A multi-scenario multi-language automatic recognition method, device, equipment and storage medium

Abstract

The present invention discloses a multi-scenario multi-language automatic recognition method, device, equipment and storage medium, which relates to the field of information processing technology. The method includes: establishing a language conversion module for the first and second users, and embedding a sentence length matching module. After obtaining the language type and real-time scenario type of both parties, these information are input into the sentence length matching module for step matching, and a sentence segmentation strategy is generated, including the sentence segmentation step of both parties. The language conversion module identifies and converts the user's language according to the corresponding sentence segmentation step. The technical problems of translation delay and semantic incoherence in existing multi-language cross-scenario communication are solved, and the technical effect of improving the real-time, accuracy and semantic coherence of multi-language communication is achieved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a multi-scenario multi-language automatic recognition method, device, equipment and storage medium. Background Art

[0002] With the advancement of globalization, the demand for multilingual communication has increased rapidly in scenarios such as cross-border business, online meetings, and international education, but existing translation and speech recognition technologies have difficulty in achieving efficient and coherent real-time translation in multi-scenario applications. Different scenarios (such as daily conversations, video calls, speeches, and meetings) have different requirements for the real-time, fluency, and semantic accuracy of translation. The differences in grammatical structure, speaking speed, and sentence complexity of each language make it difficult for traditional translation systems to provide adaptable segmentation and fluent translation, and often cause delays and semantic incoherence. Traditional translation technology lacks the ability to automatically recognize and adapt to scenarios, and is insufficient in terms of semantic integrity, coherence, and timeliness, making it difficult to meet the increasingly diverse needs of cross-language communication.

[0003] At the current stage, relevant technologies still have technical problems such as translation delay and semantic incoherence in multi-language cross-scenario communication. Summary of the invention

[0004] The present application provides a multi-scenario multi-language automatic recognition method, device, equipment and storage medium, adopts a language conversion module for establishing the first and second users, and embeds a sentence length matching module. After obtaining the language type and real-time scenario type of both parties, this information is input into the sentence length matching module for step matching, and a sentence segmentation strategy is generated, including the sentence segmentation step lengths of both parties. The language conversion module recognizes and converts the user's language according to the corresponding sentence segmentation step length, thereby achieving the technical effect of improving the real-time, accuracy and semantic coherence of multi-language communication.

[0005] The present application provides a multi-scenario multi-language automatic recognition method, comprising:

[0006] A language conversion module for a first user party and a second user party is established, wherein the language conversion module is embedded with a sentence length matching module; a first language type corresponding to the first user party and a second language type corresponding to the second user party are obtained; a real-time scene type between the first user party and the second user party is obtained; the first language type, the second language type and the real-time scene type are input into the sentence length matching module for step length matching to obtain a sentence segmentation strategy, wherein the sentence segmentation strategy includes a first sentence segmentation step length corresponding to the first user party and a second sentence segmentation step length corresponding to the second user party; the language conversion module performs language recognition conversion on the first user party and the second user party according to the first sentence segmentation step length and the second sentence segmentation step length, respectively.

[0007] The present application also provides a multi-scenario multi-language automatic recognition device, comprising:

[0008] A language conversion module establishment module, the language conversion module establishment module is used to establish a language conversion module for a first user party and a second user party, wherein the language conversion module is embedded with a sentence length matching module; a language type acquisition module, the language type acquisition module is used to acquire a first language type corresponding to the first user party and a second language type corresponding to the second user party; a real-time scene type acquisition module, the real-time scene type acquisition module is used to acquire a real-time scene type between the first user party and the second user party; a sentence segmentation strategy acquisition module, the sentence segmentation strategy acquisition module is used to input the first language type, the second language type and the real-time scene type into the sentence segmentation length matching module for step length matching, and acquire a sentence segmentation strategy, the sentence segmentation strategy includes a first sentence segmentation step length corresponding to the first user party and a second sentence segmentation step length corresponding to the second user party; a language recognition conversion module, the language recognition conversion module is used for the language conversion module to perform language recognition conversion on the first user party and the second user party according to the first sentence segmentation step length and the second sentence segmentation step length respectively.

[0009] The present application also provides an electronic device, including:

[0010] The memory is used to store executable instructions; the processor is used to implement a multi-scenario multi-language automatic recognition method when executing the executable instructions stored in the memory.

[0011] The present application also provides a computer-readable storage medium, comprising:

[0012] A computer program is stored thereon, and when the program is executed by a processor, a multi-scenario and multi-language automatic recognition method is implemented.

