User dialogue intention recognition method and device, storage medium and electronic device

Abstract

The application discloses a user dialogue intention recognition method and device, a storage medium and electronic equipment. A sentence intention recognition result of a current sentence in a current round of dialogue is obtained, wherein the sentence intention recognition result comprises a classification name and a classification score. According to the classification name, a corresponding intention state coefficient is determined. According to the classification score and the intention state coefficient, a corresponding confidence is determined. According to the confidence and a preset intention determination rule, a dialogue intention corresponding to the current sentence of the current round of dialogue is determined, wherein parameters involved in the preset intention determination rule comprise a historical dialogue intention corresponding to a historical sentence of the current round of dialogue. The application can combine the dialogue intention of the foregoing text, accurately recognize the dialogue intention of each sentence, further improve the fluency of the response, and improve the user experience.

Description

Technical Field

[0001] This invention relates to the field of semantic recognition, and in particular to a method, apparatus, storage medium, and electronic device for recognizing user dialogue intent. Background Technology

[0002] The core function of an intelligent response system is to identify user intent and respond based on that intent. Currently, intelligent response systems mainly use natural language processing modules to identify the intent of a single sentence in a conversation, that is, only classifying the sentence that the user has input.

[0003] However, specific business operations require user information to be collected and filled in before a response can be given. During the filling process, the system will ask the user questions, and the sentences entered by the user will be recognized again by the natural language recognition module. At this time, there may be problems such as failure to recognize the dialogue intent or incorrect recognition of the dialogue intent, which may further lead to problems such as unsmooth response and poor user experience. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method, apparatus, storage medium and electronic device for recognizing user dialogue intent that overcomes or at least partially solves the above problems.

[0005] Firstly, a method for identifying user dialogue intent includes:

[0006] Obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score;

[0007] Based on the classification name, determine the corresponding intent state coefficient;

[0008] The corresponding confidence level is determined based on the classification score and the intent state coefficient;

[0009] Based on the confidence level and the preset intent determination rules, the dialogue intent corresponding to the current sentence of the current round of dialogue is determined. The parameters involved in the preset intent determination rules include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue.

[0010] In conjunction with the first aspect, in some optional implementations, determining the corresponding intent state coefficient based on the classification name includes:

[0011] Based on the category name, a corresponding intent state coefficient is determined using a preset intent state determination rule, wherein the intent state coefficient matches the category name.

[0012] In conjunction with the previous implementation method, in some optional implementation methods, determining the corresponding intent state coefficient based on the category name and a preset intent state determination rule includes:

[0013] If the category name matches the newly added intent in the intent state determination rule, then the newly added coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the newly added coefficient corresponds to the newly added intent, and the newly added intent represents that the intent involved in the current sentence has never appeared in the history of the current round of conversation;

[0014] If the category name matches the exclusive intent in the intent state determination rule, then the exclusive coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the exclusive coefficient corresponds to the exclusive intent, and the exclusive intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold.

[0015] If the classification name matches the modified intent in the intent state determination rule, then the modification coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the modification coefficient corresponds to the modified intent, and the modified intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold and has no relation to the intents involved in the historical dialogue of this round of call;

[0016] If the category name matches the coexisting intent in the intent state determination rule, then the coexistence coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the coexistence coefficient corresponds to the coexisting intent, and the coexisting intent indicates that the intent involved in the current sentence is strongly correlated with at least one intent in the historical dialogue of the current round of the call.

[0017] If the category name matches the invalid intent in the intent state determination rule, then the invalid coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the invalid coefficient corresponds to the invalid intent, and the invalid intent indicates that the intent involved in the current sentence has a clear intent switch.

[0018] In conjunction with the first aspect, in some optional implementations, determining the corresponding confidence level based on the classification score and the intent state coefficient includes:

[0019] According to formula 1: C = log(score) i ×Z)×ratio i The confidence level is calculated, where C is the confidence level, i is the sentence number, and the score is... iThe classification score is given by Z, which is a preset scaling factor, and the ratio is... i The intention state coefficient is denoted as .

