Man-machine collaborative interactive translation and optimization system and method for long and difficult sentences in English

Abstract

The application provides a man-machine collaborative English long difficult sentence interactive translation and optimization system and method, relates to the English translation technical field, first carries out syntax semantic double-layer initial deconstruction processing to the English long difficult sentence, generates initial deconstruction text containing basic syntax constituent unit splitting result information and the like, then receives the structured adjustment instruction of interactive subjects, cooperatively adjusts the initial deconstruction text, generates intermediate deconstruction text, receives the semantic optimization opinion of interactive subjects again, fuses the related information in the opinion and intermediate deconstruction text, executes translation text semantic adaptation checking and logic coherence iterative optimization, and finally generates a translation text meeting target language expression habits and complete semantics, so that the quality and efficiency of English long difficult sentence translation are effectively improved through man-machine collaboration.

Description

Technical Field

[0001] This invention relates to the field of English translation technology, and more specifically, to a human-computer collaborative interactive translation and optimization system and method for long and complex English sentences. Background Technology

[0002] In the field of English translation, especially the translation of long and complex English sentences, this has always been a challenge. Traditional translation methods often focus on the surface vocabulary and simple grammatical structures of sentences, lacking in-depth analysis of the deeper syntactic and semantic relationships within long and complex sentences. Long and complex English sentences have complex syntactic structures, containing multiple clauses, modifiers, etc., making it difficult to accurately grasp the overall structure of the sentence using only conventional grammatical analysis. Furthermore, implicit semantic connections exist between the basic syntactic units within long and complex sentences; these connections are easily overlooked in traditional translation methods, leading to inaccurate and incomplete semantic expression in the translated text.

[0003] Furthermore, most existing translation methods are one-way, one-off processes, lacking effective interaction with translators. When translators encounter problems during the translation process, they cannot adjust and optimize the translation results in a timely manner, making it difficult to meet the needs of scenarios with high translation quality requirements. Moreover, the post-translation optimization stage often lacks a systematic verification and iterative optimization mechanism, resulting in insufficient semantic coherence and adaptability to the target language's expression habits in the final translation. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a human-computer collaborative interactive translation and optimization method for long and complex English sentences, the method comprising:

[0005] The system performs a two-layer initial deconstruction process on long and complex English sentences, which involves parsing syntactic rules to identify the basic syntactic units in the sentences and capturing the implicit semantic connections between these units through semantic association analysis. This process generates an initial deconstructed text that includes the decomposition results of the basic syntactic units, implicit semantic connection markers, and cross-unit semantic association indexes.

[0006] The system receives structured adjustment instructions based on the initial deconstructed text input from the interactive subject. These structured adjustment instructions include instructions for merging the results of splitting the basic syntactic constituent units, instructions for splitting the results, instructions for adjusting the unit order, and instructions for correcting the association type and adjusting the association strength of implicit semantic association context markers.

[0007] Based on the structured adjustment instructions, the boundaries of basic syntactic constituent units, the order of unit arrangement, and the implicit semantic association context in the initial deconstructed text are coordinated and adjusted. The cross-unit semantic association index is updated synchronously to generate the structurally adjusted intermediate deconstructed text. The intermediate deconstructed text includes the adjusted secondary syntactic constituent units, the updated explicit semantic association context markers, and the corrected cross-unit semantic association index.

[0008] The system receives semantic optimization suggestions from the interactive subject for the intermediate deconstructed text input. These suggestions include word substitution and sentence reconstruction suggestions for the initial translation corresponding to a specific sub-syntactic unit, as well as suggestions for strengthening transition logic and semantic coherence adaptation for connecting translations of adjacent sub-syntactic units.

[0009] By integrating semantic optimization opinions with explicit semantic relationship markers in intermediate deconstructed texts and corrected cross-unit semantic relationship indexes, semantic adaptation verification and logical coherence iterative optimization of the translated text are performed to generate a final translated text that conforms to the expression habits of the target language and is semantically complete.

[0010] Furthermore, embodiments of the present invention also provide a human-computer collaborative interactive translation and optimization system for long and complex English sentences, characterized in that it includes:

[0011] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to perform the above-described human-machine collaborative interactive translation and optimization method for long and complex English sentences by executing the machine-executable instructions.

[0012] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a computer device reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the computer device to execute the above-described human-computer collaborative interactive translation and optimization method for long and difficult English sentences.

[0013] Based on the above, through a two-tiered initial deconstruction process involving syntax and semantics, the basic syntactic units in long and complex English sentences can be accurately identified, and implicit semantic connections can be captured. This generates a comprehensive and detailed initial deconstructed text. Then, by receiving structured adjustment instructions from the interactive subject, the initial deconstructed text can be collaboratively adjusted according to these instructions, generating an intermediate deconstructed text containing information such as the adjusted secondary syntactic units. This achieves human-computer interaction during the translation process, improving the flexibility and accuracy of translation. Furthermore, by receiving semantic optimization suggestions from the interactive subject and integrating them with relevant information in the intermediate deconstructed text, semantic adaptation verification and logical coherence iterative optimization of the translated text are performed. This generates a final translated text that conforms to the expression habits of the target language and is semantically complete. Thus, from syntactic and semantic deconstruction and structural adjustment to semantic optimization, this method effectively solves the problems of semantic inaccuracy and poor coherence in traditional translation methods when dealing with long and complex English sentences, significantly improving the quality and efficiency of translating long and complex English sentences. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the execution flow of the human-computer collaborative interactive translation and optimization method for long and difficult English sentences provided in this embodiment of the invention. Detailed Implementation

[0015] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a human-computer collaborative interactive translation and optimization method for long and complex English sentences according to an embodiment of the present invention. The following is a detailed description of this human-computer collaborative interactive translation and optimization method for long and complex English sentences.

[0016] Step S110: Perform a two-layer initial deconstruction process on long and complex English sentences, using syntax and semantics. Identify the basic syntactic units in the long and complex English sentences through syntactic rule parsing, capture the implicit semantic connections between the basic syntactic units through semantic association analysis, and generate an initial deconstruction text that includes the decomposition results of the basic syntactic units, implicit semantic connection markers, and cross-unit semantic association indexes.

[0017] First, obtain the complete character sequence of the long and complex English sentence. Then, traverse this character sequence, locating the positions of all punctuation marks. Using these punctuation marks as the primary segmentation criterion, the long and complex English sentence is divided into multiple primary segments. For example, if the long and complex sentence contains commas, periods, etc., the sentence is split into several relatively independent parts based on their positions. Next, using prepositions and conjunctions as secondary segmentation criterions, each primary segment is further divided into multiple basic segments. Prepositions and conjunctions function as connectors or modifiers in a sentence; their positions allow for the subdivision of primary segments into smaller, semantically independent basic segments.

[0018] Step S111: Call the pre-trained syntactic rule model to perform part-of-speech tagging on each basic segment, identify noun word combinations, verb word combinations, adjective word combinations and adverb word combinations in the basic segment, and record the starting character position and ending character position of each type of word combination.

[0019] Each basic segment is input into a pre-trained syntactic rule model. This model, trained on a large corpus of English language, is capable of recognizing word combinations of different parts of speech within the basic segments. For each basic segment, the model analyzes its vocabulary, determining which words belong to the noun, verb, adjective, or adverb class, and identifying their combination forms. Simultaneously, it records the starting and ending character positions of these word combinations within the basic segment. For example, in a basic segment, the model identifies "beautifulflowers," where "beautiful" is an adjective combination and "flowers" is a noun combination. It records the starting and ending character positions of both "beautiful" and "flowers." Through this process, part-of-speech tagging and position recording are completed for all basic segments.

[0020] Step S112: Analyze the grammatical modification relationships between various word combinations. When a noun word combination and an adjective word combination are adjacent and have a describing and being described relationship, integrate them into a noun-type basic syntactic unit. When a verb word combination and an adverb word combination are adjacent and have a modifying and being modified relationship, integrate them into a verb-type basic syntactic unit. When two word combinations of the same type are connected by a coordinating conjunction, integrate them into a coordinating basic syntactic unit, forming the basic syntactic unit decomposition result.

[0021] For each basic segment, the grammatical modification relationships between various word combinations are analyzed in detail. If a noun combination and an adjacent adjective combination have a descriptive relationship, meaning the adjective combination describes the characteristics of the noun combination, then these two combinations are integrated into a single noun-based basic syntactic unit. For example, in "red apple," "red" describes the color of "apple," and since they are adjacent, they are integrated into the noun-based basic syntactic unit "red apple." If a verb combination and an adjacent adverb combination have a modifying relationship, meaning the adverb combination modifies the manner or degree of the verb combination, then they are integrated into a verb-based basic syntactic unit. For example, in "run quickly," "quickly" modifies the manner of "run," and they are integrated into the verb-based basic syntactic unit "run quickly." If two word combinations of the same type are connected by a coordinating conjunction, indicating a semantic parallel relationship, then they are integrated into a parallel basic syntactic unit. For example, in "apples and oranges," both "apples" and "oranges" are noun phrases connected by "and," forming a parallel basic syntactic unit "applesand oranges." By analyzing and integrating all word combinations, the breakdown of basic syntactic units is obtained.

[0022] Step S113: Call the pre-trained semantic association model to perform semantic feature extraction on each basic syntactic unit in the basic syntactic unit splitting result, and generate a semantic feature vector for each basic syntactic unit. The semantic feature vector includes the unit's core semantics, semantic category attributes, and semantic dependency tendency.

[0023] Step S1131: Convert the text content of each basic syntactic unit into a character-level sequence, input it into the embedding layer of the pre-trained semantic association model, and map the character-level sequence into a character embedding vector through the embedding layer. The character embedding vector contains the semantic encoding and positional encoding of each character.

[0024] For each basic syntactic unit, its text content is converted into a character-level sequence. For example, the text of a basic syntactic unit, "smart student," is converted into the character-level sequence ['s','m','a','r','t','','s','t','u','d','e','n','t']. This character-level sequence is then input into the embedding layer of a pre-trained semantic association model. Based on the mapping relationships learned during model training, the embedding layer generates a vector for each character, namely a character embedding vector. This character embedding vector contains not only the semantic encoding of the character but also the positional encoding of the character in the sequence, reflecting the influence of the character's position on the semantics. For example, "s" will have different positional encodings in different positions, thus taking positional factors into account in subsequent processing.

