Intelligent Chinese and Russian bilingual electrical automation interaction system

Abstract

The invention relates to the technical field of electrical automation interaction, and discloses an intelligent Chinese and Russian bilingual electrical automation interaction system, and a module for bilingual electrical interaction comprises a voice recognition and processing module which captures the voice input of a user, converts the voice input into text information for subsequent processing, and carries out the noise reduction of a voice signal; the text preprocessing module is used for carrying out word segmentation, stop word removal and word form reduction on the obtained text information and providing standardized input for subsequent translation and analysis; the situation understanding and adapting module is used for identifying situation characteristics, dialogue themes, participant roles and emotional colors of the Chinese Russian dialogues and dynamically adjusting translation strategies; and the culture adaptability translation module is used for carrying out culture adaptability adjustment on the translation result to avoid culture misunderstanding or fraudulent caused by direct translation.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical automation interaction, and particularly to an intelligent Chinese-Russian bilingual electrical automation interaction system. Background Art

[0002] With the rapid development of the global economy, the cooperation between China and Russia in many fields such as energy, manufacturing, and infrastructure construction has become increasingly close. As one of the core technologies of modern industry, electrical automation requires efficient cross-language communication and control in cross-border projects, which has promoted the demand for bilingual interaction systems;

[0003] The electrical automation industry has experienced a transformation from traditional control to digital, networked, and intelligent. The application of technologies such as smart factories, the Internet of Things (IoT), and big data analysis requires electrical automation systems to have a higher level of intelligent interaction capabilities, including functions such as cross-language processing, complex data analysis, and autonomous decision-making.

[0004] At present, due to the strong professionalism in the field of electrical automation and the presence of a large number of specific terms, existing systems may have inaccurate or context understanding errors when translating these professional terms due to the differences between Chinese and Russian, which affects the effective communication between the two parties and leads to interaction differences. Therefore, an intelligent Chinese-Russian bilingual electrical automation interaction system is proposed here. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem of inaccurate or context understanding errors existing in the prior art due to the differences between Chinese and Russian, and to propose an intelligent Chinese-Russian bilingual electrical automation interaction system.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] An intelligent Chinese-Russian bilingual electrical automation interaction system, the modules for bilingual electrical interaction include:

[0008] Speech recognition and processing module: Capture the user's speech input, convert it into text information for subsequent processing, and at the same time, perform noise reduction processing on the speech signal;

[0009] Text preprocessing module: Perform word segmentation, stop word removal, and word form reduction on the obtained text information to provide a standardized input for subsequent translation and analysis;

[0010] Situation understanding and adaptation module: Identify the situational characteristics of Chinese-Russian conversations, such as conversation topics, participant roles, and emotional colors, and dynamically adjust the translation strategy;

[0011] Cultural Adaptability Translation Module: This module establishes a cultural difference database and adjusts translation results based on contextual feature information provided by the context understanding and adaptation module. For known culturally specific expressions, it applies corresponding replacement rules from the cultural difference database and assigns a cultural adaptability score to the pre-processed text. A threshold is set for the cultural adaptability score. Score scores below the threshold are prioritized for literal translation, while scores above the threshold are translated to incorporate the target culture's expression conventions.

[0012] Real-time terminology query and update module: connects to online professional terminology databases or cloud services, accepts professional terminology query requests encountered by the cultural adaptability translation module during the translation process, and updates the professional terminology information based on the query requests;

[0013] Electrical automation data parsing module: This module receives translation text containing electrical automation data from the real-time terminology query and update module, parses and processes the data format in the electrical automation field, and transmits the parsed electrical automation data to the electrical interaction execution module.

[0014] Electrical interaction execution module: Based on the translation and verification of control instructions, they are converted into operation instructions that can be recognized by electrical equipment after parsing, and sent to the target device through the communication protocol to achieve remote control.

