Error correction method and device for speech recognition navigation domain entity, equipment and storage medium

By using the map entity database to correct the speech recognition results, the problem of low recognition accuracy of the speech recognition navigation system when dealing with complex place names is solved, the navigation accuracy and user experience are improved, and the system reliability is enhanced.

CN118936505APending Publication Date: 2024-11-12VOYAH AUTOMOBILE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410990450.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing voice recognition navigation system has low recognition accuracy when dealing with near-sound, homophone place names and rare place names, and relies on local databases to cause limited information, which cannot cover all place names, affecting the accuracy and user experience of the navigation system.

Method used

By obtaining speech recognition results, extracting navigation entities, and using the map entity database for error correction, comparing them based on the path score threshold and the highest path score, the final navigation results are determined to ensure the accuracy and practicality of the recognition.

Benefits of technology

It improves the accuracy of entity recognition in vehicle navigation, optimizes the user experience, makes navigation smoother and more convenient, and enhances the reliability and user satisfaction of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118936505A_ABST
    Figure CN118936505A_ABST
Patent Text Reader

Abstract

The invention discloses an error correction method and device for a voice recognition navigation domain entity, equipment and a storage medium, and relates to the technical field of vehicle-mounted voice recognition navigation, and the error correction method for the voice recognition navigation domain entity comprises the steps: obtaining a voice recognition result; extracting a field of the voice recognition result and an entity to be corrected; and when the field belongs to the navigation field and the word number of the entity to be corrected is greater than or equal to a preset word number, correcting the entity to be corrected according to a map entity database to complete navigation entity correction. According to the method, selection errors caused by voice recognition errors are avoided, the accuracy of voice recognition on entity recognition in the field of vehicle navigation is improved, the accuracy and practicability of navigation are further improved, the navigation experience of a user is optimized, the navigation experience is smoother and more convenient, and the reliability of the system and the satisfaction degree of the user are remarkably enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of vehicle-mounted speech recognition navigation technology, and in particular to an error correction method, device, equipment and storage medium for a speech recognition navigation domain entity. Background Art

[0002] With the advent of the era of intelligent cars, people are increasingly using voice interaction for in-car navigation. Navigation destinations are diverse, covering a variety of locations from shopping malls, restaurants, hotels to scenic spots. However, existing voice recognition systems often have difficulty accurately recognizing these place names, especially when faced with near-sounding or homophonic place names between different cities, as well as uncommon or incomplete place names, the recognition accuracy drops significantly. This situation challenges the accuracy and practicality of in-car navigation systems in entity recognition.

[0003] One of the current solutions is to use the local database for pinyin matching. The speech recognition system first recognizes the user's voice command and extracts the navigation entity information. Then, the system tries to find homophonic or near-homophonic alternative results in the local database. Finally, based on third-party map data, the distance of these alternative results from the current location is calculated, and the nearest result is selected as the navigation destination. Although this method works well in some common situations, it is prone to inaccurate matching or complete failure to match when dealing with complex situations such as multiple words, fewer words, front-to-back swaps, or incompletely similar sounds.

[0004] There are several significant problems with the existing speech recognition navigation entity error correction methods. First, since speakers often do not pronounce the complete place names on the map strictly, pinyin matching is difficult to handle complex situations such as multiple words, fewer words, and changes in tone, resulting in a significant reduction in the accuracy of the match. Secondly, compared with comprehensive third-party maps such as Amap or Baidu Maps, the local database has limited information and cannot cover all rare or uncommon place names, resulting in these place names cannot be correctly matched. In addition, the calculation of third-party map data is dependent on the chain is too long may cause network delays or uncontrollable risks, affecting user experience and real-time performance. Therefore, inaccurate entity recognition of speech recognition in the field of in-vehicle navigation has become a problem that needs to be solved urgently.

[0005] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The purpose of the present application is to provide a method, device, equipment and storage medium for correcting errors in speech recognition of navigation domain entities, aiming to solve the technical problem of inaccurate entity recognition in the field of vehicle-mounted navigation by speech recognition.

[0007] To achieve the above objectives, the present application proposes a method for correcting errors in speech recognition navigation domain entities, the method comprising:

[0008] Get the speech recognition results;

[0009] Extracting the domain and the entity to be corrected of the speech recognition result;

[0010] When the field belongs to the navigation field and the number of characters of the entity to be corrected is greater than or equal to a preset number of characters, the entity to be corrected is corrected according to the map entity database to complete navigation entity correction.

[0011] In one embodiment, the step of correcting the entity to be corrected according to the map entity database to complete the navigation entity correction includes:

[0012] Obtaining a path score threshold and a maximum path score according to a map entity database;

[0013] Comparing the highest path score with the path score threshold to obtain a comparison result;

[0014] When the comparison result is greater than, taking the speech recognition result as the final navigation result;

[0015] When the comparison result is less than or equal to, the path corresponding to the highest path score is used as the final navigation result to complete navigation entity error correction.

[0016] In one embodiment, the step of obtaining the path score threshold and the maximum path score according to the map entity database includes:

[0017] Divide the map entity database into sub-databases by region;

[0018] According to the sub-database, a finite state converter of the sub-database, a path score threshold of the finite state converter, and a target state converter error correction model are obtained;

[0019] The highest path score is obtained according to the current position of the vehicle, the entity to be corrected and the target state converter correction model.

