A sentence intention recognition method, device, electronic equipment and program product

Abstract

The application is suitable for the field of artificial intelligence technology, and provides a sentence intention recognition method and device, an electronic device and a program product. The method comprises the following steps: obtaining a to-be-recognized sentence; determining the intention quantity of the to-be-recognized sentence; according to the intention quantity of the to-be-recognized sentence, adopting a sentence splitting model corresponding to the intention quantity to perform sentence splitting processing on the to-be-recognized sentence, and obtaining one or more to-be-recognized sub-sentences corresponding to the to-be-recognized sentence; and determining the first intention recognition result of the to-be-recognized sentence according to the sub-intention recognition result of each to-be-recognized sub-sentence in the one or more to-be-recognized sub-sentences. Through the method, a suitable sentence splitting model can be selected according to the intention quantity of the to-be-recognized sentence to perform sentence splitting processing on the to-be-recognized sentence, the effect of the sentence splitting processing is improved, and the accuracy of the intention recognition of the sentence is improved.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to a method, device, electronic device and program product for recognizing the intent of a statement. Background Technology

[0002] In the field of voice artificial intelligence, electronic devices typically need to first process the user's input statement to be recognized through a sentence segmentation model to obtain several sub-statements to be recognized, and then execute the intent contained in each sub-statement to be recognized.

[0003] Current sentence segmentation models are generally only good at segmenting sentences that contain a specific number of intents. For example, some models perform well in segmenting sentences that contain only one intent, but poorly in segmenting sentences that contain multiple intents; other models perform poorly in segmenting sentences that contain only one intent, but well in segmenting sentences that contain multiple intents.

[0004] However, current electronic devices typically use only one sentence segmentation model to process the sentence to be recognized. When this sentence segmentation model processes sentences that it is not good at, the sentence segmentation effect is poor, which ultimately leads to a low accuracy rate of sentence intent recognition. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, electronic device, and program product for identifying the intent of a statement, in order to solve the technical problem that the accuracy of existing statement intent identification is low.

[0006] In a first aspect, embodiments of this application provide a method for identifying the intent of a statement, comprising:

[0007] Obtain the statement to be recognized;

[0008] Determine the number of intents in the statement to be identified;

[0009] Based on the number of intents of the statement to be identified, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be identified, thereby obtaining one or more sub-statements to be identified corresponding to the statement to be identified.

[0010] The first intent recognition result of the statement to be recognized is determined based on the sub-intent recognition result of each of the one or more sub-statements to be recognized.

[0011] Optionally, based on the number of intents in the statement to be identified, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be identified, resulting in one or more sub-statements to be identified, including:

[0012] If the number of intents is greater than or equal to the first intent number threshold, the sentence to be identified is segmented using the first sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified.

[0013] If the number of intents is less than the first intent number threshold, the sentence to be identified is segmented using the second sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. The difference between the first sentence segmentation model and the second sentence segmentation model is that when the number of intents is greater than or equal to the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is higher than the second parameter value of the first evaluation parameter of the second sentence segmentation model; when the number of intents is less than the first intent number threshold, the first parameter value is lower than the second parameter value.

[0014] Optionally, the first intent quantity threshold is determined in the following way:

[0015] Retrieve one or more test statements;

[0016] Determine one or more distinct thresholds for the number of second intentions;

[0017] For each second intent quantity threshold, a third parameter value and a fourth parameter value of the second evaluation parameter corresponding to the second intent quantity threshold are determined. The third parameter value is obtained by the first sentence segmentation model using the second intent quantity threshold to segment each of the test statements. The fourth parameter value is obtained by the second sentence segmentation model using the second intent quantity threshold to segment each of the test statements.

[0018] The first intent quantity threshold is determined based on the third parameter value and the fourth parameter value corresponding to each of the second intent quantity thresholds.

[0019] Optionally, determining the number of intents for the statement to be identified includes:

[0020] Determine the keyword information of the statement to be identified;

[0021] The second sentence segmentation model determines the number of intents based on the sentence to be identified and the keyword information.

[0022] Optionally, determining the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the plurality of sub-statements to be recognized includes:

[0023] The results of each sub-intent recognition are concatenated to obtain the intent concatenation result.

[0024] The intent splicing result is post-processed to obtain the first intent recognition result; the post-processing method includes at least one or more of the following: preset intent deletion processing, slot inheritance processing, intent deduplication processing, and intent quantity correction processing.

[0025] Optionally, after determining the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the plurality of sub-statements to be recognized, the method further includes:

[0026] The intent recognition process is performed on the statement to be recognized using a target recognition model to obtain a second intent recognition result for the statement to be recognized; the target recognition model is a recognition model used to determine the sub-intent recognition result for each sub-statement to be recognized.