[0013] The multi-scenario multi-language automatic recognition method, device, equipment and storage medium proposed in this application first establish a language conversion module for the first and second users, and embed a sentence length matching module. After obtaining the language type and real-time scenario type of both parties, this information is input into the sentence length matching module for step matching, and a sentence segmentation strategy is generated, including the sentence segmentation step lengths of both parties. The language conversion module recognizes and converts the user's language according to the corresponding sentence segmentation step length, achieving the technical effect of improving the real-time, accuracy and semantic coherence of multi-language communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A flowchart of a multi-scenario multi-language automatic recognition method provided in an embodiment of the present application;

[0016] Figure 2 A schematic diagram of the structure of a multi-scenario multi-language automatic recognition device provided in an embodiment of the present application;

[0017] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0018] Description of reference numerals: language conversion module establishment module 10 , language type acquisition module 20 , real-time scene type acquisition module 30 , sentence segmentation strategy acquisition module 40 , language recognition conversion module 50 . DETAILED DESCRIPTION

[0019] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0020] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0021] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0022] The present application embodiment provides a multi-scenario multi-language automatic recognition method, such as Figure 1 As shown, the method includes:

[0023] Step S100, establish a language conversion module for the first user and the second user, wherein the language conversion module is embedded with a sentence length matching module. Specifically, it is necessary to first clearly establish the purpose of the language conversion module for the first user and the second user to achieve cross-language barrier-free communication. Taking into account the characteristics of different languages ​​and scenarios, it is necessary to plan its architecture and determine the functions of receiving voice, processing and output and the placeholder area of ​​each submodule. Then build the main framework and select appropriate tools to create the basic structure and input and output interface. Then design the sentence length matching module separately, determine the input and output requirements and build the internal data structure and algorithm mechanism, and then embed it into the language conversion module to ensure effective data interaction. After completing the construction and embedding, prepare to collect different language voice samples and scene simulation data for preliminary testing, determine the test indicators such as language recognition accuracy, translation accuracy, sentence segmentation rationality, analyze the causes of the problems found in the test, such as speech feature extraction, translation algorithm, sentence segmentation module parameters, etc., and then optimize and adjust accordingly, repeat the test optimization until a satisfactory effect is achieved.

[0024] In a possible implementation, a language conversion module is established for the first user party and the second user party, wherein the language conversion module is embedded with a sentence length matching module, and step S100 further includes step S110, determining the user relationship between the first user party and the second user party. Specifically, it is necessary to determine the user relationship between the first user party and the second user party, and the determination involves consideration of multiple factors, such as the interaction mode, communication frequency, and correspondence of information transmission between the two parties in a specific application scenario. Through comprehensive analysis, it is clear what kind of association state the two parties are in, laying the foundation for subsequent targeted channel configuration and language conversion operations.

[0025] Step S120, if the first user party and the second user party are in a one-to-one conversion relationship, configure a one-to-one dual language conversion channel, wherein the dual language conversion channel includes a channel for the first user party to convert to the second user party, and a channel for the second user party to convert to the first user party. Specifically, once it is determined that the first user party and the second user party are in a one-to-one conversion relationship, the matching channel configuration work is performed. Specifically, to configure a one-to-one dual language conversion channel, the composition of the channel has a clear directionality, including a channel for converting from the first user party to the second user party, and a channel for converting from the second user party to the first user party. During the configuration process, it is necessary to ensure that the channels in these two directions are functionally complete and independent according to the rules set by the system and relevant technical requirements, and can accurately carry and transmit the language information of the corresponding direction, so as to provide a stable and effective transmission path for subsequent language conversion.

[0026] Step S130, the language conversion module performs one-to-one language conversion according to the one-to-one dual language conversion channel. Specifically, after the one-to-one dual language conversion channel is successfully configured, the language conversion module performs a one-to-one language conversion operation based on this channel. First, the voice content of both parties is recognized through the AI ​​voice training model, and voice features are extracted from it. The voice is converted into text form with the help of natural language processing technology. According to the rules of the configured dual language conversion channel, the text language of the first user party is accurately converted into a language form that the second user party can understand through the corresponding channel, and vice versa. During the conversion process, the conversion logic and parameter requirements set by the channel are strictly followed to ensure the accuracy and fluency of the language conversion, thereby realizing efficient and accurate cross-language communication between the first user party and the second user party.

[0027] In a possible implementation, the user relationship between the first user party and the second user party is determined, and step S110 further includes step S111: if the first user party and the second user party are in a one-to-many conversion relationship, N dual-language conversion channels are configured, wherein N is the number of users of the larger party of the first user party and the second user party, and N is a positive integer greater than or equal to 2. Specifically, after determining that the first user party and the second user party are in a one-to-many conversion relationship, a corresponding channel configuration operation is performed to specify that the number of dual-language conversion channels to be configured is N, wherein the value of N is based on the number of users of the larger party of the first user party and the second user party, and N is a positive integer greater than or equal to 2. For example, if the first user party is 1 person and the second user party is 5 people, then 5 dual-language conversion channels need to be configured. When configuring the channels, it is necessary to ensure that each channel can independently complete the conversion function setting from one language to another, so that it has complete conversion capabilities and is prepared for subsequent synchronous conversion.

