Method, device and equipment for determining reply information in voice navigation service system

Abstract

The invention discloses a method, a device and equipment for determining reply information in a voice navigation service system, and relates to the technical field of data processing. The method comprises the following steps: acquiring user question information from a voice navigation service system; respectively determining the similarity between the user question information and multiple pieces of preset standard question information; under the condition that the similarity is smaller than a first preset value, multiple pieces of candidate standard question information with the similarity with the user question information larger than a second preset value are selected; inputting the user question information and the multiple pieces of candidate standard question information into an information matching model, and determining the candidate standard question information matched with the user question information as target standard question information by utilizing a matching relationship between preset user question information in the information matching model and multiple pieces of preset standard question information; and determining reply information corresponding to the target standard question information by using the target result determination mode, and taking the reply information as reply information of the user question information. And the determination accuracy of the reply information is improved.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and in particular, relates to a method, device and equipment for determining reply information in a voice navigation service system. Background Art

[0002] With the development of artificial intelligence technology, voice navigation service systems can interact with users in natural language. In the voice navigation scenario, users interact with the voice navigation service system to obtain responses to the questions they inquired about. In order to improve the user experience, it is crucial to accurately determine the responses to the questions users inquired about.

[0003] In the prior art, the similarity between the user's question information and the standard question information is usually calculated. If the similarity is higher than a certain threshold, the standard question information is considered to be hit, and then the reply information corresponding to the standard question information is returned to the user. This method of determining the reply information can only answer some user question information with a fixed expression form and similar to the standard question information. For user question information with richer expression forms and vague content, the determined standard question information similar to the user question information will be inaccurate, and the final reply information will also be inaccurate. Summary of the invention

[0004] The embodiment of the present application provides a method, device and equipment for determining reply information in a voice navigation service system. First, the method improves the accuracy of reply information by setting two similarity thresholds, that is, when the maximum similarity between user question information and multiple preset standard question information is less than a first preset value, first select multiple candidate standard questions with a similarity greater than a second preset value, and then use an information matching model to select the most matching standard question information from the multiple candidate standard questions, thereby improving the accuracy of standard question information selection, and further improving the accuracy of reply information determination; secondly, set different result determination methods, and use the target result determination method corresponding to the target standard question information to determine the reply information, so that the method for determining the reply information is as close to the target standard question information as possible, which can also improve the accuracy of reply information determination.

[0005] In a first aspect, an embodiment of the present application provides a method for determining reply information in a voice navigation service system, comprising:

[0006] Obtain user question information from the voice navigation service system;

[0007] Determine the similarity between the user question information and multiple preset standard question information respectively;

[0008] When the largest target similarity among the similarities is less than a first preset value, selecting from the plurality of preset standard question information, a plurality of candidate standard question information whose similarity to the user question information is greater than a second preset value, the second preset value being less than the first preset value;

[0009] Inputting the user question information and the plurality of candidate standard question information into the information matching model, and using the matching relationship between the preset user question information and the plurality of preset standard question information in the information matching model, determining the candidate standard question information matching the user question information as the target standard question information;

[0010] Obtain the target result determination method corresponding to the target standard information;

[0011] The target result determination method is used to determine the reply information corresponding to the target standard question information, and the reply information is used as the reply information of the user question information.

[0012] In a possible implementation, obtaining user question information from the voice navigation service system includes:

[0013] Acquire user voice information from the voice navigation service system;

[0014] Convert user voice information into business text information, and use the business text information as user question information.

[0015] In a possible embodiment, it also includes:

[0016] When the target similarity is not less than the first preset value, it is determined that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

[0017] In a possible embodiment, it also includes:

[0018] When all the similarities are smaller than the second preset value, the preset reply information is determined as the reply information of the user's question information.

[0019] In a possible implementation, the response information corresponding to the target standard question information is determined by using the target result determination method, including:

[0020] When the target result determination method is the first result determination method, the reply information corresponding to the target standard question information is obtained from the reply information database.

[0021] In a possible implementation, the response information corresponding to the target standard question information is determined by using the target result determination method, including:

[0022] When the target result determination method is the second result determination method, query the result information corresponding to the target standard information;

[0023] Combine the result information with the preset speech template corresponding to the target standard question to obtain the reply information corresponding to the target standard question information.

[0024] In a possible implementation, after determining the reply information corresponding to the target standard question information by using the target result determination method and using the reply information as the reply information of the user question information, the method further includes:

[0025] Convert reply messages into reply voice messages;

[0026] Use the voice navigation service system to play reply voice messages.