[0027] Based on the first intent recognition result and the second intent recognition result, the target intent recognition result of the statement to be recognized is determined.

[0028] Optionally, determining the target intent recognition result of the statement to be recognized based on the first intent recognition result and the second intent recognition result includes:

[0029] Determine the first weight and first confidence level of the first intent recognition result;

[0030] Determine the second weight and second confidence level of the second intent recognition result;

[0031] The target intent recognition result is determined based on the first intent recognition result, the second intent recognition result, the first weight, the first confidence level, the second weight, and the second confidence level.

[0032] Secondly, embodiments of this application provide a statement intent recognition device, comprising:

[0033] The statement acquisition unit is used to acquire the statement to be recognized.

[0034] An intent quantity determination unit is used to determine the intent quantity of the statement to be identified;

[0035] The sentence segmentation unit is used to segment the sentence to be identified according to the number of intents of the sentence to be identified, using a sentence segmentation model corresponding to the number of intents, to obtain one or more sub-sentences to be identified corresponding to the sentence to be identified;

[0036] The first intent recognition unit is configured to determine the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the one or more sub-statements to be recognized.

[0037] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the intent recognition method for statements as described in any of the first aspects above.

[0038] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the intent recognition method for statements as described in any of the first aspects above.

[0039] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute the steps of the intent recognition method for statements as described in any of the first aspects above.

[0040] The intent recognition method, apparatus, electronic device, and program product provided in this application have the following beneficial effects:

[0041] In the intent recognition method for statements provided in this application embodiment, the statement to be recognized is first obtained, then the number of intents of the statement to be recognized is determined, and then, based on the number of intents of the statement to be recognized, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be recognized, resulting in one or more sub-statements to be recognized. Finally, the first intent recognition result of the statement to be recognized is determined based on the sub-intent recognition result of each of the one or more sub-statements to be recognized. This method allows for the selection of a suitable sentence segmentation model based on the number of intents of the statement to be recognized, improving the effectiveness of sentence segmentation and thus increasing the accuracy of intent recognition. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating the implementation of a statement intent recognition method provided in this application embodiment;

[0044] Figure 2 A schematic diagram illustrating evaluation parameters corresponding to different model thresholds provided in an embodiment of this application;

[0045] Figure 3 This is a schematic diagram illustrating the training of a sentence segmentation model according to an embodiment of this application;

[0046] Figure 4 A flowchart illustrating a method for identifying the intent of a statement provided in an embodiment of this application;

[0047] Figure 5 A schematic diagram of the structure of a statement intent recognition device provided in an embodiment of this application;

[0048] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0049] It should be noted that the terminology used in the embodiments of this application is only for explaining specific embodiments of this application and is not intended to limit this application. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, "at least one" or "one or more" means one, two or more. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0050] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0051] The execution subject of the statement intent recognition method provided in this application embodiment can be an electronic device, which may include, but is not limited to, statement intent recognition devices, mobile phones, tablets, laptops, desktop computers, vehicle control devices, etc.

[0052] The statement intent recognition method provided in this application can be applied to any application scenario that requires statement intent recognition. For example, the statement intent recognition method provided in this application can be applied to the application scenario of a vehicle's smart cockpit. Users can instruct the in-vehicle control device to execute each step of the statement intent recognition method provided in this application, thereby improving the accuracy of intent recognition of the vehicle user's statement.

[0053] Some statements to be identified include multiple intents. For example, the statement "open the sunroof and heat the steering wheel" includes the intents of "open the sunroof" and "heat the steering wheel". Some statements to be identified include only one intent. For example, the statement "please open the passenger side window" only includes the intent of "open the passenger side window".

[0054] Current sentence segmentation models typically excel at segmenting sentences containing a specific number of intents. For example, the BERT (Bidirectional Encoder Representations from Transformers) model performs well on sentences with only one intent, but poorly on sentences with multiple intents. Similarly, the Large Language Model (LLM) performs poorly on sentences with only one intent, but well on sentences with multiple intents. The effectiveness of sentence segmentation can be determined by any one or more of the following parameters: segmentation efficiency, accuracy of intent recognition after segmentation, and recall rate of intent recognition after segmentation.

[0055] Because current electronic devices typically use only one sentence segmentation model to process the sentence to be recognized, when this model processes sentences it is not good at, the sentence segmentation effect is poor, resulting in a low accuracy of sentence intent recognition. For example, the BERT sentence segmentation model performs poorly when processing the sentence "open the sunroof and heat the steering wheel".