[0020] In one embodiment, the step of obtaining the finite state converter of the sub-database, the path score threshold of the finite state converter, and the target state converter error correction model according to the sub-database includes:

[0021] Segmenting the sub-database to obtain respective language models and a first path score for each entity;

[0022] Constructing a finite state converter of the sub-database according to the language model and the first path score, wherein the modeling unit corresponding to the finite state converter is pinyin;

[0023] Obtaining an entity score of the finite state converter according to the first path score calculation;

[0024] Cutting the first path score according to a preset percentile to obtain a path score threshold of the finite state converter;

[0025] Obtaining a compensation value of the finite state converter according to the entity score;

[0026] A target state switch error correction model is obtained according to each region name, the finite state switch, the entity score and the compensation value.

[0027] In one embodiment, the step of obtaining a target state converter error correction model according to each region name, the finite state converter, the entity score, and the compensation value comprises:

[0028] Obtaining a first sub-state converter according to the finite state converter, the entity score and the compensation value;

[0029] Obtain a second sub-state converter according to each region name and the first sub-state converter;

[0030] Obtain a third sub-state converter according to the region name;

[0031] The first sub-state converter, the second sub-state converter and the third sub-state converter are serially combined to obtain a regional state converter;

[0032] The regional state converters are combined and merged to obtain a primary state converter;

[0033] The first-level state converters are combined in parallel to obtain a second-level state converter;

[0034] The secondary state converter is minimized to obtain a target state converter error correction model.

[0035] In one embodiment, the step of obtaining the highest path score according to the current position of the vehicle, the entity to be corrected, and the target state converter correction model comprises:

[0036] Generate an input sequence of the target state converter error correction model according to the entity to be corrected;

[0037] Inputting the input sequence into the target state converter error correction model for decoding to obtain a preset number of paths and corresponding second path scores and regional labels;

[0038] Matching the current position of the vehicle with the region label to obtain a matching path and a matching path score;

[0039] The result obtained by adding the matching path score to the preset matching score replaces the matching path score of the matching path;

[0040] The second path scores are sorted from largest to smallest to obtain the highest path score.

[0041] In one embodiment, the step of generating an input sequence of the target state converter error correction model according to the entity to be corrected comprises:

[0042] Generate a penalty score set according to the phonetic modeling unit in the target state converter error correction model;

[0043] Obtaining the corresponding entity pinyin according to the entity to be corrected;

[0044] An input sequence of the target state converter error correction model is generated according to the penalty score set and the entity pinyin.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a speech recognition navigation domain entity error correction device, the device comprising:

[0046] A data acquisition module, used to obtain speech recognition results;

[0047] An extraction module, used to extract the domain and the entity to be corrected of the speech recognition result;

[0048] The error correction module is used to correct the entity to be corrected according to the map entity database to complete navigation entity error correction when the field belongs to the navigation field and the number of words in the entity to be corrected is greater than or equal to a preset number of words.

[0049] In addition, to achieve the above-mentioned purpose, the present application also proposes an error correction device for speech recognition navigation domain entities, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the error correction method for speech recognition navigation domain entities as described above.

[0050] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the error correction method of the speech recognition navigation domain entity as described above are implemented.

[0051] One or more technical solutions proposed in this application have at least the following technical effects:

[0052] By obtaining speech recognition results; extracting the domain and the entity to be corrected of the speech recognition results; when the domain belongs to the navigation domain and the number of words of the entity to be corrected is greater than or equal to the preset number of words, the entity to be corrected is corrected according to the map entity database to complete the navigation entity correction. The present application first obtains the user's navigation instructions through speech recognition technology and converts them into text form, thereby reducing the time and error rate of manual input; secondly, the system analyzes the text, extracts the task domain and the navigation destination entity to be processed, effectively identifies the user's intention and reduces the possibility of misunderstanding; after confirming that the task domain is navigation and the number of words of the destination entity reaches the preset threshold, the system uses the map entity database for error correction processing to accurately find the most matching place name, avoiding selection errors caused by speech recognition errors, and improving the accuracy of speech recognition in the field of vehicle navigation. Entity recognition, thereby improving the accuracy and practicality of navigation, not only optimizes the user's navigation experience, making it more smooth and convenient, but also significantly enhances the reliability of the system and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0055] Figure 1 A flowchart diagram of the first embodiment of the error correction method for speech recognition navigation domain entity of the present application is provided;

[0056] Figure 2 A schematic diagram of a spoken language understanding module extraction process provided in Embodiment 1 of the error correction method for speech recognition navigation domain entities of the present application;

[0057] Figure 3 A flowchart diagram of the second embodiment of the error correction method for speech recognition navigation domain entity of the present application is provided;

[0058] Figure 4 This is a schematic diagram of the module structure of the error correction device of the speech recognition navigation domain entity in the embodiment of the present application;

[0059] Figure 5 This is a schematic diagram of the device structure of the hardware operating environment involved in the error correction method of the speech recognition navigation domain entity in the embodiment of the present application.

[0060] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0062] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0063] With the advent of the era of intelligent cars, people are increasingly using voice interaction for in-car navigation, covering a variety of destinations from shopping malls, restaurants to scenic spots. However, existing voice recognition systems face many challenges in recognizing these place names, especially for near-sounding, homophonic place names, and rare or incomplete place names. The currently used pinyin matching method is effective in common situations, but it often cannot accurately match complex situations such as multiple words, few words, and tone changes. At the same time, relying on a limited local database also makes it difficult to correctly handle uncommon place names, affecting the accuracy of the navigation system and user experience.