[0020] In conjunction with the first aspect, in some optional implementations, determining the dialogue intent corresponding to the current sentence of the current round of dialogue based on the confidence level and a preset intent determination rule includes:

[0021] The confidence level is compared with a first preset threshold, and the confidence level is compared with a second preset threshold, wherein the first preset threshold is greater than the second preset threshold;

[0022] If the confidence level is not less than the first preset threshold, then the dialogue intent is determined according to the preset first intent determination rule;

[0023] If the confidence level is less than the first preset threshold and greater than the second preset threshold, then the user's dialogue intent is determined according to the preset second intent determination rule;

[0024] If the confidence level is not greater than the second preset threshold, then the user's dialogue intent is determined according to the preset third intent determination rule.

[0025] In conjunction with the first aspect, in some alternative embodiments, the method further includes:

[0026] Based on the dialogue intent corresponding to the current sentence in the current round of dialogue, update the historical dialogue intent corresponding to the historical sentences in the current round of dialogue.

[0027] Secondly, a device for recognizing user dialogue intent includes: an intent state management module and a confidence calculation module;

[0028] The intent state management module includes: a receiving unit, a coefficient determination unit, and an intent determination unit;

[0029] The receiving unit is configured to obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score;

[0030] The coefficient determination unit is configured to determine the corresponding intent state coefficient based on the category name;

[0031] The confidence calculation module is configured to determine the corresponding confidence level based on the classification score and the intent state coefficient.

[0032] The intent determination unit is configured to determine the dialogue intent corresponding to the current sentence of the current round of dialogue based on the confidence level and a preset intent determination rule, wherein the parameters involved in the preset intent determination rule include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue.

[0033] In conjunction with the second aspect, in some optional embodiments, the coefficient determining unit includes: a coefficient determining subunit;

[0034] The coefficient determination subunit is configured to determine the corresponding intent state coefficient based on the category name and a preset intent state determination rule, wherein the intent state coefficient matches the category name.

[0035] Thirdly, a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the method for recognizing user dialogue intent as described in any of the preceding claims.

[0036] Fourthly, an electronic device includes at least one processor, and at least one memory and bus connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is configured to invoke program instructions in the memory to execute the user dialogue intent recognition method described in any of the preceding claims.

[0037] By employing the above technical solutions, the present invention provides a method, apparatus, storage medium, and electronic device for identifying user dialogue intent. This method can obtain the sentence intent identification result for the current sentence in the current round of dialogue. The sentence intent identification result includes: a category name and a category score; determining a corresponding intent state coefficient based on the category name; determining a corresponding confidence level based on the category score and the intent state coefficient; and determining the dialogue intent corresponding to the current sentence in the current round of dialogue based on the confidence level and a preset intent determination rule. The preset intent determination rule involves parameters including: historical dialogue intents corresponding to historical sentences in the current round of dialogue. Therefore, it can be seen that the present invention can accurately identify the dialogue intent of each sentence by combining the aforementioned dialogue intent, further improving the fluency of responses and enhancing the user experience.

[0038] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0039] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0040] Figure 1 A flowchart of a method for recognizing user dialogue intent provided by the present invention is shown;

[0041] Figure 2 A schematic diagram of an intent state determination rule provided by the present invention is shown;

[0042] Figure 3 A schematic diagram of the first intent determination rule provided by the present invention is shown;

[0043] Figure 4 A schematic diagram of the second intent determination rule provided by the present invention is shown;

[0044] Figure 5 A schematic diagram of the third intent determination rule provided by the present invention is shown;

[0045] Figure 6 A schematic diagram of the structure of a user dialogue intent recognition device provided by the present invention is shown;

[0046] Figure 7 A schematic diagram of the structure of an electronic device provided by the present invention is shown. Detailed Implementation

[0047] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0048] like Figure 1 As shown, the present invention provides a method for recognizing user dialogue intent, including: S100, S200, S300 and S400;

[0049] S100. Obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score;

[0050] Optionally, the classification name can, to some extent, characterize the dialogue intent corresponding to the current sentence obtained by the natural language recognition module, but the present invention does not impose any restrictions on this.

[0051] Optionally, the classification score can, to some extent, characterize the credibility of the dialogue intent corresponding to the current sentence obtained by the natural language recognition module. The higher the classification score, the higher the credibility of the dialogue intent corresponding to the current sentence obtained by the natural language recognition module. This invention does not impose any limitations on this.