[0027] In a possible implementation, before inputting the user question information and the plurality of candidate standard question information into the information matching model, the method further includes:

[0028] Obtaining a user question information sample, actual standard question information matching the user question information sample, and multiple candidate standard question information samples;

[0029] Inputting the user question information sample and multiple candidate standard question information samples into the initial information matching model, and using the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model, determining the candidate standard question information sample matching the user question information sample as the predicted standard question information;

[0030] Determine the error value of the initial information matching model based on the actual standard question information and the predicted standard question information;

[0031] When the error value does not meet the training stop condition, the initial matching relationship between the preset user question information and multiple preset standard question information in the initial information matching model is updated to obtain an updated initial matching relationship. The updated initial matching relationship is used to update the predicted standard question information that matches the user question information sample. The error value is updated according to the actual standard question information and the updated predicted standard question information until the error value meets the training stop condition to obtain the information matching model.

[0032] In a second aspect, an embodiment of the present application provides a device for determining reply information in a voice navigation service system, including:

[0033] An acquisition module is used to obtain user question information from the voice navigation service system;

[0034] A determination module, used to respectively determine the similarity between the user question information and a plurality of preset standard question information;

[0035] A selection module, configured to select, when the largest target similarity among the similarities is less than a first preset value, a plurality of candidate standard question information having a similarity with the user question information greater than a second preset value from the plurality of preset standard question information, wherein the second preset value is less than the first preset value;

[0036] The determination module is further used to input the user question information and the multiple candidate standard question information into the information matching model, and determine the candidate standard question information matching the user question information as the target standard question information by using the matching relationship between the preset user question information and the multiple preset standard question information in the information matching model;

[0037] The acquisition module is also used to obtain the target result determination method corresponding to the target standard information;

[0038] The determination module is also used to determine the reply information corresponding to the target standard question information by using the target result determination method, and use the reply information as the reply information of the user question information.

[0039] In a possible implementation, the acquisition module is specifically used for:

[0040] Acquire user voice information from the voice navigation service system;

[0041] Convert user voice information into business text information, and use the business text information as user question information.

[0042] In a possible embodiment, the device further includes:

[0043] The determination module is further configured to determine, when the target similarity is not less than a first preset value, that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

[0044] In a possible embodiment, the device further includes:

[0045] The determination module is further used to determine the preset reply information as the reply information of the user's question information when all the similarities are less than the second preset value.

[0046] In one possible implementation, the determination module is specifically used for:

[0047] When the target result determination method is the first result determination method, the reply information corresponding to the target standard question information is obtained from the reply information database.

[0048] In one possible implementation, the determination module is specifically used for:

[0049] When the target result determination method is the second result determination method, query the result information corresponding to the target standard information;

[0050] Combine the result information with the preset speech template corresponding to the target standard question to obtain the reply information corresponding to the target standard question information.

[0051] In a possible embodiment, the device further includes:

[0052] A conversion module, for determining the reply information corresponding to the target standard question information by using the target result determination method, and converting the reply information into a reply voice information after using the reply information as the reply information of the user's question information;

[0053] The playing module is used to play the reply voice message using the voice navigation service system.

[0054] In a possible embodiment, the device further includes:

[0055] The acquisition module is further used to acquire a user question information sample, actual standard question information matching the user question information sample, and a plurality of candidate standard question information samples before inputting the user question information and the plurality of candidate standard question information into the information matching model;

[0056] The determination module is further used to input the user question information sample and the multiple candidate standard question information samples into the initial information matching model, and use the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model to determine the candidate standard question information sample that matches the user question information sample as the predicted standard question information;

[0057] The determination module is also used to determine the error value of the initial information matching model according to the actual standard question information and the predicted standard question information;

[0058] The determination module is also used to update the initial matching relationship between the preset user question information and multiple preset standard question information in the initial information matching model when the error value does not meet the training stop condition, obtain the updated initial matching relationship, use the updated initial matching relationship to update the predicted standard question information that matches the user question information sample, and update the error value according to the actual standard question information and the updated predicted standard question information until the error value meets the training stop condition to obtain the information matching model.

[0059] In a third aspect, an embodiment of the present application provides an electronic device, the device comprising:

[0060] a processor and a memory storing computer program instructions;

[0061] When the processor executes the computer program instructions, the method for determining the reply information in the voice navigation service system according to any one of the above items is implemented.

[0062] In a fourth aspect, an embodiment of the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, any one of the above-mentioned methods for determining reply information in a voice navigation service system is implemented.

[0063] In a fifth aspect, an embodiment of the present application provides a computer program product, characterized in that when the instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute any one of the above-mentioned methods for determining reply information in a voice navigation service system.

[0064] The method, device, equipment, medium and product for determining reply information in a voice navigation service system of an embodiment of the present application include: obtaining user question information from a voice navigation service system; respectively determining the similarity between the user question information and a plurality of preset standard question information; when the largest target similarity among the similarities is less than a first preset value, selecting a plurality of candidate standard question information whose similarity to the user question information is greater than a second preset value from the plurality of preset standard question information, and the second preset value is less than the first preset value; inputting the user question information and the plurality of candidate standard question information into an information matching model, and determining the candidate standard question information matching the user question information as the target standard question information by using the matching relationship between the preset user question information and the plurality of preset standard question information in the information matching model; obtaining a target result determination method corresponding to the target standard question information; determining the reply information corresponding to the target standard question information by using the target result determination method, and using the reply information as the reply information to the user question information.