[0015] The speech recognition and processing module uses an array of multiple microphones to capture sounds from a specific direction through beamforming while suppressing noise from other directions;

[0016] Split the continuous speech signal into short time frames and apply a window function to each frame;

[0017] A feature vector reflecting the speech characteristics is extracted from each frame. Using a deep neural network, through training with a large amount of noisy speech and corresponding clean speech, the noise model and speech characteristics are learned, the language is denoised, and the text preprocessing module is activated.

[0018] The text preprocessing module performs Chinese and Russian word segmentation;

[0019] Chinese word segmentation: The optimal segmentation point is obtained based on the state transition probability of the hidden Markov model to achieve word segmentation;

[0020] Russian word segmentation: Use spaces as word separators, consider affix processing, and use machine learning-based methods to identify word boundaries and morphological changes;

[0021] Create a bilingual stop word list for Chinese and Russian, including common words like "的" and "和" in Chinese, and "и" and "в" in Russian that don't contribute much to the meaning. Simply traverse the word segmentation results and remove the stop words from the list.

[0022] Perform word lemmatization on Russian, and set the observation sequence O = o1, o2,..., o n and the hidden state sequence S = s1, s2,..., s n , as well as the state transition probability a i,j , from state s i to s j and the observation probability b j (o i ) observed at state s j when observing o i . The steps for the hidden state path S * are as follows:

[0023]

[0024] Among them, δ j (t) represents the cumulative probability of the best path ending with state s j up to time t. The initial condition is δ j (0) = 1 or 0, and backtracking is used to find the optimal path.

[0025] The situation understanding and adaptation module is based on the results of cleaning, sentence splitting, and word segmentation of Chinese and Russian data in the text preprocessing module;

[0026] Use the multilingual version of BERT for word embedding;

[0027] Based on the mBERT multilingual pre-training model, train a bilingual classifier for identifying specific scenarios in the field of electrical automation, including equipment fault reporting, operation guidance, and technical consultation;

[0028] According to the characteristics of bilingual data, adjust the classification threshold, perform sentiment annotation on the bilingual data, and the model identifies the emotions of users in different languages;

[0029] Train a bilingual emotion recognition model, combining mBERT embedding and specific emotion recognition algorithms;

[0030] Use the TF-IDF method to extract keywords from bilingual data. Considering the professional terms in the field of electrical automation, a customized dictionary is used to optimize the quality of keyword extraction;

[0031] Design a state machine for processing bilingual input and output, and dynamically adjust the dialogue strategy according to the identified scenarios, emotions, and keywords;

[0032] Develop or integrate a generation model that can generate Chinese-Russian bilingual responses.

[0033] A cultural difference database is set up in the cultural adaptation translation module;

[0034] The cultural adaptability translation module collects idioms, proverbs commonly used in the field of electrical automation in China and Russia and their corresponding meanings;

[0035] Sort out the polite expressions of the two countries in the scenarios of technical exchanges, asking for help, and expressing gratitude;

[0036] Record specific idiomatic expressions that may affect understanding in daily communication;

[0037] Structurize the data and design a database model, including original text, translation, usage scenarios, cultural notes, and emotional color fields;

[0038] Design a cultural adaptability scoring function, including scoring elements:

[0039] Context fit: Evaluate the applicability of the translation in a specific professional scenario;

[0040] Emotion retention: Measure whether the emotional expression after translation is consistent with the original text;

[0041] Cultural sensitivity: Check whether the translation avoids cultural taboos or misunderstandings;

[0042] Language fluency: Evaluate the natural fluency of the translated text;

[0043] Let Sc be the total score of cultural adaptability, and the weights of each element are w1, w2, w3, w4, and w1, w2, w3, w4 = 1, then:

[0044] Sc = w1·S context + w2·S emotion + w3·S sensitivity + w4·S fluency

[0045] Where S context ,S emotion ,S sensitivity ,S fluency are the scores of context fit, emotion retention, cultural sensitivity, and language fluency, respectively, between 0 and 1;

[0046] For known culturally specific expressions, directly apply the corresponding replacement rules in the database;

[0047] According to the score of Sc, dynamically adjust the translation strategy:

[0048] If Sc is lower than the set threshold, adopt a more conservative translation strategy, focusing on literal translation to reduce cultural misunderstandings;

[0049] If Sc is high, adopt a more flexible strategy, integrating more expression habits of the target culture to improve the naturalness and intimacy of communication.

