Multi-language semantic analysis system

Abstract

The invention relates to the technical field of semantic analysis, in particular to a multi-language semantic analysis system, which comprises a context monitoring module, a syntactic regulation adjustment module, a semantic structure matching module, a syntactic error detection module and a syntactic optimization module, and analyzes an applicable language range and an analysis effect of a rule based on a context change classification result. According to the method, the processing precision of multi-language input is improved by monitoring context changes in real time and specifically adjusting syntactic rules, the semantic analysis system can sensitively adapt to tiny changes in languages by implementing similarity calculation of fine-grained semantic units and historical dialogue data, and the accuracy of multi-language input processing is improved. Dynamic adjustment of syntactic rules is changed according to specific context of texts, adaptability to complex language structures is optimized, a real-time error evaluation and adjustment mechanism is introduced, analysis errors caused by language diversity are effectively reduced, and cross-language application efficiency and user experience are improved especially in the aspects of automatic translation and internationalization of language content.

Description

Technical Field

[0001] The present invention relates to the technical field of semantic analysis, and in particular to a multilingual semantic analysis system. Background Art

[0002] Semantic parsing is a core technology in natural language processing (NLP), which focuses on understanding and interpreting the meaning in natural language. This technology involves converting language input into a form that can be understood by computers, usually a structured data format such as logical expressions, dependency trees or semantic frameworks. The goal of semantic parsing is to accurately capture the intent and information content of sentences or texts, so that computers can perform specific tasks based on this information, such as question-answering systems, machine translation, speech recognition, and information retrieval.

[0003] Among them, the multilingual semantic parsing system is an application system of semantic parsing technology used to understand and process multiple languages. The system can accept input in multiple languages, parse the unique semantic structure and grammatical rules of each language through special algorithms, and convert them into standardized semantic representations. Its main uses include improving the efficiency and effectiveness of cross-language information retrieval, content management, automatic translation and international software development. It is particularly suitable for multilingual environments such as international enterprises, multilingual social media analysis and global news aggregation, allowing users to access, analyze and understand large amounts of text data across language boundaries.

[0004] Existing technologies are often limited by their reliance on fixed syntactic rules and a single semantic structure processing strategy, which makes it difficult to effectively adapt to the natural changes and diversity of language. In particular, when dealing with less-used languages ​​or dialects, they are unable to accurately capture language expressions in specific regions and cultures, and the parsing accuracy is insufficient. Existing systems exhibit significant processing delays and strategy rigidity in the face of real-time updated language data, especially in the real-time analysis of multilingual social media and global news content. As a result, information retrieval and content management are unable to meet the rapidly developing international needs in terms of accuracy and efficiency, and increase the difficulty for users in cross-cultural communication and data parsing. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a multilingual semantic parsing system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a multilingual semantic parsing system, the system comprising: The context monitoring module obtains the target language category, core semantic units, syntactic structure features, context-related words and historical conversation data of the input text, calculates the similarity between the core semantic units and the historical conversation data, screens the fluctuating key texts, and obtains the context change classification results; The syntactic rule adjustment module analyzes the applicable language scope and parsing effect of the rule based on the context change classification result, calculates the adaptation degree of the current text, adjusts the priority of the syntactic rule, and obtains the syntactic rule update set; The semantic structure matching module calls the syntactic rule update set, analyzes the text structure according to the part-of-speech distribution, semantic dependency and syntactic boundary of the input text, matches the sentence pattern, and calculates the text structure matching degree to obtain the grammatical structure matching degree; The syntactic error detection module analyzes the parsing error range according to the grammatical structure matching degree, compares the error amount between the parsing result and the expected syntactic pattern, identifies abnormal syntactic rules, and obtains a syntactic error evaluation value; The syntax optimization module calls the syntax error evaluation value, calculates the feedback impact ratio according to the user's error marking and correction information, optimizes the applicable direction of the syntax rule library, and obtains the syntax rule tuning result.

[0007] The improvements of the present invention are that the context change classification results include language style, semantic variation, and interaction mode; the syntactic rule update set specifically includes rule adaptability, scope of application, and priority setting; the grammatical structure matching degree specifically refers to matching accuracy, pattern consistency, and structural integrity; the syntactic error evaluation value includes error degree, error type, and impact interval; and the syntactic rule tuning results include rule adjustment effect, application frequency change, and user feedback results.