[0064] The main solutions of the embodiments of the present application are: first, the user's navigation instructions are quickly converted into text through voice recognition technology, which significantly reduces the time and error rate of manual input; second, the system identifies the navigation field of the task and the navigation destination entity to be processed based on text analysis, effectively understands the user's intention, and reduces the possibility of misunderstanding. After confirming that the task is navigation and the destination entity meets the preset word count threshold, the system uses the map entity database to perform error correction processing, accurately finds the most matching place name, and effectively avoids selection errors caused by voice recognition errors. This process significantly improves the accuracy of voice recognition for entity recognition in in-vehicle navigation, thereby improving the accuracy and practicality of navigation.

[0065] It should be noted that the execution subject of the embodiment of the present application may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device, a vehicle system, a voice recognition system, etc. that can realize the above functions. The voice recognition system is taken as an example to illustrate this embodiment and the following embodiments.

[0066] Based on this, the embodiment of the present application provides a method for correcting errors in speech recognition navigation domain entities, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the error correction method for speech recognition navigation domain entity of the present application.

[0067] In this embodiment, the error correction method of the speech recognition navigation domain entity includes steps S10 to S30:

[0068] Step S10, obtaining speech recognition results;

[0069] It should be noted that the speech recognition result refers to the text result obtained from the user's voice input after being processed by the speech recognition system, that is, the user's voice is converted into text output. In the in-vehicle navigation system, this result is usually the place name or address and other information contained in the navigation request made by the user, which is used to determine the navigation destination.

[0070] It can be understood that, first, the system receives voice commands input by the user through a microphone, such as setting a navigation destination; secondly, the received voice signal is preprocessed to remove noise and enhance signal quality; then, the preprocessed voice data is sent to the voice recognition engine and converted into recognition results in text form through acoustic models and language models. The system extracts navigation-related entity information, such as place names or addresses, based on the recognition results; finally, the obtained recognition results are passed to the navigation system for calculating the navigation route or confirming the destination, ensuring that the user's voice input commands can accurately realize the in-vehicle navigation function.

[0071] Step S20, extracting the domain and the entity to be corrected of the speech recognition result;

[0072] It should be noted that the domain refers to the functional or subject range to which the voice command belongs, which is usually used to determine the user's intention and behavior. In the in-vehicle navigation scenario, the domain can be navigation guidance, destination setting, etc. The entity to be corrected refers to the specific entity information extracted from the voice recognition results that needs to be corrected or processed, such as the name or address of the navigation destination.

[0073] It can be understood that after the user's voice command is first input into the system, the system converts the voice into text form through speech recognition technology; then, according to the specific needs of the in-vehicle navigation field, the system analyzes and determines whether the recognition result belongs to the navigation field through the spoken language understanding module (SLU), such as the module composed of Joint Bert+Crf; if it is confirmed to belong to the navigation field, the specific navigation entity information is further extracted from the recognition result, which usually includes the name or description of the navigation destination. This process ensures that the system can accurately understand the user's navigation intention and lays the foundation for subsequent error correction processing. Among them, Joint Bert+Crf is a model that combines the pre-trained model of bidirectional encoder representations from transformers (Bert) and conditional random fields (CRF). It is commonly used in the field of natural language processing, especially in named entity recognition (NER) tasks. Bert is a pre-trained language representation model that can capture semantic information in text, while CRF is a sequence labeling model that is often used to solve sequence labeling problems, such as named entity recognition. The Joint Bert+Crf model combines the context understanding ability of Bert and the sequence labeling ability of CRF, and can effectively identify and label entities in text, such as names, place names, time, etc.

[0074] Please refer to Figure 2 , Figure 2The schematic diagram of the extraction process of the spoken language understanding module provided in the first embodiment of the error correction method of the speech recognition navigation domain entity of the present application shows the process from speech recognition results to semantic understanding and error correction. First, the speech recognition system obtains the user's navigation instructions and converts them into text form, reducing the time and error rate of manual input; then, the recognition result is sent to the SLU module as input data, and the Joint Bert model and CRF (conditional random field) are used for intent classification and slot filling, so as to effectively identify the user's intention and reduce the possibility of misunderstanding; the output of the SLU module includes domain, slot and intent, which clarifies the user's intention and identifies the relevant navigation entities. Next, it is determined whether the domain is a point of interest (POI) and whether there is a valid slot, ensuring that only navigation-related domains and valid navigation entities enter the next step of processing. If the conditions are met, the finite state transducer (FST) model is used for entity error correction to improve the accuracy of entity recognition and navigation accuracy, thereby optimizing the user navigation experience; otherwise, the entity error correction is skipped to avoid unnecessary processing and improve system efficiency. The overall process combines natural language processing technology and entity error correction models, significantly enhancing the accuracy and reliability of speech recognition in in-vehicle navigation.

[0075] Step S30, when the field belongs to the navigation field and the number of characters of the entity to be corrected is greater than or equal to a preset number of characters, the entity to be corrected is corrected according to the map entity database to complete navigation entity correction.

[0076] It should be noted that the navigation field refers to the scope of the field involved in the speech recognition results, specifically the place or position to which the user requires navigation in the navigation task; the preset number of words refers to a pre-set word count threshold, which is used to determine whether the length of the entity to be corrected meets the correction conditions. This threshold may be preset by the system to filter out entities that are too short or too simple. In this embodiment, the preset number of words is 1 or 2; the map entity database refers to the map data stored in the system, which contains a database of various places, location names and related information. This database is used to correct the navigation entities extracted from the speech recognition results to ensure that the most suitable place is selected as the navigation destination.