[0050] The real-time query and update module of the system establishes a real-time or regular synchronization mechanism with the cloud term database, automatically pulls the latest term data, and uploads the newly recognized professional vocabulary in the system to the cloud at the same time to achieve two-way updates.

[0051] The electrical automation data parsing module includes a control signal translation module;

[0052] The electrical automation data parsing module receives the original data from the electrical automation device through the network interface or serial port and detects the communication protocol type of the data packet;

[0053] Combined with the specific data structure model in the field of electrical automation, the received original data is parsed into structured information;

[0054] For the parsed key data tags and descriptions, through the cultural adaptation translation module, combined with the professional term database and the cultural difference database, the tags causing cultural misunderstandings are adjusted or annotated;

[0055] According to industry standards, the parsed data is standardized to eliminate the specific representation differences of equipment manufacturers;

[0056] The control signal translation module receives the control instructions from the user or the upper system, analyzes the semantic structure and operation intention of the instructions using NLP technology, evaluates the cultural fit of the translation according to the cultural adaptation scoring function, and adjusts the translation strategy to avoid misunderstandings caused by cultural differences.

[0057] The electrical interaction execution module receives the translated and verified control instructions, which come from the commands input by the user through voice or text and have undergone cultural adaptation translation and accurate matching of professional terms;

[0058] Parse the control instructions and convert them into the internal instruction format recognizable by the electrical automation system.

[0059] The present invention has the following beneficial effects:

[0060] 1. In the present invention, by collecting and sorting out the specific idioms, proverbs, courtesy expressions and idiomatic expressions in the field of electrical automation, the database provides rich cultural context information for translation, ensuring that the translation is not only accurate in language but also culturally appropriate, and avoiding misunderstandings caused by literal translation.

[0061] 2. In the present invention, by introducing the cultural adaptation scoring function, the system can evaluate the translation quality from multiple dimensions (context fit, emotion retention, cultural sensitivity and language fluency), ensuring the comprehensiveness and appropriateness of information transmission, and enhancing the effective communication between users with different cultural backgrounds.

[0062] 3. In the present invention, through the integrated innovation of NLP technology and bilingual interaction optimization, the system can accurately identify and understand the instructions of Chinese-Russian bilingual users, reduce language barriers, and improve the efficiency and convenience of cross-border and cross-cultural work. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a system block diagram of an intelligent Chinese-Russian bilingual electrical automation interaction system proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] Embodiment 1

[0066] As Figure 1 shown, an intelligent Chinese-Russian bilingual electrical automation interaction system proposed by the present invention, the modules for bilingual electrical interaction include:

[0067] Speech recognition and processing module: Capture the user's voice input, convert it into text information for subsequent processing, and at the same time, perform noise reduction processing on the voice signal;

[0068] Text preprocessing module: Perform word segmentation, stop word removal, and word form reduction on the obtained text information to provide a standardized input for subsequent translation and analysis;

[0069] Situation understanding and adaptation module: Identify the situation characteristics of Chinese-Russian conversations, such as conversation topics, participant roles, and emotional colors, and dynamically adjust the translation strategy;

[0070] Cultural adaptation translation module: Establish a cultural difference database, adjust the translation results according to the situation characteristic information provided by the situation understanding and adaptation module, for known culturally specific expressions, apply the corresponding replacement rules in the cultural difference database, perform a cultural adaptation score on the preprocessed text, set a threshold for the cultural adaptation score, focus on literal translation for those below the threshold, and incorporate the expression habits of the target culture for translation for those above the threshold;

[0071] Term real-time query and update module: Connect to an online professional term database or cloud service, accept the professional vocabulary query requests encountered in the translation process by the cultural adaptation translation module, and update the professional vocabulary information according to the query requests;

[0072] Electrical automation data parsing module: This module receives translation text containing electrical automation data from the real-time terminology query and update module, parses and processes the data format in the electrical automation field, and transmits the parsed electrical automation data to the electrical interaction execution module.

