Method for enhancing translation precision of place name address by using context awareness and geographic library

Abstract

The invention discloses a method for enhancing geographical name and address translation precision by using context awareness and a geographic library, relates to the technical field of intelligent translation, and constructs a dynamic collaborative ambiguity resolution framework by fusing a context awareness mechanism and multi-dimensional attributes of a geographic knowledge base so as to break through semantic limitation of traditional geographical name and address translation. Specifically, firstly, dynamic association information implied in an address text is captured by utilizing context semantic coding of a non-geographical name text fragment, and structured feature embedding is performed on candidate entities in combination with a geographic knowledge base to extract multi-dimensional attributes of the candidate entities, so that cross-modal disambiguation decision is realized in a semantic aggregation stage; through a geographic feature cross validation mechanism guided by context semantics, the problem of translation ambiguity caused by static rule dependence and geographic attribute segmentation in the prior art is effectively solved, so that a more accurate multi-language address translation effect is achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent translation technology, and more specifically, to a method for enhancing the accuracy of place name and address translation by utilizing context awareness and a geographic knowledge base. Background Art

[0002] In today's globalized context, address translation is an important part of cross-language communication, and its accuracy is crucial to ensure the correct transmission of information. In the field of traditional place name and address translation, existing technologies generally face the problem of insufficient translation accuracy due to language ambiguity and the complexity of geographical entities. Conventional methods usually rely on static dictionary matching or rule-based translation strategies, which are difficult to cope with the challenge of ambiguity of the same place name in different contexts. For example, when there are administrative divisions, natural geographical entities or cultural landmarks with the same name in the source language, existing systems often lack the ability to dynamically identify contextual associations, resulting in deviations between the translation results and the target geographical location.

[0003] In addition, the application of geographic knowledge base is not deep enough in the existing technology. The existing translation model does not fully integrate the spatial hierarchical relationship, administrative division affiliation and historical naming characteristics in the structured geographic knowledge base, which makes the disambiguation process limited to the analysis of local text fragments and cannot be cross-validated through multi-dimensional geographic attributes, thus affecting the accuracy of the final translation results.

[0004] Therefore, a method that utilizes context awareness and geographic database to enhance the accuracy of place name and address translation is needed to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, a method for enhancing the accuracy of place name and address translation using context awareness and a geographic library is provided, which includes: S1: receiving a source language address text to be translated; S2: Perform standardized preprocessing and structured semantic parsing on the source language address text to obtain a candidate list of place name entities and non-place name text fragments; S3: extracting a first place-name entity to be disambiguated from the place-name entity candidate list; S4: extracting multiple candidate place name entity entries matching the first place name entity to be disambiguated from the geographic knowledge base; S5: inputting a plurality of candidate place name entity entries and non-place name text segments matching the first place name entity to be disambiguated into an ambiguity resolution model to obtain a disambiguated place name entity of the first place name entity to be disambiguated; Step 6: Loop through S3 to S5 to obtain a list of disambiguated place name entities; Step 7: Based on the disambiguated place name entity list, obtain the target language address text.

[0007] The present application has at least the following technical effects: Compared with the prior art, the present application provides a method for enhancing the accuracy of place name and address translation by using context awareness and geographic library, which integrates the context awareness mechanism with the multi-dimensional attributes of the geographic knowledge base to construct a dynamic and collaborative ambiguity resolution framework to break through the semantic limitations of traditional place name and address translation. Specifically, the context semantic encoding of non-place name text fragments is first used to capture the dynamic association information implicit in the address text, and the candidate entities are embedded with structured features in combination with the geographic knowledge base to extract the multi-dimensional attributes of the candidate entities, thereby realizing cross-modal disambiguation decisions in the semantic aggregation stage. The present application can effectively solve the translation ambiguity problem caused by static rule dependence and geographic attribute fragmentation in the prior art through a geographic feature cross-validation mechanism guided by context semantics, thereby achieving a more accurate multi-language address conversion effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 A flowchart of a method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to an embodiment of the present application; Figure 2 A data flow diagram of a method for enhancing the accuracy of place name and address translation by using context awareness and a geographic library according to an embodiment of the present application; Figure 3 It is a flowchart of sub-step S5 of the method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to an embodiment of the present application; Figure 4 This is a flowchart of sub-step S51 of the method for enhancing the accuracy of place name and address translation using context awareness and geographic library according to an embodiment of the present application. DETAILED DESCRIPTION

[0010] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0011] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0012] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0013] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0014] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0015] It should be noted in advance that the acquisition and processing of all information or data in this application are carried out in compliance with the relevant national data protection laws and policies and with the authorization of the authority administrator.

[0016] In the technical solution of the present application, a method is proposed for enhancing the translation accuracy of place names and addresses by using context awareness and geographic database. Figure 1 The present invention is a flowchart of a method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to an embodiment of the present application. Figure 2 The data flow diagram of the method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to an embodiment of the present application is shown in FIG. Figure 1 and Figure 2As shown, according to an embodiment of the present application, a method for enhancing the accuracy of place name and address translation by using context perception and geographic library includes the following steps: S1, receiving a source language address text to be translated; S2, performing standardized preprocessing and structured semantic parsing on the source language address text to obtain a candidate list of place name entities and non-place name text fragments; S3, extracting a first place name entity to be disambiguated from the place name entity candidate list; S4, extracting a plurality of candidate place name entity entries matching the first place name entity to be disambiguated from the geographic knowledge base; S5, inputting a plurality of candidate place name entity entries matching the first place name entity to be disambiguated and non-place name text fragments into an ambiguity resolution model to obtain a disambiguated place name entity of the first place name entity to be disambiguated; S6, looping through S3 to S5 to obtain a disambiguated place name entity list; S7, obtaining a target language address text based on the disambiguated place name entity list.