[0052] Optionally, the natural language recognition module of this invention can identify the sentence intent of a specific sentence to obtain the corresponding sentence intent recognition result, but it cannot combine the preceding text, that is, it cannot combine the historical sentences in the current round of dialogue to identify the sentence intent recognition result of the current sentence. Therefore, it is necessary to obtain the sentence intent recognition result of the current sentence in the current round of dialogue so as to further determine the sentence intent recognition result of the current sentence by combining it with the historical dialogue intent corresponding to the historical sentences in the current round of dialogue. This can improve the accuracy of recognizing the user's dialogue intent to a certain extent, and this invention does not limit this.

[0053] Optionally, based on the characteristics of providing services through online dialogue, when a user needs service, they will initiate a dialogue with the entity executing this invention. Once the user deems it appropriate to end the service, they may choose not to reply or instruct the entity executing this invention to end the current dialogue. The starting point of this current dialogue in this invention is the moment the user initiates the dialogue. Correspondingly, the history of sentences in this current dialogue can also be understood as all historical sentences entered by the user from the starting point of the current dialogue until the user enters the current sentence; this invention does not impose any limitations on this.

[0054] Optionally, this invention does not limit the range of classification scores, and any feasible method falls within the scope of protection of this invention. For example, a maximum score of 10 points can be used, or a thousand-point system or a hundred-point system can be used. For example, the range of classification scores is from 0 to 1.

[0055] Optionally, this invention does not impose specific restrictions on the different classification methods; any feasible method falls within the protection scope of this invention. For example, classification names may include five major categories: new category, exclusive category, modified category, coexisting category, and invalid category.

[0056] Specifically, the newly added class can include the case where the user's dialogue intent is in the newly added state by default.

[0057] The exclusive category can specifically include: (1) the user's core business dialogue intent enters the exclusive state; (2) there is only one exclusive state in the history of the current call; (3) when there is an exclusive intent in the history of the current call, there are no other states except for invalid intent, etc.

[0058] The modification category can specifically include: (1) multiple intentions that may be modified when the user's intention may need to be switched; (2) when there is a modification intention in the history of the current call, except for invalid intentions, there are no other state intentions, etc.

[0059] The coexisting classes can specifically include: (1) when the user expresses a scheme-type intent with the same slot; (2) when there are coexisting intents in the dialogue, except for invalid intents, there are no other state intents, etc.

[0060] Solution-based services refer to: collecting user information on demand to fill in slots, and then pushing the response solution to the user after the slots are filled. Task-based services refer to: collecting user information to fill in slots, executing the task after multiple rounds of interaction and confirmation of all slots, and then pushing the result after the task is completed.

[0061] Invalid categories can specifically include: (1) after the user's intent is switched, all the intents of the entire historical dialogue in the current round of the call are invalid; (2) invalid intents can coexist with any state intents, etc.

[0062] Optionally, invalid intent can be understood as intent that has no history and has a low degree of confidence; this invention does not limit this.

[0063] Optionally, this invention does not impose specific limitations on the natural language recognition module; any feasible approach falls within the scope of protection of this invention. For example, the intent recognition model can be a semantic recognition model trained based on an RNN deep learning algorithm.

[0064] S200. Determine the corresponding intent state coefficient based on the classification name;

[0065] Optionally, as mentioned above, different categories correspond to different situations. Therefore, based on the five categories of newly added, exclusive, modified, coexisting, and invalid categories described above, corresponding intent state determination rules can be set in advance.

[0066] For example, optional implementation 1, combined with Figure 1 In some optional embodiments of the illustrated implementation, step S200 includes:

[0067] Based on the category name, a corresponding intent state coefficient is determined using a preset intent state determination rule, wherein the intent state coefficient matches the category name.

[0068] Optionally, the intent state determination rules can be set in the form of a table, for example, Figure 2 As shown, through Figure 2 The table shown can determine the intent state coefficient corresponding to the current sentence, and this invention does not impose any restrictions on this.