[0073] Electrical interaction execution module: Based on the translation and verification of control instructions, they are converted into operation instructions that can be recognized by electrical equipment after parsing, and sent to the target device through the communication protocol to achieve remote control.

[0074] The speech recognition and processing module uses an array of multiple microphones to capture sounds from a specific direction through beamforming while suppressing noise from other directions;

[0075] Split the continuous speech signal into short time frames and apply a window function to each frame;

[0076] A feature vector reflecting the speech characteristics is extracted from each frame. Using a deep neural network, through training with a large amount of noisy speech and corresponding clean speech, the noise model and speech characteristics are learned, the language is denoised, and the text preprocessing module is activated.

[0077] The text preprocessing module performs Chinese and Russian word segmentation;

[0078] Chinese word segmentation: The optimal segmentation point is obtained based on the state transition probability of the hidden Markov model to achieve word segmentation;

[0079] Russian word segmentation: Use spaces as word separators, consider affix processing, and use machine learning-based methods to identify word boundaries and morphological changes;

[0080] Create a bilingual stop word list for Chinese and Russian, including common words like "的" and "和" in Chinese, and "и" and "в" in Russian that don't contribute much to the meaning. Simply traverse the word segmentation results and remove the stop words from the list.

[0081] Perform word restoration on Russian text and set the observation sequence O=o1,o2,...,o n and hidden state sequence S=s1,s2,...,s n , and the state transition probability a i,j , from state s i Transfer to s j The probability and observation probability b j (o i ) in state s j Observed below o i The probability of hidden state path S * The steps are:

[0082]

[0083] Among them, δ j (t) represents the cumulative probability of the best path ending with state s j up to time t. The initial condition is δ j (0) = 1 or 0, and backtracking is used to find the optimal path.

[0084] The situation understanding and adaptation module is based on the results of cleaning, sentence splitting, and word segmentation of Chinese and Russian data in the text preprocessing module;

[0085] Use the multilingual version of BERT for word embedding;

[0086] Based on the mBERT multilingual pre-training model, train a bilingual classifier for identifying specific scenarios in the field of electrical automation. The scenarios include equipment failure repair requests, operation guides, and technical consultations;

[0087] According to the characteristics of bilingual data, adjust the classification threshold, perform sentiment annotation on the bilingual data, and the model identifies the emotions of users in different languages;

[0088] Train a bilingual emotion recognition model by combining mBERT embedding and specific emotion recognition algorithms;

[0089] Use the TF-IDF method to extract keywords from bilingual data. Considering the professional terms in the field of electrical automation, a customized dictionary is used to optimize the quality of keyword extraction;

[0090] Design a state machine for processing bilingual input and output, and dynamically adjust the dialogue strategy according to the identified scenarios, emotions, and keywords;

[0091] Develop or integrate a generation model capable of generating Chinese-Russian bilingual responses.

[0092] In this embodiment, the Chinese-Russian bilingual electrical automation interaction system needs to accurately identify the commands of users regarding equipment control or fault reports. Users express the same intention in multiple ways, such as "turn on the motor" and "start the motor equipment", and these expressions have corresponding multiple versions in both Russian and Chinese. It is necessary to infer the most likely hidden intention sequence (i.e., the true intention of the user) from a series of observed sequences of user command expressions;

[0093] Hidden state: S = s1, s2,..., s n represents the intention categories predefined by the system, such as "turn on the device", "turn off the device", "query status";

[0094] Observation sequence: O = o1, o2,..., o m Each command or word sequence input by the user, whether in Chinese or Russian;

[0095] State transition probability: Set according to prior knowledge, the probability of transitioning from the "query state" intention to "shut down the device";

[0096] Observation probability: Based on training data, calculate the probability of observing a certain word or phrase in a specific intention state;

[0097] Initialization: δ j (1) = b j (o1), that is, the probability of each state under the first observation;

[0098] Iterative calculation: For each time step t > 1 and each state j, calculate:

[0099]

[0100] At the same time, record the predecessor state of each state j for backtracking the best path;

[0101] Termination and backtracking: Finally, find the state with the maximum probability at time t = n, and then backtrack along the recorded predecessor states to obtain the most likely hidden state sequence S * .