[0025] Step S1132: Input the character embedding vector into the convolutional layer of the semantic association model, perform convolution operation on the character embedding vector using multi-scale convolution kernels, extract local semantic features of text fragments of different lengths, and generate local semantic feature maps.

[0026] Character embedding vectors are input into the convolutional layers of the semantic association model. Multi-scale convolutional kernels, such as those of size 2, 3, and 4, are used in these layers. Different scale kernels extract local semantic features from text fragments of different lengths. For example, a size 2 kernel extracts semantic features from text fragments of length 2, such as "sm" and "ma," while a size 3 kernel extracts semantic features from text fragments of length 3, such as "sma" and "mar." These convolutional operations generate local semantic feature maps that contain local semantic feature information from text fragments of different lengths representing the basic syntactic building blocks.

[0027] Step S1133: Input the local semantic feature map into the pooling layer of the semantic association model, perform downsampling on the local semantic feature map using adaptive pooling operation, retain key local semantic features, and generate local semantic vectors.

[0028] The local semantic feature map is input into the pooling layer. The pooling layer employs adaptive pooling, automatically adjusting the pooling method based on the content of the local semantic feature map, preserving key local semantic features while reducing feature dimensionality. For example, pooling retains more information for highly important feature regions in the local semantic feature map, while compressing less important regions. After pooling, a local semantic vector is generated, containing the key local semantic features of the basic syntactic building blocks.

[0029] Step S1134: Input the local semantic vector into the recurrent layer of the semantic association model, and perform temporal semantic modeling on the local semantic vector using a bidirectional recurrent structure to capture the contextual semantic association in the text content and generate the contextual semantic vector.

[0030] Local semantic vectors are input into a recurrent layer. This recurrent layer employs a bidirectional recurrent structure, simultaneously performing temporal semantic modeling on the local semantic vectors from both the beginning to the end of the text and from the end to the beginning. This allows for a more comprehensive capture of the contextual semantic relationships within the text content, as the bidirectional recurrent structure considers the mutual influence between the preceding and following text. For example, for "smart student," the bidirectional recurrent structure simultaneously analyzes the influence of "smart" on "student" and the influence of "student" on "smart," generating a contextual semantic vector that includes these contextual semantic relationships.

[0031] Step S1135: Input the context semantic vector into the attention layer of the semantic association model, calculate the attention weight of each semantic component, perform weighted aggregation on the context semantic vector based on the attention weight, highlight the core semantic components, and generate the core semantic vector.

[0032] The context semantic vector is fed into the attention layer. The attention layer calculates an attention weight for each semantic component, the magnitude of which reflects the importance of that semantic component in the overall semantics. For example, in "smart student," "student" is likely the core semantic component and would be assigned a higher attention weight. Based on these attention weights, the context semantic vector is weighted and aggregated to highlight the core semantic components, generating a core semantic vector.

[0033] Step S1136: Extract semantic codes from the core semantic vector that can represent the core meaning of the basic syntactic constituent units, and use them as the core semantics of the units.

[0034] From the core semantic vector, a specific extraction algorithm is used to extract semantic codes that represent the core meaning of the basic syntactic constituent units. For example, the core semantic vector is a high-dimensional vector, where combinations of certain dimensions represent the core meaning of the basic syntactic constituent units. Extracting information from these dimensions yields the core semantics of the unit. For "smart student," the core semantics of the unit might be semantic codes related to "student" and "smart."

[0035] Step S1137: Based on the core semantic vector query, the preset semantic category dictionary is used to determine the semantic category to which the basic syntactic constituent unit belongs. The semantic category includes entity category, action category, and attribute category.

[0036] The core semantic vector is used to query a pre-defined semantic category dictionary. This dictionary stores feature vectors for different semantic categories (such as entity, action, and attribute categories). By calculating the similarity between the core semantic vector and the feature vectors of each semantic category in the dictionary, the semantic category with the highest similarity is found, thus determining the semantic category to which the basic syntactic unit belongs. For example, the core semantic vector of "smart student" has the highest similarity to the feature vector of the entity category, therefore it is determined to belong to the entity category.

[0037] Step S1138: Analyze the matching degree between the core semantic vector and the preset dependency relation template, and determine the other unit types that the basic syntactic constituent units may depend on in the sentence and the unit types that are depended on, as semantic dependency tendency.

[0038] The core semantic vector is matched with a predefined dependency relation template, and their matching degree is analyzed. The dependency relation template contains dependency relation patterns between different types of basic syntactic constituent units. Through matching degree analysis, it is determined which other unit types the basic syntactic constituent unit may depend on in the sentence, and which unit types it may depend on. For example, the core semantic vector of "smart student" has a high matching degree with the dependency relation template representing "action," indicating that it may depend on action-type units, and may also be depended on attribute-type units describing its state. This information constitutes the semantic dependency tendency of this basic syntactic constituent unit.

[0039] Step S1139: Sort the core semantics, semantic category attributes and semantic dependency tendency of the unit according to a preset dimension, and concatenate them into a vector of fixed length to generate the semantic feature vector of each basic syntactic unit. The dimension of the semantic feature vector is consistent with the output dimension of the semantic association model.

[0040] The core semantics, semantic category attributes, and semantic dependency tendencies of each unit are sorted according to a preset dimensional order, and then concatenated into a fixed-length vector. This vector is the semantic feature vector of the basic syntactic constituent units, and its dimension is consistent with the output dimension of the semantic association model to ensure consistency and compatibility in subsequent processing.

[0041] Step S114: Calculate the semantic similarity between the semantic feature vectors of any two basic syntactic constituent units. When the semantic similarity reaches the preset association threshold, it is determined that there is an implicit semantic association between the two. Record the identifier of the associated basic syntactic constituent unit, the association type and the association strength to form an implicit semantic association context marker.

[0042] Step S1141: Read all basic syntactic constituent units in the basic syntactic constituent unit splitting result, generate a basic syntactic constituent unit identifier list, select basic syntactic constituent units in pairs according to the identifier order, and form a unit pair set.

[0043] Read all basic syntactic constituent units from the basic syntactic constituent unit decomposition results, assign a unique identifier to each unit, and generate a list of basic syntactic constituent unit identifiers. Then, select basic syntactic constituent units in pairs according to the identifier order to form a set of unit pairs. For example, if the identifier list is [1,2,3,4], then the set of unit pairs is [(1,2),(1,3),(1,4),(2,3),(2,4),(3,4)].

[0044] Step S1142: For each unit pair in the unit pair set, extract the semantic feature vectors of the two basic syntactic constituent units. The semantic feature vectors contain vector components corresponding to the unit's core semantics, semantic category attributes, and semantic dependency tendencies.

[0045] For each unit pair in the unit pair set, semantic feature vectors of the two basic syntactic constituent units are extracted. Each semantic feature vector contains vector components corresponding to the unit's core semantics, semantic category attributes, and semantic dependency tendencies. These components will be used for subsequent similarity calculations.

[0046] Step S1143: Calculate the cosine similarity between two semantic feature vectors and use the cosine similarity as the initial semantic similarity. The calculation of the cosine similarity is based on the ratio of the sum of the products of the vector components to the product of the vector magnitude.

[0047] Calculate the cosine similarity between two semantic feature vectors. The method for calculating the cosine similarity is as follows: multiply each component of the two vectors separately, then add the products together and use the sum as the numerator; multiply the magnitudes of the two vectors and use the product as the denominator; finally, divide the numerator by the denominator to obtain the cosine similarity, which is used as the initial semantic similarity.

[0048] Step S1144: Adjust the initial semantic similarity based on the semantic category attributes of the two basic syntactic constituent units to obtain the adjusted semantic similarity.

[0049] The initial semantic similarity is adjusted based on the semantic category attributes of the two basic syntactic constituent units. For example, if the two units have the same semantic category attribute, it indicates that they are consistent in semantic category, which may increase the semantic similarity, so the initial semantic similarity is appropriately increased; if the semantic category attributes are different, the initial semantic similarity is appropriately decreased. The adjustment method follows preset rules, which are determined during model training or method design.

[0050] Step S1145: Adjust the adjusted semantic similarity based on the semantic dependency tendency of the two basic syntactic constituent units to obtain the final semantic similarity.

[0051] The adjusted semantic similarity is then adjusted based on the semantic dependency tendencies of the two basic syntactic units. If the semantic dependency tendencies of the two units match—for example, one unit's dependency tendency is to depend on another type of unit, while the other unit's dependency tendency is to be depended on by the aforementioned type of unit—then their semantic similarity may increase, requiring an increase based on the adjusted semantic similarity; conversely, it decreases. The adjustment method also follows the preset rules.

[0052] Step S1146: Compare the final semantic similarity with the preset association threshold. When the final semantic similarity is greater than or equal to the preset association threshold, it is determined that there is an implicit semantic association between the two basic syntactic constituent units in the unit pair.

[0053] The final semantic similarity is compared with a preset association threshold. If the final semantic similarity is greater than or equal to the threshold, it indicates that there is an implicit semantic association between the two basic syntactic units.

[0054] Step S1147: Determine the association type based on the unit type and semantic dependency tendency of the two basic syntactic constituent units. The association types include entity-action association, action-attribute association, entity-attribute association, and parallel entity association.

[0055] The association type is determined based on the unit type and semantic dependency tendency of the two basic syntactic constituent units. For example, if one unit is an entity category and the other is an action category, and their semantic dependency tendencies match, the association type may be entity-action association; if one is an action category and the other is an attribute category, the association type may be action-attribute association.

[0056] Step S1148: Map the final semantic similarity to the association strength value. The mapping rule is that the final semantic similarity and the association strength value are positively correlated, and the association strength is generated.

[0057] According to the preset mapping rules, the final semantic similarity is mapped to the association strength value. The mapping rule is that the final semantic similarity and the association strength value are positively correlated, that is, the higher the semantic similarity, the greater the association strength value.

[0058] Step S1149: Record the identifiers, determined association types, and association strength values ​​of the two basic syntactic constituent units that have implicit semantic associations, forming an entry for an implicit semantic association network.

[0059] Record the identifiers, association types, and association strength values ​​of two basic syntactic units that have implicit semantic associations to form an entry for an implicit semantic association network.

[0060] Step S11410: Sort all implicit semantic association entries from high to low according to the association strength value, add a timestamp of the association generation, and form an implicit semantic association marker.