[0056] Based on this, embodiments of this application provide a method for identifying the intent of a statement. Please refer to... Figure 1 , Figure 1The flowchart illustrates an implementation method for intent recognition of a statement, as provided in this application embodiment. This intent recognition method may include steps S101 to S104, detailed below:

[0057] In S101, the statement to be recognized is obtained.

[0058] In this embodiment of the application, the electronic device can first obtain the statement to be recognized that requires intent recognition. For example, the user can output the statement to be recognized, so that the electronic device can obtain the statement output by the user.

[0059] In S102, the number of intents of the statement to be identified is determined.

[0060] In this embodiment of the application, after obtaining the statement to be identified, the electronic device can first determine the number of intents of the statement to be identified.

[0061] In one possible implementation, the electronic device can first determine the keyword information (prompt) of the statement to be identified, and then use the second sentence segmentation model provided in this application embodiment to determine the intent quantity of the statement to be identified based on the statement to be identified and the keyword information.

[0062] It should be noted that the sentence segmentation model provided in this application embodiment may include a first sentence segmentation model and a second sentence segmentation model. The difference between the first sentence segmentation model and the second sentence segmentation model is that when the number of intents of the sentence to be identified is greater than or equal to the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is higher than the second parameter value of the first evaluation parameter of the second sentence segmentation model; when the number of intents is less than the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is lower than the second parameter value of the first evaluation parameter of the second sentence segmentation model.

[0063] The first parameter value of the first evaluation parameter of the first sentence segmentation model can be: the first precision and the first recall rate for intent recognition of the sub-sentences to be identified obtained after sentence segmentation by the first sentence segmentation model.

[0064] The second parameter value of the first evaluation parameter of the second sentence segmentation model can be: the second precision and the second recall rate for intent recognition of the sub-statements to be identified obtained after sentence segmentation by the second sentence segmentation model.

[0065] For example, the first sentence segmentation model provided in this application embodiment can be an LLM model, and the second sentence segmentation model provided in this application embodiment can be a BERT model.

[0066] In one possible implementation, the accuracy of the second sentence model in determining the number of intents of the sentence to be identified can also be improved by adjusting the model threshold of the second sentence model.

[0067] Specifically, different model thresholds can be determined first, and then the single-intent accuracy, single-intent recall, multi-intent accuracy, and multi-intent recall corresponding to each different model threshold can be determined. Finally, the target model threshold can be determined based on the single-intent accuracy, single-intent recall, multi-intent accuracy, and multi-intent recall corresponding to each different model threshold.

[0068] Please see Figure 2 , Figure 2 This is a schematic diagram illustrating the evaluation parameters corresponding to different model thresholds provided in an embodiment of this application. For example... Figure 2 As shown, the lower the model threshold of the second sentence segmentation model, the higher the multi-intent recall rate. Based on this, the model threshold of the second sentence segmentation model can be set to a smaller number to improve the accuracy of the second sentence segmentation model in determining the number of intents of the sentence to be identified, thereby improving the accuracy of intent recognition of the sentence.

[0069] It should be noted that in some alternative implementations, the number of intents of the sentence to be identified can be determined using models other than the second sentence model. However, this method requires an additional model to determine the number of intents of the sentence to be identified, thus increasing the cost.

[0070] In this implementation, the method of determining the number of intents of the sentence to be identified through the second sentence model does not require the use of an additional model for determining the number of intents of the sentence to be identified, thus reducing the cost of sentence intent recognition.

[0071] In S103, based on the number of intents of the statement to be identified, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be identified, resulting in one or more sub-statements to be identified.

[0072] In this embodiment of the application, after determining the number of intents of the statement to be identified, the electronic device can perform sentence segmentation on the statement to be identified using a sentence segmentation model corresponding to the number of intents, thereby obtaining one or more sub-statements to be identified corresponding to the statement to be identified.

[0073] In one possible implementation, the electronic device can first determine the relationship between the number of intents of the statement to be recognized and a first threshold number of intents, and then select a sentence segmentation model to segment the statement to be recognized based on the relationship between the number of intents of the statement to be recognized and the first threshold number of intents.

[0074] For example, if the electronic device determines that the number of intents of the statement to be identified is greater than or equal to the first intent number threshold, then the statement to be identified is segmented using the first sentence segmentation model to obtain multiple sub-statements to be identified corresponding to the statement to be identified; correspondingly, if the electronic device determines that the number of intents of the statement to be identified is less than the first intent number threshold, then the statement to be identified is segmented using the second sentence segmentation model to obtain multiple sub-statements to be identified corresponding to the statement to be identified.