[0065] In this way, by setting two similarity thresholds, the accuracy of the reply information is improved, that is, when the maximum similarity between the user question information and multiple preset standard question information is less than the first preset value, first select multiple candidate standard questions with a similarity greater than the second preset value, and then use the information matching model to select the most matching standard question information from the multiple candidate standard questions, thereby improving the accuracy of the standard question information selection, and then improving the accuracy of the reply information determination; secondly, set different result determination methods, and use the target result determination method corresponding to the target standard question information to determine the reply information, so that the method of determining the reply information is as close as possible to the target standard question information, which can also improve the accuracy of the reply information determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0067] Figure 1It is a flowchart of a method for determining reply information in a voice navigation service system provided by an embodiment of the present application;

[0068] Figure 2 is a flow chart of a method for determining reply information in a voice navigation service system provided in another embodiment of the present application;

[0069] Figure 3 is a flowchart of a method for determining reply information in a voice navigation service system provided by another embodiment of the present application;

[0070] Figure 4 is a flowchart of a method for determining reply information in a voice navigation service system provided in yet another embodiment of the present application;

[0071] Figure 5 is a flowchart of a method for determining reply information in a voice navigation service system provided in yet another embodiment of the present application;

[0072] Figure 6 is a structural diagram of a device for determining reply information in a voice navigation service system provided in yet another embodiment of the present application;

[0073] Figure 7 It is a structural diagram of an electronic device provided in yet another embodiment of the present application. DETAILED DESCRIPTION

[0074] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, 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 and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0075] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0076] It should be noted that the acquisition, storage, use and processing of data in the embodiments of the present application are in compliance with the relevant provisions of national laws and regulations.

[0077] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0078] With the development of artificial intelligence technology, voice navigation service systems can interact with users in natural language. In the voice navigation scenario, users interact with the voice navigation service system to obtain responses to the questions they inquired about. In order to improve the user experience, it is crucial to accurately determine the responses to the questions users inquired about.

[0079] In the prior art, the similarity between the user question information and the standard question information is usually calculated. If the similarity is higher than a certain threshold, the standard question information is considered to be hit, and then the reply information corresponding to the standard question information is returned to the user. This method of determining the reply information can only answer some user question information with a fixed expression form and similar to the standard question information. For user question information with richer expression forms and vaguer content, such as the user question information "Why is my card frozen?" and the standard question information "The card is frozen", there is a high similarity. If it is "My card cannot be used", "The card is frozen and the money is urgently needed but cannot be withdrawn", etc., the user question information is very different from the standard question information in expression form and has a low similarity, it is easy to fail to match. At this time, the standard question information similar to the user question information determined will be inaccurate, and the reply information finally obtained will also be inaccurate.

[0080] In order to solve the problems of the prior art, the embodiments of the present application provide a method, device and equipment for determining reply information in a voice navigation service system. The method for determining reply information in a voice navigation service system provided by the embodiments of the present application can be applied to a question-and-answer system, and the accuracy of reply information is improved by setting two similarity thresholds, that is, when the maximum similarity between user question information and multiple preset standard question information is less than a first preset value, multiple candidate standard questions with a similarity greater than a second preset value are first selected, and then the information matching model is used to select the most matching standard question information from the multiple candidate standard questions, thereby improving the accuracy of standard question information selection, and further improving the accuracy of reply information determination; secondly, different result determination methods are set, and the reply information is determined using the target result determination method corresponding to the target standard question information, so that the method for determining the reply information is as close to the target standard question information as possible, which can also improve the accuracy of reply information determination.

[0081] The following is an introduction to the method for determining reply information in the voice navigation service system provided in the embodiment of the present application. Figure 1 As shown, the method for determining reply information in the voice navigation service system provided in the embodiment of the present application includes the following steps.

[0082] S110: Obtain user question information from the voice navigation service system.

[0083] Here, the voice navigation service system can interact with the user in natural language, and the user question information is the question information that the user wants to consult, such as checking the balance, etc. As an example, the user question information may include but is not limited to "I don't want to cancel my account, I want to check the balance", "Why is my card frozen", "My card cannot be used", "I don't want to check the account first, I want to check the balance", etc.

[0084] In some embodiments, the voice navigation service system can be applied to multiple scenarios, such as finance, communications, etc.

[0085] S120: Determine the similarity between the user question information and a plurality of preset standard question information respectively.

[0086] Here, the plurality of preset standard question information may be pre-set, for example, the plurality of preset standard question information is standard question information obtained from a voice navigation standard question information library.

[0087] In some embodiments, the similarity includes vector similarity. First, the user question information and a plurality of preset standard question information are vector-converted to obtain a first vector corresponding to the user question information and a second vector corresponding to each preset standard question information. The similarity of the first vector and each second vector is determined respectively.

[0088] As an example, the preset standard question information may include, but is not limited to, "how much money is left", "check outlets", "check accounts", etc.