[0061] All implicit semantic association entries are sorted from high to low according to their association strength values, and a generation timestamp is added to form implicit semantic association markers, which show the implicit semantic associations between basic syntactic constituent units.

[0062] Step S115: Construct a cross-unit semantic association index based on implicit semantic association context markers. The cross-unit semantic association index includes a list of associated units corresponding to each basic syntactic constituent unit, an association type sorting, and a numerical range of association strength. The list of associated units is arranged from high to low association strength.

[0063] For each basic syntactic unit, other basic syntactic units with implicit semantic relationships are identified from the implicit semantic relationship context markers. Their identifiers, relationship types, and relationship strength values ​​are compiled to form a list of related units. The relationship types in the list of related units are sorted, the range of relationship strength values ​​is determined, and the list of related units is arranged from high to low relationship strength to construct a cross-unit semantic relationship index.

[0064] Step S116: Integrate the basic syntactic constituent unit decomposition results, implicit semantic association context markers, and cross-unit semantic association indexes into structured data. Arrange the structured data according to the original order of the basic syntactic constituent units in long and difficult English sentences to generate an initial deconstructed text containing the basic syntactic constituent unit decomposition results, implicit semantic association context markers, and cross-unit semantic association indexes.

[0065] The results of basic syntactic unit decomposition, implicit semantic association context markers, and cross-unit semantic association indexes are integrated into structured data. The basic syntactic units are arranged in their original order in long and complex English sentences to generate an initial deconstructed text, which shows the syntactic structure and semantic associations of long and complex sentences.

[0066] Step S120: Receive structured adjustment instructions from the interactive subject based on the initial deconstructed text input. The structured adjustment instructions include instructions for merging the results of splitting the basic syntactic constituent units, instructions for splitting, instructions for adjusting the unit order, and instructions for correcting the association type and adjusting the association strength of implicit semantic association context markers.

[0067] Step S121: Convert the initial deconstructed text into a multi-level visual interactive interface. The multi-level visual interactive interface includes a unit display layer, an association display layer, and an index display layer. The unit display layer displays basic syntactic constituent units in the original order in the form of nodes. Each node is labeled with the basic syntactic constituent unit identifier, unit type, and core vocabulary. The association display layer connects nodes with implicit semantic associations with attributed lines. The lines are labeled with the association type and association strength. The index display layer displays cross-unit semantic association indexes in the form of a list. The list includes the basic syntactic constituent unit identifier, the list of associated units, and the association strength sorting.

[0068] The initial deconstructed text is transformed into a multi-level visual interactive interface. The unit display layer presents basic syntactic constituent units in their original order as nodes, each labeled with an identifier, unit type (e.g., nominal, verbal), and core vocabulary, facilitating user interaction. The association display layer connects nodes with implicit semantic relationships using attributed lines, labeling the association type and strength for a clear visual representation. The index display layer presents a cross-unit semantic association index in list form, including basic syntactic constituent unit identifiers, a list of associated units, and a ranking by association strength, allowing users to easily verify the information.

[0069] Step S122: The interactive subject selects at least two adjacent nodes through the unit display layer, triggers the merge operation entry, enters the merged unit name and merged unit type in the merge operation entry, and generates a merge operation instruction for the split results of the basic syntax constituent units after submission. The merge operation instruction includes the basic syntax constituent unit identifier of the merged node, the merged unit name, the merged unit type and the merge order.

[0070] The interactive entity selects at least two adjacent nodes in the unit display layer to trigger the merge operation entry. The merged unit name and type are entered in the entry. After submission, a merge operation instruction is generated, containing the identifiers of the merged nodes, the merged unit name, type, and merge order, for use in subsequent merge operations.

[0071] Step S123: The interactive subject selects a single node through the unit display layer, triggers the split operation entry, defines the split position, inputs the name and type of the split sub-unit, and after submission, generates a split operation instruction for the split result of the basic syntax constituent unit. The split operation instruction includes the basic syntax constituent unit identifier of the split node, the split position coordinates, the name and type of the split sub-unit.

[0072] The interactive subject selects a single node in the unit display layer, triggering the split operation entry. The split location is defined in the entry, and the name and type of the resulting sub-unit are entered. Upon submission, a split operation instruction is generated, containing the identifier of the split node, the coordinates of the split location, and the name and type of the sub-unit, used for subsequent split operations.

[0073] Step S124: The interactive subject drags the nodes in the unit display layer to adjust the arrangement order. After confirming the adjustment, a unit order adjustment instruction is generated for the basic syntax constituent unit splitting result. The unit order adjustment instruction includes the unit arrangement sequence before adjustment, the unit arrangement sequence after adjustment, and the adjustment position identifier.

[0074] The interactive entity drags nodes in the unit display layer to adjust their arrangement order. After confirmation, a unit order adjustment instruction is generated. The instruction includes the unit arrangement sequence before and after adjustment and the adjustment position identifier, which is used to adjust the unit arrangement order in the future.

[0075] Step S125: The interactive subject selects the connection with attributes through the association display layer, triggering the association correction entry. In the association correction entry, the association type is modified and the association strength value is adjusted. After submission, an association type correction instruction and an association strength adjustment instruction for the implicit semantic association context mark are generated. The association type correction instruction includes the identifiers of the two basic syntactic constituent units corresponding to the corrected connection, the original association type, and the new association type. The association strength adjustment instruction includes the identifiers of the two basic syntactic constituent units corresponding to the corrected connection, the original association strength value, and the new association strength value.

[0076] When the interactive subject selects a connection with attributes in the association display layer, it triggers the association correction entry. In the entry, the association type and the association strength value are modified and adjusted. After submission, an association type correction instruction and an association strength adjustment instruction are generated. The instruction includes the identifiers of the two basic syntactic constituent units corresponding to the corrected connection, the original association type, the new association type, the original association strength value, and the new association strength value, which are used for subsequent correction of implicit semantic association context markers.

[0077] Step S126: The interactive subject checks the cross-unit semantic association index through the index display layer. When it finds that the index is inconsistent with the association display layer, it triggers the index synchronization correction entry. After submission, an index synchronization instruction is generated. The index synchronization instruction includes the identifier of the basic syntactic constituent unit to be corrected, the list of associated units before correction, and the list of associated units after correction.

[0078] The interactive entity checks the cross-unit semantic association index in the index display layer. When an inconsistency is found, the index synchronization correction entry is triggered. After submission, an index synchronization instruction is generated. The instruction contains the identifier of the basic syntactic constituent unit to be corrected, and the list of associated units before and after correction, which is used for subsequent synchronous correction of the cross-unit semantic association index.

[0079] Step S127: Receive merge operation instructions, split operation instructions, unit order adjustment instructions, association type correction instructions, association strength adjustment instructions and index synchronization instructions submitted by the interactive subject, classify them according to operation type, add instruction submission timestamps, and integrate them into structured adjustment instructions.

[0080] It receives various instructions submitted by the interactive subject, classifies them according to operation type, adds submission timestamps, and integrates them into structured adjustment instructions for convenient subsequent processing and execution.

[0081] Step S130: Based on the structured adjustment instructions, the boundaries of the basic syntactic constituent units, the order of the units, and the implicit semantic association relationships in the initial deconstructed text are coordinated and adjusted. The cross-unit semantic association index is updated synchronously to generate the intermediate deconstructed text after structural adjustment. The intermediate deconstructed text includes the adjusted secondary syntactic constituent units, the updated explicit semantic association network markers, and the corrected cross-unit semantic association index.

[0082] Step S131: Parse the merge operation instruction in the structured adjustment instruction, extract the basic syntactic constituent unit identifier of the merged node, and locate the corresponding basic syntactic constituent unit in the basic syntactic constituent unit splitting result of the initial deconstructed text.

[0083] Parse the merge operation instructions and extract the basic syntactic unit identifiers of the nodes to be merged. Based on the identifiers in the initial deconstructed text's basic syntactic unit splitting results, locate the corresponding basic syntactic unit and determine the merge target.

[0084] Step S132: Read the text content of the located basic syntactic constituent unit, concatenate the text content according to the merging order in the merging operation instruction, replace it with the text fragment corresponding to the merged unit name in the merging operation instruction, generate the merged secondary syntactic constituent unit, assign a new unit identifier to the merged secondary syntactic constituent unit, and record the merged unit type.

[0085] Read the text content of the located basic syntactic constituent units, concatenate them in the merging order, replace them with text fragments corresponding to the merged unit names, and generate merged secondary syntactic constituent units. Assign a new identifier to each unit and record the merged unit type.

[0086] Step S133: Parse the splitting operation instructions in the structured adjustment instructions, extract the basic syntactic constituent unit identifiers and splitting position coordinates of the splitting nodes, and locate the corresponding basic syntactic constituent units in the basic syntactic constituent unit splitting results of the initial deconstructed text.

[0087] The splitting operation instructions are parsed to extract the basic syntactic unit identifiers and splitting position coordinates of the split nodes. The corresponding basic syntactic unit is located in the initial deconstructed text's basic syntactic unit splitting results to determine the splitting object and its position.

[0088] Step S134: Read the text content of the located basic syntactic constituent unit, divide the text content according to the split position coordinates, generate the split secondary syntactic constituent unit according to the sub-unit name in the split operation instruction, assign a new unit identifier to each split secondary syntactic constituent unit, and record the type of the split sub-unit.

[0089] Read the text content of the located basic syntactic unit, segment it according to the split position coordinates, and generate the corresponding sub-syntactic units based on the sub-unit names. Assign a new identifier to each sub-syntactic unit and record the sub-unit type.

[0090] Step S135: Parse the unit order adjustment instruction in the structured adjustment instruction, extract the adjusted unit arrangement sequence, adjust the arrangement order of the basic syntactic constituent unit splitting results in the initial deconstructed text to the adjusted unit arrangement sequence, and form the adjusted secondary syntactic constituent unit arrangement result.

[0091] The parsing unit order adjustment instruction is used to extract the adjusted unit arrangement sequence. The order of the basic syntactic constituent units in the initial deconstructed text is adjusted to this sequence to form the adjusted arrangement result of the secondary syntactic constituent units.

[0092] Step S136: Parse the association type correction instruction in the structured adjustment instruction, extract the two basic syntactic constituent unit identifiers and the new association type corresponding to the corrected connection, locate the corresponding association entry in the implicit semantic association context marker of the initial deconstructed text, and update the association type of the association entry to the new association type.