[0028] Step S112, the language conversion module performs one-to-many language synchronous conversion according to the N dual-language conversion channels. Specifically, after completing the configuration of the N dual-language conversion channels, the language conversion module starts a one-to-many language synchronous conversion operation according to the configured channels. Since each dual-language conversion channel operates independently, each channel can synchronously start the conversion task during the conversion process. For example, for the 5 configured dual-language conversion channels, each channel can perform a conversion operation in real time for its corresponding input language, and synchronously convert the language of the first user party into the language form corresponding to different individuals of the second user party, and vice versa. Moreover, the conversion is performed in real time, and the conversion action can be completed quickly after receiving the voice input (after undergoing the pre-processing process such as AI voice training model recognition, voice feature extraction, and conversion into text in combination with natural language processing technology), so as to ensure that in a one-to-many communication scenario, the first user party and multiple second users can achieve efficient, accurate and synchronous cross-language communication results.

[0029] In a possible implementation, the user relationship between the first user party and the second user party is determined, and step S110 further includes step S113, if the first user party and the second user party are in a many-to-many conversion relationship, M×N dual-language conversion channels are configured, where M is the number of the first user party and N is the number of the second user party. Specifically, when it is determined that the first user party and the second user party present a many-to-many conversion relationship, the configuration of the dual-language conversion channels adapted thereto is started, and the number of dual-language conversion channels to be configured is M×N, where M represents the number of the first user party and N represents the number of the second user party. For example, if the first user party has 3 people and the second user party has 4 people, then 3×4=12 dual-language conversion channels need to be configured. When configuring the channels, it is necessary to ensure that each channel can be accurately set with the corresponding function according to the established rules and relevant technical requirements of the system, so that it has the complete ability to convert from one language to another, and provide a reliable channel foundation for the subsequent synchronous conversion of many-to-many languages.

[0030] Step S114, the language conversion module performs multi-to-multi-language synchronous conversion according to the N dual-language conversion channels. Specifically, after completing the configuration of the M×N dual-language conversion channels, the language conversion module performs multi-to-multi-language synchronous conversion operations according to the configured channels. Each dual-language conversion channel is set to be able to independently complete the language conversion task, so in the actual conversion process, the channels will be started and operated at the same time. Specifically, for each dual-language conversion channel, the voice content from different first users is first received. Before this, the voice content needs to be recognized by the AI ​​voice training model, and voice features are extracted from it. Then, the voice is converted into text form in combination with natural language processing technology. The channel will accurately convert the received text-based voice content into a language form that can be understood by the corresponding second user according to the pre-set conversion rules, and vice versa. Through such synchronous conversion operations, in many-to-many communication scenarios, each first user party and each second user party can achieve efficient, accurate and synchronous cross-language communication with the help of these dual-language conversion channels, ensuring that information can be smoothly transmitted and converted between different users, effectively meeting the language conversion needs in many-to-many communication scenarios.

[0031] Step S200, obtain the first language type corresponding to the first user party and the second language type corresponding to the second user party. Specifically, in the process of obtaining the first language type corresponding to the first user party and the second language type corresponding to the second user party, the acquisition target must be clarified first, that is, the language type of both parties must be accurately determined, because it is an important basis for subsequent effective language conversion and cross-language communication. Then, it can be achieved through a variety of acquisition methods. It can be directly read from the user's setting information on the application or device. If it cannot be directly acquired, it can also be used with the help of feature analysis of the user's voice, using the AI ​​voice training model to recognize the collected voice and extract voice features such as intonation and speech speed, and infer the language type in combination with the language voice feature database, or analyze the vocabulary, grammatical structure and other characteristics of the text content input by the user, and determine it in combination with relevant algorithms and databases. Finally, verification and confirmation work must be carried out, by matching the acquired language type with the known language pattern for verification, and confirming with the interactive feedback with the user. If there is an error, it will be re-acquired and analyzed until the language type corresponding to both parties is accurately acquired, thereby providing accurate basic data for subsequent processes.

[0032] Step S300, obtaining the real-time scene type between the first user party and the second user party. Specifically, it is extremely important to accurately obtain the real-time scene type between the first user party and the second user party, because it will affect language use, expression habits and processing requirements. The main ways to obtain are: based on device sensor information, such as using a microphone to collect environmental sound features and a camera to identify surrounding visual features to infer the scene type; analyzing user behavior data, judging the scene based on features such as user operation software, text input speed and pause time; combining application usage scenarios, determining the scene based on the specific application used and its functional characteristics. After that, the obtained scene type needs to be verified and confirmed. It can be cross-validated through multi-source data, and information from different channels can be combined to verify each other. If there is any discrepancy, further analysis and determination are required; it can also be confirmed with the help of user feedback. If the user feedback is inconsistent with the acquired information, it is necessary to re-evaluate and analyze, so as to accurately obtain the real-time scene type, and provide a reliable basis for subsequent language processing and other operations to achieve barrier-free communication across languages.