[0102] Example 2

[0103] As Figure 1 shown, based on Example 1, a cultural adaptation database is set in the cultural adaptation translation module;

[0104] The cultural adaptation translation module collects common idioms, proverbs and their corresponding meanings in the field of electrical automation in China and Russia;

[0105] Sort out the polite expressions in the scenarios of technical communication, asking for help and expressing gratitude between the two countries;

[0106] Record specific idiomatic expressions that may affect understanding in daily communication;

[0107] Structurize the data and design a database model, including fields for the original text, translation, usage scenarios, cultural annotations, and emotional colors;

[0108] Conduct the design of a cultural adaptation scoring function, including scoring elements:

[0109] Context fitness: Evaluate the applicability of the translation in a specific professional scenario;

[0110] Emotion retention: Measure whether the emotional expression after translation is consistent with the original text;

[0111] Cultural sensitivity: Check whether the translation avoids cultural taboos or misunderstandings;

[0112] Language fluency: Evaluate the natural fluency of the translated text;

[0113] Let Sc be the total score of cultural adaptability, and the weights of each element be w1, w2, w3, w4, and w1 + w2 + w3 + w4 = 1. Then:

[0114] Sc = w1·S context + w2·S emotion + w3·S sensitivity + w4·S fluency

[0115] Where S context , S emotion , S sensitivity , S fluency are the scores of context fit, emotion retention, cultural sensitivity, and language fluency respectively, ranging from 0 to 1;

[0116] For known culturally specific expressions, directly apply the corresponding replacement rules in the database;

[0117] According to the score of Sc, dynamically adjust the translation strategy:

[0118] If Sc is lower than the set threshold, adopt a more conservative translation strategy, focusing on literal translation to reduce cultural misunderstandings;

[0119] If Sc is high, adopt a more flexible strategy, integrating more expression habits of the target culture to improve the naturalness and intimacy of communication.

[0120] A real-time query and update module system for terms establishes a real-time or regular synchronization mechanism with the cloud term library, automatically pulls the latest term data, and at the same time uploads the newly recognized professional vocabulary in the system to the cloud to achieve two-way updates.

[0121] The electrical automation data parsing module includes a control signal translation module;

[0122] The electrical automation data parsing module receives the original data from electrical automation devices through a network interface or a serial port and detects the communication protocol type of the data packet;

[0123] Combined with the specific data structure model in the field of electrical automation, the received original data is parsed into structured information;

[0124] For the key data tags and descriptions parsed out, through the cultural adaptability translation module, combined with the professional term library and the cultural difference database, adjust or annotate the tags that cause cultural misunderstandings;

[0125] According to industry standards, standardize the parsed data to eliminate the specific representation differences of equipment manufacturers;

[0126] The control signal translation module receives control instructions from the user or the upper-level system, analyzes the semantic structure and operation intention of the instructions using NLP technology, evaluates the cultural fit of the translation according to the cultural adaptability scoring function, and adjusts the translation strategy to avoid misunderstandings caused by cultural differences.

[0127] The electrical interaction execution module receives the translated and verified control instructions, which are commands input by the user through voice or text and have undergone cultural adaptability translation and accurate matching of professional terms;

[0128] Parse the control instructions and convert them into an internal instruction format that can be recognized by the electrical automation system.

[0129] In this embodiment, based on Sc = w1·S context +w2·S emotion +w3·S sensitivity +w4·S fluency Formula:

[0130] Suppose in a conversation between Chinese and Russian engineers about the maintenance of electrical equipment, the Chinese engineer asks, "Is the overload protection function of this equipment normal?" The system needs to translate this question into Russian;

[0131] Context fit S context : The system recognizes that "overload protection" is a professional term and needs to ensure that the accurate Russian expression in the field of electrical automation, such as "защита от перегрузок", is used in the translation.