[0093] The association type correction instruction is parsed, and the identifiers of the two basic syntactic constituent units corresponding to the corrected connection and the new association type are extracted. The corresponding association entry is located in the implicit semantic association context markers of the initial deconstructed text, and the association type is updated to the new association type.

[0094] Step S137: Parse the association strength adjustment instruction in the structured adjustment instruction, extract the two basic syntactic constituent unit identifiers and the new association strength value corresponding to the modified connection, locate the corresponding association item in the implicit semantic association context marker of the initial deconstructed text, update the association strength of the association item to the new association strength value, and form the updated explicit semantic association context marker.

[0095] The association strength adjustment instruction is parsed, and the identifiers of the two basic syntactic constituent units corresponding to the modified connection and the new association strength value are extracted. The corresponding association entries are located in the implicit semantic association context markers of the initial deconstructed text, and the association strength is updated to the new association strength value, forming the updated explicit semantic association context markers.

[0096] Step S138: Parse the index synchronization instruction in the structured adjustment instruction, extract the basic syntactic constituent unit identifiers to be corrected and the corrected list of related units, locate the corresponding index entries in the cross-unit semantic association index of the initial deconstructed text, update the list of related units of the index entries to the corrected list of related units, reorder the list of related units according to the new association strength value, and form the corrected cross-unit semantic association index.

[0097] The index synchronization command is parsed to extract the identifiers of the basic syntactic constituent units that need to be corrected and the corrected list of related units. The corresponding index entries are located in the cross-unit semantic association index of the initial deconstructed text, the list of related units is updated to the corrected list, and reordered according to the new association strength values ​​to form the corrected cross-unit semantic association index.

[0098] Step S139: Detect whether there are new semantic associations between the adjusted secondary syntactic constituent units, calculate the semantic feature vector similarity of adjacent secondary syntactic constituent units, and when the similarity reaches a preset threshold, add explicit semantic association context marker entries and cross-unit semantic association index entries.

[0099] Detect new semantic associations between the adjusted secondary syntactic constituent units. Calculate the semantic feature vector similarity of adjacent secondary syntactic constituent units. If the similarity reaches a preset threshold, add explicit semantic association context marker entries and cross-unit semantic association index entries.

[0100] Step S1310: Integrate the adjusted secondary syntactic constituent unit arrangement results, the updated explicit semantic association context markers, and the corrected cross-unit semantic association index, add an adjustment completion timestamp, and generate the structurally adjusted intermediate deconstructed text.

[0101] Integrate the adjusted arrangement of secondary syntactic constituent units, the updated explicit semantic association context markers, and the revised cross-unit semantic association index, add an adjustment completion timestamp, and generate intermediate deconstructed text.

[0102] Step S140: Receive semantic optimization opinions from the interactive subject for the intermediate deconstructed text input. The semantic optimization opinions include word substitution suggestions and sentence reconstruction suggestions for the preliminary translation corresponding to specific sub-syntactic constituent units, as well as transition logic strengthening suggestions and semantic coherence adaptation suggestions for the connection of translations of adjacent sub-syntactic constituent units.

[0103] Step S141: Call a preset translation generation interface to independently translate each secondary syntactic constituent unit in the intermediate deconstructed text, and generate a preliminary translation corresponding to each secondary syntactic constituent unit. The preliminary translation includes a literal translation version, a free translation version, and a sentence pattern variant version.

[0104] Call a preset translation generation interface to independently translate each secondary syntactic constituent unit in the intermediate deconstructed text. The translation generation interface will generate a preliminary translation of the literal translation version, the free translation version, and the sentence pattern variant version according to the semantic features of the unit. For example, if the text of a secondary syntactic constituent unit is "advanced technology", the literal translation version is "advanced technology", the free translation version may be "frontier technology", and the sentence pattern variant version may be "technologically advanced".

[0105] Step S142: Convert the intermediate deconstructed text into a translation interaction optimization interface. The translation interaction optimization interface includes a unit translation display area, a connection and cohesion display area, and an optimization opinion input area. The unit translation display area displays the original content, the preliminary translation, and the unit identifier of each secondary syntactic constituent unit in the order of arrangement of the secondary syntactic constituent units. The preliminary translation includes a literal translation version, a free translation version, and a sentence pattern variant version. The connection and cohesion display area displays the connection relationship of the preliminary translations of adjacent secondary syntactic constituent units in the form of a flowchart, marking the connection position and the connection type. The optimization opinion input area includes a diction replacement input box, a sentence pattern reconstruction input box, a transition logic input box, and a semantic coherence input box.

[0106] Convert the intermediate deconstructed text into a translation interaction optimization interface. The unit translation display area displays the original content, the preliminary translation (including the literal translation, free translation, and sentence pattern variant versions), and the unit identifier in the order of arrangement of the secondary syntactic constituent units. The connection and cohesion display area displays the connection relationship of the preliminary translations of adjacent secondary syntactic constituent units in the form of a flowchart, marking the connection position and the connection type. The optimization opinion input area includes multiple input boxes for the interaction subject to input optimization opinions.

[0107] Step S143: The interaction subject selects the preliminary translation of a specific secondary syntactic constituent unit in the unit translation display area. When it is considered that the diction of the literal translation version is inaccurate, enter the replacement word, the replaced phrase, and the replacement reason in the diction replacement input box to generate a diction replacement suggestion for the preliminary translation corresponding to the specific secondary syntactic constituent unit. The diction replacement suggestion includes the target secondary syntactic constituent unit identifier, the original word, the replacement word, the replaced phrase, and the replacement reason.

[0108] The interactive subject selects the preliminary translation of a specific subsyntactic unit in the unit translation display area. If the subject believes that the wording of the literal translation is inaccurate, it enters the replacement word, the replaced phrase, and the reason for the replacement in the wording replacement input box to generate wording replacement suggestions. The suggestions include the target subsyntactic unit identifier, the original word, the replacement word, the replaced phrase, and the reason for the replacement.

[0109] Step S144: The interactive subject selects the preliminary translation of a specific sub-syntactic unit in the unit translation display area. When it is believed that the sentence structure does not conform to the target language habit, it enters the reconstructed sentence structure, the reconstructed sentence, and the reason for reconstruction in the sentence reconstruction input box. This generates a sentence reconstruction suggestion for the preliminary translation corresponding to the specific sub-syntactic unit. The sentence reconstruction suggestion includes the target sub-syntactic unit identifier, the original sentence structure, the reconstructed sentence structure, the reconstructed sentence, and the reason for reconstruction.

[0110] The interactive subject selects a preliminary translation of a specific subsyntactic unit in the unit translation display area. If the subject believes that the sentence structure does not conform to the target language habits, it enters the reconstructed sentence structure, sentence, and reason for reconstruction in the sentence reconstruction input box. The system generates a sentence reconstruction suggestion, which includes the target subsyntactic unit identifier, the original sentence structure, the reconstructed sentence structure, sentence, and reason for reconstruction.

[0111] Step S145: The interactive subject selects the connection position of the translation of the adjacent sub-syntactic unit in the connection display area. When the transition logic is considered unclear, the subject enters the suggested transition words, transition sentence patterns and transition logic types in the transition logic input box to generate transition logic strengthening suggestions for the connection of the translation of the adjacent sub-syntactic unit. The transition logic strengthening suggestions include connection position identifier, adjacent sub-syntactic unit identifier, suggested transition words, transition sentence patterns and transition logic types.

[0112] The interactive subject selects the connection position of the translation of the adjacent secondary syntactic unit in the connection display area. If the transition logic is not clear, the subject enters the suggested transition words, sentence patterns and transition logic types in the transition logic input box to generate transition logic reinforcement suggestions. The suggestions include connection position identifiers, adjacent secondary syntactic unit identifiers, suggested transition words, sentence patterns and transition logic types.

[0113] Step S146: The interactive subject browses the connection content of the translations of adjacent secondary syntactic units in the connection display area. When it believes that the semantic coherence is insufficient, it enters semantic supplementation suggestions, word order adjustment suggestions and coherence adaptation directions in the semantic coherence input box to generate semantic coherence adaptation suggestions for the connection of translations of adjacent secondary syntactic units. The semantic coherence adaptation suggestions include connection position identifiers, adjacent secondary syntactic unit identifiers, semantic supplementation suggestions, word order adjustment suggestions and coherence adaptation directions.

[0114] The interactive subject browses the connection content of the translation of adjacent secondary syntactic units in the connection display area. If the semantic coherence is insufficient, the subject enters semantic supplementation suggestions, word order adjustment suggestions and coherence adaptation direction in the semantic coherence input box to generate semantic coherence adaptation suggestions. The suggestions include connection position identifiers, adjacent secondary syntactic unit identifiers, semantic supplementation suggestions, word order adjustment suggestions and coherence adaptation direction.

[0115] Step S147: After the interactive subject submits all the input optimization opinions, the format of the wording replacement suggestions, sentence restructuring suggestions, transition logic strengthening suggestions and semantic coherence adaptation suggestions is checked. The check content is whether each suggestion contains a complete target identifier, original content, new content and the reason for the suggestion.

[0116] After the interaction subject submits all optimization suggestions, the above suggestions are formatted and checked to see if each suggestion contains a complete target identifier, original content, new content, and reason for the suggestion.

[0117] Step S148: Receive verified suggestions for word replacement, sentence restructuring, transition logic enhancement, and semantic coherence adaptation. Categorize the suggestions by type, add a submission timestamp, and integrate them into semantic optimization opinions.

[0118] Receive verified suggestions, categorize them by suggestion type, add submission timestamps, and integrate them into semantic optimization opinions.

[0119] Step S150: Integrate the semantic optimization opinions with the explicit semantic relationship markers in the intermediate deconstructed text and the corrected cross-unit semantic relationship index, perform semantic adaptation verification and logical coherence iterative optimization of the translation, and generate the final translated text that conforms to the expression habits of the target language and is semantically complete.

[0120] Step S151: Analyze the word substitution suggestions in the semantic optimization opinions, extract the target secondary syntactic unit identifier, replacement words and the replaced phrases, and locate the corresponding secondary syntactic unit and its preliminary translation in the intermediate deconstructed text.

[0121] The text analyzes the suggested word substitutions, extracting the identifiers of the target secondary syntactic units, the replacement words, and the replaced phrases. The corresponding secondary syntactic units and their preliminary translations are then located within the intermediate deconstructed text.

