corresponding translation pointer for each translation priority.
[0076] Specifically, in step S1, after receiving multiple translation sequences, the translation sequences are stored in storage units. Assume the received translation sequences include SeQ1, SeQ2, ..., SeQN, and their corresponding storage units are MeQ1, MeQ2, ..., MeQN;
[0077] If the translation sequence SeQi contains plain text translation priority and plain audio translation priority;
[0078] Then the pointer P0 for plain text translation priority and the pointer P1 for plain audio translation priority are defined as the address of the storage unit MeQi where the translation sequence SeQi is located;
[0079] The following is represented using computer pseudocode (C++):
[0080] P0 = &MeQi; P1 = &MeQi; where & represents taking the address.
[0081] Accessing data via pointers avoids the need for frequent readings of the data itself.
[0082] Therefore, in step S2, the translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority.
[0083] Next, obtaining the resource parameters of the translation engine array in step S3 includes:
[0084] Obtain the real-time resource parameters of the translation engine array, and predict the available resource parameters of the translation engine array within a future preset time period based on the real-time resource parameters;
[0085] The translation engine array includes multiple translation engine models, each of which can achieve translation enhancement results corresponding to different translation priorities.
[0086] As a specific example, the resource parameters include the schedulable resource parameters of the physical host where the translation array engine is located, including CPU cores, GPU cores, graphics cards, sound cards, network resources, etc.
[0087] In this embodiment, not only real-time resource parameters are considered, but also available resource parameters within a preset time period in the future. This is because, according to statistics, when users expect diversified translation enhancement results for multimedia, they can tolerate a certain degree of delay.
[0088] For example, when a user expects the translation result of the currently input sequence to be translated to be a video translation, the user can tolerate a wait of up to 500ms. Therefore, in this embodiment, not only is the real-time resource parameter of the translation engine array obtained, but also the available resource parameter of the translation engine array in the next 500ms is predicted based on the real-time resource parameter.
[0089] Next, proceed to step S4: Based on the resource parameters, determine the schedulable translation pointers for each translation engine model.
[0090] Specifically, determine the real-time resource parameters and available resource parameters for each translation engine model within a preset future time period;
[0091] When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.
[0092] As a specific example, for the translation engine model F1, if it can schedule enough CPU and sound card resources in real time, and / or can schedule enough CPU and sound card resources in a future preset time period, then the translation engine model F1 supports translation enhancement with plain text translation priority and plain audio translation priority.
[0093] Assuming that the sequence SeQ1 contains plain text translation priorities and SeQ2 contains plain audio translation priorities, then the first translation pointers &MeQ1 and &MeQ2 corresponding to the plain text translation priorities and the plain audio translation priorities are used as the schedulable translation pointers of the translation engine model F1.
[0094] At this point, step S5 takes the translation sequence including the first translation priority as the sequence to be translated; that is, sequences SeQ1 and SeQ2 are taken as the sequence to be translated.
[0095] It should be noted that after the sequence to be translated is determined at this point, it is accessed through pointers;
[0096] Finally, step S6 includes: inputting the sequence to be translated into a first translation engine model, and the first translation engine model outputting a translation enhancement result corresponding to a first translation priority.
[0097] As another example, for the translation engine model F2, if it can schedule sufficient CPU resources, GPU resources and sound card resources in real time, and / or can schedule sufficient CPU resources, GPU resources and sound card resources in a future preset time period, then the translation engine model F2 supports translation enhancement with priority for plain text translation, priority for plain audio translation, priority for image and text translation, and priority for video translation.
[0098] Assuming that sequence SeQ4 contains text-image translation priority and SeQ7 contains video translation priority, then the first translation pointers &MeQ4 and &MeQ7 corresponding to the text-image translation priority and the video translation priority, respectively, are used as the schedulable translation pointers of the translation engine model F2.
[0099] The other steps are similar to those in the above embodiments, and therefore will not be repeated.
[0100] Figures 1-2 The method described herein can be represented as computer program instructions, stored in the memory of an electronic device, and then executed by the processor of the electronic device to implement all or part of the steps of the method.
[0101] Taking step S3 as an example, the resource parameters of the translation engine array can be the number of available CPU cores NumCPU, the number of available GPU cores NumGPU, the amount of available graphics card resources MemV, etc.; then, predict the minimum number of available CPU cores NumTCPU, the minimum number of available GPU cores NumTGPU, the minimum amount of available graphics card resources MemTV, etc. within a preset time period T in the future. The prediction can be based on a variety of prediction models, such as statistical fitting methods, time series prediction models, etc., which will not be elaborated in detail in this invention.
[0102] The preset future time period T can be based on the user's maximum tolerance time T calculated empirically. max set up.