[0033] Step S400, input the first language type, the second language type and the real-time scene type into the sentence length matching module for step matching, and obtain the sentence segmentation strategy, which includes the first sentence segmentation step corresponding to the first user party and the second sentence segmentation step corresponding to the second user party. Specifically, after accurately obtaining the first language type corresponding to the first user party, the second language type corresponding to the second user party and the real-time scene type between the two parties, these information need to be accurately input into the sentence length matching module according to the format requirements specified by the sentence length matching module. Then, the sentence length matching module carries out the step length matching operation according to the internal preset algorithm and rules. It will deeply analyze the grammar and sentence characteristics of different language types and the influence of the real-time scene type, such as the language communication in daily dialogue scenes is fast and casual, tending to be short step length, and the language expression in formal scenes such as speeches and meetings is rigorous, tending to be long step length. Through comprehensive consideration and data comparison, the most suitable sentence segmentation step length for the first and second users is determined. Finally, the module generates a segmentation strategy including a first segmentation step corresponding to the first user and a second segmentation step corresponding to the second user. The second segmentation step can be selected to adapt to different contexts according to the scene type and user language. The short step is suitable for daily conversation scenes, and the long step is suitable for speeches and meetings, etc., thereby providing important segmentation basis for subsequent language recognition conversion and other operations, and improving the accuracy and fluency of cross-language communication.

[0034] In a possible implementation, the first language type, the second language type, and the real-time scene type are input into the sentence segmentation length matching module for step matching to obtain a sentence segmentation strategy, wherein the sentence segmentation strategy includes a first sentence segmentation step corresponding to the first user party and a second sentence segmentation step corresponding to the second user party. Step S400 further includes step S410, constructing a step reference library, wherein the step reference library stores a sentence segmentation step range corresponding to each language type under each scene type. Specifically, first, a step length reference library is constructed with the aim of providing comprehensive and accurate data support for subsequent sentence segmentation step length matching analysis. During the construction process, detailed data collection and organization are required for each language type that may be involved, as well as the situation of each language type in different scenario types. For example, for common languages ​​​​such as English, Chinese, and French, their sentence segmentation characteristics in various scenario types such as daily conversations, business meetings, academic lectures, and speeches should be considered separately, and the sentence segmentation step length range corresponding to each language type in each scenario type should be accurately stored in the step length reference library. The determination of the sentence segmentation step length range may be based on a large amount of language sample analysis, previous language processing experience, and related linguistic research results, etc., to construct a step length reference library covering multiple languages ​​and scenario combinations, so that the system can quickly and accurately obtain the required sentence segmentation step length reference information in subsequent processing.

[0035] Step S420, extract the complexity features of each language type, and configure the step adjustment factor of each language type according to the complexity features, wherein the complexity features include grammatical complexity and sentence structure complexity. Specifically, after completing the construction of the step reference library, the complexity features of each language type need to be extracted. The complexity features mainly include grammatical complexity and sentence structure complexity. For grammatical complexity, it is necessary to analyze the complexity of the grammatical rules of the language. For example, there are many tenses, voices, clauses and other grammatical forms in English. Although Chinese grammar is relatively flexible, it also has its own unique features in the usage of function words and word order, which will also affect the grammatical complexity. For sentence structure complexity, it is necessary to examine the composition of the sentence. For example, English sentences may have very long attributive clauses, adverbial clauses and other complex structures. Chinese sentences also have special sentence structures such as conjunctive sentences and conjunctive sentences. In terms of accuracy, After extracting the complexity characteristics of each language type, the step adjustment factor of each language type is configured according to the complexity characteristics. The configuration of the step adjustment factor is a process of comprehensively considering various factors of the complexity characteristics. If the grammatical complexity and sentence structure complexity of a language are high, then a relatively large step adjustment factor may be configured, which means that in the subsequent segmentation step matching analysis, the segmentation step needs to be adjusted to a greater extent to adapt to the segmentation requirements of this complex language structure; conversely, if the complexity is low, the step adjustment factor may be relatively small. According to the actual complexity characteristics of different language types, a suitable step adjustment factor is tailored for each language type.

[0036] Step S430, the sentence length matching module performs step length matching analysis in the step length reference library based on the step length adjustment factor. Specifically, after the construction of the step length reference library is completed and the step length adjustment factor is configured for each language type, the sentence length matching module begins to perform step length matching analysis in the step length reference library based on the step length adjustment factor. First, the sentence length matching module receives the language information to be processed, including the language type and related scene type, etc. According to the received language type, the previously configured step length adjustment factor is obtained. With the step length adjustment factor, the sentence length matching module searches the step length reference library for a sentence length range that matches the current language type and scene type. For example, if the current processing is the sentence segmentation of English in a business meeting scene. If there is a problem, the sentence segmentation step range of English in the business meeting scenario will be found in the step reference library, and the step adjustment factor of English configured before will be combined to further analyze and adjust this step range. By comprehensively considering the step adjustment factor and the sentence segmentation step range in the step reference library, the sentence segmentation length matching module will finally determine the sentence segmentation step that best suits the current language type and scenario type. The sentence segmentation step will serve as an important basis for sentence segmentation operations on the language in the subsequent language processing process, thereby ensuring reasonable and accurate sentence segmentation in different language types and scenario types, and improving the quality and effect of language processing.