[0069] For example, in optional implementation 2, combined with optional implementation 1, in some optional implementations, the step of determining the corresponding intent state coefficient according to the category name through a preset intent state determination rule, wherein the intent state coefficient matches the category name, includes: steps 1.1, 1.2, 1.3, 1.4 and 1.5;

[0070] Step 1.1: If the category name matches the newly added intent in the intent state determination rule, then determine the newly added coefficient in the intent state determination rule as the intent state coefficient, wherein the newly added coefficient corresponds to the newly added intent, and the newly added intent represents that the intent involved in the current sentence has never appeared in the history of the current round of conversation;

[0071] Step 1.2: If the category name matches the exclusive intent in the intent state determination rule, then the exclusive coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the exclusive coefficient corresponds to the exclusive intent, and the exclusive intent represents that the confidence level of the intent involved in the current sentence is higher than a certain threshold.

[0072] Step 1.3: If the classification name matches the modified intent in the intent state determination rule, then the modification coefficient in the intent state determination rule is determined as the intent state coefficient. The modification coefficient corresponds to the modified intent, and the modified intent indicates that the confidence of the intent involved in the current sentence is higher than a certain threshold and has no relation to the intents involved in the historical dialogue of this round of call.

[0073] Step 1.4: If the classification name matches the coexisting intent in the intent state determination rule, then determine the coexistence coefficient in the intent state determination rule as the intent state coefficient, wherein the coexistence coefficient corresponds to the coexisting intent, and the coexisting intent indicates that the intent involved in the current sentence is strongly correlated with at least one intent in the historical dialogue of the current round of the call.

[0074] Step 1.5: If the category name matches the invalid intent in the intent state determination rule, then determine the invalid coefficient in the intent state determination rule as the intent state coefficient, wherein the invalid coefficient corresponds to the invalid intent, and the invalid intent indicates that the intent involved in the current sentence has a clear intent switch.

[0075] Optionally, steps 1.1, 1.2, 1.3, 1.4, and 1.5 are performed in parallel and there is no necessary order of execution. This invention does not impose any restrictions on this.

[0076] Optionally, the new class, exclusive class, modified class, coexisting class, and invalid class mentioned in this invention can correspond sequentially to the new intent, exclusive intent, modified intent, coexisting intent, and invalid intent, or can be understood as sequentially equivalent. That is, the new class, exclusive class, modified intent, coexisting intent, and invalid class are used for description on the natural language recognition module side, while the new intent, exclusive intent, modified intent, coexisting intent, and invalid intent are used for corresponding description on the intent state management module side, for the purpose of distinction. This invention does not impose any limitations on this.

[0077] Optionally, this invention does not impose specific restrictions on the values ​​of the new coefficient, exclusive coefficient, modified coefficient, concurrent coefficient, and invalid coefficient. These values ​​can be set according to actual needs, and this invention does not impose any restrictions on them. For example, the following settings can be made: new coefficient = 2.5, exclusive coefficient = 3.5, modified coefficient = 2.5, concurrent coefficient = 3, and invalid coefficient = 2.

[0078] S300. Determine the corresponding confidence level based on the classification score and the intent state coefficient;

[0079] Optionally, as mentioned above, although the classification score can, to some extent, characterize the credibility of the dialogue intent corresponding to the current sentence identified by the natural language recognition module, the confidence score can be further calculated by the confidence calculation module since the natural language recognition module does not incorporate the historical dialogue of the current round. This invention does not limit this.

[0080] Optionally, this invention does not limit the process of calculating the confidence level; any feasible method falls within the scope of protection of this invention. For example, optional embodiment 3, combined with... Figure 1 In some optional embodiments of the illustrated implementation, S300 includes:

[0081] According to formula 1: C = log(score) i ×Z)×ratio i The confidence level is calculated, where C is the confidence level, i is the sentence number, and the score is... i The classification score is given by Z, which is a preset scaling factor, and the ratio is... i The intention state coefficient is denoted as .

[0082] S400. Based on the confidence level and the preset intent determination rule, determine the dialogue intent corresponding to the current sentence of the current round of dialogue, wherein the parameters involved in the preset intent determination rule include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue.

[0083] Optionally, the present invention may record the historical dialogue intentions of the current round of dialogue so that the intention determination rules in S400 can refer to the historical dialogue intentions to determine the dialogue intention of the current sentence. The present invention does not limit this.

[0084] Optionally, the preset intent determination rules can record different intent determination rules corresponding to different confidence levels. An intent determination rule can also record in detail the different dialogue intents corresponding to different situations, so as to determine the dialogue intent corresponding to the current sentence in the current round of dialogue. This invention does not impose any limitations on this.