[0077] It is understandable that first, the system will determine whether the field to which the speech recognition result belongs is the navigation field, that is, the user is requesting a navigation task; secondly, the system will check whether the number of words in the entity to be corrected reaches the preset word count threshold to ensure the complexity of the entity and the necessity of error correction; once it is confirmed that the field and word count conditions are met, the system will use the map entity database for error correction. This process includes comparing the entity to be corrected in the speech recognition result with the location information in the map entity database, and trying to find the most matching place name. Through such error correction steps, the system can improve the accuracy of understanding the user's intention in the navigation task, ensure that the most suitable destination is selected for navigation, and thus optimize the user experience and navigation accuracy.

[0078] This embodiment provides a method for correcting the error of speech recognition navigation domain entities, by obtaining speech recognition results; extracting the domain of the speech recognition results and the entity to be corrected; when the domain belongs to the navigation domain and the number of words of the entity to be corrected is greater than or equal to the preset number of words, the entity to be corrected is corrected according to the map entity database to complete the navigation entity correction. This embodiment first obtains the user's navigation instructions through speech recognition technology and converts them into text form, thereby reducing the time and error rate of manual input; secondly, the system analyzes the text, extracts the task domain and the navigation destination entity to be processed, effectively identifies the user's intention and reduces the possibility of misunderstanding; after confirming that the task domain is navigation and the number of words of the destination entity reaches a preset threshold, the system uses the map entity database for error correction processing, accurately finds the most matching place name, avoids selection errors caused by speech recognition errors, improves the accuracy of speech recognition in the field of vehicle navigation, and thus improves the accuracy and practicality of navigation, which not only optimizes the user's navigation experience, making it more smooth and convenient, but also significantly enhances the reliability of the system and user satisfaction.

[0079] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can refer to the above introduction, and will not be repeated later. Figure 3 , Figure 3 This is a flow chart of a second embodiment of the error correction method for speech recognition navigation domain entity of the present application. Step S30 of the error correction method for speech recognition navigation domain entity includes steps S31 to S34:

[0080] Step S31, obtaining a path score threshold and a maximum path score according to a map entity database;

[0081] It should be noted that the path score threshold refers to a set score limit used to evaluate and filter the path scores in the map entity database. This threshold determines which navigation entities with path scores higher than the threshold will be considered reasonable choices. The highest path score is the highest score calculated for each navigation entity in the map entity database. This score reflects the matching degree and practicality of the navigation entity with the user's voice input, and is usually used to determine the best navigation destination suggestion.

[0082] It can be understood that, first, the system obtains the navigation entity information input by the user, such as the destination name, through voice recognition. Secondly, the system uses the information in the map entity database to match and score the recognized destination names. In this process, the system compares the recognized entities with the entities in the database to evaluate their matching degree and rationality. Then, the system calculates the score of each entity based on the scoring rules and algorithms, reflecting its matching degree with the user's voice input and navigation practicality. Finally, the system determines the entity with the highest score from the scored entities as the optimal navigation suggestion to ensure that the user obtains the most suitable navigation destination. This process effectively improves the accuracy and user experience of the voice navigation system.

[0083] As an example, the steps of obtaining a path score threshold and a maximum path score based on a map entity database include: dividing the map entity database into sub-databases by region; obtaining a finite state converter of the sub-database, a path score threshold of the finite state converter, and a target state converter error correction model based on the sub-database; and obtaining the maximum path score based on the current position of the vehicle, the entity to be corrected, and the target state converter error correction model.

[0084] The sub-database refers to a smaller database set obtained by dividing the map entity database according to other specific standards such as regions, districts, counties or cities. Each sub-database contains navigation entity information in a specific area. Finite State Transducer (FST) is a mathematical model used to describe the state of a system and the transition between states. In this embodiment, the finite state transducer is used to represent the possible paths and state transitions of navigation entities, as well as the associations between them. The path score threshold is a set score limit used to evaluate and screen the path score of each navigation entity in the map entity database. Paths exceeding the threshold are considered to be reasonable navigation destination choices. The target state transducer error correction model is a model used to describe and correct possible errors or inconsistencies in navigation entities. It is based on the framework of the finite state transducer and infers the most appropriate navigation destination by comparing the current position of the vehicle, the entity to be corrected, and the preset model. The current position of the vehicle refers to the actual geographical location or coordinates of the vehicle, which is usually obtained through the global positioning system (GPS) or other location services. The highest path score is the highest score calculated for each navigation entity. This score reflects the matching degree and practicality of the navigation entity with the user's voice input, and is used to determine the optimal navigation destination suggestion. Firstly, the map entity database is divided into sub-databases by region, and each sub-database contains navigation entity information of a specific region; then, for each sub-database, a corresponding finite state converter model is constructed, which describes the possible paths and state transitions of the navigation entity, and determines the path score threshold of each state converter to evaluate the rationality of the navigation entity; at the same time, a target state converter error correction model is established, which infers the most suitable navigation destination option based on the current position of the vehicle, the entity to be corrected, and the preset rules; finally, by comparing the path scores of different navigation entities, the highest path score is determined and recommended to provide the best voice recognition navigation experience.