[0037] In a possible implementation, the first language type, the second language type, and the real-time scene type are input into the sentence length matching module for step matching to obtain a sentence segmentation strategy, wherein the sentence segmentation strategy includes a first sentence segmentation step corresponding to the first user party and a second sentence segmentation step corresponding to the second user party. Step S400 further includes step S440, wherein the first language type, the second language type, and the real-time scene type are input into the sentence length matching module, and a first initial step corresponding to the first language type and a second initial step corresponding to the second language type under the real-time scene type are obtained from the step reference library. Specifically, it is necessary to ensure that the first language type corresponding to the first user, the second language type corresponding to the second user, and the real-time scenario type between the two parties have been accurately obtained, which reflect the characteristics and context of language communication from different angles. The first language type and the second language type clarify the specific language used by both parties. Different languages ​​differ in grammar, vocabulary, sentence structure, etc., which will affect the way of segmentation. The real-time scenario type covers different communication scenarios such as daily conversations, business meetings, speeches, etc. The expression rhythm and segmentation habits of the language will also be different in different scenarios. The prepared first language type, second language type and real-time scenario type should be matched according to the input format and requirements specified by the segmentation length matching module to ensure accuracy. The information is input into the module. After receiving the input information, the sentence length matching module will immediately interact with the step reference library. The step reference library stores rich data, including the sentence step range corresponding to each language type in each scene type. The module will accurately search and obtain the first initial step corresponding to the first language type and the second initial step corresponding to the second language type in the real-time scene type in the step reference library according to the input real-time scene type. For example, if the real-time scene type is daily conversation and the first language type is English, the first initial step corresponding to English in the daily conversation scene will be found from the step reference library; similarly, the corresponding second initial step will be obtained for the second language type.

[0038] In a possible implementation, the first language type, the second language type, and the real-time scenario type are input into the sentence length matching module for step matching to obtain a sentence segmentation strategy, wherein the sentence segmentation strategy includes a first sentence segmentation step corresponding to the first user party and a second sentence segmentation step corresponding to the second user party. Step S400 further includes step S450, in which the first initial step and the second initial step are adjusted based on the step adjustment factor until the current recognition accuracy reaches a preset recognition accuracy, and the first sentence segmentation step and the second sentence segmentation step are output.Specifically, the step adjustment factor is configured based on the analysis of the complexity characteristics of each language type. The complexity characteristics mainly include grammatical complexity and sentence structure complexity. The complexity of different language types is different. For example, some languages ​​have numerous grammatical rules and complex sentence structures, and their corresponding step adjustment factors may be relatively large; while relatively simple languages ​​may have relatively small step adjustment factors. The role of the step adjustment factor is to reasonably adjust the initial step size according to the actual complexity of the language to adapt to the sentence segmentation requirements of different languages ​​in different scenarios. Based on the first initial step size and the second initial step size that have been obtained, as well as the corresponding step adjustment factors, the two initial step sizes are adjusted. The first initial step length is adjusted according to the step length adjustment factor of the first language type and the current real-time scene type. For example, if the step length adjustment factor of the first language type indicates that the sentence segmentation step length of the language needs to be appropriately increased in the current scene, the first initial step length will be increased and adjusted accordingly. The second initial step length is adjusted according to the step length adjustment factor of the second language type and based on the current real-time scene type. For example, if the step length adjustment factor of the second language type indicates that the sentence segmentation step length should be appropriately shortened in the current scene, the second initial step length will be shortened accordingly. In the process of adjusting the step length, it is not enough to complete the adjustment once and for all, but it is necessary to continuously pay attention to the current recognition accuracy, input the language content processed by the adjusted step length into the relevant language recognition module, obtain its current recognition accuracy, and continuously adjust the first initial step length and the second initial step length until the current recognition accuracy reaches the preset recognition accuracy. The preset recognition accuracy is a standard value set in advance, which represents the degree of accurate recognition that is expected to be achieved in the process of language recognition conversion. Only when the current recognition accuracy reaches this standard value can the accuracy and reliability of subsequent language conversion be ensured. After repeated adjustment of the step length, it is finally determined that the current recognition accuracy has reached the preset recognition accuracy. When the preset recognition accuracy is reached, the final segmentation step length can be prepared to be output. After a series of adjustments, the first initial step length and the second initial step length have become suitable step lengths that meet the language recognition accuracy requirements, namely the first segmentation step length and the second segmentation step length. The adjusted first segmentation step length and the second segmentation step length are output from the segmentation length matching module. The two segmentation step lengths will serve as the key basis for the subsequent language conversion module to perform accurate segmentation operations when performing language recognition conversion on the first user and the second user. This can ensure efficient and accurate language processing under different language types and real-time scenario types, thereby improving the smoothness and accuracy of cross-language communication.

[0039] In a possible implementation, the first language type, the second language type and the real-time scene type are input into the sentence length matching module for step matching to obtain a sentence segmentation strategy, wherein the sentence segmentation strategy includes a first sentence segmentation step corresponding to the first user party and a second sentence segmentation step corresponding to the second user party. Step S400 further includes step S460, and the sentence length matching module also includes a speech speed detection unit. Specifically, in the architecture of the entire language processing solution, the sentence length matching module plays a vital role. The module is provided with a speech speed detection unit. The speech speed detection unit is a functional component specifically used for real-time monitoring of the speech speed during language communication. It is integrated in the sentence length matching module and works in conjunction with other parts of the module to provide accurate data support for the subsequent reasonable adjustment of the sentence segmentation step according to the speech speed.