[0089] S130. When the largest target similarity among the similarities is less than a first preset value, select multiple candidate standard question information from multiple preset standard question information, whose similarity to the user question information is greater than a second preset value, where the second preset value is less than the first preset value.

[0090] Here, the first preset value is a higher similarity threshold value set in advance, such as 0.9. The second preset value is a relatively higher similarity threshold value set in advance, but is smaller than the first preset value, such as 0.6. It can be understood that the second preset value is a threshold value that is appropriately lowered, and multiple standard question information with a certain degree of similarity is retained as candidates. When facing a complex expression of the user, the possibility of hitting the correct standard question information with low similarity is retained.

[0091] S140: Input the user question information and multiple candidate standard question information into the information matching model, and use the matching relationship between the preset user question information and the multiple preset standard question information in the information matching model to determine the candidate standard question information matching the user question information as the target standard question information.

[0092] Here, the information matching model is a model trained in advance. The preset user question information includes the user question information obtained from the voice navigation service system. The plurality of preset standard question information includes the target standard question information.

[0093] In some embodiments, inputting the user question information and the plurality of candidate standard question information into the information matching model includes: constructing prompt information using the user question information and the plurality of candidate standard question information, and inputting the prompt information into the information matching model.

[0094] As an example, the prompt information is: "Please select the standard question that best matches the user question information from the given candidate standard question information; user question information: XX; candidate standard question information: standard question information 1; standard question information 2; standard question information 3".

[0095] In some embodiments, the information matching model can be a large model. The large model is essentially a type of statistical mathematical model that converts the input text into a mathematical form, performs calculations, and finally generates a result. This process can be expressed as formula (1), which is as follows:

[0096] y=φ(embedding(x)) (1)

[0097] Among them, x is the input prompt information, embedding means converting the input into mathematical form, φ represents the calculation process of the large model, including the parameters of the large model, and y represents the output target standard information.

[0098] The large model can be a ChatGLM model, a model with large-scale parameters and complex computing structure, which can handle various complex natural language tasks. It continuously updates the model parameters through the three-stage training process of pre-training, instruction fine-tuning and reinforced feedback training, and finally enables the model to have dialogue capabilities. Although the large model has certain dialogue capabilities, in specific downstream tasks, it is necessary to design appropriate prompt information to allow the large model to output ideal results. Here, the task that the large model needs to complete is to find the one that best matches the user's question information from multiple candidate standard question information and output it, and use prompt engineering to let the large model complete the standard question selection work. For example, the candidate standard question information and the user's question information are assembled into prompt information and input to the large model. The prompt information is: "Please select the standard question that best matches the user's question information from the given candidate standard question information; user question information: I don't want to cancel the account, I want to check the balance; candidate standard question information: check balance; cancel; check account", and the input result is "check balance".

[0099] S150: Obtain a target result determination method corresponding to the target standard information.

[0100] In some embodiments, the target result determination method corresponding to the target standard information is acquired according to the relationship information between the preset standard information and the preset result determination method. Here, the preset standard information includes the target standard information, and the preset result determination method includes the target result determination method.

[0101] S160: Determine the reply information corresponding to the target standard question information by using the target result determination method, and use the reply information as the reply information of the user question information.

[0102] It can be understood that, since the target standard question information is the standard question information that best matches the user question information, the reply information corresponding to the target standard question information is used as the reply information of the user question information.

[0103] In this way, by setting two similarity thresholds, the accuracy of the reply information is improved, that is, when the maximum similarity between the user question information and multiple preset standard question information is less than the first preset value, first select multiple candidate standard questions with a similarity greater than the second preset value, and then use the information matching model to select the most matching standard question information from the multiple candidate standard questions, thereby improving the accuracy of the standard question information selection, and then improving the accuracy of the reply information determination; secondly, set different result determination methods, and use the target result determination method corresponding to the target standard question information to determine the reply information, so that the method of determining the reply information is as close as possible to the target standard question information, which can also improve the accuracy of the reply information determination.

[0104] Based on this, in some embodiments, Figure 2 As shown, the above S110 may specifically include:

[0105] S111, obtaining user voice information from the voice navigation service system;

[0106] S112: Convert the user voice information into business text information, and use the business text information as user question information.

[0107] In some embodiments, the user dials a phone to access the voice navigation service system and starts to interact with the voice navigation service system. The voice navigation service system can obtain user voice information through a voice recognition module, obtain user voice information from the voice navigation service system, convert the user voice information into service text information, and use the service text information as user question information.

[0108] In some embodiments, the user voice information is question information with a question-and-answer intention, which can be identified by a conversation management module. The specific implementation method is not specifically limited here.

[0109] In this way, no matter the user's question information is in voice or text form, the reply information can be determined. If it is a voice message, the reply information can also be determined after being converted into text, which is convenient for users and improves the user experience.