[0017] In particular, the S1 receives a source language address text to be translated. In practical applications, the source language address text may come from cross-border logistics documents, multilingual map service interfaces, or fragmented information manually input by users, and its expression often contains spelling variations, abbreviations, non-standard separators, and even mixed punctuation marks.

[0018] In particular, the S2 performs standardized preprocessing and structured semantic parsing on the source language address text to obtain a candidate list of place name entities and non-place name text fragments. In the technical solution of the present application, first, the source language address text is text cleaned to obtain a purified source language address text. It should be understood that the original source language address text has multimodal noise interference, such as unnecessary punctuation, spelling errors or irrelevant characters, etc., and these factors will interfere with the subsequent processing steps. Therefore, in the technical solution of the present application, the source language address text is text cleaned to obtain a purified source language address text; specifically, a structured preprocessing channel is established to eliminate interference information outside the address elements through a patterned noise filtering mechanism to ensure that the input of the subsequent named entity recognition module meets the geographic entity parsing specification. By removing irrelevant characters and correcting possible errors, named entity recognition can be made more accurate, which helps to more accurately locate and parse the geographic entity names and their contextual information in the text to be translated. In an example, the source language address text can be text cleaned by the following steps: a series of predefined rules are used to remove or correct unnecessary elements in the source language address text. For example, all non-alphanumeric characters are filtered out (except for specific reserved characters, such as spaces), common spelling errors are corrected, and the format is unified (such as dates, phone numbers, etc.). In addition, natural language processing technology is used to identify and correct potential language errors or non-standard expressions. In particular, by combining machine learning models, it is possible to automatically learn how to effectively clean the input text based on a large number of annotated data sets to ensure that the output is a purified source language address text with a reasonable structure and clear content. In this way, the system can more efficiently extract useful information from the source language address text and provide high-quality input data for the next steps.

[0019] Next, the purified source language address text is segmented to obtain a set of source language address text words. It should be understood that although the original address text has been cleaned and denoised, it still needs to be further decomposed into smaller information units so that subsequent steps such as named entity recognition and place name disambiguation can be performed more accurately. Therefore, in the technical solution of the present application, the purified source language address text is segmented to obtain a set of source language address text words. The generated set of source language address text words can better support the system to identify which parts belong to place name entities (such as city and street names) and which are non-place name information (such as house numbers, building names, etc.). This is crucial for correctly parsing the address structure and determining the relationship between the various components in the subsequent steps.

[0020] Then, each source language address text word in the set of source language address text words is subjected to named entity recognition and classification to obtain a place name entity candidate list and a non-place name text fragment. It should be understood that the address text usually contains rich geographical information and other non-place name information (such as house number, unit name, etc.), and this information is essential for accurately translating the entire address. Therefore, in the technical solution of the present application, each source language address text word in the set of source language address text words is subjected to named entity recognition and classification to obtain a place name entity candidate list and a non-place name text fragment. That is, by performing named entity recognition and classification on each source language address text word in the set of source language address text words, the geographical entity and the non-geographic entity in the address text are accurately distinguished, and the geographical entity is further classified and identified to facilitate subsequent disambiguation and translation work. Specifically, through effective named entity recognition, the system can accurately extract the place name entities involved in the address, and provide basic data support for the subsequent use of the geographic knowledge base for fuzzy query and ambiguity elimination. At the same time, the separated non-place name text fragment can also be used as part of the context information to help better understand and resolve the ambiguity of the place name entity. In one example, the address text after cleaning and segmentation can be deeply analyzed by using advanced natural language processing technology. For example, using a pre-trained language model, the system can identify which words represent geographical entities such as countries, cities, streets, and which words belong to non-place name entities (such as house numbers composed of numbers). In this process, the model not only relies on the information of the words themselves, but also considers the context around them to improve recognition accuracy. In addition, this step may also be combined with an external knowledge base (such as a geographic database) to enhance the credibility of the recognition results and ensure that the recognized place name entities are consistent with the actual situation. In this way, the system can efficiently extract all relevant place name entities from the address text to form a candidate list, and treat the remaining parts as non-place name text fragments, thereby improving the overall accuracy and reliability of address translation.

[0021] In particular, the S3 and S4 extract the first place name entity to be disambiguated from the place name entity candidate list, and use the first place name entity to be disambiguated as a keyword to perform a fuzzy query in the geographic knowledge base to obtain multiple candidate place name entity entries that match the first place name entity to be disambiguated. That is, the ability to understand and resolve place names is enhanced by utilizing the rich geographic information provided by the geographic knowledge base, wherein the geographic knowledge base contains detailed geographic information, including but not limited to the spatial hierarchical relationship, historical changes and cultural background of place names, etc., which are essential for correctly understanding and resolving place names. In the technical solution of the present application, by matching and querying the place name entity to be disambiguated with the data in the geographic knowledge base, more background information about the place name can be obtained, such as the administrative division to which it belongs, the surrounding geographical features, etc. This not only helps to eliminate ambiguity, but also provides more comprehensive and accurate translation results. In addition, fuzzy query allows the system to find possible matches even when faced with spelling errors or similar names, further improving the robustness and practicality of the system. In one example, the following steps can be used to perform a fuzzy query in the geographic knowledge base using the first place name entity to be disambiguated as a keyword: First, a place name entity to be disambiguated is selected from the place name entity candidate list obtained through word segmentation and named entity recognition. The selection criteria can be determined based on factors such as the importance of the place name in the text or its frequency of occurrence; Then, the place name entity to be disambiguated is used as a keyword to perform a fuzzy query operation in the pre-built geographic knowledge base. The query results are usually a series of possible matching place name entity entries, each of which carries detailed geographic attribute information. This information will then be used in the subsequent ambiguity resolution process. By combining contextual information and other non-place name text fragments, the system can more accurately determine the place name entity that best suits the current context, thereby improving the accuracy and reliability of the entire address translation process.