[0040] Step S470, the speech speed of the first user and the second user is detected in real time according to the speech speed detection unit. Specifically, when the language communication activity is started, the speech speed detection unit begins to play its role and detects the speech speed of the first user and the second user in real time. In the actual operation process, when the first user starts to speak, the speech speed detection unit will use speech recognition technology to collect and analyze its speech signal, accurately capture the key elements such as the utterance frequency and syllable interval time of the speech, and then convert these elements into quantifiable speech speed data through a specific algorithm. For the second user, the speech speed detection unit will also obtain its speech speed information in real time during its speech in the same way. Through the real-time detection process, the actual speech speed of the first user and the second user during the communication process can be accurately and timely grasped, laying the foundation for the subsequent adjustment of the sentence segmentation step.

[0041] Step S480, adjusting the first segmentation step and the second segmentation step according to the speaking speed of the first user and the second user. Specifically, after obtaining the accurate speaking speed information of the first user and the second user, the first segmentation step and the second segmentation step that have been determined should be adjusted according to these speaking speed data. For the first segmentation step, if it is detected that the speaking speed of the first user is fast, it means that the content he says is more in a unit time. In order to ensure the accuracy of language processing and the rationality of segmentation, it may be necessary to appropriately increase the first segmentation step so that the sentence after segmentation can still maintain a clear semantic structure in the fast language flow, so that subsequent language recognition conversion and other operations can proceed smoothly. On the contrary, if the speaking speed of the first user is slow, it may be necessary to appropriately reduce it. The first segmentation step makes the segmented sentence more consistent with the relatively slow language rhythm to avoid the sentence being too long and affecting understanding. For the second segmentation step, when it is detected that the second user party speaks faster, it is also necessary to appropriately increase the second segmentation step according to the speaking speed; if the second user party speaks slower, the second segmentation step is appropriately reduced accordingly. By making targeted adjustments to the first segmentation step and the second segmentation step according to the respective speaking speeds of the first user party and the second user party, it can be ensured that in language communication scenarios with different speaking speeds, the segmentation step can always match the speaking speed, thereby further improving the overall effect of language processing and achieving smoother and more accurate cross-language communication.

[0042] Step S500, the language conversion module performs language recognition conversion on the first user party and the second user party according to the first segmentation step and the second segmentation step. Specifically, when performing language recognition conversion, the language conversion module first obtains the first segmentation step and the second segmentation step determined by a series of complex steps, and these steps fully consider the language type of the first user party and the second user party, the real-time scene type of both parties, and the speech speed. When the first user party performs language expression (in the form of voice or text input, etc.), the module first uses the corresponding technology (voice recognition or text processing technology) to preliminarily process the input content, such as voice input is converted into text and features are extracted, and then the preliminarily processed text is segmented according to the first segmentation step, and then the sentence fragments after segmentation are deeply analyzed to identify their vocabulary, grammatical structure, semantic relationship, etc. to determine the meaning and grammatical function information, and then according to this information, the sentence is converted from the source language to the target language in combination with the grammatical rules, vocabulary library and conversion algorithm of the target language, and the sentence structure, vocabulary selection, etc. are appropriately adjusted. Similarly, when the second user expresses himself in language, the module also adopts a similar processing method, first performing preliminary processing, then segmenting according to the second segmentation step, then analyzing and identifying sentence fragments, and finally converting the sentence from the source language to the target language and adjusting it as needed. By performing language recognition and conversion on both parties according to different segmentation steps, accurate and efficient cross-language communication can be achieved in different language environments and communication scenarios, meeting the language conversion needs of both parties.

[0043] The embodiment of the present application adopts a language conversion module for establishing the first and second users, and embeds a sentence length matching module. After obtaining the language type and real-time scene type of both parties, this information is input into the sentence length matching module for step matching, and a sentence segmentation strategy is generated, including the sentence segmentation step lengths of both parties. The language conversion module recognizes and converts the user's language according to the corresponding sentence segmentation step length, achieving the technical effect of improving the real-time, accuracy and semantic coherence of multilingual communication.

[0044] In the above, refer to Figure 1 A multi-scenario multi-language automatic recognition method according to an embodiment of the present invention is described in detail. Figure 2 A multi-scenario multi-language automatic recognition device according to an embodiment of the present invention is described.

[0045] A multi-scenario multi-language automatic recognition device according to an embodiment of the present invention is used to solve the technical problems of translation delay and semantic incoherence in existing multi-language cross-scenario communication, and achieve the technical effect of improving the real-time, accuracy and semantic coherence of multi-language communication. A multi-scenario multi-language automatic recognition device includes: a language conversion module establishment module 10, a language type acquisition module 20, a real-time scene type acquisition module 30, a sentence segmentation strategy acquisition module 40, and a language recognition conversion module 50.

[0046] The language conversion module establishment module 10 is used to establish the language conversion modules of the first user party and the second user party, wherein the language conversion module is embedded with a sentence length matching module.

[0047] The language type acquisition module 20 is used to acquire a first language type corresponding to the first user party and a second language type corresponding to the second user party.