[0110] Based on this, in some embodiments, corresponding to the above S130, the method may further include:

[0111] When the target similarity is not less than the first preset value, it is determined that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

[0112] It is understandable that, since the first preset value is a high similarity threshold value set in advance, if the target similarity is not less than the first preset value, it means that the preset standard question information corresponding to the target similarity is already very similar to the user question information. At this time, the preset standard question information corresponding to the target similarity is directly determined as the target standard question information matching the user question information, the target result determination method corresponding to the target standard question information is obtained, the response information corresponding to the target standard question information is determined using the target result determination method, and the response information is used as the response information of the user question information.

[0113] In this way, if the preset standard question information corresponding to the target similarity is already very similar to the user question information, then there is no need to use the information matching model to match the standard question information and the most matching standard question information can be obtained. While ensuring the accuracy of the standard question information selection, it can reduce the waste of computing resources and improve the efficiency of determining the reply information.

[0114] Based on this, in some embodiments, the method may further include:

[0115] When all the similarities are smaller than the second preset value, the preset reply information is determined as the reply information of the user's question information.

[0116] Here, the preset reply message is set in advance. For example, the preset reply message is "no reply message can be obtained".

[0117] In some embodiments, since the second preset value is a threshold that has been appropriately lowered, that is, the similarity threshold has been lowered, the similarity between the user question information and multiple preset standard question information still cannot exceed the threshold, indicating that the standard question information matching the user question information cannot be hit, and the preset reply information is determined to be the reply information of the user question information.

[0118] In this way, when the standard question information that matches the user's question information cannot be determined, the preset reply information is returned to the user, so that the user knows that the question cannot be answered, and the user does not need to wait too much, thereby improving the user experience.

[0119] Based on this, in some embodiments, Figure 3 As shown, in the above S160, the response information corresponding to the target standard question information is determined by using the target result determination method, which may include:

[0120] S210: When the target result determination method is the first result determination method, obtain response information corresponding to the target standard question information from a response information database.

[0121] Here, the first result determination method represents that there is fixed reply information for the target standard question information.

[0122] In some embodiments, the reply information database stores "standard question-reply" question-answer pairs with fixed reply information in advance. When the target result determination method is the first result determination method, the "standard question-reply" question-answer pair corresponding to the target standard question information in the reply information database is located, and the reply information of the question-answer pair is obtained.

[0123] In this way, if there is fixed reply information for the target standard question information, the reply information is directly returned to the user, which ensures the accuracy of the reply information and improves the efficiency of determining the reply information.

[0124] Based on this, in some embodiments, Figure 4 As shown, in the above S160, the response information corresponding to the target standard question information is determined by using the target result determination method, which may include:

[0125] S220: When the target result determination method is the second result determination method, query result information corresponding to the target standard information;

[0126] S230. Combine the result information and the preset speech template corresponding to the target standard question to obtain reply information corresponding to the target standard question information.

[0127] Here, the second result determination method represents that there is no fixed reply information for the target standard question information, and it is necessary to query in combination with the actual situation.

[0128] In some embodiments, when the target result determination method is the second result determination method, the result information corresponding to the target standard question information is queried. For example, if the target standard question information is "check balance", the result information corresponding to the target standard question information is queried to obtain the result information, which is 6. The result information is combined with the preset speech template corresponding to the target standard question to obtain the reply information corresponding to the target standard question information.

[0129] It is understandable that, as in the above example, "6" cannot be directly returned to the user. It is necessary to combine "6" with the preset speech template corresponding to the target standard question and then return it to the user. For example, "Your balance is 6 yuan"

[0130] In this way, accurate responses can also be obtained for some user questions that need to be answered based on actual conditions. For example, for user questions related to time, fixed response information cannot be used. It is necessary to query the actual result information and combine the preset speech templates to obtain the response information, thereby improving the accuracy of the response information.

[0131] Based on this, in some embodiments, Figure 5 As shown, after the above S160, the method may further include:

[0132] S170, converting the reply message into a reply voice message;

[0133] S180: Use the voice navigation service system to play a reply voice message.

[0134] In some embodiments, the reply message is synthesized into a voice message and then broadcast.

[0135] In this way, the user does not need to open the display page to view the reply message or play the reply voice message, which improves the user experience.

[0136] Based on this, in some embodiments, before the above S140, the method may further include:

[0137] Obtaining a user question information sample, actual standard question information matching the user question information sample, and multiple candidate standard question information samples;

[0138] Inputting the user question information sample and multiple candidate standard question information samples into the initial information matching model, and using the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model, determining the candidate standard question information sample matching the user question information sample as the predicted standard question information;

[0139] Determine the error value of the initial information matching model based on the actual standard question information and the predicted standard question information;

[0140] When the error value does not meet the training stop condition, the initial matching relationship between the preset user question information and multiple preset standard question information in the initial information matching model is updated to obtain an updated initial matching relationship. The updated initial matching relationship is used to update the predicted standard question information that matches the user question information sample. The error value is updated according to the actual standard question information and the updated predicted standard question information until the error value meets the training stop condition to obtain the information matching model.