[0008] The present invention is improved in that the context monitoring module comprises: The text feature extraction submodule obtains the input text, detects the target language category of the input text, extracts the core semantic units, and identifies the syntactic structure features. It also collects context-related words and historical conversation data to establish a text feature dataset. The context change analysis submodule calculates the similarity between the core semantic unit and the historical dialogue data based on the text feature dataset, and analyzes the context change of the input text in the target language category using the formula: ; Calculate the contextual deviation of each semantic unit , identify the key points of context change, where Represents the input text The context weight value of the core semantic unit, Represents the historical conversation data The context weight value of the core semantic unit, Representative The contextual criticality weight of the core semantic unit, Represents the total number of core semantic units; The fluctuation key screening submodule analyzes the fluctuation key input text based on the context shift degree, compares the change trend of the difference input text, screens the fluctuation key input text, and obtains the context change classification result.

[0009] The present invention is improved in that the syntactic rule adjustment module comprises: The rule applicability analysis submodule analyzes the applicable language scope of each rule based on the context change classification result and the rules in the syntactic rule library, analyzes the rule analysis effect, and obtains the rule adaptability data; The rule priority adjustment submodule calculates the degree of adaptation between the current text context category and each rule based on the rule adaptation data, using the formula: ; Get the syntax rule application priority ,in, Representative The fitness of the syntactic rules, Representative The weight of the syntactic rule, represents the key feature value of the current text context category, Representative The feature value of the context category covered by the syntactic rule, Represents the total number of syntactic rules; The application ratio update submodule calculates the application ratio of the difference rules based on the syntactic rule application priority, adjusts the application frequency of the low-adaptation rules, increases the application proportion of the high-adaptation rules, and screens the priority-critical rules to obtain the syntactic rule update set.

[0010] The present invention is improved in that the semantic structure matching module comprises: The text structure parsing submodule calls the syntactic rule update set, parses the part-of-speech categories in the text according to the part-of-speech distribution, semantic dependency and syntactic boundaries of the input text, analyzes the hierarchical relationship of the syntactic structure, and extracts the key information of the core verb, subject and object to establish text structure data; The sentence pattern matching submodule extracts syntactic structure information based on the text structure data and matches the corresponding sentence pattern using the formula: ; Calculating sentence matching deviation , get the sentence matching error data, where, Represents the grammatical structure feature value of the input text, The feature value representing the standard sentence pattern, represents the weight of the grammatical feature item, The number of dimensions representing the syntactic features of the text; The structural matching degree calculation submodule calculates the text structure matching degree based on the sentence matching error data, sets a matching degree threshold range, analyzes the matching conditions of the difference sentence patterns, and obtains the grammatical structure matching degree.

[0011] The present invention is improved in that the syntax error detection module comprises: The syntactic structure matching submodule obtains the parsing error range of the input text according to the grammatical structure matching degree, calculates the syntactic matching degree of the input text through the syntactic rule library, and obtains the syntactic matching degree value; The parsing error calculation submodule compares the error amount between the input text and the expected syntactic pattern based on the syntactic matching value, and screens out the syntactic rules that are beyond the normal range, using the formula: ; Calculate the analytical error value HE, where Represents the number of syntactic structures in the text, Representative The syntactic matching degree of the input text, Representative The expected syntactic match, Representative The weight of a syntactic structure; The syntax evaluation submodule calls the parsing error value, classifies the error items, determines the syntax error type, analyzes the error impact range, and obtains a syntax error evaluation value.