[0132] Emotion retention S emotion : The original sentence is a neutral question. Maintain this neutral tone during translation and avoid adding unnecessary urgency or dissatisfaction.

[0133] Cultural sensitivity S sensitivity : Considering the acceptance of the questioning method by both China and Russia, ensure that the translation is polite without losing professionalism and avoid direct or rude expressions.

[0134] Language fluency S fluency : Ensure that the Russian sentence structure of the translation is natural, conforms to the Russian expression habit, and is easy for Russian engineers to understand;

[0135] Based on the above analysis, the system assigns a cultural adaptability score Sc to the translation result. If the score of a certain dimension is relatively low, such as insufficient cultural sensitivity, the system may adjust the translation strategy and adopt a more formal or euphemistic expression. For example, if the original literal translation may lead to misunderstanding, the system may change it to: "Эmοm кοмпοнент οбеспечивает надежную защиту οm перегрузοк, вернο?" This not only accurately conveys the query intention but also reflects politeness and professionalism.

[0136] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent Chinese-Russian bilingual electrical automation interaction system, characterized in that, Modules for bilingual electrical interaction include: Speech recognition and processing module: captures user voice input, converts it into text information for subsequent processing, and performs noise reduction on the speech signal; Text preprocessing module: performs word segmentation, stop word removal, and lemmatization on the acquired text information to provide standardized input for subsequent translation and analysis; Contextual Understanding and Adaptation Module: Identify the contextual characteristics of Sino-Russian dialogues, including the topic, roles of participants, and emotional overtones, and dynamically adjust translation strategies; Cultural Adaptability Translation Module: This module establishes a cultural difference database and adjusts translation results based on contextual feature information provided by the context understanding and adaptation module. For known culturally specific expressions, it applies corresponding replacement rules from the cultural difference database and assigns a cultural adaptability score to the pre-processed text. A threshold is set for the cultural adaptability score. Score scores below the threshold are prioritized for literal translation, while scores above the threshold are translated to incorporate the target culture's expression conventions. Real-time terminology query and update module: connects to online professional terminology databases or cloud services, accepts professional terminology query requests encountered by the cultural adaptability translation module during the translation process, and updates the professional terminology information based on the query requests; Electrical automation data parsing module: This module receives translation text containing electrical automation data from the real-time terminology query and update module, parses and processes the data format in the electrical automation field, and transmits the parsed electrical automation data to the electrical interaction execution module. Electrical interaction execution module: Based on the translation and verification of control instructions, they are converted into operation instructions that can be recognized by electrical equipment after parsing, and sent to the target device through the communication protocol to achieve remote control.

2. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, The speech recognition and processing module uses an array of multiple microphones to capture sounds from a specific direction through beamforming while suppressing noise from other directions; Split the continuous speech signal into short time frames and apply a window function to each frame; A feature vector reflecting the speech characteristics is extracted from each frame. Using a deep neural network, through training with a large amount of noisy speech and corresponding clean speech, the noise model and speech characteristics are learned, the language is denoised, and the text preprocessing module is activated.

3. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, The text preprocessing module performs Chinese and Russian word segmentation; Chinese word segmentation: The optimal segmentation point is obtained based on the state transition probability of the hidden Markov model to achieve word segmentation; Russian word segmentation: Use spaces as word separators, consider affix processing, and use machine learning-based methods to identify word boundaries and morphological changes; Create a bilingual stop word list for Chinese and Russian, including common words like "的" and "和" in Chinese, and "и" and "в" in Russian that don't contribute much to the meaning. Simply traverse the word segmentation results and remove the stop words from the list. Perform word restoration on Russian text and set the observation sequence O=o1,o2,...,o n and hidden state sequence S=s1,s2,...,s n , and the state transition probability a i,j , from state s i Transfer to s j The probability and observation probability b j (o i ) in state s j Observed below o i The probability of hidden state path S * The steps are: Among them, δ j (t) represents the cumulative probability of the best path ending with state s j up to time t. The initial condition is δ j (0) = 1 or 0, and backtracking is used to find the optimal path.

4. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, The context understanding and adaptation module cleans, segments, and segment Chinese and Russian data based on the text preprocessing module. Use multilingual versions of BERT for word embedding; Based on the mBERT multilingual pre-trained model, a bilingual classifier is trained to identify specific scenarios in the electrical automation field, including equipment fault reporting, operation guidance, and technical consultation. Adjust the classification threshold according to the characteristics of bilingual data, perform sentiment annotation on the bilingual data, and the model identifies the emotions of users in different languages; Train a bilingual emotion recognition model, combining mBERT embeddings and specific emotion recognition algorithms; Use the TF-IDF method to extract keywords from bilingual data. Considering the professional terms in the field of electrical automation, a customized dictionary is used to optimize the quality of keyword extraction; Design a state machine for processing bilingual input and output, and dynamically adjust the dialogue strategy according to the identified scenarios, emotions, and keywords; Develop or integrate a generation model that can generate Chinese-Russian bilingual responses.

5. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, A cultural difference database is set up in the cultural adaptation translation module; The cultural adaptation translation module collects common idioms, proverbs and their corresponding meanings in the field of electrical automation in China and Russia; Sort out the polite expressions in the scenarios of technical communication, asking for help, and expressing gratitude between the two countries; Record specific idiomatic expressions that may affect understanding in daily communication; Structurize the data and design a database model, including the original text, translation, usage scenarios, cultural annotations, and emotional color fields; Design a cultural adaptation scoring function, including scoring elements: Context fit: Evaluate the applicability of the translation in a specific professional scenario; Emotion retention: Measure whether the emotional expression after translation is consistent with the original text; Cultural sensitivity: Check whether the translation avoids cultural taboos or misunderstandings; Language fluency: Evaluate the natural fluency of the translated text; Let Sc be the total score of cultural adaptation, and the weights of each element are w1, w2, w3, w4, and w1, w2, w3, w4 = 1, then: Sc = w1·S context + w2·S emotion + w3·S sensitivity + w4·S fluency Among them, S context , S emotion , S sensitivity , S fluency are the scores for context fitness, emotion retention, cultural sensitivity, and language fluency respectively, ranging from 0 to 1; For known culturally specific expressions, directly apply the corresponding replacement rules in the database; Dynamically adjust the translation strategy according to the score of Sc: If Sc is lower than the set threshold, adopt a more conservative translation strategy, focusing on literal translation to reduce cultural misunderstandings; If Sc is high, adopt a more flexible strategy, integrating more expression habits of the target culture to improve the naturalness and intimacy of communication.

6. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, A real-time or regular synchronization mechanism is established between the term real-time query and update module system and the cloud term database to automatically pull the latest term data, and at the same time upload the newly recognized professional vocabulary of the system to the cloud to achieve two-way updates.

7. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, The electrical automation data parsing module includes a control signal translation module; The electrical automation data parsing module receives the original data from electrical automation devices through a network interface or a serial port, and detects the communication protocol type of the data packet; Combined with the specific data structure model in the field of electrical automation, the received original data is parsed into structured information; For the parsed key data tags and descriptions, through the cultural adaptation translation module, combined with the professional term database and the cultural difference database, adjust or annotate the tags that cause cultural misunderstandings; According to industry standards, standardize the parsed data to eliminate the specific representation differences of equipment manufacturers; The control signal translation module receives control instructions from users or upper-level systems, uses NLP technology to analyze the semantic structure and operation intention of the instructions, evaluates the cultural fit of the translation according to the cultural adaptation scoring function, and adjusts the translation strategy to avoid misunderstandings caused by cultural differences.

8. An intelligent Chinese-Russian bilingual electrical automation interaction system according to claim 1, characterized in that, The electrical interaction execution module receives the translated and verified control instructions, which are derived from the commands input by the user through voice or text and have undergone cultural adaptation translation and accurate matching of professional terms; Parse the control instructions and convert them into an internal instruction format recognizable by the electrical automation system.

Images