[0022] In particular, in S5, multiple candidate place name entity entries and non-place name text segments matching the first place name entity to be disambiguated are input into the ambiguity resolution model to obtain the disambiguated place name entity of the first place name entity to be disambiguated. In particular, in a specific example of the present application, Figure 3 As shown, the S5 includes: S51, using non-place name text fragments as context information, performing semantic-level disambiguation aggregation analysis on multiple candidate place name entity entries to obtain semantic aggregation coding features of disambiguated place name entity query responses; S52, semantically decoding the semantic aggregation coding features of the disambiguated place name entity query responses to obtain the disambiguated place name entity of the first place name entity to be disambiguated.

[0023] Specifically, the S51 uses the non-place name text segment as context information to perform semantic level disambiguation aggregation analysis on multiple candidate place name entity entries to obtain semantic aggregation encoding features of disambiguation place name entity query response. In particular, in a specific example of the present application, Figure 4 As shown, the S51 includes: S511, using a non-place name text fragment as context information, and performing semantic understanding on the context information to obtain a semantic coding vector of the context information of the place name entity to be disambiguated; S512, using a candidate place name entity entry embedding matrix to perform structured coding on each of multiple candidate place name entity entries to obtain semantic embedding coding vectors of multiple candidate place name entity entries; S513, performing place name entity semantic disambiguation aggregation coding on the semantic coding vector of the context information of the place name entity to be disambiguated and the semantic embedding coding vectors of multiple candidate place name entity entries to obtain a disambiguated place name entity query response semantic aggregation coding vector as a disambiguated place name entity query response semantic aggregation coding feature.

[0024] More specifically, the S511 uses the non-place name text fragment as context information, and performs semantic understanding on the context information to obtain the semantic coding vector of the place name entity context information to be disambiguated. That is, in an embodiment of the present application, first, the place name entity context information to be disambiguated is semantically embedded and coded using the place name entity embedding matrix to be disambiguated, so as to obtain the context sequence of the semantic embedding coding vector of the place name entity context information to be disambiguated. Because the traditional method often ignores the rich background information provided by the non-place name text fragment, it is difficult to make the right choice when facing the place names with the same name but different meanings. By converting the non-place name text fragment into a semantic embedding vector, the system can better capture the association between the non-place name text fragment and the place name entity to be disambiguated, thereby improving the accuracy of the disambiguation process. Doing so can not only help identify the place name that best fits the current context, but also enhance the robustness and adaptability of the entire translation process. In one example, the context information can be semantically embedded and encoded through the following steps: first, an embedding matrix of non-place name text fragments is constructed; then, for each address text containing a place name entity to be disambiguated, the non-place name part is extracted and converted into a corresponding semantic embedding encoding vector of the context information of the place name entity to be disambiguated through the above embedding matrix; in this process, the language model will consider the relationship within and between text fragments to ensure that the final semantic embedding encoding vector of the context information of the place name entity to be disambiguated can accurately reflect the meaning of the non-place name text fragment and its relevance to the place name entity to be disambiguated.

[0025] Next, the context sequence of the semantic embedding coding vector of the context information of the place name entity to be disambiguated is subjected to context semantic encoding based on the bidirectional LSTM model to obtain the semantic encoding vector of the context information of the place name entity to be disambiguated. Considering that it is difficult to capture the complex semantic relationship and dynamic changes of context in non-place name text fragments by relying solely on static word vectors or simple embedding methods. Especially when dealing with polysemous place names, relying solely on the information of the place name itself cannot effectively distinguish place name entities with different meanings, while the context information provides additional clues to help identify the correct place name. In addition, in a complex address structure, the key place name may be far away from the vocabulary that provides important context information. Since the bidirectional LSTM model combines the forward and backward information flow, it can overcome the gradient vanishing problem encountered by traditional RNN or LSTM when dealing with long-term dependencies to a certain extent, thereby more effectively utilizing the useful information in the entire sentence to assist the disambiguation of place name entities. Therefore, in the technical solution of the present application, the context sequence of the semantic embedding coding vector of the context information of the place name entity to be disambiguated is subjected to context semantic encoding based on the bidirectional LSTM model to obtain the semantic encoding vector of the context information of the place name entity to be disambiguated. Among them, the bidirectional LSTM model can process the context sequence of the semantic embedding encoding vector of the context information of the disambiguated place name entity from both the forward and backward directions, and even those words that are far away from the place name entity to be disambiguated but still provide important clues can be taken into account, which greatly enhances the model's ability to understand complex contexts. The bidirectional LSTM model captures the contextual semantic connection in non-place name text fragments, so as to provide a rich semantic background for each place name entity to be disambiguated more accurately.