[0122] Step S152: Replace the original words in the preliminary translation of the located secondary syntactic units with the replacement words, replace the original phrases with the replacement phrases, generate the unit translation after wording adjustment, and record the basis for wording adjustment and the text comparison before and after adjustment.

[0123] Replace the original words and phrases in the preliminary translation of the identified secondary syntactic units with alternative words, generating the unit translation with adjusted wording. Record the basis for the wording adjustment (i.e., the reasons for the replacement in the wording replacement suggestions) and a comparison of the text before and after the adjustment.

[0124] Step S153: Analyze the sentence reconstruction suggestions in the semantic optimization opinions, extract the target secondary syntactic unit identifier, the reconstructed sentence structure and the reconstructed sentence, and locate the corresponding secondary syntactic unit and its preliminary translation in the intermediate deconstructed text.

[0125] The analysis process involves reconstructing the sentence structure, extracting the identifiers of the target secondary syntactic units, the reconstructed sentence structure, and the reconstructed sentences. The corresponding secondary syntactic units and their preliminary translations are then located within the intermediate deconstructed text.

[0126] Step S154: Adjust the preliminary translation of the located secondary syntactic units according to the reconstructed sentence structure, replace it with the reconstructed sentence, generate the unit translation after sentence structure adjustment, and record the basis for sentence structure adjustment and the comparison of sentence structure before and after adjustment.

[0127] The initial translations of the identified secondary syntactic units are adjusted according to the reconstructed sentence structure, and replaced with the reconstructed sentences to generate the unit translations after sentence structure adjustment. The basis for sentence structure adjustment (i.e., the reasons for reconstruction in the sentence structure reconstruction suggestions) and the comparison of sentence structure before and after adjustment are recorded.

[0128] Step S155: Integrate the unit translations after wording adjustment and sentence structure adjustment, remove duplicate adjustments, and form a preliminary optimized set of unit translations. Establish a one-to-one correspondence between each secondary syntactic unit and the corresponding preliminary optimized unit translation, and record the correspondence table.

[0129] Integrate the unit translations with adjusted wording and sentence structure, remove duplicate adjustments, and form a preliminary optimized set of unit translations. Establish a one-to-one correspondence between each sub-syntactic unit and its corresponding preliminary optimized unit translation, and record the correspondence table.

[0130] Step S156: Analyze the transition logic strengthening suggestions in the semantic optimization opinions, extract the connection position identifier, the identifier of adjacent secondary syntactic constituent units, the suggested transition words and transition sentence patterns, and locate the corresponding adjacent unit translations in the preliminarily optimized unit translation set.

[0131] The analysis process strengthens transition logic suggestions by extracting linking position identifiers, identifiers of adjacent secondary syntactic units, suggested transition words, and transition sentence structures. The corresponding adjacent unit translations are then located within the preliminarily optimized set of unit translations.

[0132] Step S157: Insert suggested transition words or phrases at the connection points between the located adjacent translation units, generate the translation sequence after adding the transition, and record the basis for adding the transition and the connection comparison before and after adding the transition.

[0133] Insert suggested transition words or phrases at the connection points between the located adjacent translation units, generating a sequence of translations with the transitions added. Record the basis for adding the transitions (i.e., the reasons in the transition logic reinforcement suggestions) and a comparison of the connection before and after the addition.

[0134] Step S158: Analyze the semantic coherence adaptation suggestions in the semantic optimization opinions, extract the cohesion position identifiers, adjacent secondary syntactic unit identifiers, semantic supplementation suggestions and word order adjustment suggestions, and locate the corresponding adjacent unit translations in the translated sequence after adding transitions.

[0135] The semantic coherence adaptation suggestions are analyzed, and cohesion position identifiers, adjacent secondary syntactic unit identifiers, semantic supplementation suggestions, and word order adjustment suggestions are extracted. The corresponding adjacent unit translations are then located in the translated sequence after the transitions are added.

[0136] Step S159: Add necessary semantic supplementary content according to the semantic supplementation suggestions, adjust the arrangement order of adjacent unit translations according to the word order adjustment suggestions, generate a coherence-optimized translation sequence, and record the basis for coherence optimization and the semantic comparison before and after optimization.

[0137] Add necessary semantic supplements based on semantic supplementation suggestions, adjust the order of adjacent translation units based on word order adjustment suggestions, and generate a coherence-optimized translation sequence. Record the basis for coherence optimization (i.e., the reasons in the semantic coherence adaptation suggestions) and the semantic comparison before and after optimization.

[0138] Step S1510: Call the pre-trained semantic adaptation model to perform semantic adaptation verification on the coherence-optimized translation sequence. The verification content is the semantic consistency between each unit translation and the corresponding subsyntactic unit original text, the logical coherence between adjacent unit translations, and the standardization of the target language expression of the overall translation.

[0139] The pre-trained semantic adaptation model is invoked to perform semantic adaptation verification on the coherence-optimized translation sequence. The verification includes: the semantic consistency between each unit translation and the corresponding sub-syntactic unit original text, i.e., checking whether the translation accurately conveys the core semantics of the original text; the logical coherence between adjacent unit translations, i.e., checking whether the logical relationship between adjacent translations is clear and reasonable; and the target language expression norms of the overall translation, i.e., checking whether the translation conforms to the grammar, vocabulary usage habits, etc. of the target language.

[0140] Step S15101: Divide the coherence-optimized translation sequence into individual unit translations. Each unit translation corresponds to a secondary syntactic unit. Extract the text content of each unit translation and the original text content of the corresponding secondary syntactic unit.

[0141] The coherence-optimized translation sequence is split into individual unit translations based on the division of secondary syntactic units. For example, if a coherence-optimized translation sequence consists of translations from multiple secondary syntactic units, it is split into unit translations corresponding to each secondary syntactic unit. Simultaneously, the text content of each unit translation and the original text content of the corresponding secondary syntactic unit are extracted for subsequent semantic consistency verification.

[0142] Step S15102: Call the semantic consistency verification module of the pre-trained semantic adaptation model and input the unit translation text content and the corresponding original text content into the module.

[0143] The semantic consistency verification module in the pre-trained semantic adaptation model is invoked. The text content of each unit's translation and the original text content of the corresponding secondary syntactic constituent units are input into this semantic consistency verification module. This semantic consistency verification module has been trained on a large amount of bilingual corpus and is able to analyze the semantic relationship between the translation and the original text.

[0144] Step S15103: The semantic consistency verification module performs semantic parsing on the original text content and extracts the core semantic elements of the original text. The core semantic elements include core nouns, core verbs, core adjectives and semantic relationships.

[0145] The semantic consistency verification module performs semantic parsing on the input source text. For example, for the source text "After years of dedicated research, Dr. Smith achieved a breakthrough," the module will identify the core nouns "Dr. Smith" and "breakthrough," the core verb "achieved," the core adjective "dedicated," and the semantic relationship "Dr. Smith" achieving "breakthrough" through "dedicated research." Through this parsing, the core semantic elements of the source text are extracted.

[0146] Step S15104: Perform semantic analysis on the unit translation text content to extract the core semantic elements of the translation. The core semantic elements include core nouns, core verbs, core adjectives and semantic relationships.

[0147] The module also performs semantic analysis on the input unit translation text. For example, if the corresponding translation is "After years of focused research, Dr. Smith has achieved a breakthrough," the module will identify the core nouns "Dr. Smith" and "breakthrough," the core verb "achieve," the core adjective "focused," and the semantic relationship "Dr. Smith" achieves "breakthrough" through "focused research." This extracts the core semantic elements of the translation.

[0148] Step S15105: Compare the overlap between the core semantic elements of the original text and the core semantic elements of the translation, calculate the overlap percentage, and determine that the semantic consistency standard is met when the overlap percentage is greater than or equal to the semantic consistency standard threshold.

[0149] The core semantic elements of the source text and the target text are compared, and their overlap percentage is calculated. The overlap is calculated based on the matching of core semantic elements, including core nouns, core verbs, core adjectives, and semantic relationships. For example, if the core nouns, core verbs, core adjectives, and semantic relationships in both the source and target texts match, the overlap percentage will be high. When the overlap percentage is greater than or equal to a preset semantic consistency standard threshold (e.g., 80%), the semantic consistency of that unit of translation is considered to have met the standard.

[0150] Step S15106: Extract the text content of adjacent unit translations in the coherence-optimized translation sequence and form adjacent translation pairs in sequence.

[0151] From the coherence-optimized translation sequence, extract the text content of adjacent unit translations and form a set of adjacent translation pairs according to their order in the sequence. For example, if the translation sequence is "After years of focused research, Dr. Smith has made a breakthrough that may revolutionize the way we use clean energy", then the set of adjacent translation pairs is [("After years of focused research, Dr. Smith has made a breakthrough", "This breakthrough may revolutionize the way we use clean energy")].

[0152] Step S15107: Analyze the logical connectors, referential relationships, and semantic continuity relationships between adjacent translation pairs, and calculate the logical coherence score. The logical coherence score is calculated based on the combination of connector fit, referential clarity, and semantic continuity naturalness.

[0153] For each pair of adjacent translations, analyze their logical connectors, referential relationships, and semantic coherence. The fit of logical connectors refers to whether the connector accurately expresses the logical relationship between adjacent translations, such as the appropriate use of "and," "but," and "however." The clarity of referential relationships refers to whether the referential elements in the translation (such as "this" and "that") clearly refer to the preceding content. The naturalness of semantic coherence refers to whether the semantic transition between adjacent translations is natural and fluent. Based on these three aspects of analysis, calculate the logical coherence score. For example, if the connector fit score is 0.8, the referential clarity score is 0.9, and the semantic coherence score is 0.8, then the logical coherence score might be (0.8 + 0.9 + 0.8) / 3 = 0.833 (this is just an example calculation method; the actual calculation method may be more complex and will consider the weight of each factor).

[0154] Step S15108: When the logical coherence score is greater than or equal to the logical coherence standard threshold, it is determined that the logical coherence between adjacent unit translations meets the standard.

[0155] The calculated logical coherence score is compared with a preset logical coherence standard threshold (e.g., 0.7). When the logical coherence score is greater than or equal to the threshold, the logical coherence between the adjacent translation units is deemed to meet the standard.

[0156] Step S15109: Treat the coherence-optimized translation sequence as a whole text and perform a grammar rule check on the whole text. The grammar rule check includes subject-verb agreement check, tense consistency check, sentence collocation check, and punctuation usage check.