[0048] The real-time scene type acquisition module 30 is used to acquire the real-time scene type between the first user party and the second user party.

[0049] The segmentation strategy acquisition module 40 is used to input the first language type, the second language type and the real-time scene type into the segmentation length matching module for step matching to obtain the segmentation strategy, wherein the segmentation strategy includes a first segmentation step corresponding to the first user and a second segmentation step corresponding to the second user.

[0050] The language identification conversion module 50 is used for the language conversion module to perform language identification conversion on the first user party and the second user party according to the first segmentation step length and the second segmentation step length respectively.

[0051] The specific configuration of the language conversion module establishment module 10 will be described in detail below. As described above, a language conversion module for the first user party and the second user party is established, wherein the language conversion module is embedded with a sentence length matching module, and the language conversion module establishment module 10 further includes: a user relationship judgment unit, the user relationship judgment unit is used to judge the user relationship between the first user party and the second user party; a dual language conversion channel configuration unit, the dual language conversion channel configuration unit is used to configure a one-to-one dual language conversion channel if the first user party and the second user party are in a one-to-one conversion relationship, wherein the dual language conversion channel includes a channel for the first user party to convert to the second user party, and a channel for the second user party to convert to the first user party; a language conversion unit, the language conversion unit is used for the language conversion module to perform one-to-one language conversion according to the one-to-one dual language conversion channel.

[0052] Among them, the user relationship between the first user party and the second user party is determined, and the user relationship determination unit further includes: a dual language conversion channel configuration subunit, the dual language conversion channel configuration subunit is used to configure N dual language conversion channels if the first user party and the second user party are in a one-to-many conversion relationship, wherein N is the number of users of the larger party between the first user party and the second user party, and N is a positive integer greater than or equal to 2; a synchronous conversion subunit, the synchronous conversion subunit is used for the language conversion module to perform one-to-many language synchronous conversion according to the N dual language conversion channels.

[0053] Among them, the user relationship between the first user party and the second user party is determined, and the user relationship determination unit further includes: a conversion relationship determination subunit, the conversion relationship determination subunit is used to configure M×N dual language conversion channels if the first user party and the second user party are in a many-to-many conversion relationship, wherein M is the number of first user parties and N is the number of second user parties; a many-to-many language synchronous conversion subunit, the many-to-many language synchronous conversion subunit is used for the language conversion module to perform many-to-many language synchronous conversion according to the N dual language conversion channels.

[0054] The specific configuration of the segmentation strategy acquisition module 40 will be described in detail below. As described above, the first language type, the second language type and the real-time scene type are input into the segmentation length matching module for step length matching to obtain the segmentation strategy, the segmentation strategy includes the first segmentation step length corresponding to the first user party and the second segmentation step length corresponding to the second user party, and the segmentation strategy acquisition module 40 further includes: a step length reference library construction unit, the step length reference library construction unit is used to construct a step length reference library, wherein the step length reference library stores the segmentation step length range corresponding to each language type under each scene type; a complexity feature extraction unit, the complexity feature extraction unit is used to extract the complexity feature of each language type, and configure the step length adjustment factor of each language type according to the complexity feature, wherein the complexity feature includes grammatical complexity and sentence structure complexity; a step length matching analysis unit, the step length matching analysis unit is used for the segmentation length matching module to perform step length matching analysis in the step length reference library based on the step length adjustment factor.

[0055] Among them, the sentence segmentation strategy acquisition module 40 further includes: a step length matching unit, the step length matching unit is used to input the first language type, the second language type and the real-time scene type into the sentence segmentation length matching module, and obtain the first initial step length corresponding to the first language type and the second initial step length corresponding to the second language type under the real-time scene type in the step length reference library; a step length adjustment unit, the step length adjustment unit is used to adjust the first initial step length and the second initial step length based on the step length adjustment factor until the current recognition accuracy reaches the preset recognition accuracy, and output the first sentence segmentation step length and the second sentence segmentation step length.

[0056] Among them, the sentence segmentation strategy acquisition module 60 further includes: a sentence segmentation length matching module component unit, the sentence segmentation length matching module component unit is used for the sentence segmentation length matching module also includes a speech speed detection unit; a speech speed detection unit, the speech speed detection unit is used to detect the speech speed of the first user party and the second user party in real time according to the speech speed detection unit; a sentence segmentation step adjustment unit, the sentence segmentation step adjustment unit is used to adjust the first sentence segmentation step and the second sentence segmentation step according to the speech speed of the first user party and the second user party respectively.

[0057] The multi-scenario multi-language automatic recognition device provided in the embodiment of the present invention can execute the multi-scenario multi-language automatic recognition method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0058] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0059] Based on the foregoing embodiments, the embodiments of the present application further provide an electronic device and a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of the electronic device, it can implement the method described in any of the foregoing embodiments.

[0060] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing an implementation of the present invention. Figure 3The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. The electronic device is in the form of a general computing device, and its components may include but are not limited to an input device 301, a processor 302, a memory 303, and an output device 304. Among them, the processor 302 may be one or more; the memory 303 may include a computer-readable medium and at least one program product, and the program product has a set (at least one) of program modules, which are configured to perform the functions of each embodiment of the present application.