[0103] Preferably, if And if NumGPU≥4 and NumCPU≥2, then
[0104]
[0105] like And if NumGPU≥4 and NumCPU≥2, then
[0106]
[0107] In other cases, T = T max .
[0108] See next. Figures 3-4 .
[0109] Figure 3 This diagram illustrates the architecture of a priority-based translation enhancement system according to an embodiment of the present invention.
[0110] exist Figure 3 In this context, the translation enhancement system includes a translation engine array, which comprises multiple translation engine models; each translation engine model can achieve translation enhancement results corresponding to different translation priorities.
[0111] The system includes several functional units that can implement the above method steps, as follows:
[0112] A sequence receiving unit, wherein the receiving unit is used to receive and store the translated sequence;
[0113] A priority identification unit is used to identify the translation priority contained in each translation sequence, wherein the translation priority includes plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority;
[0114] A pointer determination unit determines a corresponding translation pointer based on each translation priority. The translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority.
[0115] A real-time resource parameter acquisition unit is used to acquire the real-time resource parameters of the translation engine array;
[0116] Available resource parameter prediction unit, which predicts available resource parameters of the translation engine array within a future preset time period based on real-time resource parameter prediction;
[0117] A schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters.
[0118] The sequence to be translated is determined by a unit that determines at least one translation sequence as a sequence to be translated based on a schedulable translation pointer for each translation engine model.
[0119] The translation enhancement output unit takes all the arrays to be translated as input to the translation engine array, and the translation engine array outputs the translation enhancement result.
[0120] In a specific implementation, the priority identification unit determines at least one translation priority contained in each translation sequence based on the sequence label or sequence keyword of each translation sequence.
[0121] exist Figure 3 The system also includes a storage unit for storing sequences to be translated received by the sequence receiving unit. Each sequence to be translated has a storage address, which can be used as a pointer.
[0122] Each translation priority corresponds to a translation pointer that points to the storage address of the translation sequence with that priority.
[0123] Figure 4 The diagram further illustrates the architecture of the translation pointers corresponding to the translation priority determination in various embodiments of the present invention.
[0124] The translation priorities in the translation sequence include: plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority;
[0125] Therefore, the translation pointers corresponding to each translation priority include plain text translation priority pointer P0, plain audio translation priority pointer P1, image and text translation priority pointer P2, and video translation priority pointer P3.
[0126] Figure 4In the process, the received translation sequences include SeQ1, SeQ2, ..., SeQN, and their corresponding storage units are MeQ1, MeQ2, ..., MeQN;
[0127] If the translation sequence SeQ1 includes plain text translation priority and image-text translation priority;
[0128] Then the pointer P0 for plain text translation priority and the pointer P2 for image-text translation priority are defined as the address of the storage unit MeQ1 where the translation sequence SeQ1 is located;
[0129] The following is represented using computer pseudocode (C++):
[0130] P0 = &MeQ1; P2 = &MeQ1; where & indicates taking the address.
[0131] like Figure 4 As shown, the plain text translation priority pointer P0 and the image-text translation priority pointer P2 point to the storage unit MeQ1 of SeQ1.
[0132] The relationships between other pointers are similar and need not be elaborated upon. That is, the translation pointer corresponding to each translation priority points to the memory address of the translation sequence with that priority.
[0133] In this way, when the corresponding translation priority can be executed later, for example, when the translation engine model supports enhanced translation output with text-image translation priority, the pointer P1 of text-image translation priority can be found, and at least one translation sequence can be obtained as the sequence to be translated based on P1.
[0134] The schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters, specifically including:
[0135] When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.
[0136] For specific implementation examples, please refer to the foregoing method implementation examples, i.e. Figures 1-2 The steps are explained below.
[0137] As can be seen, compared with the prior art, the improvements of the present invention include at least the following:
[0138] (1) After receiving the translation sequence, first determine at least one translation priority contained in each translation sequence; then determine the corresponding translation pointer for each translation priority, and subsequently call the translation array available for the translation priority through the translation pointer, thus avoiding frequent data writing and reading;
[0139] (2) After obtaining the resource parameters of the translation engine array, the schedulable translation pointers of each translation engine model are determined based on the resource parameters, which can select the translation priority supported by the system resources and avoid translation waiting or lag when resources are insufficient;
[0140] (3) The system resources not only take into account the current real-time resource parameters, but also the available resource parameters within a preset time period in the future, so that the available resource parameters within the user's tolerable waiting time period are also taken into account, which conforms to the objective reality of the scenario and maximizes the use of system resources.