[0026] More specifically, the S512 uses the candidate place name entity entry embedding matrix to perform structured encoding on each of the multiple candidate place name entity entries to obtain multiple candidate place name entity entry semantic embedding encoding vectors. That is, by converting each candidate place name entity into a vector representation in a high-dimensional space, the system can better understand and compare the differences between these candidate entities, thereby improving the accuracy of the disambiguation process. This structured encoding not only takes into account the characteristics of the place name itself, but also includes other geographical attributes related to it (such as the administrative division to which it belongs, nearby landmarks, etc.), which helps to evaluate the fitness of the candidate place name entity from multiple dimensions and ultimately select the translation result that best fits the current context. In one example, each candidate place name entity entry in a plurality of candidate place name entity entries can be structuredly encoded by the following steps: first, construct an embedding matrix of candidate place name entity entries, which can map the input place name entity into a high-dimensional vector space, where each dimension represents the characteristics of the place name in a specific aspect. For each candidate place name entity entry from the geographic knowledge base, the system will calculate a corresponding semantic embedding encoding vector of the candidate place name entity entry based on the specific information it contains (such as geographic location coordinates, administrative region, historical background, etc.), and the semantic embedding encoding vector of the candidate place name entity entry can accurately reflect the core characteristics of the original place name entity and its relationship with other entities. In this way, the system can effectively convert complex geographic information into an easy-to-process numerical form, which is convenient for semantic analysis and disambiguation aggregation encoding in subsequent steps.

[0027] More specifically, the S513 performs place name entity semantic disambiguation aggregation coding on the semantic coding vector of the context information of the place name entity to be disambiguated and the semantic embedding coding vectors of multiple candidate place name entity entries, so as to obtain the semantic aggregation coding vector of the disambiguated place name entity query response as the semantic aggregation coding feature of the disambiguated place name entity query response. It should be understood that when the prior art processes geographic entities with the same name, due to insufficient context association and lack of multi-dimensional attribute integration of the geographic knowledge base, it is impossible to effectively distinguish the deep semantic differences of the fields with the same name. For example, when there is a lack of explicit geographic level identification in the source text, the traditional method relies on surface feature matching and is prone to cross-domain mapping errors. Therefore, in order to achieve the deep coupling of text context features and geographic knowledge features, in the technical solution of the present application, the semantic coding vector of the context information of the place name entity to be disambiguated and the semantic embedding coding vectors of multiple candidate place name entity entries are subjected to place name entity semantic disambiguation aggregation coding, so as to obtain the semantic aggregation coding vector of the disambiguated place name entity query response. That is, through data-driven feature enhancement and dynamic weight allocation mechanism, the dimensional deviation of the heterogeneous feature space is eliminated, and the semantic weight of the candidate entity is optimized based on global context perception. Specifically, firstly, deconvolution coding is used to enhance the features of the semantic encoding vector of the context to be disambiguated, so as to improve its representation ability through adaptive learning and ensure that it is aligned with the semantic embedding encoding vector of the candidate item in terms of feature scale, so as to eliminate the dimensionality bias in the cross-domain comparison process; then, the enhanced context encoding vector and the encoding vectors of each candidate item are input into the monomer semantic query unit, and a set of monomer semantic query score encoding vectors of the entity to be disambiguated-candidate place name containing multi-dimensional semantic association information is generated through a non-linear transformation layer, so as to capture the expression of semantic relevance at different levels of abstraction, such as implicit features such as administrative division affiliation and spatial proximity; further, the competition and synergy relationship between candidate items is modeled through the relationship gating agent module, which uses a gating mechanism to dynamically suppress noise candidates and strengthen highly correlated items; finally, the set of monomer semantic query score encoding vectors of the entity to be disambiguated-candidate place name is weightedly aggregated based on weights. In this process, dynamic information screening is achieved through the self-attention mechanism to ensure that the aggregation results focus on the geographical entity attributes that are strongly associated with the context. The generated semantic aggregation encoding vector of the disambiguation place name entity query response can simultaneously capture the implicit context constraints in the non-place name text fragments (such as the hierarchical implications of the administrative division modifiers "province" and "city") and the interaction between the candidate entry's geographical attributes (such as spatial topological relationships and administrative affiliation). In this way, the semantic sensitivity and geographical knowledge integration capabilities of the disambiguation model are improved.

[0028] Specifically, firstly, the semantic encoding vector of the context information of the place name entity to be disambiguated is enhanced based on deconvolution coding to obtain the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated, wherein the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated has the same feature scale as the semantic embedding encoding vector of each candidate place name entity entry in the semantic embedding encoding vector of multiple candidate place name entity entries. It should be understood that the semantic encoding vector of the context information of the place name entity to be disambiguated generated by the bidirectional LSTM may focus on local language patterns (such as the position characteristics of the address entity in the text sequence), while the semantic vector generated by the embedding matrix of the candidate entity entry is based on the encoding of the structured attributes of the geographic knowledge base (such as administrative division level, spatial topological relationship, historical evolution information). The difference in feature distribution and dimension between the two directly restricts the accuracy of semantic alignment. The traditional method leads to a semantic gap in cross-modal matching due to the heterogeneity of feature space. Through the deconvolution coding operation, the system uses a learnable deconvolution kernel network to perform nonlinear feature expansion on the context encoding, adaptively improve its representation ability, and at the same time force the output feature dimension to align with the scale of the candidate entry embedding encoding to ensure that the two are comparable in a unified semantic space. In this process, not only the limitations of manually designed interpolation rules are avoided, but also the multi-granularity associations of contextual semantics are captured through multi-channel convolution operations, such as implicit administrative hierarchy relationships or functional scenario features. The semantically enhanced coding vector of the context information of the place name entity to be disambiguated after feature enhancement can establish deep associations with the candidate entry encoding in terms of attributes such as spatial proximity and membership while retaining the original semantic topological structure. In a specific example, the semantic coding vector of the context information of the place name entity to be disambiguated is subjected to feature enhancement based on deconvolution coding using the following feature enhancement formula to obtain the semantically enhanced coding vector of the context information of the place name entity to be disambiguated; wherein the feature enhancement formula is: ; in, is the semantic encoding vector of the context information of the place name entity to be disambiguated, represents deconvolution coding, represents the deconvolution weight matrix, represents the one-norm of a vector, A semantically enhanced encoding vector for the context information of the place-name entity to be disambiguated.