[0061] The memory 303 shown in the embodiment of the present invention may adopt any combination of one or more computer-readable media; the computer-readable storage medium may be, but is not limited to, an infrared, semiconductor system, device or component, or any combination of the above, for storing software programs, computer executable programs and modules, such as program instructions / modules corresponding to a multi-scenario multi-language automatic recognition method in an embodiment of the present invention. The processor 302 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 303, thereby realizing the above-mentioned multi-scenario multi-language automatic recognition method.

[0062] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-scenario multi-language automatic recognition method, characterized in that: The method comprises: Establishing a language conversion module for the first user party and the second user party, wherein the language conversion module is embedded with a sentence length matching module; Acquire a first language type corresponding to the first user party and a second language type corresponding to the second user party; Acquire a real-time scene type between the first user party and the second user party; Inputting the first language type, the second language type, and the real-time scenario type into the segmentation length matching module for step length matching, and acquiring a segmentation strategy, wherein the segmentation strategy includes a first segmentation step length corresponding to the first user party and a second segmentation step length corresponding to the second user party; The language conversion module performs language recognition conversion on the first user party and the second user party according to the first segmentation step length and the second segmentation step length respectively.

2. The method according to claim 1, characterized in that Establishing a language conversion module for the first user party and the second user party, the method further includes: determining a user relationship between the first user party and the second user party; If the first user party and the second user party are in a one-to-one conversion relationship, a one-to-one dual language conversion channel is configured, wherein the dual language conversion channel includes a channel for the first user party to convert to the second user party, and a channel for the second user party to convert to the first user party; The language conversion module performs one-to-one language conversion according to the one-to-one dual language conversion channel.

3. The method according to claim 2, characterized in that Determining the user relationship between the first user party and the second user party further includes: If the first user party and the second user party are in a one-to-many conversion relationship, configure N dual language conversion channels, where N is the number of users of the larger party between the first user party and the second user party, and N is a positive integer greater than or equal to 2; The language conversion module performs one-to-multi-language synchronous conversion according to the N dual-language conversion channels.

4. The method according to claim 2, characterized in that Determining the user relationship between the first user party and the second user party further includes: If the first user party and the second user party are in a many-to-many conversion relationship, configure M×N dual language conversion channels, where M is the number of first user parties and N is the number of second user parties; The language conversion module performs multi-to-multi-language synchronous conversion according to the N dual-language conversion channels.

5. The method according to claim 1, characterized in that Inputting the first language type, the second language type and the real-time scene type into the sentence length matching module for step length matching, the method further includes: Constructing a step length reference library, wherein the step length reference library stores the sentence segmentation step length range corresponding to each language type in each scenario type; Extracting complexity features of each language type, and configuring a step adjustment factor of each language type according to the complexity features, wherein the complexity features include grammatical complexity and sentence structure complexity; The sentence length matching module performs step length matching analysis in the step length reference library based on the step length adjustment factor.

6. The method according to claim 5, characterized in that Inputting the first language type, the second language type and the real-time scene type into the sentence length matching module, and acquiring a first initial step length corresponding to the first language type and a second initial step length corresponding to the second language type under the real-time scene type from the step length reference library; The first initial step length and the second initial step length are adjusted based on the step length adjustment factor until the current recognition accuracy reaches a preset recognition accuracy, and a first sentence segmentation step length and a second sentence segmentation step length are output.

7. The method according to claim 5, characterized in that The sentence length matching module also includes a speech speed detection unit; detecting the speech speed of the first user party and the second user party in real time according to the speech speed detection unit; The first sentence segmentation step length and the second sentence segmentation step length are adjusted according to the speech speeds of the first user and the second user, respectively.

8. A multi-scenario multi-language automatic recognition device, characterized in that: The device is used to implement the multi-scenario multi-language automatic recognition method according to any one of claims 1 to 7, and the device comprises: A language conversion module establishment module, the language conversion module establishment module is used to establish a language conversion module of the first user party and the second user party, wherein the language conversion module is embedded with a sentence length matching module; a language type acquisition module, the language type acquisition module being used to acquire a first language type corresponding to the first user party and a second language type corresponding to the second user party; A real-time scene type acquisition module, the real-time scene type acquisition module is used to acquire the real-time scene type between the first user party and the second user party; a sentence segmentation strategy acquisition module, the sentence segmentation strategy acquisition module being used to input the first language type, the second language type and the real-time scene type into the sentence segmentation length matching module for step length matching, and acquire a sentence segmentation strategy, the sentence segmentation strategy including a first sentence segmentation step length corresponding to the first user party and a second sentence segmentation step length corresponding to the second user party; A language identification conversion module is used by the language conversion module to perform language identification conversion on the first user party and the second user party according to the first segmentation step length and the second segmentation step length respectively.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement the multi-scenario multi-language automatic recognition method as described in any one of claims 1 to 7 when executing the executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a multi-scenario multi-language automatic recognition method according to any one of claims 1 to 7 is implemented.

Images