[0085] As an example, the steps of obtaining the finite state converter of the sub-database, the path score threshold of the finite state converter and the target state converter error correction model according to the sub-database include: performing word segmentation on the sub-database to obtain respective language models and a first path score for each entity; constructing the finite state converter of the sub-database according to the language model and the first path score, wherein the modeling unit corresponding to the finite state converter is pinyin; obtaining the entity score of the finite state converter according to the first path score calculation; truncating the first path score according to a preset percentile to obtain the path score threshold of the finite state converter; obtaining the compensation value of the finite state converter according to the entity score; and obtaining the target state converter error correction model according to the names of the regions, the finite state converter, the entity score and the compensation value.

[0086] Word segmentation refers to the segmentation or decomposition of place names in the map entity database, usually according to linguistic or phonetic rules, for subsequent processing and modeling. Language model refers to a mathematical model used to describe language structure and rules, usually used to predict or evaluate the rationality or probability distribution of a text. The first path score refers to the initial score calculated for each place name entity according to predefined rules or models in the process of establishing a finite state converter model. Modeling unit refers to the basic unit used for description and modeling in the finite state converter, which can be a phonetic, character sequence or other linguistic unit. Entity score refers to a score or score calculated for each place name entity in the finite state converter according to its phonetic sequence or other characteristics. Preset percentile refers to a pre-set percentile value used to intercept the first path score, such as 70%, to determine the path score threshold method. Compensation value refers to an adjustment or correction value calculated based on the entity score or other measurement criteria, used to optimize the accuracy or adaptability of the path score. Region name refers to the name of each area or region divided in the map entity database, a unit used to segment and manage entity information. First, the map entity database is segmented. This step aims to extract the language model and initial first path score of each place name from the database, which will serve as the basis for subsequent steps. Then, the language model and first path score of each place name are used to build the corresponding finite state transducer (FST), in which pinyin is used as the basic unit of modeling, which helps to structure the modeling and analysis of place names. In the process of building the finite state transducer, the initial entity score in the finite state transducer is determined by calculating the first path score of each place name. Subsequently, the path score threshold of the finite state transducer is determined according to the preset percentile interception method, which determines which place names may be selected as the target of error correction. Furthermore, according to the entity score of each place name, the compensation value is calculated and applied. These compensation values ​​are used to adjust and optimize the error correction model of the finite state transducer, thereby improving the system's ability to handle various complex place names. Finally, the target state transducer error correction model is constructed by combining the specific names of each region, the corresponding finite state transducer, the entity score and the compensation value. This model will serve as the core of the final navigation entity error correction process, ensuring that the system can accurately and efficiently identify and correct place name errors in user voice input, thereby significantly improving the user experience and practicality of the in-vehicle navigation system.

[0087] As an example, the step of obtaining a target state converter error correction model according to each region name, the finite state converter, the entity score and the compensation value includes: obtaining a first sub-state converter according to the finite state converter, the entity score and the compensation value; obtaining a second sub-state converter according to each region name and the first sub-state converter; obtaining a third sub-state converter according to the region name; serially merging the first sub-state converter, the second sub-state converter and the third sub-state converter to obtain a region state converter; parallel-merging the region state converters to obtain a first-level state converter; parallel-merging the first-level state converters to obtain a second-level state converter; minimizing the second-level state converter to obtain a target state converter error correction model.

[0088] The first sub-state converter refers to the initial state converter obtained according to the finite state converter, entity score and compensation value. This sub-state converter forms an initial error correction model by combining the structured modeling of place names and the corresponding scoring information. The second sub-state converter is a sub-state converter obtained based on the names of each region and the first sub-state converter. It combines the language characteristics of the place names and the structure of the first sub-state converter, and is used to more finely process and correct specific speech recognition errors that may exist in different regions. The third sub-state converter is a state converter directly generated according to the specific region name. This sub-state converter is closely related to the speech input mode and place name characteristics of the specific region, which helps to further optimize and personalize the error correction model. The regional state converter is the result of serially merging the first sub-state converter, the second sub-state converter and the third sub-state converter in sequence. This merged state converter combines information at different levels, including general place name processing logic and customized processing for specific regions. Parallel merging refers to merging multiple state converters together in parallel. In this step, the serial combination of regional state converters will be combined in parallel with other regional state converters to form a more comprehensive processing framework. The first-level state converter is the result of further merging and optimizing the regional state converter. In this embodiment, it refers to the provincial state converter, which is used to handle place name recognition and error correction tasks within a specific province. The second-level state converter is a further combination of the first-level state converter. In this embodiment, it specifically refers to the national state converter, which is used to handle place name recognition and error correction needs across the country. Minimization refers to optimizing the state converter to reduce redundancy and complexity in the state converter, thereby improving the operating efficiency and response speed of the system. In the construction of the target state converter error correction model, this step ensures that the state converter can process the speech recognition results in the most efficient way and achieve accurate place name error correction function.