[0141] In this way, the information matching model is trained through samples in advance, so that the target standard question information determined by the information matching model is more accurate, thereby improving the accuracy of the reply information determination.

[0142] The embodiments provided in this application can be applied to the question-answering system to enhance the ability of the question-answering system to retrieve difficult samples and improve the overall performance. The similarity threshold judgment is increased, and a large model is introduced to determine the relevance between the preset standard question information and the user question information, so as to select the target standard question information that best matches the user question information from the candidate standard question information. In the process of the large model determining the relevance between the preset standard question information and the user question information, the prompt engineering technology is used to enable the large model to better complete the selection of the standard question.

[0143] The embodiment provided in the present application can effectively improve the answer rate of the question-answering system by lowering the similarity threshold, screening out some potential candidate standard question information, and then using a large model for accurate judgment. With the help of the excellent language understanding ability of the large model to make judgments, accurately identify the user's intentions, and well solve the problem of inaccurate determination of reply information for user question information with richer expressions and vague content, effectively enhancing the question-answering ability of the question-answering system. Design prompt information so that the large model can effectively complete downstream tasks, including correctly identifying standard question information that matches user question information. Through the design of the threshold range, both effect and efficiency can be taken into account. For example, some high-frequency and large-volume problems such as "checking phone bills" can be solved quickly, and large models can also be used to solve some long-tail problems.

[0144] Based on the method for determining reply information in the voice navigation service system provided in the above embodiment, the present application also provides a specific implementation of a device for determining reply information in the voice navigation service system. Please refer to the following embodiment.

[0145] See also Figure 6 , the apparatus 300 for determining reply information in the voice navigation service system provided in the embodiment of the present application includes:

[0146] The acquisition module 310 is used to acquire user question information from the voice navigation service system;

[0147] A determination module 320, for respectively determining similarities between the user question information and a plurality of preset standard question information;

[0148] A selection module 330 is used to select, from the plurality of preset standard question information, a plurality of candidate standard question information whose similarity to the user question information is greater than a second preset value when the largest target similarity among the similarities is less than a first preset value, and the second preset value is less than the first preset value;

[0149] The determination module 320 is further configured to input the user question information and the plurality of candidate standard question information into the information matching model, and determine the candidate standard question information matching the user question information as the target standard question information by using the matching relationship between the preset user question information and the plurality of preset standard question information in the information matching model;

[0150] The acquisition module 310 is also used to acquire a target result determination method corresponding to the target standard information;

[0151] The determination module 320 is further used to determine the reply information corresponding to the target standard question information by using the target result determination method, and use the reply information as the reply information of the user question information.

[0152] Based on this, in some embodiments, the acquisition module 310 may be specifically used for:

[0153] Acquire user voice information from the voice navigation service system;

[0154] Convert user voice information into business text information, and use the business text information as user question information.

[0155] Based on this, in some embodiments, the apparatus 300 may further include:

[0156] The determination module 320 is further configured to determine, when the target similarity is not less than a first preset value, that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

[0157] Based on this, in some embodiments, the apparatus 300 may further include:

[0158] The determination module 320 is further configured to determine, when all similarities are less than a second preset value, that the preset reply information is the reply information to the user's question information.

[0159] Based on this, in some embodiments, the determination module 320 may be specifically used to:

[0160] When the target result determination method is the first result determination method, the reply information corresponding to the target standard question information is obtained from the reply information database.

[0161] Based on this, in some embodiments, the determination module 320 may be specifically used to:

[0162] When the target result determination method is the second result determination method, query the result information corresponding to the target standard information;

[0163] Combine the result information with the preset speech template corresponding to the target standard question to obtain the reply information corresponding to the target standard question information.

[0164] Based on this, in some embodiments, the apparatus 300 may further include:

[0165] A conversion module, for determining the reply information corresponding to the target standard question information by using the target result determination method, and converting the reply information into a reply voice information after using the reply information as the reply information of the user's question information;

[0166] The playing module is used to play the reply voice message using the voice navigation service system.

[0167] Based on this, in some embodiments, the apparatus 300 may further include:

[0168] The acquisition module 310 is further used to acquire a user question information sample, actual standard question information matching the user question information sample, and a plurality of candidate standard question information samples before inputting the user question information and the plurality of candidate standard question information into the information matching model;

[0169] The determination module 320 is further configured to input the user question information sample and the plurality of candidate standard question information samples into the initial information matching model, and determine the candidate standard question information sample matching the user question information sample as the predicted standard question information by using the initial matching relationship between the preset user question information and the plurality of preset standard question information in the initial information matching model;

[0170] The determination module 320 is further used to determine the error value of the initial information matching model according to the actual standard question information and the predicted standard question information;

[0171] The determination module 320 is also used to update the initial matching relationship between the preset user question information and multiple preset standard question information in the initial information matching model when the error value does not meet the training stop condition, to obtain an updated initial matching relationship, and to update the predicted standard question information that matches the user question information sample using the updated initial matching relationship, and to update the error value according to the actual standard question information and the updated predicted standard question information until the error value meets the training stop condition to obtain the information matching model.