[0157] The optimized translation sequence is treated as a whole text and subjected to grammatical rule checks. Subject-verb agreement checks whether the subject and verb agree in person and number; for example, if "he" is the subject, the verb should be "has" instead of "have". Tense consistency checks whether the tenses in the text are consistent; for example, in a text describing past events, the past tense should be used primarily. Sentence collocation checks whether the sentence structure and vocabulary collocation conform to the grammatical conventions of the target language; for example, "make a decision" is the correct collocation, while "do adecision" is incorrect. Punctuation checks verify the correct use of punctuation marks, such as periods, commas, and quotation marks.

[0158] Step S151010: Perform an expression habit check on the entire text. The expression habit check includes a check for the matching degree of common sentence patterns, a check for the rationality of word collocation, and a check for the suitability of language style.

[0159] The entire text undergoes an expression habit check. The common sentence structure matching check examines whether the sentence structures in the text conform to the common expressions of the target language. For example, the common English structure "it is + adj. + to do sth." is correctly converted into a common sentence structure in the target language in the translation. The lexical collocation check examines whether the collocations between words are reasonable. For example, "fertile sunshine" is an unreasonable collocation, while "ample sunshine" is reasonable. The register check examines whether the register of the text is appropriate for the usage scenario. For example, formal academic texts should use formal registers rather than colloquial expressions.

[0160] Step S151011: Combine the results of the grammar rule check and the expression habit check to calculate the expression standard score. When the expression standard score is greater than or equal to the expression standard threshold, it is determined that the overall translation meets the target language expression standard.

[0161] The expression standardization score is calculated based on the results of both the grammar rule check and the expression habit check. The calculation method may be a comprehensive assessment based on the number and severity of errors found in the grammar rule check and the number and severity of problems found in the expression habit check. For example, if there are no errors in the grammar rule check and no problems in the expression habit check, then the expression standardization score will be high. When the expression standardization score is greater than or equal to a preset expression standardization threshold (e.g., 0.85), the overall translation is deemed to meet the target language expression standardization requirements.

[0162] Step S151012: Summarize the semantic consistency verification results, logical coherence verification results, and target language expression standardization verification results to generate a semantic adaptation verification report. The semantic adaptation verification report includes the semantic consistency results of each unit, the logical coherence results of each pair of adjacent units, and the overall expression standardization results.

[0163] The semantic consistency verification results (whether the semantic consistency of each unit meets the standard), the logical coherence verification results (whether the logical coherence of each pair of adjacent units meets the standard), and the target language expression standardization verification results (whether the expression standardization of the overall translation meets the standard) are summarized to generate a semantic adaptation verification report. This semantic adaptation verification report records in detail the semantic consistency of each unit, the logical coherence of each pair of adjacent units, and the expression standardization of the overall translation.

[0164] Step S151013: When all verification results meet the standard, the translated sequence after coherence optimization is determined to have passed the semantic adaptation verification; when there are non-compliant results, the non-compliant positions and reasons are marked, and the process is returned to the coherence optimization stage for adjustment.

[0165] Check all verification results in the semantic adaptation verification report. If all verification results meet the standards—that is, the semantic consistency of each unit meets the standards, the logical coherence of each pair of adjacent units meets the standards, and the overall expression standardization of the translation meets the standards—then the translation sequence after coherence optimization is deemed to have passed the semantic adaptation verification. If there are non-compliant results, mark the non-compliant locations (e.g., which unit's semantic consistency does not meet the standards, which pair of adjacent units' logical coherence does not meet the standards, and which part of the overall translation does not meet the standards for expression standardization) and the reasons for non-compliance (e.g., semantic inconsistency is due to incorrect translation of core verbs, logical incoherence is due to missing conjunctions, and expression standardization is due to incorrect word collocation). Then return to the coherence optimization stage, make adjustments based on the marked issues, and then perform semantic adaptation verification again until all verification results meet the standards.

[0166] Step S1511: When semantic inconsistency is found during semantic adaptation verification, the corresponding suggestions in the semantic optimization opinions are traced back, and the unit translation is adjusted in combination with the explicit semantic relationship markers in the intermediate deconstructed text; when logical incoherence is found, the transition logic strengthening suggestions and semantic coherence adaptation suggestions are traced back, and the connecting content is adjusted in combination with the corrected cross-unit semantic relationship index.

[0167] If semantic inconsistencies are found during semantic adaptation verification, the corresponding suggestions in the semantic optimization opinions are reviewed, and the unit translations are adjusted in conjunction with the explicit semantic relationship markers in the intermediate deconstructed text to ensure that the translated semantics are consistent with the original text. If logical incoherence is found, the transition logic strengthening suggestions and semantic coherence adaptation suggestions are reviewed, and the connecting content is adjusted in conjunction with the revised cross-unit semantic relationship index to enhance logical coherence.

[0168] Step S1512: Repeat the semantic adaptation verification and adjustment until the semantic adaptation verification result meets the preset standard, forming the final optimized translation sequence.

[0169] Repeat the semantic adaptation verification and adjustment operations until the semantic adaptation verification results meet the preset standards, such as semantic consistency, logical coherence, and expression standardization, to form the final optimized translation sequence.

[0170] Step S1513: Integrate the final optimized translation sequence according to the unit arrangement order, remove redundant transition content, add translation generation timestamps and optimization basis explanations, and generate the final translated text that conforms to the expression habits of the target language and is semantically complete.

[0171] The final optimized translation sequence is integrated and arranged in unit order, removing redundant transitional content. A translation generation timestamp and optimization rationale are added to generate the final translated text that conforms to the target language's expression habits and is semantically complete.

[0172] Based on the same inventive concept, embodiments of this application provide a human-computer collaborative interactive translation and optimization system for English long and difficult sentences for performing the above-described human-computer collaborative interactive translation and optimization method. The human-computer collaborative interactive translation and optimization system for English long and difficult sentences may include a communication unit, a machine-readable storage medium, and a processor.

[0173] In this embodiment, both the machine-readable storage medium and the processor are located separately within the human-machine collaborative interactive translation and optimization system for complex English sentences. However, it should be understood that the machine-readable storage medium can also be independent of the human-machine collaborative interactive translation and optimization system for complex English sentences and can be accessed by the processor via a bus interface. Alternatively, the machine-readable storage medium can be integrated into the processor and can communicate and interact with external systems through a communication unit.

[0174] The processor is the control center of this human-machine collaborative interactive translation and optimization system for complex English sentences. It connects various parts of the system via various interfaces and lines, and executes software programs and / or modules stored in a machine-readable storage medium, as well as calling data stored in the same medium. This allows for the overall monitoring of the system. Optionally, the processor may include one or more processing cores; for example, it may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may not be integrated into the processor. The machine-readable storage medium stores machine-executable instructions for implementing the scheme of this application, and the processor executes the machine-executable instructions stored in the machine-readable storage medium to implement the human-machine collaborative interactive translation and optimization method for complex English sentences provided in the foregoing method embodiments.

[0175] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A human-computer collaborative interactive translation and optimization method for long and complex English sentences, characterized in that, The method includes: The system performs a two-layer initial deconstruction process on long and complex English sentences, which involves parsing syntactic rules to identify the basic syntactic units in the sentences and capturing the implicit semantic connections between these units through semantic association analysis. This process generates an initial deconstructed text that includes the decomposition results of the basic syntactic units, implicit semantic connection markers, and cross-unit semantic association indexes. The system receives structured adjustment instructions from the interactive subject based on the initial deconstructed text input. These instructions include instructions for merging, splitting, and adjusting the unit order of the basic syntactic constituent units, as well as instructions for correcting the association type and adjusting the association strength of implicit semantic relationship markers. The system further includes: converting the initial deconstructed text into a multi-level visual interactive interface, comprising a unit display layer, an association display layer, and an index display layer. The unit display layer displays basic syntactic constituent units in their original order as nodes, with each node labeled with its basic syntactic constituent unit identifier, unit type, and core vocabulary. The association display layer connects nodes with implicit semantic relationships using attributed lines, labeled with the association type and association strength. The index display layer displays a cross-unit semantic relationship index in list form, containing basic syntactic constituent unit identifiers, a list of associated units, and a ranking of association strength. Based on the structured adjustment instructions, the boundaries of basic syntactic constituent units, the order of unit arrangement, and the implicit semantic association context in the initial deconstructed text are coordinated and adjusted. The cross-unit semantic association index is updated synchronously to generate the structurally adjusted intermediate deconstructed text. The intermediate deconstructed text includes the adjusted secondary syntactic constituent units, the updated explicit semantic association context markers, and the corrected cross-unit semantic association index. The system receives semantic optimization suggestions from the interactive subject for the intermediate deconstructed text input. These suggestions include word substitution and sentence reconstruction suggestions for the initial translation corresponding to a specific sub-syntactic unit, as well as suggestions for strengthening transition logic and semantic coherence adaptation for connecting translations of adjacent sub-syntactic units. By integrating semantic optimization opinions with explicit semantic relationship markers in intermediate deconstructed texts and corrected cross-unit semantic relationship indexes, semantic adaptation verification and logical coherence iterative optimization of the translated text are performed to generate a final translated text that conforms to the expression habits of the target language and is semantically complete.

2. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 1, characterized in that, The process involves performing a two-tiered initial deconstruction of long and complex English sentences, which uses syntactic rules to identify the basic syntactic units within the sentences and semantic association analysis to capture the implicit semantic connections between these units. This generates an initial deconstructed text containing the decomposition results of the basic syntactic units, implicit semantic connection markers, and cross-unit semantic association indexes. Read the complete character sequence of a long and complex English sentence, locate the punctuation marks in the long and complex English sentence according to the character sequence order, divide the long and complex English sentence into multiple first-level segments based on the punctuation marks as the first-level segmentation basis, and then divide each first-level segment into multiple basic segments based on prepositions and conjunctions as the second-level segmentation basis. The pre-trained syntactic rule model is invoked to perform part-of-speech tagging on each basic segment, identifying noun word combinations, verb word combinations, adjective word combinations, and adverb word combinations in the basic segment, and recording the start and end character positions of each type of word combination; Analyzing the grammatical modification relationships between various word combinations, when a noun word combination and an adjective word combination are adjacent and have a descriptive-described relationship, they are integrated into a noun-type basic syntactic unit; when a verb word combination and an adverb word combination are adjacent and have a modifying-modified relationship, they are integrated into a verb-type basic syntactic unit; when two word combinations of the same type are connected by a coordinating conjunction, they are integrated into a coordinating basic syntactic unit, forming the basic syntactic unit decomposition result; The pre-trained semantic association model is invoked to perform semantic feature extraction on each basic syntactic unit in the basic syntactic unit splitting result, generating a semantic feature vector for each basic syntactic unit. The semantic feature vector includes the unit's core semantics, semantic category attributes, and semantic dependency tendency. Calculate the semantic similarity between the semantic feature vectors of any two basic syntactic constituent units. When the semantic similarity reaches a preset association threshold, it is determined that there is an implicit semantic association between the two. Record the identifier of the basic syntactic constituent unit, the association type and the association strength to form an implicit semantic association context marker. A cross-unit semantic association index is constructed based on implicit semantic association context markers. The cross-unit semantic association index includes a list of associated units corresponding to each basic syntactic unit, an association type sorting, and a numerical range of association strength. The list of associated units is arranged from high to low association strength. The basic syntactic constituent unit decomposition results, implicit semantic association context markers, and cross-unit semantic association indexes are integrated into structured data. The structured data is arranged according to the original order of the basic syntactic constituent units in long and difficult English sentences to generate an initial deconstructed text containing the basic syntactic constituent unit decomposition results, implicit semantic association context markers, and cross-unit semantic association indexes.

3. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 1, characterized in that, The receiving interactive subject provides structured adjustment instructions based on the initial deconstructed text input. These structured adjustment instructions include instructions for merging the results of splitting the basic syntactic constituent units, instructions for splitting, instructions for adjusting the unit order, and instructions for correcting the association type and adjusting the association strength of implicit semantic association context markers. It also includes: The interactive subject selects at least two adjacent nodes through the unit display layer, triggers the merge operation entry, enters the merged unit name and merged unit type in the merge operation entry, and generates a merge operation instruction for the split results of the basic syntax constituent units after submission. The merge operation instruction includes the basic syntax constituent unit identifier of the merged node, the merged unit name, the merged unit type and the merge order. The interactive subject selects a single node through the unit display layer, triggering the split operation entry. In the split operation entry, the split position is defined, the name and type of the split sub-unit are entered, and after submission, a split operation instruction is generated for the split result of the basic syntax constituent unit. The split operation instruction includes the basic syntax constituent unit identifier of the split node, the split position coordinates, the name and type of the split sub-unit; The interactive entity can drag nodes in the unit display layer to adjust their arrangement order. After confirming the adjustment, it generates a unit order adjustment instruction for the basic syntax constituent unit splitting result. The unit order adjustment instruction includes the unit arrangement sequence before adjustment, the unit arrangement sequence after adjustment, and the adjustment position identifier. The interactive entity selects a connection with attributes through the associated display layer, triggering the association correction entry. In the association correction entry, the association type is modified and the association strength value is adjusted. After submission, an association type correction instruction and an association strength adjustment instruction are generated for the implicit semantic association context marker. The association type correction instruction includes the identifiers of the two basic syntactic constituent units corresponding to the corrected connection, the original association type, and the new association type. The association strength adjustment instruction includes the identifiers of the two basic syntactic constituent units corresponding to the corrected connection, the original association strength value, and the new association strength value. The interactive entity checks the cross-unit semantic association index through the index display layer. When it finds that the index is inconsistent with the association display layer, it triggers the index synchronization correction entry. After submission, an index synchronization instruction is generated. The index synchronization instruction includes the identifier of the basic syntactic constituent unit to be corrected, the list of association units before correction, and the list of association units after correction. It receives merge operation instructions, split operation instructions, unit order adjustment instructions, association type correction instructions, association strength adjustment instructions, and index synchronization instructions submitted by the interactive subject, and classifies them according to operation type, adds instruction submission timestamps, and integrates them into structured adjustment instructions.

4. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 1, characterized in that, The process involves collaboratively adjusting the boundaries of basic syntactic constituent units, the order of unit arrangement, and the implicit semantic relationships in the initial deconstructed text based on structured adjustment instructions, synchronously updating the cross-unit semantic relationship index, and generating the structurally adjusted intermediate deconstructed text, including: Parse the merge operation instructions in the structured adjustment instructions, extract the basic syntactic constituent unit identifiers of the merged nodes, and locate the corresponding basic syntactic constituent units in the basic syntactic constituent unit splitting results of the initial deconstructed text. Read the text content of the located basic syntactic constituent unit, concatenate the text content according to the merging order in the merging operation instruction, replace it with the text fragment corresponding to the merged unit name in the merging operation instruction, generate the merged secondary syntactic constituent unit, assign a new unit identifier to the merged secondary syntactic constituent unit, and record the merged unit type. The splitting operation instructions in the structured adjustment instructions are analyzed, the basic syntactic constituent unit identifiers and splitting position coordinates of the split nodes are extracted, and the corresponding basic syntactic constituent units are located in the basic syntactic constituent unit splitting results of the initial deconstructed text. Read the text content of the located basic syntactic constituent units, split the text content according to the split position coordinates, generate the split secondary syntactic constituent units according to the sub-unit names in the split operation instructions, assign a new unit identifier to each split secondary syntactic constituent unit, and record the type of the split sub-unit. The unit order adjustment instruction in the structured adjustment instruction is parsed, the adjusted unit arrangement sequence is extracted, and the arrangement order of the basic syntactic constituent unit splitting results in the initial deconstructed text is adjusted to the adjusted unit arrangement sequence to form the adjusted secondary syntactic constituent unit arrangement result. The association type correction instruction in the structured adjustment instruction is parsed, the identifiers of the two basic syntactic constituent units corresponding to the corrected connection and the new association type are extracted, the corresponding association entries are located in the implicit semantic association context markers of the initial deconstructed text, and the association type of the association entries is updated to the new association type. The association strength adjustment instruction in the structured adjustment instruction is parsed, and the identifiers of the two basic syntactic constituent units corresponding to the modified connection and the new association strength value are extracted. The corresponding association items are located in the implicit semantic association context markers of the initial deconstructed text, and the association strength of the association items is updated to the new association strength value to form the updated explicit semantic association context markers. The index synchronization instruction in the structured adjustment instruction is parsed, the basic syntactic constituent unit identifiers to be corrected and the corrected list of related units are extracted, the corresponding index entries are located in the cross-unit semantic association index of the initial deconstructed text, the list of related units of the index entries is updated to the corrected list of related units, and the list of related units is reordered according to the new association strength value to form the corrected cross-unit semantic association index. The system detects whether there are new semantic associations between the adjusted secondary syntactic units, calculates the similarity of semantic feature vectors of adjacent secondary syntactic units, and adds explicit semantic association context marker entries and cross-unit semantic association index entries when the similarity reaches a preset threshold. Integrate the adjusted arrangement of secondary syntactic constituent units, the updated explicit semantic association context markers, and the corrected cross-unit semantic association index, add an adjustment completion timestamp, and generate the structurally adjusted intermediate deconstructed text.

5. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 1, characterized in that, The semantic optimization opinions of the receiving interactive subject for the intermediate deconstructed text input include: The preset translation generation interface is invoked to perform independent translation on each secondary syntactic unit in the intermediate deconstructed text, generating a preliminary translation corresponding to each secondary syntactic unit. The preliminary translation includes a literal translation, a free translation, and a sentence variation version. The intermediate deconstructed text is converted into a translation interaction optimization interface, which includes a unit translation display area, a correlation and connection display area, and an optimization suggestion input area. The unit translation display area displays the original text, preliminary translation, and unit identifier of each sub-syntactic unit in the order of sub-syntactic units. The preliminary translation includes a literal translation, a free translation, and a sentence variation version. The correlation and connection display area shows the connection relationship of the preliminary translations of adjacent sub-syntactic units in the form of a flowchart, and marks the connection position and connection type. The optimization suggestion input area includes a word substitution input box, a sentence reconstruction input box, a transition logic input box, and a semantic coherence input box. The interactive subject selects a preliminary translation of a specific sub-syntactic unit in the unit translation display area. When the subject believes that the wording of the literal translation is inaccurate, it enters the replacement word, the replaced phrase, and the reason for the replacement in the wording replacement input box. This generates wording replacement suggestions for the preliminary translation corresponding to the specific sub-syntactic unit. The wording replacement suggestions include the target sub-syntactic unit identifier, the original word, the replacement word, the replaced phrase, and the reason for the replacement. The interactive subject selects a preliminary translation of a specific sub-syntactic unit in the unit translation display area. When the subject believes that the sentence structure does not conform to the target language habit, it enters the reconstructed sentence structure, the reconstructed sentence, and the reason for the reconstruction in the sentence reconstruction input box. This generates a sentence reconstruction suggestion for the preliminary translation of the specific sub-syntactic unit. The sentence reconstruction suggestion includes the target sub-syntactic unit identifier, the original sentence structure, the reconstructed sentence structure, the reconstructed sentence, and the reason for the reconstruction. The interactive subject selects the connection position of the translation of adjacent secondary syntactic constituent units in the connection display area. When the transition logic is considered unclear, the subject enters the suggested transition words, transition sentence patterns and transition logic types in the transition logic input box to generate transition logic strengthening suggestions for the connection of translations of adjacent secondary syntactic constituent units. The transition logic strengthening suggestions include connection position identifiers, adjacent secondary syntactic constituent unit identifiers, suggested transition words, transition sentence patterns and transition logic types. The interactive subject browses the connection content of the translations of adjacent sub-syntactic units in the connection display area. When the semantic coherence is deemed insufficient, the subject enters semantic supplementation suggestions, word order adjustment suggestions, and coherence adaptation directions in the semantic coherence input box. This generates semantic coherence adaptation suggestions for the connection of translations of adjacent sub-syntactic units. The semantic coherence adaptation suggestions include connection position identifiers, adjacent sub-syntactic unit identifiers, semantic supplementation suggestions, word order adjustment suggestions, and coherence adaptation directions. After the interactive subject submits all input optimization suggestions, the system performs format verification on the wording replacement suggestions, sentence restructuring suggestions, transition logic strengthening suggestions, and semantic coherence adaptation suggestions. The verification content is whether each suggestion contains a complete target identifier, original content, new content, and the reason for the suggestion. Receive verified suggestions for word replacement, sentence restructuring, transition logic enhancement, and semantic coherence adaptation. Categorize the suggestions by type, add a submission timestamp, and integrate them into semantic optimization opinions.

6. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 1, characterized in that, The integrated semantic optimization opinions, explicit semantic relationship markers in the intermediate deconstructed text, and the corrected cross-unit semantic relationship index are used to perform semantic adaptation verification and logical coherence iterative optimization of the translated text, generating a final translated text that conforms to the expression habits of the target language and is semantically complete, including: The wording replacement suggestions in the semantic optimization opinions are analyzed, and the target secondary syntactic unit identifiers, replacement words and replacement phrases are extracted. The corresponding secondary syntactic units and their preliminary translations are located in the intermediate deconstructed text. Replace the original words with alternative words in the preliminary translation of the located secondary syntactic units, replace the original phrases with the alternative phrases, generate the unit translation after wording adjustment, and record the basis for wording adjustment and the text comparison before and after adjustment. The sentence reconstruction suggestions in the semantic optimization opinions are analyzed, and the target secondary syntactic unit identifiers, reconstructed sentence structures and reconstructed sentences are extracted. The corresponding secondary syntactic units and their preliminary translations are located in the intermediate deconstructed text. The initial translations of the located secondary syntactic units are adjusted according to the reconstructed sentence structure, replaced with the reconstructed sentences, and the unit translations after sentence structure adjustment are generated. The basis for sentence structure adjustment and the comparison of sentence structure before and after adjustment are recorded. Integrate the unit translations after wording adjustment and the unit translations after sentence structure adjustment, remove duplicate adjustments, and form a preliminary optimized set of unit translations. Establish a one-to-one correspondence between each sub-syntactic unit and the corresponding preliminary optimized unit translation, and record the correspondence table. The transition logic strengthening suggestions in the semantic optimization opinions are analyzed, and the joint position identifiers, adjacent secondary syntactic unit identifiers, suggested transition words and transition sentence patterns are extracted. The corresponding adjacent unit translations are located in the preliminarily optimized unit translation set. Insert suggested transition words or phrases at the connection points between adjacent translated units, generate a sequence of translated texts after adding transitions, and record the basis for adding transitions and the connection comparison before and after adding transitions. The semantic coherence adaptation suggestions in the semantic optimization opinions are analyzed, and the cohesion position markers, adjacent secondary syntactic unit markers, semantic supplementation suggestions and word order adjustment suggestions are extracted. The corresponding adjacent unit translations are located in the translated sequence after the transition is added. Add necessary semantic supplements based on semantic supplement suggestions, adjust the order of adjacent unit translations based on word order adjustment suggestions, generate a coherence-optimized translation sequence, and record the basis for coherence optimization and semantic comparison before and after optimization. The pre-trained semantic adaptation model is invoked to perform semantic adaptation verification on the coherence-optimized translation sequence. The verification content includes the semantic consistency between each unit translation and the corresponding subsyntactic unit original text, the logical coherence between adjacent unit translations, and the standardization of the target language expression of the overall translation. When semantic inconsistencies are found during semantic adaptation verification, the corresponding suggestions in the semantic optimization opinions are reviewed, and the unit translation is adjusted in combination with the explicit semantic relationship markers in the intermediate deconstructed text; when logical incoherence is found, the transition logic strengthening suggestions and semantic coherence adaptation suggestions are reviewed, and the connecting content is adjusted in combination with the corrected cross-unit semantic relationship index. Repeat the semantic adaptation verification and adjustment until the semantic adaptation verification result meets the preset standard, forming the final optimized translation sequence; The final optimized translation sequence is integrated according to the unit order, redundant transition content is removed, translation generation timestamps and optimization basis explanations are added, and a final translated text that conforms to the expression habits of the target language and is semantically complete is generated.

7. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 2, characterized in that, The pre-trained semantic association model is invoked to perform semantic feature extraction on each basic syntactic unit in the basic syntactic unit decomposition result, generating a semantic feature vector for each basic syntactic unit, including: The text content of each basic syntactic unit is converted into a character-level sequence and input into the embedding layer of a pre-trained semantic association model. The embedding layer maps the character-level sequence into a character embedding vector, which contains the semantic encoding and positional encoding of each character. The character embedding vector is input into the convolutional layer of the semantic association model. Multi-scale convolution kernels are used to perform convolution operations on the character embedding vector to extract local semantic features of text fragments of different lengths and generate local semantic feature maps. The local semantic feature map is input into the pooling layer of the semantic association model. Adaptive pooling operation is used to downsample the local semantic feature map, retain key local semantic features, and generate local semantic vectors. The local semantic vector is input into the recurrent layer of the semantic association model. A bidirectional recurrent structure is used to perform temporal semantic modeling on the local semantic vector, capture the contextual semantic association in the text content, and generate contextual semantic vectors. The context semantic vector is input into the attention layer of the semantic association model, the attention weight of each semantic component is calculated, and the context semantic vector is weighted and aggregated based on the attention weight to highlight the core semantic components and generate the core semantic vector. Extract semantic codes from the core semantic vector that can represent the core meaning of the basic syntactic constituent units, and use them as the core semantics of the units; Based on the core semantic vector query, the semantic category of the basic syntactic unit is determined. The semantic category includes entity category, action category, and attribute category. Analyze the matching degree between the core semantic vector and the preset dependency relation template to determine the other unit types that the basic syntactic constituent units may depend on in the sentence and the unit types that are depended on, as semantic dependency tendency; The core semantics, semantic category attributes, and semantic dependency tendencies of each unit are sorted according to a preset dimension and concatenated into a fixed-length vector to generate a semantic feature vector for each basic syntactic unit. The dimension of the semantic feature vector is consistent with the output dimension of the semantic association model.

8. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 2, characterized in that, The semantic similarity between the semantic feature vectors of any two basic syntactic constituent units is calculated. When the semantic similarity reaches a preset association threshold, it is determined that there is an implicit semantic association between the two. The identifiers of the associated basic syntactic constituent units, the association type, and the association strength are recorded to form an implicit semantic association context marker, including: Read all basic syntactic constituent units from the basic syntactic constituent unit decomposition result, generate a list of basic syntactic constituent unit identifiers, select basic syntactic constituent units in pairs according to the identifier order, and form a set of unit pairs; For each unit pair in the unit pair set, extract two basic syntaxes to form the semantic feature vector of each unit. The semantic feature vector contains vector components corresponding to the unit's core semantics, semantic category attributes, and semantic dependency tendency. Calculate the cosine similarity between two semantic feature vectors and use the cosine similarity as the initial semantic similarity. The calculation of the cosine similarity is based on the ratio of the sum of the products of the vector components to the product of the vector magnitudes. The initial semantic similarity is adjusted based on the semantic category attributes of the two basic syntactic units to obtain the adjusted semantic similarity. The final semantic similarity is obtained by adjusting the semantic dependency tendency of the two basic syntactic units. The final semantic similarity is compared with a preset association threshold. When the final semantic similarity is greater than or equal to the preset association threshold, it is determined that there is an implicit semantic association between the two basic syntactic constituent units in the unit pair. The association type is determined based on the unit type and semantic dependency tendency of the two basic syntactic constituent units. The association types include entity-action association, action-attribute association, entity-attribute association, and parallel entity association. The final semantic similarity is mapped to a numerical value of association strength. The mapping rule is that the final semantic similarity and the numerical value of association strength are positively correlated, and the association strength is generated accordingly. Record the identifiers, defined association types, and association strength values ​​of two basic syntactic units that have implicit semantic relationships, forming an entry for an implicit semantic relationship network. Sort all implicit semantic association entries from highest to lowest according to their association strength values, add a timestamp of the association generation, and form implicit semantic association markers.

9. The human-computer collaborative interactive translation and optimization method for long and complex English sentences according to claim 6, characterized in that, The pre-trained semantic adaptation model is invoked to perform semantic adaptation verification on the coherence-optimized translation sequence. The verification includes the semantic consistency between each unit translation and the corresponding sub-syntactic constituent unit original text, the logical coherence between adjacent unit translations, and the standardization of the target language expression of the overall translation, including: The coherence-optimized translation sequence is split into individual translation units, each translation unit corresponds to a secondary syntactic unit, and the text content of each translation unit and the original text content of the corresponding secondary syntactic unit are extracted. Call the semantic consistency verification module of the pre-trained semantic adaptation model and input the translated text content and the corresponding original text content into the module; The semantic consistency verification module performs semantic parsing on the original text content to extract the core semantic elements of the original text. The core semantic elements include core nouns, core verbs, core adjectives, and semantic relationships. Semantic analysis is performed on the translated text of each unit to extract the core semantic elements of the translation. The core semantic elements include core nouns, core verbs, core adjectives, and semantic relationships. Compare the overlap between the core semantic elements of the original text and the core semantic elements of the translation, calculate the overlap percentage, and determine that semantic consistency is met when the overlap percentage is greater than or equal to the semantic consistency standard threshold. Extract the text content of adjacent translation units in the coherence-optimized translation sequence and form a set of adjacent translation pairs in sequence; Analyze the logical connectors, referential relationships, and semantic connections between adjacent translation pairs, and calculate the logical coherence score. The logical coherence score is calculated based on the combination of connector fit, referential clarity, and semantic naturalness. When the logical coherence score is greater than or equal to the logical coherence standard threshold, the logical coherence between adjacent unit translations is deemed to meet the standard. The coherence-optimized translation sequence is treated as a whole text, and a grammar rule check is performed on the whole text. The grammar rule check includes subject-verb agreement check, tense consistency check, sentence collocation check, and punctuation usage check. The entire text is checked for expression habits, which includes checking the matching degree of common sentence patterns, the reasonableness of word collocation, and the suitability of language style. Based on the combined results of the grammar rule check and the expression habit check, the expression standard score is calculated. When the expression standard score is greater than or equal to the expression standard threshold, the overall translation is judged to meet the target language expression standard. Summarize the semantic consistency verification results, logical coherence verification results, and target language expression standardization verification results to generate a semantic adaptation verification report. The semantic adaptation verification report includes the semantic consistency result of each unit, the logical coherence result of each pair of adjacent units, and the overall expression standardization result. When all verification results meet the standards, the translated sequence after coherence optimization is deemed to have passed the semantic adaptation verification; when there are non-compliant results, the non-compliant positions and reasons are marked, and the process is returned to the coherence optimization stage for adjustment.

10. A human-computer collaborative interactive translation and optimization system for long and complex English sentences, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the human-machine collaborative interactive translation and optimization method for long and difficult English sentences according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images