[0012] The present invention is improved in that the syntax optimization module comprises: The error assessment submodule calls the syntactic error assessment value, compares the user error annotation information, and uses the formula: ; Calculate the syntactic error deviation of user annotations , screening impact factors, among which, represents the error value of the evaluation, Represents the error value marked by the user, Represents the total number of sentences; The rule adjustment submodule calculates the impact ratio of user feedback based on the syntactic error deviation and user correction information, optimizes the applicable direction of the syntactic rule library according to the impact ratio, and adjusts the application ratio of the syntactic rules to obtain the syntactic rule tuning result.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the processing accuracy of multilingual input is improved by real-time monitoring of context changes and targeted adjustment of syntactic rules. By implementing fine-grained semantic unit and historical conversation data similarity calculation, the semantic parsing system can sensitively adapt to slight changes in the language. The dynamic adjustment of syntactic rules changes according to the specific context of the text, optimizes the adaptability to complex language structures, and introduces a real-time error evaluation and adjustment mechanism, which effectively reduces parsing errors caused by language diversity and improves the efficiency of cross-language applications and user experience, especially in automatic translation and internationalized language content. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a system flow chart of the present invention; Figure 2 is a flow chart of the context monitoring module in the present invention; Figure 3 It is a flow chart of the syntax rule adjustment module in the present invention; Figure 4 It is a flowchart of the semantic structure matching module in the present invention; Figure 5 is a flow chart of a syntax error detection module in the present invention; Figure 6 It is a flowchart of the syntax optimization module in the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0017] Example: See Figure 1 , the present invention provides a technical solution: a multilingual semantic parsing system comprising: The context monitoring module obtains the target language category, core semantic units, syntactic structure features, contextual related words and historical conversation data of the input text, analyzes the context changes of the input text in the target language category, calculates the similarity between the core semantic units and the historical conversation data, identifies the key points of context changes, and screens the input text with key fluctuations to obtain the context change classification results; The syntactic rule adjustment module analyzes the applicable language scope and parsing effect of each rule based on the results of context change classification and the rules in the syntactic rule library, calculates the degree of adaptation to the current text context category, adjusts the application priority of the syntactic rules, and updates the application ratio of the syntactic rules according to the degree of adaptation to obtain the syntactic rule update set; The semantic structure matching module calls the syntactic rule update set, performs text structure analysis according to the part-of-speech distribution, semantic dependency and syntactic boundary of the input text, matches the corresponding sentence pattern, and calculates the structural matching degree of the text to obtain the grammatical structure matching degree; The syntactic error detection module obtains the parsing error range of the input text based on the grammatical structure matching degree, compares the error amount of the parsing result with the expected syntactic pattern, identifies the syntactic rules whose parsing error exceeds the normal range, and analyzes the error type and its impact range to obtain the syntactic error evaluation value; The syntax optimization module calls the syntax error evaluation value, calculates the impact ratio of user feedback based on user error marking and correction information, optimizes the applicable direction of the syntax rule library, and adjusts the application ratio of the syntax rules to obtain the syntax rule tuning results.

[0018] The results of context change classification include language style, semantic variation, and interaction mode. The syntactic rule update set specifically includes rule adaptability, scope of application, and priority setting. The grammatical structure matching degree specifically refers to matching accuracy, pattern consistency, and structural integrity. The syntactic error assessment value includes error degree, error type, and impact range. The syntactic rule tuning results include rule adjustment effects, changes in application frequency, and user feedback results.

[0019] See also Figure 2 , the context monitoring module includes: The text feature extraction submodule obtains the input text, detects the target language category of the input text, extracts the core semantic units, and identifies the syntactic structure features. It also collects context-related words and historical conversation data to establish a text feature dataset. First, detect the target language category of the input text. Through comparative analysis based on character distribution, lexical features, and grammatical structures, if the character set contains a high proportion of pinyin characters and conforms to the spelling structure of English, it can be determined as an English text. If the character set contains Chinese characters and the word collocations conform to Chinese grammar, it can be determined as a Chinese text. For multilingual texts, they are classified according to the language category with the highest proportion. For example, if the input text is "Hello, 你好", with the English character proportion being 50% and the Chinese character proportion being 50%, the dominant language can be determined based on the context information, and the core semantic units can be extracted. For different language categories, different lexical analysis strategies are adopted. For example, in Chinese, the core semantic units can be extracted through word segmentation technology. For example, "人工智能技术" can be split into "人工智能" and "技术". In English, they are extracted based on part-of-speech tagging. For example, the core semantic units in "machinelearningmodel" can be "machinelearning" and "model". Identify syntactic structure features and establish syntactic tree structures for different languages. In Chinese text, "我喜欢人工智能" can be parsed into a subject-predicate-object structure, where "我" is the subject, "喜欢" is the verb, and "人工智能" is the object. In English text, "IloveAI" can also be parsed into a subject-predicate-object structure. At the same time, collect context-related words and historical dialogue data, and based on the information of the previous and subsequent sentences, judge the context relevance. For example, in the historical dialogue data "你喜欢科技吗?我喜欢人工智能", there is a semantic association between "人工智能" and "科技", which can be used as context-related words to ensure that the text feature data can fully reflect the semantics and syntactic characteristics of the input text, and generate a text feature dataset.