[0029] Next, the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of each candidate place name entity item in the multiple candidate place name entity item semantic embedding encoding vectors are respectively input into the place name entity monomer semantic query unit to obtain the set of initial to-be-disambiguated candidate place name entity monomer semantic query score encoding vectors. Considering that the traditional method relies on surface feature similarity measurement (such as cosine similarity) and cannot effectively capture the deep semantic association between the context and the candidate items. For example, the context of "Paris" in the source text may imply educational attributes (such as "University"), while the candidate items with the same name in the geographic knowledge base (such as Paris in France and Paris in Texas) have essential differences in structural attributes such as administrative levels and functional scenarios. Through the monomer semantic query unit, the system constructs an implicit semantic representor based on the deep neural network architecture, measures the correlation between the context semantic enhancement encoding and the candidate item embedding encoding from multiple dimensions, and breaks through the semantic gap of traditional methods. Specifically, the single semantic query unit uses a nonlinear transformation layer and a cross-attention mechanism to perform cross-modal interaction between the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated (representing the functional scenario such as "University") and the semantic embedding encoding vector of the candidate place name entity item (including attributes such as administrative affiliation and historical evolution), and generates a multi-dimensional initial single semantic query score encoding vector of the candidate place name entity to be disambiguated. Among them, different dimensions represent the spatial proximity matching degree, functional scenario consistency or historical name association strength, rather than a single scalar score. By capturing deep semantic associations, it provides a highly discriminative input for the relationship gated proxy module, and ultimately achieves the topological consistency and geographic logical integrity of cross-language address translation. In a specific example, the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of each candidate place name entity item in the multiple candidate place name entity item semantic embedding encoding vectors are respectively input into the place name entity single semantic query unit according to the following semantic query formula to obtain a set of initial single semantic query score encoding vectors of the candidate place name entity to be disambiguated; wherein, the semantic query formula is: ; ; in, Semantic embedding encoding vectors for the multiple candidate place name entity entries, are the first, second, and third semantic embedding encoding vectors of multiple candidate place name entity entries. and The semantic embedding encoding vector of candidate place name entity entries, express function, represents the semantic weight matrix, represents the bias term, for The corresponding initial to-be-disambiguated-candidate place name entity monomer semantic query score encoding vector.

[0030] In particular, it should be understood that when the set of initial semantic query score encoding vectors of candidate place name entities to be disambiguated is obtained by feature concatenation and simple interactive calculation, its distribution in the semantic query encoding space may not accurately reflect the complex relationship between the context and the geographic knowledge base due to insufficient dynamic characteristics of the features or spatial alignment deviation. For example, when entities with the same name but different geographical locations appear in the address text (such as "Paris" may correspond to Paris, France or Paris, Texas, USA), traditional static encoding methods are difficult to capture the dynamic impact of non-place name contexts (such as adjacent words "Texas" or "France") on entity disambiguation, resulting in deviations in the matching weight distribution of candidate entities. Therefore, preferably, the set of initial semantic query score encoding vectors of candidate place name entities to be disambiguated is subjected to feature dynamic optimization mapping to obtain a set of semantic query score encoding vectors of candidate place name entities to be disambiguated. That is, a differential manifold alignment mechanism for cross-modal features is established through dynamic optimization mapping of features to transform the set of the initial semantic query score encoding vectors of the candidate place name entity monomers to be disambiguated from the static feature space to the dynamically optimized semantic query space, thereby improving the semantic alignment ability of the feature distribution, so that it can more flexibly reflect the association between the context and the geographic knowledge base. Specifically, the semantic dynamic eigenvector representation of the candidate place name entity monomers to be disambiguated is constructed by cascading feature ontologies, and the semantic interaction potential vector of the candidate place name entity monomers to be disambiguated is calculated, and the dynamic characteristics of the mapping process are introduced to enable the feature representation to adapt to context changes while retaining the original semantic information through the combination of linear transformation and residual connection. Furthermore, the inherent alignment mechanism adjusts the feature projection path under canonical symmetry conditions through coupling constants and covariance matrices to ensure that the alignment of the feature space and the semantic query space not only meets the dynamic optimization requirements, but also avoids instability caused by noise amplification. In addition, the introduction of the covariance matrix enables the feature projection process to follow the spatial hierarchical relationship (such as the "province-city-district" structure) and historical naming features (such as the old name "New Towne") in the geographic knowledge base, and realize cross-validation of multi-dimensional geographic attributes. In this way, the optimized set of initial semantic query score encoding vectors for candidate place name entities to be disambiguated not only reduces the translation deviation caused by language polysemy, but also enables the model to maintain high robustness through the synergy of dynamic characteristics and inherent alignment, ensuring strict consistency between the target language address text and the real geographic location.

[0031] Specifically, in the place name entity monomer semantic query unit, the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated is obtained. and the corresponding candidate place name entity entry semantic embedding encoding vector When encoding the score of a single semantic query by direct feature concatenation, it is expected that the representation accuracy of the single semantic query score can be improved by optimizing the dynamic feature representation from the feature space to the semantic query encoding space and improving the inherent alignment characteristics between the feature space and the semantic query encoding space.

[0032] Based on this, if the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated is and the corresponding candidate place name entity entry semantic embedding encoding vector The cascade feature is recorded as ,Right now First, the semantic interaction potential vector of the entity monomer to be disambiguated-candidate place name under the mapping representation is calculated: ; That is, the cascade feature ontology is used as the semantic dynamic eigenvector representation of the entity monomer to be disambiguated-candidate place name to obtain the projection representation to the eigenstate under the spatial mapping process.