[0085] For example, optional implementation 4, combined with Figure 1 In some optional embodiments, the S400 shown includes steps 2.1, 2.2, 2.3, and 2.4.

[0086] Step 2.1: Compare the confidence level with the first preset threshold, and compare the confidence level with the second preset threshold, wherein the first preset threshold is greater than the second preset threshold;

[0087] Step 2.2: If the confidence level is not less than the first preset threshold, then determine the dialogue intent according to the preset first intent determination rule;

[0088] Optionally, the present invention does not specifically limit the process by which the intent state management module determines the dialogue intent according to a preset first intent determination rule; specific details can be found in the first intent determination rule. For example, according to... Figure 3 The first intent determination rule shown can determine the dialogue intent corresponding to the current sentence, but this invention does not impose any limitations on it.

[0089] Optional, such as Figure 3 The table shown can be a first intent determination rule with a confidence level of not less than 10. Using the first intent determination rule, the dialogue intent corresponding to the current sentence can be returned, i.e. Figure 3 The present invention does not limit the scope of the "Return to Dialogue Intent" field shown.

[0090] Step 2.3: If the confidence level is less than the first preset threshold and greater than the second preset threshold, then determine the user's dialogue intent according to the preset second intent determination rule;

[0091] Optionally, the present invention does not specifically limit the process by which the intent state management module determines the dialogue intent according to a preset second intent determination rule; specific details can be found in the second intent determination rule. For example, according to... Figure 4 The second diagram shown illustrates the determination rules, which can determine the dialogue intent corresponding to the current sentence. This invention does not impose any limitations on this.

[0092] Optional, such as Figure 4 The table shown represents a second intent determination rule with a confidence level between 10 and 5. Using this second intent determination rule, the dialogue intent corresponding to the current sentence can be returned, i.e. Figure 4 The present invention does not limit the scope of the "Return to Dialogue Intent" field shown.

[0093] Step 2.4: If the confidence level is not greater than the second preset threshold, then determine the user's dialogue intent according to the preset third intent determination rule.

[0094] Optionally, the present invention does not specifically limit the process by which the intent state management module determines the dialogue intent according to a preset third intent determination rule; specific details can be found in the third intent determination rule. For example, according to... Figure 5 The third intent determination rule shown can determine the dialogue intent corresponding to the current sentence, but this invention does not impose any limitations on it.

[0095] Optional, such as Figure 5 The table shown can represent third-party intent determination rules with a confidence level no greater than 5. These rules allow us to return the dialogue intent corresponding to the current sentence. Figure 5 The present invention does not limit the scope of the "Return to Dialogue Intent" field shown.

[0096] Optionally, steps 2.2, 2.3, and 2.4 are performed in parallel and there is no necessary order of execution. This invention does not impose any restrictions on this.

[0097] Optional implementation 5, combined with Figure 1 In some optional embodiments, prior to step S100, the method further includes steps 3.1, 3.2, 3.3, and 3.4.

[0098] Step 3.1: The natural language recognition module in this system calls the preset intent recognition model to perform intent recognition on the current sentence, thereby determining the sentence intent recognition result corresponding to the current sentence. The intent recognition model is a model trained based on the RNN deep learning algorithm.

[0099] Step 3.2: The natural language recognition module sends the sentence intent recognition result to the dialogue management module in this system;

[0100] Optionally, the dialogue management module can adopt different response methods based on different dialogue intentions: 1. Solution-based business: collect user information to fill slots as needed, and push the response solution to the user after the slots are filled; 2. Task-based business: collect user information to fill slots, and after multiple rounds of interaction, all slots are filled and confirmed, the task is executed, and the result is pushed after the task is completed. This invention does not limit this.

[0101] Step 3.3: The dialogue management module sends the sentence intent recognition result to the dialogue intent decision module in this system to invoke the dialogue intent decision module;

[0102] Optionally, the dialogue intent decision module can be embedded in the dialogue management module as a sub-module, or it can be called by the dialogue management module as an independent module. This invention does not limit this.

[0103] Optionally, the dialogue intent decision module can complete the process of receiving the sentence intent recognition result and sending it to the intent state management module, as well as the subsequent process of pushing the result to the user. This invention does not limit this process.