[0075] Examples and differences between the first clause model and the second clause model can be found in the implementation examples corresponding to S102, and will not be repeated here.

[0076] In this implementation, the electronic device can first acquire one or more test statements and first determine one or more different second intent quantity thresholds. The more test statements, the better, and the number of intents in the test statements does not have to be exactly the same. The second intent quantity thresholds can include, but are not limited to, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, etc.

[0077] After obtaining the test statement and one or more different second intent quantity thresholds, the electronic device can determine the third parameter value of the second evaluation parameter and the fourth parameter value of the second evaluation parameter for each second intent quantity threshold. The third parameter value is determined by the first sentence segmentation model after segmenting each test statement using the second intent quantity threshold, and the fourth parameter value is determined by the second sentence segmentation model after segmenting each test statement using the second intent quantity threshold.

[0078] The third parameter value of the second evaluation parameter corresponding to any second intent quantity threshold can be determined in the following way: First, the second intent quantity threshold is used to process each test statement into several sub-statements to be identified by the first sentence segmentation model. Then, intent identification is performed on these sub-statements to be identified. Finally, the third precision and third recall of intent identification are determined based on the intent identification results.

[0079] The fourth parameter value of the second evaluation parameter corresponding to any second intent quantity threshold can be determined in the following way: First, the second intent quantity threshold is used to process each test statement into several sub-statements to be identified by the second sentence segmentation model. Then, intent identification is performed on these sub-statements to be identified. Finally, the fourth precision and fourth recall of intent identification are determined based on the intent identification results.

[0080] The above method allows us to determine the third and fourth parameter values ​​of the second evaluation parameter corresponding to each second intent quantity threshold. Then, the electronic device can determine the first intent quantity threshold based on these third and fourth parameter values. The specific method for determining the first intent quantity threshold based on the third and fourth parameter values ​​of each second intent quantity threshold can be set according to actual needs and is not limited here.

[0081] It should be noted that, in other possible implementations, the electronic device may also determine the sentence segmentation model to be used in the following ways: if the number of intents of the sentence to be identified is less than the intent number threshold 1, then sentence segmentation model 1 is used; if the number of intents of the sentence to be identified is greater than or equal to the intent number threshold 1 and less than the intent number threshold 2, then sentence segmentation model 2 is used; if the number of intents of the sentence to be identified is greater than or equal to the intent number threshold 2 and less than the intent number threshold 3, then sentence segmentation model 3 is used.

[0082] Based on this, it is understood that the threshold for the number of intents used in this application can be several, and the sentence segmentation model used for different intent intervals can also be set according to actual needs. Any solution corresponding to the idea of ​​selecting different sentence segmentation models according to different intent intervals is within the protection scope of this application.

[0083] In S104, the first intent recognition result of the statement to be recognized is determined based on the sub-intent recognition result of each statement to be recognized in one or more statements to be recognized.

[0084] In this embodiment of the application, after obtaining one or more sub-statements corresponding to the statement to be identified, the electronic device can first determine the sub-intent recognition result of each sub-statement to be identified, and then determine the first intent recognition result of the statement to be identified based on the sub-intent recognition result of each sub-statement to be identified.

[0085] In one possible implementation, after obtaining one or more sub-statements corresponding to the statement to be identified, the electronic device can input each sub-statement to be identified into the target recognition model, so as to instruct the target recognition model to generate and output the sub-intent recognition result corresponding to each sub-statement to be identified, and the electronic device can obtain the sub-intent recognition result corresponding to each sub-statement to be identified output by the target recognition model.

[0086] In practical applications, electronic devices can input each sub-statement to be identified into the target recognition model in parallel to improve the efficiency of obtaining the sub-intent recognition result corresponding to each sub-statement, thereby improving the efficiency of intent recognition of the statement.

[0087] The target recognition model can be used to generate and output the intent recognition result corresponding to the input statement. For example, the target recognition model can be a (domain-intent-slot) model.

[0088] After obtaining the sub-intent recognition result corresponding to each sub-statement to be recognized, the electronic device can concatenate the individual sub-intent recognition results to obtain the intent concatenation result. After obtaining the intent concatenation result, the electronic device can perform post-processing on the intent concatenation result to obtain the first intent recognition result.

[0089] The post-processing methods include at least one or more of the following: preset intent reduction processing, slot inheritance processing, intent deduplication processing, and intent quantity correction processing.

[0090] The preset intent deletion process can remove blacklisted intents and ignored intents from the intent concatenation result. Blacklisted intents are those existing in a preset blacklist, and ignored intents are those existing in a preset ignored list. For example, blacklisted intents could be intents that pose a threat to driving safety.