[0033] Then, the intrinsic alignment between the feature space and the semantic query encoding space is performed: ; in The coupling constant is calculated in the same way as in deconvolution enhancement to preserve symmetry, i.e.: ; and is the covariance matrix, that is, Come get.

[0034] Therefore, in the semantic interaction potential vector of the entity monomer between the to-be-disambiguated and candidate place names As the inherent alignment generator, the covariant path under the canonical symmetry condition is used as the mapping canonical fixation guarantee, which improves the representation accuracy of the monomer semantic query score.

[0035] Furthermore, based on the self-distribution characteristics of the feature set of the set of monomer semantic query score encoding vectors of the candidate place name entities to be disambiguated, the monomer semantic matching degree of each monomer semantic query score encoding vector of the candidate place name entities to be disambiguated in the set of monomer semantic query score encoding vectors of the candidate place name entities to be disambiguated is determined to obtain a set of monomer semantic matching degrees of the candidate place name entities to be disambiguated. It should be understood that when traditional methods evaluate the matching degree of candidate entities in isolation, the dynamic influence of the global context on individual semantic associations is ignored. For example, when the distribution of monomer semantic query score encoding vectors of the candidate place name entities to be disambiguated of multiple candidate entities shows significant differences, relying solely on absolute scores is susceptible to noise interference and cannot distinguish the relative importance of highly correlated entries in the whole. In the technical solution of the present application, adaptive optimization of matching degree is achieved through a context-aware mechanism to avoid semantic deviations caused by local evaluations. Specifically, the system uses probability density estimation or distribution fitting methods to analyze the distribution form (such as long-tail distribution, normal distribution) of the set of semantic query score encoding vectors of candidate place name entities to be disambiguated, and generates matching weights based on the degree of deviation of each semantic query score encoding vector of candidate place name entities to be disambiguated from the distribution center (such as Z-score standardization). For example, if the score vector of a candidate entity is significantly higher than the distribution mean in the multidimensional semantic space, its matching degree will be strengthened; conversely, vectors close to or below the mean will be suppressed. In this way, the misjudgment problem caused by static thresholds or isolated matching in traditional methods is solved, and the contextual consistency of disambiguation decisions is improved through dynamic weight allocation, supporting the reliable application of high-precision address translation in multilingual service scenarios. In a specific example, the following semantic matching formula is used to determine the monomer semantic matching degree of each monomer semantic query score encoding vector of the candidate place name entity to be disambiguated in the set of monomer semantic query score encoding vectors of the candidate place name entity to be disambiguated, so as to obtain a set of monomer semantic matching degrees of the candidate place name entity to be disambiguated; wherein, the semantic matching formula is: ; in, express function, is the number of semantic embedding coding vectors of candidate place name entity entries in the multiple semantic embedding coding vectors of candidate place name entity entries, represents exponential operation, represents vector multiplication, represents the square of the one-norm of a vector, is the transpose of the semantic query score encoding vector of the k-th candidate place name entity to be disambiguated, for The corresponding semantic matching degree between the entity to be disambiguated and the candidate place name.

[0036] Then, the set of semantic matching degrees of the candidate place name entities to be disambiguated is input into the relational gating proxy module to obtain the set of semantic self-attention weights of the candidate place name entities to be disambiguated. It should be understood that when the system preliminarily screens out multiple candidate place name entities (e.g., multiple "Paris" with the same name but different geographical locations) through semantic matching, directly relying on the original matching degree to allocate attention weights may face two core problems: first, the matching degree value only reflects the surface semantic association between the text fragment and the candidate entity, and cannot capture the implicit geographical hierarchical relationship in the address text (such as the state relationship in "Paris, Texas") or historical naming features (such as the geographical range differences of "New York" in different historical stages); second, the traditional self-attention mechanism has limited integration capabilities for multi-source heterogeneous information (such as street descriptions and administrative division modifiers in address texts), and is prone to biased weight allocation due to local noise interference. In the technical solution of the present application, the original matching degree is nonlinearly transformed through a gating function, and while retaining the basic semantic association strength, the predefined spatial relationship constraints (such as the affiliation between the administrative region to which the candidate entity belongs and the superior administrative division mentioned in the context) and semantic association rules (such as the coexistence probability of natural geographical entities and cultural landmarks) in the geographic knowledge base are injected. In this way, the robustness of the model to noise interference is improved, the semantic consistency of the disambiguation decision is enhanced, and the semantic deviation problem caused by static weight allocation in traditional methods is solved. The refined control of context perception is achieved through the gating mechanism, and a highly discriminative weight input is provided for subsequent self-attention aggregation, supporting the precise mapping of geographic topological integrity and administrative hierarchy logic in cross-language address translation. In a specific example, the set of semantic matching degrees of the candidate place name entity monomers to be disambiguated is input into the relationship gating proxy module using the following gating formula to obtain the set of semantic self-attention weights of the candidate place name entity monomers to be disambiguated; wherein the gating formula is: ; in, represents the gating function, is the preset threshold, express The corresponding semantic self-attention weight of the entity to be disambiguated-candidate place name.

[0037] Furthermore, based on the set of semantic self-attention weights of the monomer query semantics of the candidate place name entity to be disambiguated, the set of monomer semantic query score encoding vectors of the candidate place name entity to be disambiguated is aggregated to obtain the semantic aggregate encoding vector of the disambiguated place name entity query response. That is, through the dynamic weight allocation mechanism, i.e., the self-attention mechanism, the importance of each candidate place name entity is flexibly adjusted to achieve attention to key semantic information and ignore irrelevant or noisy information. This weighted aggregation method not only improves the model's ability to understand complex contexts, but also makes the final translation results more accurate and reliable. In a specific example, the set of monomer semantic query score encoding vectors of the candidate place name entity to be disambiguated is aggregated using the following aggregation formula to obtain the semantic aggregate encoding vector of the disambiguated place name entity query response; wherein, the aggregation formula is: ; in, A semantically aggregated encoding vector is generated for the disambiguated place name entity query response.