[0104] Step 3.4: The dialogue intent decision module sends the sentence intent recognition result to the intent state management module in this system to invoke the intent state management module;

[0105] Optionally, the dialogue intent decision module sends the sentence intent recognition result to the intent state management module. This can be understood as the intent state management module obtaining the sentence intent recognition result, and the present invention does not limit this.

[0106] Optionally, after obtaining the sentence intent recognition result, the intent state management module triggers the execution of the corresponding process, including the process of calling the confidence calculation module. This invention does not limit this process.

[0107] The intent state management module, the dialogue intent decision module, and the confidence calculation module are all modules called internally by the dialogue management module, and the dialogue intent decision module calls the intent state management module, while the intent state management module calls the confidence calculation module.

[0108] Optional implementation 6, combined with Figure 1 In some optional embodiments, as illustrated, the method further includes:

[0109] Based on the dialogue intent corresponding to the current sentence in the current round of dialogue, update the historical dialogue intent corresponding to the historical sentences in the current round of dialogue.

[0110] Optionally, as mentioned above, when the preset intent determination rules involve historical dialogue intents—that is, when the intent state management module determines the dialogue intent corresponding to the current sentence of the current round of dialogue based on the confidence level and the preset intent determination rules—historical dialogue intents are used. Therefore, the accuracy of historical dialogue intents can affect the final execution effect of this invention to some extent. Thus, the historical dialogue intents corresponding to the historical sentences of the current round of dialogue can be further updated based on the dialogue intent corresponding to the current sentence. For details, please refer to [link to relevant documentation]. Figure 3 , Figure 4 and Figure 5 The present invention does not limit the scope of the "Dialogue Intent Update Logic" section.

[0111] like Figure 6 As shown, the present invention provides a device for recognizing user dialogue intent, including: an intent state management module 100 and a confidence calculation module 200;

[0112] The intent state management module 100 includes: a receiving unit 110, a coefficient determination unit 120, and an intent determination unit 130;

[0113] The receiving unit 110 is configured to obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score;

[0114] The coefficient determination unit 120 is configured to determine the corresponding intent state coefficient based on the category name;

[0115] The confidence calculation module 200 is configured to determine the corresponding confidence level based on the classification score and the intent state coefficient.

[0116] The intent determination unit 130 is configured to determine the dialogue intent corresponding to the current sentence of the current round of dialogue based on the confidence level and a preset intent determination rule, wherein the parameters involved in the preset intent determination rule include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue.

[0117] Combination Figure 6 In some optional embodiments, the coefficient determination unit 120 shown includes a coefficient determination subunit.

[0118] The coefficient determination subunit is configured to determine the corresponding intent state coefficient based on the category name and a preset intent state determination rule, wherein the intent state coefficient matches the category name.

[0119] In conjunction with the previous implementation method, in some optional implementation methods, the coefficient determining subunit is specifically configured to perform:

[0120] If the category name matches the newly added intent in the intent state determination rule, then the newly added coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the newly added coefficient corresponds to the newly added intent, and the newly added intent represents that the intent involved in the current sentence has never appeared in the history of the current round of conversation;

[0121] If the category name matches the exclusive intent in the intent state determination rule, then the exclusive coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the exclusive coefficient corresponds to the exclusive intent, and the exclusive intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold.

[0122] If the classification name matches the modified intent in the intent state determination rule, then the modification coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the modification coefficient corresponds to the modified intent, and the modified intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold and has no relation to the intents involved in the historical dialogue of this round of call;

[0123] If the category name matches the coexisting intent in the intent state determination rule, then the coexistence coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the coexistence coefficient corresponds to the coexisting intent, and the coexisting intent indicates that the intent involved in the current sentence is strongly correlated with at least one intent in the historical dialogue of the current round of the call.

[0124] If the category name matches the invalid intent in the intent state determination rule, then the invalid coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the invalid coefficient corresponds to the invalid intent, and the invalid intent indicates that the intent involved in the current sentence has a clear intent switch.