[0020] Based on the text feature dataset, the context change analysis sub-module calculates the similarity between the core semantic units and the historical dialogue data, analyzes the context changes of the input text in the target language category, and uses the formula: ; Calculate the context offset degree of each semantic unit , identify the key points of context change, where, represents the context weight value of the th core semantic unit of the input text, represents the context weight value of the th core semantic unit in the historical dialogue data, represents the context criticality weight of the th core semantic unit, represents the total number of core semantic units; By calculating the deviation between the weight of the core semantic unit in the input text and the weight of the corresponding semantic unit in the historical conversation data, the degree of context shift is determined. If the weight of a core semantic unit in the current text is much higher than that in the historical conversation data, it means that the importance of the word in the context has changed. The value can be obtained through data monitoring and sampling. Context change calculation parameter table: ; As shown in the table above, the weight of "AI" in the input text Higher weight than historical text , its contextual criticality weight , substitute the parameters into the formula for calculation: ; ; ; ; Calculate the context deviation The larger the value, the more obvious the context change. This result shows that the importance of "AI" in the input text has increased significantly compared to the historical dialogue data, resulting in a higher overall context shift. The key points of context change are identified and the context shift is obtained.

[0021] The fluctuation key screening submodule analyzes the fluctuation key input text based on the context deviation degree, compares the change trend of the difference input text, screens the fluctuation key input text, and obtains the context change classification result; Analyze the fluctuating key input texts and calculate the context change trends of different input texts. If the context deviation of a certain input text is much higher than the average level, it indicates that the text has prominent context change characteristics. For example, in multiple dialogue data samples, the average context deviation is 0.5, while the context deviation of the current input text is 0.81, and the deviation exceeds 60%. It can be determined that the context of the text has changed significantly, and the fluctuating key input texts are screened, as shown in the following table: Context shift analysis table: ; The context shift of text C is much higher than that of other texts, so it is determined to be a fluctuating key text. By calculating the context shift trends of different texts, a difference comparison curve is established to determine which texts have a change range exceeding the set threshold. For example, if the set threshold is the average context shift + 30%, then the calculation is: , where the contextual deviation of text C is , so it is determined to be a fluctuating key text and the context change classification result is obtained.

[0022] See also Figure 3, the syntax rule adjustment module includes: The rule applicability analysis submodule analyzes the applicable language scope of each rule based on the results of context change classification and the rules in the syntactic rule library, analyzes the rule parsing effect, and obtains the rule adaptability data; For different language categories, the grammatical structure features corresponding to the rules are extracted. For example, some rules are only applicable to the subject-predicate-object structure of Chinese, while some rules are applicable to the passive voice sentence structure of English. In specific analysis, the syntactic structure data of the rules are obtained. For example, the syntactic rules of Chinese adopt the form of "subject + verb + object", while English adopts "subject + passive verb + by + agent". When analyzing the applicability of the rules, the application frequency of each rule in different language categories is calculated, and combined with the classified context change results, the context category of the current text is judged to determine the applicability of the rule. The parsing effect of each rule is then analyzed. The parsing success rate, misjudgment rate and coverage of the rules are calculated to measure their parsing capabilities in different context categories. For example, if the parsing success rate of a rule in Chinese text reaches 90%, but the misjudgment rate in English text is as high as 50%, it means that the rule has a higher adaptability in Chinese and a lower adaptability in English. To further quantify the adaptability, the matching degree between the syntactic structure after rule parsing and the actual structure of the text is compared, the matching rate is calculated, and a weighted operation is performed in combination with the misjudgment rate to obtain the rule adaptability data.

[0023] The rule priority adjustment submodule calculates the degree of adaptation between the current text context category and each rule based on the rule adaptation data, using the formula: ; Get the syntax rule application priority ,in, Representative The fitness of the syntactic rules, Representative The weight of the syntactic rule, represents the key feature value of the current text context category, Representative The feature value of the context category covered by the syntactic rule, Represents the total number of syntactic rules; In different contexts, the adaptability of rules is different. For example, a rule for subject-verb-object structure is much more adaptable in the Chinese environment than in the English environment. Therefore, it is necessary to combine the weight and category feature differences for calculation. There are five rules, and their adaptability, weight, and context feature values ​​are shown in the following table: Syntax rule priority calculation parameter table: ; Substitute the values ​​into the formula to calculate: ; ; ; Calculate the rule priority ,This result shows that the overall adaptability between the current text context category and the ,rule is high, and it can be used as a high-priority application rule to obtain the ,application priority of the syntactic rule.