[0141] (4) Each translation engine model determines at least one translation sequence as a sequence to be translated based on a schedulable translation pointer, and uses all the sequences to be translated as input to the translation engine array. The translation engine array outputs translation enhancement results, so that it is not necessary to schedule resources to achieve enhanced translation for every sequence to be translated, but only when enough resources can be scheduled to support enhanced translation. When the current translation resources do not support enhanced translation, the user can also get the minimum normal translation output.
[0142] It is understood that each embodiment of the present invention can achieve one of the effects individually, and multiple embodiments of the present invention can be combined to achieve all the above effects. However, it is not required that each embodiment of the present invention achieve all the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the prior art.
[0143] For any module structures not specifically defined in this invention, the existing technical specifications shall prevail. The existing technical specifications mentioned in the foregoing background and specific embodiments sections are considered part of this invention and are used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual contents of the claims.
Claims
1. A priority-based translation enhancement method, characterized in that, The method includes the following steps: S1: Receive multiple translation sequences and determine at least one translation priority for each translation sequence based on the sequence tag or sequence keyword of each translation sequence; The translation priorities in the translation sequence include: plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority; S2: Determine the corresponding translation pointer for each translation priority; the translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority; S3: Obtain the resource parameters of the translation engine array; S4: Based on the resource parameters, determine the schedulable translation pointers for each translation engine model; S5: Each translation engine model determines at least one translation sequence as the sequence to be translated based on schedulable translation pointers; S6: Use all arrays to be translated as input to the translation engine array, which outputs the translation enhancement result.
2. The priority-based translation enhancement method as described in claim 1, characterized in that, In step S1, the text sequence to be translated is entered through a text input device as the translation sequence, and / or the speech sequence to be translated is received through a voice input device as the translation sequence.
3. The priority-based translation enhancement method as described in claim 2, characterized in that... : In step S2, if the translation sequence includes plain text translation priority and plain audio translation priority, then the pointers P0 and P1 of the plain text translation priority are defined as the addresses of the storage units where the translation sequence is located.
4. The priority-based translation enhancement method as described in claim 2, characterized in that... : The resource parameters of the translation engine array obtained in step S3 include: Obtain the real-time resource parameters of the translation engine array, and predict the available resource parameters of the translation engine array within a future preset time period based on the real-time resource parameters; The translation engine array includes multiple translation engine models, each of which can achieve translation enhancement results corresponding to different translation priorities.
5. The priority-based translation enhancement method as described in claim 2, characterized in that... : Step S4, which determines the schedulable translation pointers for each translation engine model based on the resource parameters, specifically includes: Determine the real-time resource parameters and available resource parameters for each translation engine model within a preset future time period; When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.
6. The priority-based translation enhancement method as described in claim 5, characterized in that... : Step S5 includes: taking the translation sequence including the first translation priority as the sequence to be translated; Step S6 includes: inputting the sequence to be translated into a first translation engine model, and the first translation engine model outputting a translation enhancement result corresponding to a first translation priority.
7. A priority-based translation enhancement system, the translation enhancement system comprising a translation engine array, the translation engine array comprising multiple translation engine models; Its features are, The system also includes: A sequence receiving unit, wherein the receiving unit is used to receive and store the translated sequence; A priority identification unit is used to identify the translation priority contained in each translation sequence, wherein the translation priority includes plain text translation priority, plain audio translation priority, image-text translation priority, and video translation priority; A pointer determination unit determines a corresponding translation pointer based on each translation priority. The translation pointer corresponding to each translation priority is used to point to the storage address of the translation sequence with that translation priority. A real-time resource parameter acquisition unit is used to acquire the real-time resource parameters of the translation engine array; Available resource parameter prediction unit, which predicts available resource parameters of the translation engine array within a future preset time period based on real-time resource parameter prediction; A schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters. The unit that determines the sequence to be translated determines at least one translation sequence as the sequence to be translated based on a schedulable translation pointer for each translation engine model. The translation enhancement output unit takes all the arrays to be translated as input to the translation engine array, and the translation engine array outputs the translation enhancement result.
8. A priority-based translation enhancement system as described in claim 7, characterized in that: The priority identification unit determines at least one translation priority for each translation sequence based on the sequence tag or sequence keyword of each translation sequence.
9. A priority-based translation enhancement system as described in claim 7, characterized in that: Each translation engine model can achieve translation enhancement results corresponding to different translation priorities.
10. A priority-based translation enhancement system as described in claim 7 or 9, characterized in that: The schedulable pointer determination unit is used to determine the schedulable translation pointers for each translation engine model based on the real-time resource parameters and the available resource parameters, specifically including: When the real-time resource parameters of the first translation engine model and the available resource parameters within a preset future time period support the translation enhancement of the first translation priority, the first translation pointer corresponding to the first translation priority is used as the schedulable translation pointer of the first translation engine model.