Search

CN-121997947-A - Game translation collaborative management and monitoring method, system, equipment and medium

CN121997947ACN 121997947 ACN121997947 ACN 121997947ACN-121997947-A

Abstract

The invention discloses a method, a system, equipment and a medium for collaborative management and monitoring of game translation, wherein the method specifically comprises the steps of constructing a term library version tree according to preprocessed game resources; classifying and marking game texts according to game contexts based on a term library version tree, constructing a game context vector space by combining a game screenshot, a 3D scene model and an audio dialogue, wherein the game context vector space is used for generating unique context fingerprints for each game text, constructing a capability map based on translator historical translation data, performing intelligent task allocation by combining text types and context features identified by the context fingerprints, recommending optimal translator combinations for specific game projects, and marking high-risk texts and automatically generating alternative translation methods by combining a pre-constructed translation risk prediction model. The invention constructs a comprehensive and efficient game translation collaborative management and monitoring system, and remarkably improves the quality and efficiency of game translation.

Inventors

  • Du Hongjia
  • XU ZHIQIN

Assignees

  • 安徽三七极域网络科技有限公司

Dates

Publication Date
20260508
Application Date
20260206

Claims (10)

  1. 1. A game translation collaborative management and monitoring method is characterized by comprising the following steps: extracting and preprocessing game resources through a special analysis plug-in of a game engine, constructing a term library version tree according to the preprocessed game resources, and binding and associating the term library version tree with a game version number; classifying and marking game texts according to game contexts based on the term library version tree, and constructing a game context vector space by combining game screenshot, a 3D scene model and an audio dialogue, wherein the game context vector space is used for generating unique context fingerprints for each segment of game texts; constructing a capability map based on translator historical translation data, performing intelligent task allocation by combining text categories identified by the context fingerprints and the context features, and recommending optimal translator combinations for specific game items; in the context-aided translation process according to the optimal translator combination, marking high-risk texts and automatically generating alternative translations by utilizing visual references provided by a game context vector space and combining a pre-constructed translation risk prediction model; And performing quality inspection on the translation text through the integrated GQA model at a real-time collaborative monitoring interface, and feeding the inspection result of the GQA model back to the term library version tree.
  2. 2. The method according to claim 1, wherein the steps of extracting and preprocessing game resources through the game engine-specific parsing plug-in, constructing a term library version tree according to the preprocessed game resources, and binding and associating the term library version tree with the game version number include: extracting game resources through a special analytic plug-in of a game engine and preprocessing, and establishing association mapping of text resources and game context resources; Constructing a term library version tree based on the association mapping; Binding and associating the term library version tree with the game version number, and establishing a term change tracking mechanism between versions; When a game version update is detected, automatically identifying term differences according to a term change tracking mechanism, and triggering a translation memory update of associated text in a term library version tree based on the term differences.
  3. 3. The method according to claim 1, wherein the classifying and marking game text according to game context based on the term library version tree and constructing a game context vector space in combination with game shots, 3D scene models and audio dialogs specifically comprises: Based on the term context information stored in the term library version tree, classifying and marking the game text in a multi-dimensional manner according to the term category attribute and the text function position, and determining the game function attribute of the game text; based on game function attributes, combining a game screenshot, a 3D scene model and an audio dialogue, and constructing a game context vector space by adopting a layered coding architecture; Mapping the classified game text into a game context vector space, converting the high-dimensional context vector into fingerprint codes with fixed length through a hash algorithm, and generating unique context fingerprints for each game text.
  4. 4. The method of claim 1, wherein constructing a capability map based on translator historical translation data, and performing intelligent tasking in combination with text categories identified by context fingerprints and context features, recommends optimal translator combinations for a particular game item, comprises: performing multidimensional analysis based on the translator historical translation data, and constructing a translator capacity map reflecting the professional capacity of the translator; generating a demand vector of the translation task according to the text category and the contextual characteristics identified by the contextual fingerprint; Matching calculation is carried out on the demand vector and the translator capacity vector in the translator capacity map, and an adaptation degree result between the translator and the translation task is determined; Candidate translators are screened for the specific game item based on the fitness result, and the optimal translator combination recommendation scheme is generated according to the cooperation history and time zone distribution among the candidate translators and the workload distribution.
  5. 5. The method according to claim 1, wherein the marking high risk text and automatically generating alternative translations using visual references provided by the game context vector space in combination with pre-constructed translation risk prediction models, comprises: invoking a multi-modal visual reference provided by the game context vector space to provide an immersive translation context for the translator; Performing multidimensional risk assessment on the current translation text according to a pre-constructed translation risk prediction model, and identifying and marking high-risk text fragments; And combining strategies of a rule base, translation memory and a neural network aiming at the marked high-risk text to generate a plurality of alternative translation schemes, and sequencing the alternative translation schemes in priority through a quality evaluation algorithm.
  6. 6. The method of claim 5, wherein the pre-constructing of the translation risk prediction model comprises: Establishing a multi-level risk feature identification system, and determining various risk features in game localization through the risk feature identification system; Based on various risk characteristics, training samples are collected from a historical error database, expert annotation corpus and cross-culture comparison data and are annotated, and a basic data set is constructed; According to the basic data set, combining with the field knowledge of the game industry, training a hierarchical mixed neural model framework by adopting a multi-task learning framework to generate a translation risk prediction model, wherein the translation risk prediction model comprises a feature extraction layer, an attention mechanism layer and a multi-task output layer.
  7. 7. The method according to any one of claims 1 to 6, wherein the quality inspection of the translated text is performed on the real-time collaborative monitoring interface through an integrated GQA model, and the inspection result of the GQA model is fed back to a term library version tree, specifically comprising: Integrating a GQA model in a real-time collaborative monitoring interface, and carrying out real-time quality assessment on the translation text to obtain a quality assessment result; Based on the quality evaluation result, carrying out severity grading and classification treatment on the abnormal problems found in the translation text, and determining the treatment priority of the abnormal problems; mapping the abnormal problems to corresponding nodes of the term library version tree one by one according to the processing priority, and establishing a traceability path from the quality problem to the term root; Based on the traceability path, analyzing the influence range of the term change on the translation text, and formulating a progressive update strategy.
  8. 8. A game translation collaborative management and monitoring system, the system comprising: the first processing module is used for extracting and preprocessing game resources through the special analysis plug-in of the game engine, constructing a term library version tree according to the preprocessed game resources, and binding and associating the term library version tree with a game version number; the second processing module is used for classifying and marking game texts according to game contexts based on the term library version tree and constructing a game context vector space by combining the game screenshot, the 3D scene model and the audio dialogue, wherein the game context vector space is used for generating unique context fingerprints for each segment of game texts; The third processing module is used for constructing a capability map based on translator historical translation data, performing intelligent task allocation by combining text types identified by the context fingerprints and the context features, and recommending optimal translator combinations for specific game items; The fourth processing module is used for marking high-risk texts and automatically generating alternative translation methods by utilizing visual references provided by a game context vector space and combining a pre-constructed translation risk prediction model in the context-aided translation process according to the optimal translator combination; And the fifth processing module is used for carrying out quality inspection on the translation text through the integrated GQA model in the real-time collaborative monitoring interface and feeding back the inspection result of the GQA model to the term library version tree.
  9. 9. A computer device comprising a memory and a processor and a computer program stored on the memory, which when executed on the processor implements the game translation co-management and monitoring method according to any one of claims 1 to 7.
  10. 10. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the game translation co-management and monitoring method according to any one of claims 1 to 7.