[0024] The application ratio update submodule calculates the application ratio of difference rules based on the syntactic rule application priority, adjusts the application frequency of low-adaptation rules, increases the application ratio of high-adaptation rules, and screens the rules with key priorities to obtain the syntactic rule update set; Adopt a hierarchical update strategy to reduce the application ratio of rules with lower adaptability and increase the ratio of rules with higher adaptability. For example, if the priority score of a rule is , and the average priority score in the current context is 0.50, then the application ratio of this rule can be appropriately increased, and the application ratio of rules below 0.50 needs to be reduced. The adjustment threshold is set to That is, the application ratio of rules with a value higher than 0.50 is increased by 20%, and the application ratio of rules with a value lower than 0.50 is reduced by 20%. The rules with the highest adaptability are selected, and their adjusted application ratios are calculated to finally obtain the updated set of syntactic rules.

[0025] See also Figure 4 , the semantic structure matching module includes: The text structure parsing submodule calls the syntactic rule update set, parses the part-of-speech categories in the text according to the part-of-speech distribution, semantic dependency and syntactic boundaries of the input text, analyzes the hierarchical relationship of the syntactic structure, and extracts the key information of the core verb, subject and object to establish the text structure data; Get the word segmentation results of the input text and identify the part of speech category of each word. For example, the sentence "cat eats fish" has the word segmentation result of [cat / noun, eat / verb, fish / noun]. Then, analyze the semantic dependency relationship of each word. For example, "cat" as the subject forms a subject-predicate relationship with "eat", while "fish" as the object forms a verb-object relationship with "eat". Then, determine the syntactic boundaries and identify the complex sentence structure. For example, "cat eats fish, dog drinks water" contains two parallel clauses, each of which independently forms a complete subject-predicate-object structure. Then, analyze the syntactic boundaries. The hierarchical relationship of the structure is analyzed to determine the hierarchical relationship between the main clause and the subordinate clause of the sentence. For example, in "Because the weather is hot, cats like to stay in the shade", the main clause is "Cats like to stay in the shade", and "Because the weather is hot" is its adverbial clause. Subsequently, the key information of the core verb, subject and object is extracted and structured data is stored. For example, for the sentence "Students study mathematics seriously", the key information extracted is [Subject = student, verb = study, object = mathematics]. Finally, the data is sorted and stored as text structure data.

[0026] The sentence pattern matching submodule extracts syntactic structure information based on text structure data and matches the corresponding sentence pattern using the formula: ; Calculating sentence matching deviation , get the sentence matching error data, where, Represents the grammatical structure feature value of the input text, indicating the The numerical quantification results of dimensional grammatical features, such as the numerical encoding of subject-predicate-object relations, the depth of clause levels, etc. The characteristic value representing the standard sentence pattern indicates that the standard sentence structure is in Reference values ​​on dimensional features, such as the average dependency depth of a certain type of sentence, the proportion of syntactic components, etc. Represents the weight of the grammatical feature item, which measures the importance of different grammatical features to the matching calculation. The weight is set according to the influence of the language feature. For example, the weight of the feature with higher syntactic structure stability is larger. The number of dimensions representing the syntactic features of the text, that is, the total number of syntactic features involved in the calculation process, such as part-of-speech distribution, dependency structure, syntactic depth, and other dimensions; Get the syntactic structure data of the input text. For example, for the sentence "Students study mathematics seriously", the text structure data is [Subject = Student, Verb = Study, Object = Mathematics]. Then, select the closest standard sentence pattern from the sentence pattern library, such as the common subject-verb-object sentence pattern [Subject + Verb + Object]. The calculation involves five grammatical features, and their values ​​are as follows: Sentence matching calculation parameter table: ; Substitute the values ​​into the formula to calculate: ; ; ; Calculate the sentence matching error ,This result shows that the deviation between the input text and the standard sentence pattern is small, ,which can be used to further calculate the sentence matching error data.

[0027] The structural matching degree calculation submodule calculates the text structure matching degree based on the sentence matching error data, sets the matching degree threshold range, analyzes the matching of the difference sentence patterns, and obtains the grammatical structure matching degree; Set a matching threshold range. For example, sentences with a matching error of less than 0.05 are considered to have a high matching degree, while sentences with a matching error of more than 0.15 are considered to have a low matching degree. The setting of this threshold is based on the error distribution of different text matching samples. For example, among 5,000 sentences, more than 85% of the sentences with a matching error of less than 0.05 are manually annotated correct matches, so 0.05 is set as the high matching threshold. The error rate of sentences with a matching error of more than 0.15 exceeds 70%, so 0.15 is set as the low matching threshold. Then, compare the matching error data and analyze the matching errors of different sentence patterns. For example, for the sentence "Students study mathematics seriously", the matching error calculation value is 0.034, which is lower than the set threshold of 0.05. Therefore, the text is highly matched with the standard sentence pattern. Then, the matching weight coefficient is adjusted to make different types of grammatical features maintain a reasonable proportion in the final calculation. For example, for the matching of long sentence structures, it is necessary to increase the weight of sentence hierarchy complexity, while for short sentence matching, the depth of dependency is more critical. Finally, the overall matching degree of the text is calculated. For example, when the standard matching degree is set to 1.0, and the matching error is The value is compared with the benchmark value of 0.15, and the matching degree can be calculated by the following formula: , substitute the calculated value : , calculate the grammatical structure matching degree , indicating that the overall matching degree of the sentence is high.