[0091] Slot inheritance processing can be used for inheritance between multiple adjacent intents. For example, if the first intent is "turn on the air conditioner to 30 degrees" and the second intent is "adjust to 20 degrees", then slot inheritance processing can be performed on these two intents, so that the second intent can be adjusted to "adjust the air conditioner temperature to 20 degrees".

[0092] Intent deduplication can be used to remove duplicate intents.

[0093] The intent quantity correction process can be used to increase or decrease the number of intents to meet preset requirements. For example, if the number of intents is 7 and the intent threshold is 6, the intent quantity correction process can be performed to reduce the number of intents to 6.

[0094] As can be seen from the above, in the intent recognition method for statements provided in this application embodiment, the statement to be recognized is first obtained, then the number of intents of the statement to be recognized is determined, and then, based on the number of intents of the statement to be recognized, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be recognized, resulting in one or more sub-statements to be recognized. Finally, the first intent recognition result of the statement to be recognized is determined based on the sub-intent recognition result of each of the one or more sub-statements to be recognized. This method allows for the selection of a suitable sentence segmentation model based on the number of intents of the statement to be recognized, improving the effectiveness of sentence segmentation and thus increasing the accuracy of intent recognition.

[0095] The above embodiments describe the process of segmenting the statement to be identified into one or more sub-statements to be identified, and determining the first intent identification result of the statement to be identified based on the sub-intent identification result of each sub-statement to be identified.

[0096] In one possible implementation, the electronic device can directly perform intent recognition processing on the statement to be recognized using a target recognition model (without sentence segmentation) to obtain a second intent recognition result for the statement to be recognized. Here, the target recognition model is the aforementioned recognition model used to determine the sub-intent recognition result for each sub-statement to be recognized.

[0097] Based on this, after obtaining the first intent recognition result obtained through the sentence segmentation process and the second intent recognition result obtained without the sentence segmentation process, the electronic device can also determine the target intent recognition result of the sentence to be recognized based on the first intent recognition result and the second intent recognition result.

[0098] Specifically, the electronic device can determine the first weight and first confidence level of the first intent recognition result, and determine the second weight and second confidence level of the second intent recognition result. Then, based on the first intent recognition result, the second intent recognition result, the first weight, the first confidence level, the second weight, and the second confidence level, the target intent recognition result can be determined.

[0099] The method of obtaining two different intent recognition results through two different approaches, and then obtaining the target intent recognition result based on the two different intent recognition results, can further improve the accuracy of intent recognition of statements.

[0100] In one possible implementation, before executing the intent recognition method for the statement provided in this application embodiment, the first sentence model and the second sentence model can be trained first. For example, the second sentence model can be trained using unsupervised data, and the first sentence model can be trained using supervised data, with fine-tuning performed using labeled supervised data on both the first and second sentence models.

[0101] Please see Figure 3 , Figure 3 This is a schematic diagram illustrating the training of a sentence segmentation model, as provided in an embodiment of this application. Figure 3 As shown, unsupervised training samples can be obtained first through NLU single-round training data, and then the BERT model (second sentence segmentation model) can be trained using the unsupervised training samples. Supervised data can be obtained by annotating specific data through multi-tasks, and the LLM model (first sentence segmentation model) and the BERT model (second sentence segmentation model) can be fine-tuned using the supervised data.

[0102] Based on the intent recognition method for statements provided in the embodiments of this application, the following is a flowchart illustrating another intent recognition method for statements. Please refer to... Figure 4 , Figure 4 This is a flowchart illustrating a method for identifying the intent of a statement, as provided in an embodiment of this application.

[0103] like Figure 4 As shown, the electronic device can acquire the user's audio input and convert the audio input speech signal into text information. Then, the electronic device can search a keyword information (Prompt) database based on the text information to determine the keyword information of the text information.

[0104] Subsequently, the electronic device can input the keyword information and text information of the text information into the BERT model (i.e., the second sentence segmentation model in this embodiment) to instruct the second sentence segmentation model to determine the number of intents based on the keyword information and text information. If the number of intents is less than 2 (i.e., the first intent number threshold in this embodiment), the BERT model can segment the text information to obtain one or more sub-statements to be identified. If the number of intents is greater than or equal to 2 (i.e., the first intent number threshold in this embodiment), the LLM model (i.e., the first sentence segmentation model in this embodiment) can segment the text information to obtain one or more sub-statements to be identified.

[0105] After obtaining the sub-statement to be identified, the electronic device can input each sub-statement to be identified in parallel into the domain-intent-slot model (i.e., the target recognition model in this embodiment of the application) to instruct the domain-intent-slot model to output the intent recognition result corresponding to each sub-statement to be identified.