[0125] Combination Figure 6 In some optional embodiments of the implementation shown, the confidence calculation module 200 is specifically configured to perform:

[0126] According to formula 1: C = log(score) i ×Z)×ratio i The confidence level is calculated, where C is the confidence level, i is the sentence number, and the score is... i The classification score is given by Z, which is a preset scaling factor, and the ratio is... iThe intention state coefficient is denoted as .

[0127] Combination Figure 6 In some optional embodiments, the intent determination unit 130, as shown in the implementation, includes: a comparison subunit, a first determination subunit, a second determination subunit, and a third determination subunit;

[0128] The comparison subunit is configured to perform a comparison between the confidence level and a first preset threshold, and a comparison between the confidence level and a second preset threshold, wherein the first preset threshold is greater than the second preset threshold;

[0129] The first determining subunit is configured to determine the dialogue intent according to a preset first intent determination rule if the confidence level is not less than the first preset threshold.

[0130] The second determining subunit is configured to perform the following action: if the confidence level is less than the first preset threshold and greater than the second preset threshold, then determine the user's dialogue intent according to a preset second intent determination rule.

[0131] The third determining subunit is configured to determine the user's dialogue intent according to a preset third intent determination rule if the confidence level is not greater than the second preset threshold.

[0132] Combination Figure 6 In some optional embodiments, the apparatus further includes a natural language recognition module, a dialogue management module, and a dialogue intent decision module, as shown in the embodiments.

[0133] The natural language recognition module includes: a calling submodule and a sending subunit;

[0134] The calling submodule is configured to call a preset intent recognition model to perform intent recognition on the current sentence before the intent state management module 100 obtains the sentence intent recognition result output by the natural language recognition module for the current sentence of the current round of dialogue, thereby determining the sentence intent recognition result corresponding to the current sentence. The intent recognition model is a model trained based on the RNN deep learning algorithm.

[0135] The sending subunit is configured to send the sentence intent recognition result to the dialogue management module;

[0136] The dialogue management module is configured to send the sentence intent recognition result to the dialogue intent decision module to invoke the dialogue intent decision module;

[0137] The dialogue intent decision module is configured to send the sentence intent recognition result to the intent state management module 100 to invoke the intent state management module 100;

[0138] The intent state management module 100, the dialogue intent decision module, and the confidence calculation module 200 are all modules called internally by the dialogue management module. The dialogue intent decision module calls the intent state management module 100, and the intent state management module 100 calls the confidence calculation module 200.

[0139] Combination Figure 6 In some optional embodiments, the apparatus further includes an intent update unit;

[0140] The intent update unit is configured to update the historical dialogue intent corresponding to the historical sentences of the current round of dialogue based on the dialogue intent corresponding to the current sentence of the current round of dialogue.

[0141] The present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the method for recognizing user dialogue intent as described in any of the preceding claims.

[0142] like Figure 7 As shown, the present invention provides an electronic device 70, which includes at least one processor 701, at least one memory 702 and a bus 703 connected to the processor 701; wherein the processor 701 and the memory 702 communicate with each other through the bus 703; the processor 701 is used to call program instructions in the memory 702 to execute the user dialogue intent recognition method described above.

[0143] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0144] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0145] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0146] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for identifying user dialogue intent, characterized in that, include: Obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score; Based on the category name, a corresponding intent state coefficient is determined using a preset intent state determination rule, wherein the intent state coefficient matches the category name; The corresponding confidence level is determined based on the classification score and the intent state coefficient; Based on the confidence level and the preset intent determination rules, the dialogue intent corresponding to the current sentence of the current round of dialogue is determined, wherein the parameters involved in the preset intent determination rules include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue; The step of determining the corresponding intent state coefficient based on the category name and a preset intent state determination rule includes: If the category name matches the newly added intent in the intent state determination rule, then the newly added coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the newly added coefficient corresponds to the newly added intent, and the newly added intent represents that the intent involved in the current sentence has never appeared in the history of the current round of conversation; If the category name matches the exclusive intent in the intent state determination rule, then the exclusive coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the exclusive coefficient corresponds to the exclusive intent, and the exclusive intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold. If the classification name matches the modified intent in the intent state determination rule, then the modification coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the modification coefficient corresponds to the modified intent, and the modified intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold and has no relation to the intents involved in the historical dialogue of this round of call; If the category name matches the coexisting intent in the intent state determination rule, then the coexistence coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the coexistence coefficient corresponds to the coexisting intent, and the coexisting intent indicates that the intent involved in the current sentence is strongly correlated with at least one intent in the historical dialogue of the current round of the call. If the category name matches the invalid intent in the intent state determination rule, then the invalid coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the invalid coefficient corresponds to the invalid intent, and the invalid intent indicates that the intent involved in the current sentence has a clear intent switch; The step of determining the corresponding confidence level based on the classification score and the intent state coefficient includes: According to Formula 1: The confidence level is calculated, wherein the confidence level is... For the confidence level, the For sentence labeling, the stated For the classification score, the The preset scaling factor, the The intention state coefficient is denoted as .