[0089] First, word segmentation is performed based on the place name entity database of each region, and the respective language models are generated, and the first path score of each entity is calculated. These steps help to establish the unique language characteristics and entity recognition basis of each region, and provide data support and basic scoring for the construction of the subsequent error correction model. Secondly, the finite state converter of each region is constructed based on the language model and the first path score. These finite state converters use pinyin as the modeling unit, map the speech recognition results and entity recognition paths to specific states and transitions, thereby forming a modeled description of the structure and semantics of the place name. Then, the path score threshold of each finite state converter is obtained by intercepting the first path score according to the preset percentile. This step helps to determine the reasonable recognition range of each state converter under different path distribution conditions, thereby improving the accuracy and robustness of the error correction model. After calculating the entity score, the compensation value of each finite state converter is calculated based on the entity score. These compensation values ​​play an important role in processing various place name input forms and speech recognition errors, and can provide additional correction and adjustment mechanisms in the error correction process. Subsequently, the second sub-state converter is generated according to each region name and the first sub-state converter. This step uses the region-specific language model and finite state converter structure to further refine and optimize the entity recognition and path selection of each region to adapt to different geographical environments and voice input conditions. Then, the third sub-state converter is directly generated according to the specific region name. This state converter directly reflects the voice recognition and place name processing requirements of the specific region and provides a higher level of place name error correction and recognition functions. Next, the first sub-state converter, the second sub-state converter and the third sub-state converter are serially merged in sequence to form a regional state converter. This merging process ensures the gradual advancement between different state converters and effectively integrates the multi-level information of place name recognition and error correction to improve the overall accuracy and efficiency. Subsequently, the regional state converters are paralleled and merged to form a first-level state converter. This parallel-merging process parallelizes the information of multiple regional state converters, provides a framework and support for simultaneous processing of multiple regions, and ensures the consistency and efficiency of the system between different regions. Finally, the first-level state converters are combined and merged to obtain the second-level state converter. This step further integrates the place name processing capabilities across the country and provides a unified model and basis for national place name correction and recognition. Finally, the second-level state converter is minimized to obtain the target state converter error correction model. This process improves the system's operating efficiency and response speed by reducing redundancy and optimizing the structure in the state converter, ensuring the stability and reliability of the place name recognition and error correction functions.

[0090] As an example, the step of obtaining the highest path score based on the current position of the vehicle, the entity to be corrected and the target state converter correction model includes: generating an input sequence of the target state converter correction model based on the entity to be corrected; inputting the input sequence into the target state converter correction model for decoding to obtain a preset number of paths and corresponding second path scores and regional labels; matching the current position of the vehicle with the regional label to obtain a matching path and a matching path score; replacing the matching path score of the matching path with the result obtained by adding the matching path score to the preset matching score; and sorting the second path scores from large to small to obtain the highest path score.

[0091] The input sequence refers to the input data sequence generated according to the entity to be corrected, which is usually converted from the speech recognition result and is used to be input into the target state converter correction model for processing and decoding. Decoding refers to converting and processing the input sequence in the target state converter correction model to generate a preset number of paths and calculate the score of each path. The preset number refers to the number of paths that need to be generated during the decoding process, which is usually pre-set by the system to ensure that multiple possible correct path options can be covered. The second path score refers to the score of each generated path, which is usually calculated based on the model's analysis and evaluation of the input sequence to determine the accuracy and selectability of the path. The regional label refers to the geographical location label or regional label corresponding to each path, which is used for subsequent matching and screening with the current location of the vehicle. The matching path refers to the path that matches the current location of the vehicle among the paths obtained after decoding, which is used to determine the final navigation destination. The matching path score refers to the score of the matching path, that is, the evaluation score of the path during the decoding process, which is usually adjusted and corrected according to the matching situation of the current location of the vehicle. The preset matching score refers to the score pre-set by the system for matching path score adjustment to ensure that the final selected navigation destination is reasonable and gives priority to the actual situation of the vehicle's current location. For example, assuming that the system presets a matching score of 10 points, when the system generates multiple paths based on the target state converter error correction model and matches one of the paths based on the vehicle's current location, if the original score of the path is 80 points, the system will adjust the final matching path score of this path to 90 points (i.e. 80 points plus 10 points). Such an adjustment can ensure that the system is more inclined to select the valid navigation path closest to the vehicle's current location as the final navigation destination.

[0092] As an example, the steps of generating an input sequence of the target state converter error correction model based on the entity to be corrected include: generating a penalty score set based on the pinyin modeling unit in the target state converter error correction model; obtaining the corresponding entity pinyin based on the entity to be corrected; and generating an input sequence of the target state converter error correction model based on the penalty score set and the entity pinyin.

[0093] The pinyin modeling unit refers to a set of all possible pinyin units, including homophones with different tones, similar sounds, easily confused sounds in the model, and irrelevant pinyin. These pinyin units are used to construct the input sequence of the target state converter error correction model. The penalty score set is a score set set for each pinyin unit in the pinyin modeling unit, which is used to distinguish the target pinyin from other related or irrelevant pinyins. Generally speaking, the score of the target pinyin is 1, the score of other related pinyins is 0.9 (i.e., the penalty score), and the score of irrelevant pinyin is 0. Entity pinyin refers to the pinyin sequence extracted according to the entity to be corrected. These pinyin sequences will be used as input sequences for the generation of the target state converter error correction model. The input sequence refers to the data sequence used to input into the target state converter error correction model, usually a sequence containing the entity information to be corrected. Specifically, the input sequence can be a pinyin sequence, a text sequence, or other forms of data sequence, which is used for the model to recognize, decode and correct errors. For example, if the entity to be corrected is a place name, such as "Chaoyang District, Beijing", its corresponding input sequence can be its pinyin sequence: "bei jing shi chao yang qu". This pinyin sequence will be used as input for the target state converter error correction model to process and correct errors, so as to improve the accuracy of speech recognition and the correctness of navigation instructions.

[0094] Step S32, comparing the highest path score with the path score threshold to obtain a comparison result;

[0095] It should be noted that the "comparison result" refers to a judgment or conclusion obtained after comparing the highest path score with the path score threshold. Specifically, the comparison result can be a Boolean value (such as true or false), a numerical value (such as greater than, less than, equal to) or other forms, which is used to determine whether the set path score requirement or condition is met.