[0106] After obtaining the intent recognition result corresponding to each sub-statement to be recognized, the electronic device can perform post-processing on the intent recognition result corresponding to each sub-statement to be recognized, thereby obtaining the first intent recognition result.

[0107] In addition, after obtaining the text information, the electronic device can also input the text information that has not been segmented into the domain-intent-slot model, so as to instruct the domain-intent-slot model to output the second intent recognition result corresponding to the text information that has not been segmented into sentences, so that the electronic device can obtain the second intent recognition result.

[0108] After obtaining the first intent recognition result and the second intent recognition result, the electronic device can perform multi-result fusion arbitration based on the first intent recognition result and the second intent recognition result to obtain the target intent recognition result.

[0109] The method provided in this application embodiment can also be implemented by a smart cockpit. A smart cockpit can be implemented using a standalone server or a server cluster consisting of multiple servers. A smart cockpit system refers to an automotive electronic system equipped with intelligent and connected in-vehicle devices and services, such as in-vehicle infotainment systems, instrument panels, head-up displays (HUDs), streaming rearview mirrors, and voice interaction systems.

[0110] The electronic system of a smart cockpit can be a complete system consisting of a central control platform, a full LCD instrument panel, a central control display, an entertainment system, a smart speaker, a vehicle networking module, a streaming media rearview mirror, and a remote information processing system. The smart cockpit system can be centered around the cockpit area controller, implementing the functions of the aforementioned smart cockpit electronic systems on a unified hardware and software platform, and integrating intelligent interaction, intelligent scenarios, and personalized services.

[0111] Intelligent cockpit systems form the foundation for human-vehicle interaction and vehicle-to-external connectivity. The application scenarios of intelligent cockpit systems typically cover all scenarios in which a user uses the vehicle. Specifically, this includes time-based scenarios before, during, and after vehicle use. It can also include spatial scenarios involving the driver, front passenger, rear passengers, and relevant people or objects outside the vehicle. Compared to past command-based interactions, the human-machine interaction of intelligent cockpits combines the vehicle and user usage scenarios, based on image recognition, voice recognition, and environmental perception. This achieves more intelligent effects that better meet user needs (such as voice changing when making phone calls). In other words, an intelligent cockpit refers to an in-vehicle system equipped with intelligent and connected devices and services, such as in-vehicle infotainment systems, instrument panels, head-up displays (HUDs), streaming rearview mirrors, and voice interaction systems. In this embodiment, the in-vehicle cockpit refers to an intelligent cockpit.

[0112] Based on the statement intent recognition method provided in the above embodiments, this application further provides a statement intent recognition device that implements the above method embodiments. Please refer to [link to relevant documentation]. Figure 5 , Figure 5 This is a schematic diagram of the structure of a statement intent recognition device provided in an embodiment of this application. Figure 5 As shown, the intent recognition device 50 may include: a statement acquisition unit 51, an intent quantity determination unit 52, a sentence segmentation unit 53, and a first intent recognition unit 54. Wherein:

[0113] The statement acquisition unit 51 is used to acquire the statement to be recognized.

[0114] The intent quantity determination unit 52 is used to determine the intent quantity of the statement to be identified.

[0115] Sentence segmentation unit 53 is used to segment the sentence to be identified according to the number of intents of the sentence to be identified, using a sentence segmentation model corresponding to the number of intents, to obtain one or more sub-sentences to be identified corresponding to the sentence to be identified.

[0116] The first intent recognition unit 54 is used to determine the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each statement to be recognized in one or more statements to be recognized.

[0117] Optionally, clause unit 53 is specifically used for:

[0118] If the number of intents is greater than or equal to the first intent number threshold, the sentence to be identified is segmented using the first sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified.

[0119] If the number of intents is less than the first intent number threshold, the sentence to be identified is segmented using the second sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. The difference between the first sentence segmentation model and the second sentence segmentation model is that when the number of intents is greater than or equal to the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is higher than the second parameter value of the first evaluation parameter of the second sentence segmentation model; when the number of intents is less than the first intent number threshold, the first parameter value is lower than the second parameter value.

[0120] Optionally, the intent recognition device 50 may further include an intent quantity threshold determination unit, wherein:

[0121] The intent quantity threshold determination unit is specifically used for:

[0122] Retrieve one or more test statements;

[0123] Determine one or more distinct thresholds for the number of second intentions;

[0124] For each second intent quantity threshold, determine the third parameter value of the second evaluation parameter corresponding to the second intent quantity threshold, and the fourth parameter value of the second evaluation parameter. The third parameter value is determined by the first sentence segmentation model after segmenting each test statement using the second intent quantity threshold, and the fourth parameter value is determined by the second sentence segmentation model after segmenting each test statement using the second intent quantity threshold.