[0028] See also Figure 5 , the syntax error detection module includes: The syntactic structure matching submodule obtains the parsing error range of the input text based on the syntactic structure matching degree, calculates the syntactic matching degree of the input text through the syntactic rule library, and obtains the syntactic matching degree value; The input text is decomposed into independent sentences, and each sentence is further decomposed into words or morphemes. Then, a syntactic analyzer is used to parse each sentence to generate a corresponding syntactic tree or dependency graph. The structure represents the grammatical relationship between the words in the sentence, such as subject-predicate relationship, attributive-adverbial-complement relationship, etc. Subsequently, the generated syntactic structure is compared with a predefined syntactic rule library, which contains common syntactic patterns and structures in the target language, such as subject-predicate-object (SVO) structure, attributive clause structure, etc. By comparison, the degree of match between the syntactic structure of the input text and the expected pattern can be determined. The matching degree can be quantified by calculating the difference between the input syntactic structure and the expected syntactic structure. For example, using the edit distance or other similarity measurement methods, assuming that the edit distance between the actual syntactic structure of the input sentence and the expected structure is 3, and the sentence length is 10, the matching degree can be expressed as 1-(3 / 10)=0.7, that is, the matching degree is 70%. In this way, the parsing error range of the input text is obtained.

[0029] The parsing error calculation submodule compares the error between the input text and the expected syntactic pattern based on the syntactic matching value, and filters out the syntactic rules that are beyond the normal range, using the formula: ; Calculate the analytical error value HE, where Represents the number of syntactic structures in the text, Representative The syntactic matching degree of the input text, Representative The expected syntactic match, Representative The weight of a syntactic structure; Compare the error between the input text and the expected syntactic pattern. First, set an expected syntactic matching threshold, such as 0.8, which means that sentences with a matching degree lower than 80% have syntactic errors. For each input sentence, calculate its syntactic matching value. If the value is lower than the expected threshold, the sentence is considered to have a syntactic error. Next, filter out the sentences below the threshold, and calculate the parsing error value for each filtered sentence. If a sentence contains 5 syntactic structures, the syntactic matching degrees of the input text are 0.6, 0.7, 0.5, 0.8 and 0.9 respectively, the expected syntactic matching degrees are all 0.9, and the weights are all 1, then the parsing error value is calculated as follows: ; ; ; ; This result shows that the analytical error value It reflects the average deviation between the syntactic match of the input text and the expected syntactic match. The error value is between 0.1 and 0.3, which belongs to the medium error range, indicating that the syntactic structure of the text deviates from the standard mode, but still has a certain degree of comprehensibility. It needs to be combined with the error classification threshold (such as minor errors: , medium error: , serious error: ), which is used to classify the level of syntactic errors, and the determination of the classification thresholds of 0.1 and 0.3 for parsing error is based on a test of 500,000 sentence samples. For sentences with an error value less than 0.1, the probability that their grammatical errors affect reading comprehension is less than 5%. For sentences with an error value between 0.1 and 0.3, the probability that their errors affect reading comprehension is between 5% and 30%. For sentences with an error value greater than 0.3, the probability that their errors affect reading comprehension is more than 50%. Therefore, 0.1 and 0.3 are set as the dividing points for minor errors, moderate errors, and severe errors. This value can be further optimized as the scale of test data increases.