[0096] It is understandable that the calculated highest path score needs to be compared with the pre-set path score threshold first. The purpose of the comparison is to evaluate whether the highest path score meets or exceeds the set standard. If the highest path score meets or exceeds the path score threshold, the comparison result will be true (True), indicating that the set requirements have been met and subsequent steps can be taken, such as selecting the path as the navigation destination. On the contrary, if the highest path score does not reach the path score threshold, the comparison result will be false (False). At this time, other paths may need to be considered or error correction measures may be taken to improve the accuracy and applicability of the path. This comparison process ensures that during the navigation entity error correction process, the selected path can meet the pre-set quality standards and user expectations.

[0097] Step S33, when the comparison result is greater than, taking the speech recognition result as the final navigation result;

[0098] It should be noted that the final navigation result refers to the optimal path or destination obtained based on the voice recognition input and error correction algorithm after being processed by the entire navigation entity error correction system.

[0099] It is understood that when the comparison result is greater than, the step of using the speech recognition result as the final navigation result involves the system determining that the navigation destination after error correction has reached the expected accuracy and usability standards, which means that after the speech recognition system has identified and corrected the user's navigation destination through the aforementioned steps, the final result is determined to be sufficiently reliable and accurate. Such a determination ensures that in the in-vehicle navigation system, the navigation destination obtained by the user through voice input is the final choice that has been optimized and verified by the system, thereby improving the practicality and user experience of the overall navigation system.

[0100] Step S34, when the comparison result is less than or equal to, taking the path corresponding to the highest path score as the final navigation result, and completing navigation entity error correction.

[0101] It can be understood that the system determines the starting point of navigation based on the actual position of the current vehicle. Then, based on the target state converter error correction model generated in the previous step, the system evaluates the accuracy of each possible navigation destination in the speech recognition result and selects the entity with the highest path score as the final navigation destination. The path score here refers to the rationality evaluation score of each navigation entity relative to the current vehicle position with the help of the target state converter error correction model. In this way, the system can select the best navigation entity in the case of possible errors in speech recognition, thereby improving the user's navigation experience and accuracy.

[0102] In this embodiment, a path score threshold and a maximum path score are obtained according to a map entity database; the maximum path score is compared with the path score threshold to obtain a comparison result; when the comparison result is greater than, the speech recognition result is used as the final navigation result; when the comparison result is less than or equal to, the path corresponding to the maximum path score is used as the final navigation result to complete navigation entity error correction. In this embodiment, the path score of each navigation entity is first calculated according to the map entity database, which reflects its rationality and reliability as a navigation destination, and helps the system provide a basis when selecting a navigation destination; then the calculated maximum path score is compared with a preset path score threshold, which determines the next processing flow. If the maximum path score exceeds the threshold, the system will directly use the speech recognition result as the final navigation destination, thereby improving the system's automatic confirmation ability for the accuracy of voice input and improving user experience; conversely, if the maximum path score does not reach the threshold, the system will select the navigation entity with the highest path score as the final navigation destination. This automatic error correction mechanism effectively prevents navigation deviations caused by speech recognition errors, ensures that users can reach the most suitable destination, not only improves the accuracy and response speed of the system, but also simplifies user operations, and improves overall convenience and user satisfaction.

[0103] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the error correction method of the speech recognition navigation domain entity of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0104] This application also provides a speech recognition navigation domain entity error correction device, please refer to Figure 4 , the error correction device of the speech recognition navigation domain entity includes:

[0105] A data acquisition module 10 is used to obtain speech recognition results;

[0106] An extraction module 20, used to extract the domain and the entity to be corrected of the speech recognition result;

[0107] The error correction module 30 is used to correct the entity to be corrected according to the map entity database to complete navigation entity error correction when the field belongs to the navigation field and the number of words in the entity to be corrected is greater than or equal to a preset number of words.

[0108] The error correction device for speech recognition navigation domain entity provided by the present application adopts the error correction method for speech recognition navigation domain entity in the above embodiment, which can solve the technical problem of inaccurate entity recognition by speech recognition in the field of vehicle navigation. Compared with the prior art, the beneficial effects of the error correction device for speech recognition navigation domain entity provided by the present application are the same as the beneficial effects of the error correction method for speech recognition navigation domain entity provided by the above embodiment, and the other technical features in the error correction device for speech recognition navigation domain entity are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0109] The present application provides an error correction device for a speech recognition navigation domain entity, the error correction device for a speech recognition navigation domain entity comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the error correction method for the speech recognition navigation domain entity in the above-mentioned embodiment one.

[0110] Reference below Figure 5 , which shows a schematic diagram of the structure of the error correction device of the speech recognition navigation domain entity suitable for implementing the embodiment of the present application. The error correction device of the speech recognition navigation domain entity in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The error correction device of the speech recognition navigation domain entity shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0111] like Figure 5As shown, the error correction device of the speech recognition navigation domain entity may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the error correction device of the speech recognition navigation domain entity are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the error correction device of the speech recognition navigation domain entity to communicate wirelessly or wired with other devices to exchange data. Although the error correction device of the speech recognition navigation domain entity with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0112] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0113] The error correction device for speech recognition navigation domain entity provided by the present application adopts the error correction method for speech recognition navigation domain entity in the above embodiment, which can solve the technical problem of inaccurate entity recognition by speech recognition in the field of vehicle navigation. Compared with the prior art, the beneficial effects of the error correction device for speech recognition navigation domain entity provided by the present application are the same as the beneficial effects of the error correction method for speech recognition navigation domain entity provided by the above embodiment, and the other technical features in the error correction device for speech recognition navigation domain entity are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0114] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0115] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0116] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the error correction method for speech recognition navigation domain entities in the above-mentioned embodiment.