[0172] The various modules of the device for determining reply information in the voice navigation service system provided in the embodiment of the present application can realize the functions of the various steps of the method for determining reply information in the voice navigation service system provided above and achieve the corresponding technical effects, which will not be repeated here for the sake of brevity.

[0173] Based on the same inventive concept, an embodiment of the present application also provides an electronic device.

[0174] Figure 7 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown.

[0175] The electronic device may include a processor 401 and a memory 402 storing computer program instructions.

[0176] Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0177] Memory 402 may include a large capacity memory for data or instructions. By way of example and not limitation, memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, memory 402 may include a removable or non-removable (or fixed) medium. In appropriate cases, memory 402 may be inside or outside of an integrated gateway disaster recovery device. In a specific embodiment, memory 402 is a non-volatile solid-state memory.

[0178] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of the present disclosure.

[0179] The processor 401 reads and executes the computer program instructions stored in the memory 402 to implement any one of the methods for determining reply information in the voice navigation service system in the above embodiments.

[0180] In one example, the electronic device may further include a communication interface 403 and a bus 410. Figure 7 As shown, the processor 401, the memory 402, and the communication interface 403 are connected via a bus 410 and communicate with each other.

[0181] The communication interface 403 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0182] The bus 410 includes hardware, software or both, coupling the components of the electronic device to each other. For example, but not limitation, the bus may include an accelerated graphics port (Accelerated Graphics Port, AGP) or other graphics bus, an enhanced industry standard architecture (Extended Industry Standard Architecture, EISA) bus, a front side bus (Front Side Bus, FSB), a hypertransport (Hyper Transport, HT) interconnection, an industry standard architecture (Industry Standard Architecture, ISA) bus, an infinite bandwidth interconnection, a low pin count (Linear Predictive Coding, LPC) bus, a memory bus, a micro channel architecture (MicroChannel Architecture, MCA) bus, a peripheral component interconnect (Peripheral Component Interconnect, PCI) bus, a PCI-Express (Peripheral Component Interconnect-X, PCI-X) bus, a serial advanced technology attachment (Serial Advanced Technology Attachment, SATA) bus, a video electronics standard association local (VESA Local Bus, VLB) bus or other suitable buses or a combination of two or more of these. Where appropriate, the bus 410 may include one or more buses. Although the embodiments of the present application describe and illustrate a specific bus, the present application considers any suitable bus or interconnection. The electronic device can execute the method for determining reply information in the voice navigation service system in the embodiments of the present invention, thereby implementing the method for determining reply information in the voice navigation service system.

[0183] In addition, in combination with the method for determining reply information in the voice navigation service system in the above embodiment, the embodiment of the present application can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any method for determining reply information in the voice navigation service system in the above embodiment is implemented.

[0184] The present application also provides a computer program product. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes each process of an embodiment of a method for determining reply information in any one of the above-mentioned voice navigation service systems.

[0185] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0186] The functional blocks shown in the structural block diagram described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), appropriate firmware, plug-in, function card, etc. When implemented in software, the elements of the present application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or communication link by a data signal carried in a carrier. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (Read-Only Memory, ROM), flash memory, erasable read-only memory (Erasable ReadOnly Memory, EROM), floppy disks, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical discs, hard disks, optical fiber media, radio frequency (Radio Frequency, RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0187] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0188] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0189] The above are only specific implementation methods of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A method for determining reply information in a voice navigation service system, characterized in that: include: Obtain user question information from the voice navigation service system; Respectively determining the similarity between the user question information and a plurality of preset standard question information; When the largest target similarity among the similarities is less than a first preset value, selecting from the plurality of preset standard question information, a plurality of candidate standard question information whose similarity to the user question information is greater than a second preset value, wherein the second preset value is less than the first preset value; Inputting the user question information and the plurality of candidate standard question information into an information matching model, and using the matching relationship between the preset user question information and the plurality of preset standard question information in the information matching model, determining the candidate standard question information matching the user question information as the target standard question information; Obtaining a target result determination method corresponding to the target standard information; The target result determination method is used to determine the reply information corresponding to the target standard question information, and the reply information is used as the reply information of the user question information.

2. The method according to claim 1, characterized in that The obtaining of user question information from the voice navigation service system includes: Acquire user voice information from the voice navigation service system; The user voice information is converted into service text information, and the service text information is used as user question information.

3. The method according to claim 1, characterized in that Also includes: In a case where the target similarity is not less than the first preset value, it is determined that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

4. The method according to claim 1 or 3, characterized in that: Also includes: When all the similarities are smaller than the second preset value, the preset reply information is determined as the reply information of the user question information.

5. The method according to claim 1, characterized in that The step of determining the response information corresponding to the target standard question information by using the target result determination method includes: In the case where the target result determination method is the first result determination method, the reply information corresponding to the target standard question information is acquired from a reply information database.