[0125] The first intent quantity threshold is determined based on the third and fourth parameter values ​​corresponding to each second intent quantity threshold.

[0126] Optionally, the intent quantity determination unit 52 is specifically used for:

[0127] Determine the keyword information of the statement to be identified;

[0128] The second sentence segmentation model determines the number of intents based on the sentence to be identified and keyword information.

[0129] Optionally, the first intent recognition unit 54 is specifically used for:

[0130] The results of each sub-intent recognition are concatenated to obtain the intent concatenation result.

[0131] The intent splicing result is post-processed to obtain the first intent recognition result; the post-processing method includes at least one or more of the following: preset intent reduction processing, slot inheritance processing, intent deduplication processing, and intent quantity correction processing.

[0132] Optionally, the intent recognition device 50 may further include a second intent recognition unit. Wherein:

[0133] The second intent recognition unit is specifically used for:

[0134] The target recognition model is used to perform intent recognition processing on the statement to be recognized, and a second intent recognition result of the statement to be recognized is obtained; the target recognition model is a recognition model used to determine the sub-intent recognition result of each sub-statement to be recognized.

[0135] Based on the first intent recognition result and the second intent recognition result, the target intent recognition result of the statement to be recognized is determined.

[0136] Optionally, the second intent recognition unit is specifically used for:

[0137] Determine the first weight and first confidence level of the first intent recognition result;

[0138] Determine the second weight and second confidence level of the second intent recognition result;

[0139] The target intent recognition result is determined based on the first intent recognition result, the second intent recognition result, the first weight, the first confidence level, the second weight, and the second confidence level.

[0140] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 6 provided in this embodiment may include: a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60, such as a program corresponding to a statement intent recognition method. When the processor 60 executes the computer program 62, it implements the steps described above in the embodiment of the statement intent recognition method, for example... Figure 1 S101 to S104 are shown. Alternatively, when the processor 60 executes the computer program 62, it implements the functions of each module / unit in the above-described intent recognition device embodiment, for example... Figure 5 The functions of units 51 to 54 shown.

[0141] For example, computer program 62 can be divided into one or more modules / units, one or more of which are stored in memory 61 and executed by processor 60 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of computer program 62 in electronic device 6. For example, computer program 62 can be divided into a statement acquisition unit 51, an intent quantity determination unit 52, a sentence segmentation unit 53, and a first intent recognition unit 54. For the specific functions of each unit, please refer to... Figure 5 The relevant descriptions in the corresponding embodiments are not repeated here.

[0142] Those skilled in the art will understand that Figure 6 This is merely an example of electronic device 6 and does not constitute a limitation on electronic device 6. It may include more or fewer components than shown, or combine certain components, or use different components.

[0143] The processor 60 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0144] The memory 61 can be an internal storage unit of the electronic device 6, such as a hard disk or RAM. The memory 61 can also be an external storage device of the electronic device 6, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, or flash card. Furthermore, the memory 61 can include both internal and external storage units of the electronic device 6. The memory 61 is used to store computer programs and other programs and data required by the electronic device. The memory 61 can also be used to temporarily store data that has been output or will be output.

[0145] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units is merely an example. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the intent recognition device can be divided into different functional units to complete all or part of the functions described above. The functional units in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0146] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps in the various method embodiments described above.

[0147] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0148] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, refer to the relevant descriptions of other embodiments.

[0149] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0150] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

[0151] The user personal information involved in the various embodiments of this application is processed in strict accordance with the requirements of laws and regulations, following the principles of legality, legitimacy, and necessity, and based on the reasonable purpose of the business scenario. This processing involves personal information actively provided by users during the use of the product / service, information generated as a result of using the product / service, and personal information obtained with user authorization. The user personal information processed by the applicant may vary depending on the specific product / service scenario and should be based on the specific scenario in which the user uses the product / service. This may involve user account information, device information, driving information, vehicle information, or other related information. The applicant will treat user personal information and its processing with a high degree of diligence. The applicant attaches great importance to the security of user personal information and has taken reasonable and feasible security protection measures that comply with industry standards to protect user information and prevent unauthorized access, public disclosure, use, modification, damage, or loss of personal information.