Description

Game translation collaborative management and monitoring method, system, equipment and medium Technical Field The invention relates to the technical field of artificial intelligence, in particular to a game translation collaborative management and monitoring method, a system, equipment and a medium. Background With the vigorous development of the global game market, game localization translation has become a key link for the game industry to expand the international market and promote the global user game experience. Currently, game localization involves a variety of file types, covering UI text, scenario scripts, item descriptions, achievement systems, and the like. However, conventional translation management systems, when dealing with game localization needs, expose a number of limitations, including: 1. File type handling difficulties conventional translation management systems have difficulty in uniformly handling the multiple file types involved in a game. Different file types have unique formats and structures, such as a special format of a Unity prefab file and the like, and a traditional system cannot directly and effectively translate the file types, so that format errors or information loss occur in the translation process, and the integrity and playability of a game are affected. 2. The term management dilemma is that game terms are highly industry specific and existing systems lack specialized game-specific term library management. The inconsistent terms of various languages of games lead to the difference of the translation results in terms of semantic expression and style, destroy the integrity and consistency of games and reduce the game experience of players. 3. The multi-language progress collaborative management is lacking, the synchronous release requirement of the multi-language version is increasingly strong, but the traditional system lacks an effective multi-language progress collaborative management mechanism. The progress of the multilingual version is opaque, the information communication among the translation teams of various languages is not smooth, the work progress is difficult to coordinate, the global synchronous release of the game is affected, and the market precedent is missed. 4. The original format is destroyed, namely, the game text often contains variables and code fragments, and when the conventional translation tool processes the text, the original format is easily destroyed, so that game function abnormality or display error is caused, and the workload and cost of subsequent repair are increased. 5. The increment translation management efficiency is low, namely the increment translation management efficiency is low when the game is updated frequently and the version iterates. The traditional system can not effectively identify version change content, so that the translation repetition rate is high, a great deal of manpower and time resources are wasted, and the efficiency and quality of translation work are reduced. 6. The lack of context auxiliary translation is that auxiliary translation support for game contexts (roles and scenes) is lacking, and a translator is difficult to accurately understand the specific situation of a game text, so that a translation result is inconsistent with the actual content of the game, and the accuracy of translation and the immersion of the game are affected. 7. And the difficulty of style consistency is that the multi-type text of the game is distributed and managed, and the style consistency is difficult to maintain. The different text types lack unified style guidance in the translation process, so that the language styles of the whole game are uneven, and the cultural quality and the attraction of the game are reduced. 8. The quality inspection rule is insufficient, and the quality inspection rule special to game text, such as character length limitation, is lacked. The traditional quality inspection method cannot meet the special requirements of game localization, so that the translation result does not accord with game specifications in the aspects of format, content and the like, and normal running and user experience of the game are affected. Disclosure of Invention The invention aims to provide a game translation collaborative management and monitoring method, a system, equipment and a medium, which construct a comprehensive and efficient game translation collaborative management and monitoring system, and remarkably improve the quality and efficiency of game translation so as to solve at least one of the problems in the prior art. In a first aspect, the present invention provides a game translation collaborative management and monitoring method, where the method specifically includes: extracting and preprocessing game resources through a special analysis plug-in of a game engine, constructing a term library version tree according to the preprocessed game resources, and binding and associating the term library version tree