[0117] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0118] The computer-readable storage medium may be included in the error correction device of the speech recognition navigation domain entity; or may exist independently without being assembled into the error correction device of the speech recognition navigation domain entity.

[0119] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the error correction device of the speech recognition navigation domain entity, the error correction device of the speech recognition navigation domain entity: obtains the speech recognition result; extracts the domain and the entity to be corrected of the speech recognition result; when the domain belongs to the navigation domain and the number of words of the entity to be corrected is greater than or equal to the preset number of words, corrects the entity to be corrected according to the map entity database to complete the navigation entity correction.

[0120] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0121] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0122] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0123] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned error correction method for speech recognition navigation domain entities, and can solve the technical problem of inaccurate entity recognition by speech recognition in the field of vehicle navigation. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the error correction method for speech recognition navigation domain entities provided by the above-mentioned embodiment, and will not be repeated here.

[0124] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned method for correcting errors in speech recognition navigation domain entities when executed by a processor.

[0125] The computer program product provided by the present application can solve the technical problem of inaccurate entity recognition in the field of vehicle navigation by voice recognition. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the error correction method of voice recognition navigation domain entity provided by the above embodiment, which will not be described in detail here.

[0126] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for correcting errors in speech recognition navigation domain entities, characterized in that: The method comprises: Get the speech recognition results; Extracting the domain and the entity to be corrected of the speech recognition result; When the field belongs to the navigation field and the number of characters of the entity to be corrected is greater than or equal to a preset number of characters, the entity to be corrected is corrected according to the map entity database to complete navigation entity correction.

2. The method according to claim 1, characterized in that The step of correcting the entity to be corrected according to the map entity database to complete the navigation entity correction comprises: Obtaining a path score threshold and a maximum path score according to a map entity database; Comparing the highest path score with the path score threshold to obtain a comparison result; When the comparison result is greater than, taking the speech recognition result as the final navigation result; When the comparison result is less than or equal to, the path corresponding to the highest path score is used as the final navigation result to complete navigation entity error correction.

3. The method according to claim 2, characterized in that The step of obtaining the path score threshold and the maximum path score according to the map entity database comprises: Divide the map entity database into sub-databases by region; According to the sub-database, a finite state converter of the sub-database, a path score threshold of the finite state converter, and a target state converter error correction model are obtained; The highest path score is obtained according to the current position of the vehicle, the entity to be corrected and the target state converter correction model.

4. The method according to claim 3, characterized in that The step of obtaining the finite state converter of the sub-database, the path score threshold of the finite state converter and the target state converter error correction model according to the sub-database comprises: Segmenting the sub-database to obtain respective language models and a first path score for each entity; Constructing a finite state converter of the sub-database according to the language model and the first path score, wherein the modeling unit corresponding to the finite state converter is pinyin; Obtaining an entity score of the finite state converter according to the first path score calculation; Cutting the first path score according to a preset percentile to obtain a path score threshold of the finite state converter; Obtaining a compensation value of the finite state converter according to the entity score; A target state switch error correction model is obtained according to each region name, the finite state switch, the entity score and the compensation value.

5. The method according to claim 4, characterized in that The step of obtaining a target state converter error correction model according to each region name, the finite state converter, the entity score and the compensation value comprises: Obtaining a first sub-state converter according to the finite state converter, the entity score and the compensation value; Obtain a second sub-state converter according to each region name and the first sub-state converter; Obtain a third sub-state converter according to the region name; The first sub-state converter, the second sub-state converter and the third sub-state converter are serially combined to obtain a regional state converter; The regional state converters are combined and merged to obtain a primary state converter; The first-level state converters are combined in parallel to obtain a second-level state converter; The secondary state converter is minimized to obtain a target state converter error correction model.

6. The method according to claim 3, characterized in that The step of obtaining the highest path score according to the current position of the vehicle, the entity to be corrected and the target state converter correction model comprises: Generate an input sequence of the target state converter error correction model according to the entity to be corrected; Inputting the input sequence into the target state converter error correction model for decoding to obtain a preset number of paths and corresponding second path scores and regional labels; Matching the current position of the vehicle with the region label to obtain a matching path and a matching path score; The result obtained by adding the matching path score to the preset matching score replaces the matching path score of the matching path; The second path scores are sorted from largest to smallest to obtain the highest path score.

7. The method according to claim 6, characterized in that The step of generating an input sequence of the target state converter error correction model according to the entity to be corrected comprises: Generate a penalty score set according to the phonetic modeling unit in the target state converter error correction model; Obtaining the corresponding entity pinyin according to the entity to be corrected; An input sequence of the target state converter error correction model is generated according to the penalty score set and the entity pinyin.

8. A speech recognition navigation domain entity error correction device, characterized in that: The device comprises: A data acquisition module, used to obtain speech recognition results; An extraction module, used to extract the domain and the entity to be corrected of the speech recognition result; The error correction module is used to correct the entity to be corrected according to the map entity database to complete navigation entity error correction when the field belongs to the navigation field and the number of words in the entity to be corrected is greater than or equal to a preset number of words.

9. An error correction device for speech recognition navigation domain entity, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the error correction method for speech recognition navigation domain entities as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the method for correcting errors of speech recognition navigation domain entities according to any one of claims 1 to 7 are implemented.