2. The method according to claim 1, characterized in that, The step of determining the dialogue intent corresponding to the current sentence in the current round of dialogue based on the confidence level and preset intent determination rules includes: The confidence level is compared with a first preset threshold, and the confidence level is compared with a second preset threshold, wherein the first preset threshold is greater than the second preset threshold; If the confidence level is not less than the first preset threshold, then the dialogue intent is determined according to the preset first intent determination rule; If the confidence level is less than the first preset threshold and greater than the second preset threshold, then the user's dialogue intent is determined according to the preset second intent determination rule; If the confidence level is not greater than the second preset threshold, then the user's dialogue intent is determined according to the preset third intent determination rule.

3. The method according to claim 1, characterized in that, The method further includes: Based on the dialogue intent corresponding to the current sentence in the current round of dialogue, update the historical dialogue intent corresponding to the historical sentences in the current round of dialogue.

4. A device for recognizing user dialogue intent, characterized in that, include: Intent state management module and confidence calculation module; The intent state management module includes: a receiving unit, a coefficient determination unit, and an intent determination unit; The receiving unit is configured to obtain the sentence intent recognition result for the current sentence in the current round of dialogue, wherein the sentence intent recognition result includes: category name and category score; The coefficient determination unit is configured to determine the corresponding intent state coefficient based on the category name and a preset intent state determination rule, wherein the intent state coefficient matches the category name. The confidence calculation module is configured to determine the corresponding confidence level based on the classification score and the intent state coefficient. The intent determination unit is configured to determine the dialogue intent corresponding to the current sentence of the current round of dialogue based on the confidence level and a preset intent determination rule, wherein the parameters involved in the preset intent determination rule include: the historical dialogue intent corresponding to the historical sentences of the current round of dialogue; The coefficient determining unit, which determines the corresponding intent state coefficient based on the category name and a preset intent state determination rule, is specifically used for: If the category name matches the newly added intent in the intent state determination rule, then the newly added coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the newly added coefficient corresponds to the newly added intent, and the newly added intent represents that the intent involved in the current sentence has never appeared in the history of the current round of conversation; If the category name matches the exclusive intent in the intent state determination rule, then the exclusive coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the exclusive coefficient corresponds to the exclusive intent, and the exclusive intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold. If the classification name matches the modified intent in the intent state determination rule, then the modification coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the modification coefficient corresponds to the modified intent, and the modified intent represents that the confidence of the intent involved in the current sentence is higher than a certain threshold and has no relation to the intents involved in the historical dialogue of this round of call; If the category name matches the coexisting intent in the intent state determination rule, then the coexistence coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the coexistence coefficient corresponds to the coexisting intent, and the coexisting intent indicates that the intent involved in the current sentence is strongly correlated with at least one intent in the historical dialogue of the current round of the call. If the category name matches the invalid intent in the intent state determination rule, then the invalid coefficient in the intent state determination rule is determined as the intent state coefficient, wherein the invalid coefficient corresponds to the invalid intent, and the invalid intent indicates that the intent involved in the current sentence has a clear intent switch; The confidence calculation module, which determines the corresponding confidence level based on the classification score and the intent state coefficient, is specifically used for: According to Formula 1: The confidence level is calculated, wherein the confidence level is... For the confidence level, the For sentence labeling, the stated For the classification score, the The preset scaling factor, the The intention state coefficient is denoted as .

5. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the method for recognizing user dialogue intent as described in any one of claims 1 to 3.

6. An electronic device, characterized in that, The electronic device includes at least one processor, and at least one memory and bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the user dialogue intent recognition method as described in any one of claims 1 to 3.

Images