6. The method according to claim 1, characterized in that The step of determining the response information corresponding to the target standard question information by using the target result determination method includes: In the case where the target result determination method is the second result determination method, querying result information corresponding to the target standard information; The result information is combined with a preset speech template corresponding to the target standard question to obtain reply information corresponding to the target standard question information.

7. The method according to claim 1, characterized in that After determining the reply information corresponding to the target standard question information by using the target result determination method, and using the reply information as the reply information of the user question information, the method further includes: Converting the reply message into a reply voice message; The voice navigation service system is used to play the reply voice message.

8. The method according to claim 1, characterized in that Before inputting the user question information and the plurality of candidate standard question information into the information matching model, the method further includes: Acquire a user question information sample, actual standard question information matching the user question information sample, and a plurality of candidate standard question information samples; Inputting the user question information sample and the multiple candidate standard question information samples into an initial information matching model, and using the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model, determining the candidate standard question information sample matching the user question information sample as the predicted standard question information; Determining an error value of the initial information matching model according to the actual standard question information and the predicted standard question information; When the error value does not satisfy the training stop condition, the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model is updated to obtain an updated initial matching relationship, and the predicted standard question information matching the user question information sample is updated using the updated initial matching relationship. The error value is updated according to the actual standard question information and the updated predicted standard question information until the error value satisfies the training stop condition to obtain the information matching model.

9. A device for determining reply information in a voice navigation service system, characterized in that: include: An acquisition module is used to obtain user question information from the voice navigation service system; A determination module, used to respectively determine the similarity between the user question information and a plurality of preset standard question information; A selection module, configured to select, when the largest target similarity among the similarities is less than a first preset value, from the plurality of preset standard question information, a plurality of candidate standard question information whose similarity to the user question information is greater than a second preset value, wherein the second preset value is less than the first preset value; The determination module is further configured to input the user question information and the plurality of candidate standard question information into an information matching model, and determine the candidate standard question information matching the user question information as the target standard question information by using the matching relationship between the preset user question information and the plurality of preset standard question information in the information matching model; The acquisition module is further used to acquire a target result determination method corresponding to the target standard information; The determination module is further used to determine the reply information corresponding to the target standard question information by using the target result determination method, and use the reply information as the reply information of the user question information.

10. The device according to claim 9, characterized in that The acquisition module is specifically used for: Acquire user voice information from the voice navigation service system; The user voice information is converted into service text information, and the service text information is used as user question information.

11. The device according to claim 9, characterized in that The device also includes: The determination module is further configured to determine, when the target similarity is not less than the first preset value, that the preset standard question information corresponding to the target similarity is the target standard question information matching the user question information.

12. The device according to claim 9 or 11, characterized in that The device also includes: The determination module is further configured to determine, when all of the similarities are less than the second preset value, that the preset reply information is the reply information of the user question information.

13. The device according to claim 9, characterized in that The determination module is specifically used for: In the case where the target result determination method is the first result determination method, the reply information corresponding to the target standard question information is acquired from a reply information database.

14. The device according to claim 9, characterized in that The determination module is specifically used for: In the case where the target result determination method is the second result determination method, querying result information corresponding to the target standard information; The result information is combined with a preset speech template corresponding to the target standard question to obtain reply information corresponding to the target standard question information.

15. The device according to claim 9, characterized in that The device also includes: A conversion module, configured to convert the reply information into reply voice information after determining the reply information corresponding to the target standard question information by using the target result determination method and using the reply information as the reply information of the user question information; The playing module is used to play the reply voice information by using the voice navigation service system.

16. The device according to claim 9, characterized in that The device also includes: The acquisition module is further used to acquire a user question information sample, actual standard question information matching the user question information sample, and a plurality of candidate standard question information samples before inputting the user question information and the plurality of candidate standard question information into the information matching model; The determination module is further configured to input the user question information sample and the plurality of candidate standard question information samples into an initial information matching model, and determine the candidate standard question information sample matching the user question information sample as the predicted standard question information by using the initial matching relationship between the preset user question information and the plurality of preset standard question information in the initial information matching model; The determination module is further used to determine an error value of the initial information matching model according to the actual standard question information and the predicted standard question information; The determination module is also used to update the initial matching relationship between the preset user question information and the multiple preset standard question information in the initial information matching model when the error value does not meet the training stop condition, to obtain an updated initial matching relationship, and to update the predicted standard question information that matches the user question information sample using the updated initial matching relationship, and to update the error value according to the actual standard question information and the updated predicted standard question information until the error value meets the training stop condition, thereby obtaining the information matching model.

17. An electronic device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the method for determining reply information in the voice navigation service system according to any one of claims 1 to 8 is implemented.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the method for determining reply information in a voice navigation service system according to any one of claims 1 to 8.

19. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the method for determining reply information in a voice navigation service system as described in any one of claims 1 to 8.

Images