[0030] The syntax evaluation submodule calls the parsing error value, classifies the error items, determines the syntax error type, analyzes the error impact range, and obtains the syntax error evaluation value; According to the size of the parsing error value, the sentences are divided into different error levels. For example, an error value less than 0.1 is a minor error, between 0.1 and 0.3 is a moderate error, and greater than 0.3 is a severe error. Then, the type of syntactic error is determined and the scope of error impact is analyzed. Specifically, for each sentence, its main syntactic error type is determined, such as subject-predicate inconsistency, incorrect modifier position, incorrect clause structure, etc. This can be achieved by analyzing the specific differences between the syntactic structure and the expected pattern. For example, if the subject and predicate of a sentence are inconsistent in number or person, it can be determined to be a subject-predicate inconsistency error. Finally, the syntactic error evaluation value is obtained by combining the error type and error level. The syntactic error evaluation value can be a comprehensive score that reflects the degree of syntactic correctness of the sentence. For example, for a sentence, if its parsing error value is 0.25 and it is determined to be a subject-predicate inconsistency error, a corresponding evaluation value can be given, such as 70 points (out of 100 points), indicating that the sentence has a moderate degree of syntactic error.

[0031] See also Figure 6 , the syntax optimization module includes: The error assessment submodule calls the syntactic error assessment value and compares the user's error annotation information using the formula: ; Calculate the syntactic error deviation of user annotations , screening impact factors, among which, represents the error value of the evaluation, Represents the error value marked by the user, Represents the total number of sentences; There is a set of sentence data sets containing error values ​​automatically evaluated by the system and the error value manually marked by the user , calculate the absolute difference between the two for each sentence, and add up the average value to form the error deviation , if the data set contains five sentences, the error values ​​are as follows: ; The error deviation is calculated as follows: ; After obtaining the error deviation, it is necessary to screen the influencing factors, such as the difference in subjective judgment of different users on the same grammatical error, or the recognition accuracy of the system for a certain type of error. The screening standard can be set as the error deviation threshold. The basis for determining the threshold is the error fluctuation range marked by different users in the historical training data set. It is usually set to 1.5 times the standard deviation of the average error of the training set. If the error mean of the training data set is 0.02 and the standard deviation is 0.015, the threshold is calculated as follows: ; when When it exceeds 0.0425, it is considered that there is a large difference in the user annotations and further analysis is required, while below this value, the system error is considered to be small and no adjustment is required to obtain the syntactic error deviation amount.

[0032] The rule adjustment submodule calculates the impact ratio of user feedback based on the syntactic error deviation and user correction information, optimizes the applicable direction of the syntactic rule library according to the impact ratio, and adjusts the application ratio of the syntactic rules to obtain the syntactic rule tuning results; Combined with the user correction information, the impact ratio of user feedback is calculated, that is, the degree of adjustment of the application direction of the syntactic rule library by the user modification information is determined. Specifically, by comparing the sentence structure before and after the user modification, statistics are counted on which grammatical rules have changed in application frequency, and their impact ratios are calculated. For example, assuming that a rule was originally applicable to 100 sentences, and after adjustment through user feedback, it becomes applicable to 120 sentences, then the impact ratio is calculated as follows: ,like If it exceeds the preset threshold (such as 0.1), it is necessary to adjust the applicable direction of the rule, optimize the rule base, and adjust the application ratio of the rule accordingly. For example, if the original applicability of a rule is 50%, it needs to be increased to 60% after adjustment to obtain the syntactic rule tuning result.

[0033] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A multilingual semantic parsing system, characterized in that: The system comprises: The context monitoring module obtains the target language category, core semantic units, syntactic structure features, context-related words and historical conversation data of the input text, calculates the similarity between the core semantic units and the historical conversation data, screens the fluctuating key texts, and obtains the context change classification results; The syntactic rule adjustment module analyzes the applicable language scope and parsing effect of the rule based on the context change classification result, calculates the adaptation degree of the current text, adjusts the priority of the syntactic rule, and obtains the syntactic rule update set; The semantic structure matching module calls the syntactic rule update set, analyzes the text structure according to the part-of-speech distribution, semantic dependency and syntactic boundary of the input text, matches the sentence pattern, and calculates the text structure matching degree to obtain the grammatical structure matching degree; The syntactic error detection module analyzes the parsing error range according to the grammatical structure matching degree, compares the error amount between the parsing result and the expected syntactic pattern, identifies abnormal syntactic rules, and obtains a syntactic error evaluation value; The syntax optimization module calls the syntax error evaluation value, calculates the feedback impact ratio according to the user's error marking and correction information, optimizes the applicable direction of the syntax rule library, and obtains the syntax rule tuning result.