Claims

1. A method for recognizing the intent of a statement, characterized in that, include: Obtain the statement to be recognized; Determine the number of intents in the statement to be identified; Based on the number of intents of the statement to be identified, a sentence segmentation model corresponding to the number of intents is used to segment the statement to be identified, thereby obtaining one or more sub-statements to be identified corresponding to the statement to be identified. The first intent recognition result of the statement to be recognized is determined based on the sub-intent recognition result of each of the statements to be recognized in one or more of the sub-statements to be recognized; The step of segmenting the statement to be identified using a sentence segmentation model corresponding to the number of intents, based on the number of intents to be identified, to obtain one or more sub-statements to be identified, includes: If the number of intents is greater than or equal to the first intent number threshold, the sentence to be identified is segmented using the first sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. If the number of intents is less than the first intent number threshold, the sentence to be identified is segmented using the second sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. The difference between the first sentence segmentation model and the second sentence segmentation model is that when the number of intents is greater than or equal to the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is higher than the second parameter value of the first evaluation parameter of the second sentence segmentation model; when the number of intents is less than the first intent number threshold, the first parameter value is lower than the second parameter value.

2. The method according to claim 1, characterized in that, The first intent quantity threshold is determined in the following way: Retrieve one or more test statements; Determine one or more distinct thresholds for the number of second intentions; For each second intent quantity threshold, a third parameter value and a fourth parameter value of the second evaluation parameter corresponding to the second intent quantity threshold are determined. The third parameter value is obtained by the first sentence segmentation model using the second intent quantity threshold to segment each of the test statements. The fourth parameter value is obtained by the second sentence segmentation model using the second intent quantity threshold to segment each of the test statements. The first intent quantity threshold is determined based on the third parameter value and the fourth parameter value corresponding to each of the second intent quantity thresholds.

3. The method according to claim 1, characterized in that, Determining the number of intents for the statement to be identified includes: Determine the keyword information of the statement to be identified; The second sentence segmentation model determines the number of intents based on the sentence to be identified and the keyword information.

4. The method according to any one of claims 1 to 3, characterized in that, The step of determining the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the multiple sub-statements to be recognized includes: The results of each sub-intent recognition are concatenated to obtain the intent concatenation result. The intent splicing result is post-processed to obtain the first intent recognition result; the post-processing method includes at least one or more of the following: preset intent deletion processing, slot inheritance processing, intent deduplication processing, and intent quantity correction processing.

5. The method according to any one of claims 1 to 3, characterized in that, After determining the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the plurality of statements to be recognized, the method further includes: The intent recognition process is performed on the statement to be recognized using a target recognition model to obtain a second intent recognition result for the statement to be recognized; the target recognition model is a recognition model used to determine the sub-intent recognition result for each sub-statement to be recognized. Based on the first intent recognition result and the second intent recognition result, the target intent recognition result of the statement to be recognized is determined.

6. The method according to claim 5, characterized in that, Determining the target intent recognition result of the statement to be recognized based on the first intent recognition result and the second intent recognition result includes: Determine the first weight and first confidence level of the first intent recognition result; Determine the second weight and second confidence level of the second intent recognition result; The target intent recognition result is determined based on the first intent recognition result, the second intent recognition result, the first weight, the first confidence level, the second weight, and the second confidence level.

7. A statement intent recognition device, characterized in that, include: The statement acquisition unit is used to acquire the statement to be recognized. An intent quantity determination unit is used to determine the intent quantity of the statement to be identified; The sentence segmentation unit is used to segment the sentence to be identified according to the number of intents of the sentence to be identified, using a sentence segmentation model corresponding to the number of intents, to obtain one or more sub-sentences to be identified corresponding to the sentence to be identified. The first intent recognition unit is configured to determine the first intent recognition result of the statement to be recognized based on the sub-intent recognition result of each of the one or more sub-statements to be recognized. The clause segmentation unit is specifically used for: If the number of intents is greater than or equal to the first intent number threshold, the sentence to be identified is segmented using the first sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. If the number of intents is less than the first intent number threshold, the sentence to be identified is segmented using the second sentence segmentation model to obtain multiple sub-sentences to be identified corresponding to the sentence to be identified. The difference between the first sentence segmentation model and the second sentence segmentation model is that when the number of intents is greater than or equal to the first intent number threshold, the first parameter value of the first evaluation parameter of the first sentence segmentation model is higher than the second parameter value of the first evaluation parameter of the second sentence segmentation model; when the number of intents is less than the first intent number threshold, the first parameter value is lower than the second parameter value.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements each step of the intent recognition method for the statement as described in any one of claims 1 to 6.

9. A computer program product, characterized in that, When the computer program product is executed by a processor, it implements the steps of the intent recognition method for the statement as described in any one of claims 1 to 6.

Images