[0038] Specifically, the S52 semantically decodes the semantic aggregation coding features of the disambiguated place name entity query response to obtain the disambiguated place name entity of the first place name entity to be disambiguated. That is, in the technical solution of the present application, the semantic aggregation coding vector of the disambiguated place name entity query response is semantically decoded based on the RNN model to obtain the disambiguated place name entity of the first place name entity to be disambiguated. It should be understood that although the semantic aggregation coding vector of the disambiguated place name entity query response integrates the contextual information and the structured features of the geographic knowledge base, it is essentially still a complex sequence containing multi-dimensional semantic associations. If the traditional method directly selects candidate entities through simple classification or rule matching, it is easy to cause misjudgment due to ignoring the dynamic dependencies implicit in the sequence. The RNN model, through its recurrent network structure, can parse the semantic fragments in the aggregation code step by time, capture the strongly associated parts of the semantic aggregation coding vector of the disambiguated place name entity query response, suppress the redundant information in the vector, and thus dynamically construct a disambiguation decision path.

[0039] In particular, the S6, loops through S3 to S5 to obtain a list of disambiguated place name entities. It should be understood that the source language address text may contain multiple place name entities, each of which has its specific context and may have complex associations with other place name entities. Therefore, processing only a single place name entity cannot meet the needs of overall translation. In the technical solution of the present application, loops through S3 to S5 to obtain a list of disambiguated place name entities. A complete and unambiguous list of place name entities is constructed by gradually parsing and resolving the ambiguity of each place name entity to be translated. Among them, the list of disambiguated place name entities ensures that each place name has been thoroughly analyzed and verified, and by disambiguating all place name entities in the source language address text one by one, the exact meaning of each place name in a specific context can be effectively captured, thereby reducing the problem of geographic information distortion caused by misidentification or translation.

[0040] In particular, the S7 obtains the target language address text based on the disambiguated place name entity list. In the technical solution of the present application, the disambiguated place name entity list is input into the pre-trained neural machine translation model to obtain the target language address text. Among them, the neural machine translation model will translate each place name entity according to the learned language rules, vocabulary mapping relationships and knowledge in a specific field. It is worth noting that in the translation process, the model not only focuses on the conversion of a single place name entity, but also comprehensively considers the characteristics of the entire address structure to ensure the logical coherence and consistency between the various parts. For example, the neural machine translation model will pay attention to details such as whether the hierarchical relationship of administrative divisions is correct, whether the combination of street names and house numbers is reasonable, etc. Finally, after this series of complex calculations and adjustments, the system will output the target language address text. In this way, the comprehensive and accurate conversion from the source language address text to the target language address text is completed. Through this process, it can be ensured that users of different languages ​​can understand and use the converted address information without obstacles.

[0041] In summary, according to the embodiment of the present application, a method for enhancing the accuracy of place name and address translation by using context awareness and geographic library is explained, which constructs a dynamic and collaborative ambiguity resolution framework by integrating the context awareness mechanism and the multi-dimensional attributes of the geographic knowledge base, so as to break through the semantic limitations of traditional place name and address translation. Specifically, firstly, the context semantic encoding of non-place name text fragments is used to capture the dynamic association information implicit in the address text, and the candidate entities are embedded with structured features in combination with the geographic knowledge base to extract the multi-dimensional attributes of the candidate entities, thereby realizing cross-modal disambiguation decisions in the semantic aggregation stage. In this way, the geographic feature cross-validation mechanism guided by context semantics can effectively solve the translation ambiguity problem caused by static rule dependence and geographic attribute fragmentation in the prior art, thereby achieving a more accurate multi-language address conversion effect.

[0042] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for enhancing the accuracy of place name and address translation using context awareness and geographic database, characterized in that: include: S1: receiving a source language address text to be translated; S2: Perform standardized preprocessing and structured semantic parsing on the source language address text to obtain a candidate list of place name entities and non-place name text fragments; S3: extracting a first place-name entity to be disambiguated from the place-name entity candidate list; S4: extracting multiple candidate place name entity entries matching the first place name entity to be disambiguated from the geographic knowledge base; S5: Inputting multiple candidate place name entity entries and non-place name text fragments matching the first place name entity to be disambiguated into an ambiguity resolution model to obtain a disambiguated place name entity of the first place name entity to be disambiguated, including: using the non-place name text fragment as context information, performing semantic level disambiguation aggregation analysis on multiple candidate place name entity entries to obtain semantic aggregation encoding features of disambiguated place name entity query responses; performing semantic decoding on the semantic aggregation encoding features of the disambiguated place name entity query responses to obtain the disambiguated place name entity of the first place name entity to be disambiguated; Step 6: Loop through S3 to S5 to obtain a list of disambiguated place name entities; Step 7: Based on the disambiguated place name entity list, obtain the target language address text.

2. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 1, characterized in that S2 include: Perform text cleaning on the source language address text to obtain the purified source language address text: Performing word segmentation processing on the purified source language address text to obtain a set of source language address text words; Named entity recognition and classification are performed on each source language address text word in the set of source language address text words to obtain a place name entity candidate list and a non-place name text segment.

3. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 2, characterized in that: S4 includes: using the first place-name entity to be disambiguated as a keyword, performing a fuzzy query in a geographic knowledge base to obtain a plurality of candidate place-name entity entries that match the first place-name entity to be disambiguated.

4. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 3 is characterized in that: Using non-place name text fragments as context information, semantic-level disambiguation aggregation analysis is performed on multiple candidate place name entity entries to obtain semantic aggregation encoding features of disambiguation place name entity query responses, including: Using non-place name text fragments as context information, semantic understanding of the context information is performed to obtain a semantic encoding vector of the context information of the place name entity to be disambiguated; Using the candidate place name entity entry embedding matrix, structurally encoding each of the multiple candidate place name entity entries is performed to obtain semantic embedding encoding vectors of the multiple candidate place name entity entries; The semantic encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of multiple candidate place name entity entries are subjected to place name entity semantic disambiguation aggregation encoding to obtain a disambiguated place name entity query response semantic aggregation encoding vector as a disambiguated place name entity query response semantic aggregation encoding feature.

5. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 4, characterized in that: Using non-place name text fragments as context information, the context information is semantically understood to obtain the semantic encoding vector of the context information of the place name entity to be disambiguated, including: Using the embedding matrix of the place name entity to be disambiguated, semantic embedding encoding is performed on the context information of the place name entity to be disambiguated, so as to obtain a context sequence of the semantic embedding encoding vector of the context information of the place name entity to be disambiguated; The context sequence of the semantic embedding encoding vector of the context information of the place name entity to be disambiguated is subjected to context semantic encoding based on the bidirectional LSTM model to obtain the semantic encoding vector of the context information of the place name entity to be disambiguated.

6. The method for enhancing the accuracy of place name and address translation using context awareness and geographic library according to claim 5, characterized in that: The semantic encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of multiple candidate place name entity entries are subjected to place name entity semantic disambiguation aggregation encoding to obtain a semantic aggregation encoding vector of the disambiguated place name entity query response, including: Performing feature enhancement based on deconvolution coding on the semantic coding vector of the context information of the place name entity to be disambiguated, so as to obtain the semantic enhanced coding vector of the context information of the place name entity to be disambiguated; Performing a single semantic query on the place name entity by using the semantic enhancement encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of each candidate place name entity item in the semantic embedding encoding vectors of multiple candidate place name entity items, so as to obtain a set of single semantic query score encoding vectors of the candidate place name entity to be disambiguated; Based on the self-distribution characteristics of the feature set of the set of semantic query score encoding vectors of the to-be-disambiguated-candidate place name entities, an adaptive aggregation analysis based on gated modulation is performed on the set of semantic query score encoding vectors of the to-be-disambiguated-candidate place name entities to obtain the semantic aggregation encoding vector of the disambiguated place name entity query response.

7. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 6, characterized in that: The semantic enhancement encoding vector of the context information of the place name entity to be disambiguated and the semantic embedding encoding vectors of each candidate place name entity item in the semantic embedding encoding vectors of multiple candidate place name entity items are used for place name entity monomer semantic query to obtain a set of to-be-disambiguated-candidate place name entity monomer semantic query score encoding vectors, including: The semantic enhancement coding vector of the context information of the place name entity to be disambiguated and the semantic embedding coding vectors of each candidate place name entity item in the multiple candidate place name entity item semantic embedding coding vectors are respectively input into the place name entity monomer semantic query unit to obtain a set of initial to-be-disambiguated-candidate place name entity monomer semantic query score coding vectors; The set of initial to-be-disambiguated-candidate place name entity monomer semantic query score encoding vectors is subjected to dynamic feature optimization mapping to obtain a set of to-be-disambiguated-candidate place name entity monomer semantic query score encoding vectors.

8. The method for enhancing the accuracy of place name and address translation by using context awareness and geographic library according to claim 7, characterized in that: Based on the self-distribution characteristics of the feature set of the set of semantic query score encoding vectors of the to-be-disambiguated-candidate place name entity monomers, an adaptive aggregation analysis based on gated modulation is performed on the set of semantic query score encoding vectors of the to-be-disambiguated-candidate place name entity monomers to obtain a semantic aggregation encoding vector of the disambiguated place name entity query response, including: Based on the self-distribution characteristics of the feature set of the set of monomer semantic query score encoding vectors of the to-be-disambiguated-candidate place name entities, determine the monomer semantic matching degree of each to-be-disambiguated-candidate place name entity monomer semantic query score encoding vector in the set of monomer semantic query score encoding vectors of the to-be-disambiguated-candidate place name entities, so as to obtain a set of monomer semantic matching degrees of the to-be-disambiguated-candidate place name entities; The set of semantic matching degrees of entity monomers of to-be-disambiguated and candidate place names is input into the relation gating agent module to obtain the set of semantic self-attention weights of entity monomers of to-be-disambiguated and candidate place names; Based on the set of semantic self-attention weights of the to-be-disambiguated-candidate place name entity monomer query, the set of to-be-disambiguated-candidate place name entity monomer semantic query score encoding vectors is aggregated to obtain the disambiguated place name entity query response semantic aggregate encoding vector.

9. The method for enhancing the accuracy of place name and address translation using context awareness and geographic library according to claim 8, characterized in that: Semantically decoding the semantic aggregation encoding features of the disambiguation place name entity query response to obtain the disambiguation place name entity of the first place name entity to be disambiguated, including: The semantically aggregated encoding vector of the disambiguated place name entity query response is semantically decoded based on the RNN model to obtain the disambiguated place name entity of the first place name entity to be disambiguated.

10. The method for enhancing the accuracy of place name and address translation using context awareness and geographic library according to claim 8, characterized in that: Based on the disambiguated place name entity list, the target language address text is obtained, including: The disambiguated place name entity list is fed into a pre-trained neural machine translation model to obtain the target language address text.

Images