2. The multilingual semantic parsing system according to claim 1, characterized in that: The context change classification results include language style, semantic variation, and interaction mode. The syntactic rule update set specifically includes rule adaptability, scope of application, and priority setting. The grammatical structure matching degree specifically refers to matching accuracy, pattern consistency, and structural integrity. The syntactic error evaluation value includes error degree, error type, and impact range. The syntactic rule tuning results include rule adjustment effect, application frequency change, and user feedback results.

3. The multilingual semantic analysis system according to claim 1, characterized in that: The context monitoring module includes: The text feature extraction submodule obtains the input text, detects the target language category of the input text, extracts the core semantic units, and identifies the syntactic structure features. It also collects context-related words and historical conversation data to establish a text feature dataset. The context change analysis submodule calculates the similarity between the core semantic unit and the historical dialogue data based on the text feature dataset, and analyzes the context change of the input text in the target language category using the formula: ； Calculate the contextual deviation of each semantic unit , identify the key points of context change, where Represents the input text The context weight value of the core semantic unit, Represents the historical conversation data The context weight value of the core semantic unit, Representative The contextual criticality weight of each core semantic unit, Represents the total number of core semantic units; The fluctuation key screening submodule analyzes the fluctuation key input text based on the context shift degree, compares the change trend of the difference input text, screens the fluctuation key input text, and obtains the context change classification result.

4. The multilingual semantic parsing system according to claim 1, characterized in that: The syntax rule adjustment module includes: The rule applicability analysis submodule analyzes the applicable language scope of each rule based on the context change classification result and the rules in the syntactic rule library, analyzes the rule analysis effect, and obtains the rule adaptability data; The rule priority adjustment submodule calculates the degree of adaptation between the current text context category and each rule based on the rule adaptation data, using the formula: ； Get the syntax rule application priority ,in, Representative The fitness of the syntactic rules, Representative The weight of the syntactic rule, represents the key feature value of the current text context category, Representative The feature value of the context category covered by the syntactic rule, Represents the total number of syntactic rules; The application ratio update submodule calculates the application ratio of the difference rules based on the syntactic rule application priority, adjusts the application frequency of the low-adaptation rules, increases the application proportion of the high-adaptation rules, and screens the priority-critical rules to obtain the syntactic rule update set.

5. The multilingual semantic parsing system according to claim 1, characterized in that: The semantic structure matching module includes: The text structure parsing submodule calls the syntactic rule update set, parses the part-of-speech categories in the text according to the part-of-speech distribution, semantic dependency and syntactic boundaries of the input text, analyzes the hierarchical relationship of the syntactic structure, and extracts the key information of the core verb, subject and object to establish text structure data; The sentence pattern matching submodule extracts the syntactic structure information based on the text structure data and matches the corresponding sentence pattern using the formula: ； Calculating sentence matching deviation , get the sentence matching error data, where, Represents the grammatical structure feature value of the input text, The feature value representing the standard sentence pattern, represents the weight of the grammatical feature item, The number of dimensions representing the syntactic features of the text; The structural matching degree calculation submodule calculates the text structure matching degree based on the sentence matching error data, sets a matching degree threshold range, analyzes the matching conditions of the difference sentence patterns, and obtains the grammatical structure matching degree.

6. The multilingual semantic analysis system according to claim 1, characterized in that: The syntax error detection module comprises: The syntactic structure matching submodule obtains the parsing error range of the input text according to the grammatical structure matching degree, calculates the syntactic matching degree of the input text through the syntactic rule library, and obtains the syntactic matching degree value; The parsing error calculation submodule compares the error amount between the input text and the expected syntactic pattern based on the syntactic matching value, and screens out the syntactic rules that are beyond the normal range, using the formula: ； Calculate the analytical error value HE, where Represents the number of syntactic structures in the text, Representative The syntactic matching degree of the input text, Representative The expected syntactic match, Representative The weight of a syntactic structure; The syntax evaluation submodule calls the parsing error value, classifies the error items, determines the syntax error type, analyzes the error impact range, and obtains a syntax error evaluation value.

7. The multilingual semantic analysis system according to claim 1, characterized in that: The syntax optimization module includes: The error assessment submodule calls the syntactic error assessment value, compares the user error annotation information, and uses the formula: ； Calculate the syntactic error deviation of user annotations , screening impact factors, among which, represents the error value of the evaluation, Represents the error value marked by the user, Represents the total number of sentences; The rule adjustment submodule calculates the impact ratio of user feedback based on the syntactic error deviation and user correction information, optimizes the applicable direction of the syntactic rule library according to the impact ratio, and adjusts the application ratio of the syntactic rules to obtain the syntactic rule tuning result.

Images