Identification information generation method, device, electronic device, and computer-readable medium

Abstract

The embodiments of the present disclosure disclose a method, apparatus, electronic device, and computer-readable medium for generating recognition information. A specific implementation of the method includes: acquiring a target user's voice, wherein the target user's voice is the voice of a target conversation user in response to a question; generating recognition text corresponding to the target user's voice as the target recognition text; determining the probability of the target recognition text appearing in a historical conversation text set as the target probability; and generating recognition information indicating whether the target recognition text was correctly recognized based on the target probability. This implementation is related to intelligent customer service and can quickly and efficiently determine whether the target recognition text was correctly recognized.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and more particularly to a method, device, electronic device, and computer-readable medium for generating identification information. Background Art

[0002] Natural language conversational interactions based on question-and-answer (Q&A) are increasingly being applied in real life. Examples include logistics delivery bookings, user research follow-up, and in-vehicle robots in smart cars. Natural language conversational interactions begin with user speech recognition. This is typically achieved by leveraging automatic speech recognition (ASR) technology to generate text corresponding to the user's speech.

[0003] However, when using the above method to recognize user voice, the following technical problems often occur:

[0004] Speech recognition based on automatic speech recognition technology is the key starting point for natural language conversational interaction. However, it often results in recognition errors due to factors such as background sound, noise, or multiple people speaking, which in turn affects the subsequent progress of the entire conversation process. Summary of the Invention

[0005] The content of this disclosure is used to briefly introduce concepts that will be described in detail in the detailed description section below. The content of this disclosure is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] Some embodiments of the present disclosure propose identification information generation methods, devices, electronic devices, and computer-readable media to solve the technical problems mentioned in the above background technology section.

[0007] In a first aspect, some embodiments of the present disclosure provide a method for generating recognition information, including: obtaining a target user voice, wherein the target user voice is the voice of a target dialogue user in response to a system question; generating a recognition text corresponding to the target user voice as a target recognition text; determining the probability of the target recognition text appearing in a historical dialogue text set as a target probability; and generating recognition information according to the target probability, which represents whether the target recognition text is correctly recognized.

[0008] Optionally, the above-mentioned determination of the probability of the target recognition text appearing in the historical conversation text set as the target probability includes: determining a historical conversation text group in the historical conversation text set that includes the target recognition text; determining the number of texts of the historical conversation texts in the historical conversation text group as the first text number; determining the number of texts of the historical conversation texts in the historical conversation text set as the second text number; and determining the above-mentioned target probability based on the above-mentioned first text number and the above-mentioned second text number.

[0009] Optionally, determining the target probability based on the first number of texts and the second number of texts includes: adding the first number of texts and the second number of texts to predetermined values respectively to obtain the added first number of texts and the added second number of texts; dividing the added first number of texts by the added second number of texts to obtain the divided value as the target probability.

[0010] Optionally, generating recognition information representing whether the target recognition text is correctly recognized based on the target probability includes: in response to determining that the target probability is greater than or equal to a first threshold, generating recognition information representing that the target recognition text is correctly recognized.

[0011] Optionally, the above-mentioned generation of recognition information characterizing whether the above-mentioned target recognition text is correctly recognized based on the above-mentioned target probability includes: in response to determining that the above-mentioned target probability is less than a second threshold, generating recognition information characterizing that the above-mentioned target recognition text is incorrectly recognized, wherein the above-mentioned second threshold is less than or equal to the above-mentioned first threshold.

[0012] Optionally, the historical conversation texts in the above-mentioned historical conversation text set include: the above-mentioned system questions and recognition texts; and the above-mentioned generation of recognition information representing whether the above-mentioned target recognition text is correctly recognized based on the above-mentioned target probability includes: determining the historical conversation text group in the above-mentioned historical conversation text set that includes the above-mentioned target recognition text; determining the historical conversation texts whose corresponding recognition texts are the same as the above-mentioned target recognition texts from the above-mentioned historical conversation text set to obtain a subset of historical conversation texts; screening out historical conversation texts that meet preset text conditions from the above-mentioned historical conversation text group subset as target historical conversation texts; determining the probability corresponding to the probability that the recognition text in the above-mentioned target historical conversation text appears in at least one historical conversation text as the historical probability, wherein the above-mentioned at least one historical conversation text is the historical conversation text in the above-mentioned historical conversation text group excluding the above-mentioned target historical conversation text; generating recognition information representing whether the above-mentioned target recognition text is correctly recognized based on the above-mentioned historical probability and the above-mentioned target probability.

[0013] Optionally, the above-mentioned generation of recognition information characterizing whether the above-mentioned target recognition text is correctly recognized based on the above-mentioned historical probability and the above-mentioned target probability includes: multiplying the above-mentioned historical probability by a value α to obtain a multiplied historical probability, wherein the value range of the above-mentioned value α is [0, 1]; multiplying the above-mentioned target probability by a value 1-α to obtain a multiplied probability; adding the above-mentioned multiplied historical probability and the above-mentioned multiplied probability to obtain an added probability; in response to determining that the above-mentioned added probability is greater than or equal to a first threshold, generating recognition information characterizing that the above-mentioned target recognition text is correctly recognized.

[0014] Optionally, the above method also includes: using the above recognition information to adjust the training data set of the above speech recognition model to obtain an adjusted training data set; using the above adjusted training data set to adjust the parameters of the above speech recognition model to obtain an adjusted speech recognition model.

[0015] Optionally, the method further includes: generating semantic understanding information based on the recognition information and the target recognition text.

[0016] In a second aspect, some embodiments of the present disclosure provide a recognition information generating device, including: an acquisition unit, configured to acquire a target user voice, wherein the target user voice is the voice of the target dialogue user in response to a question; a first generation unit, configured to generate a recognition text corresponding to the target user voice as a target recognition text; a determination unit, configured to determine the probability of the target recognition text appearing in a historical dialogue text set as a target probability; and a second generation unit, configured to generate recognition information representing whether the target recognition text is correctly recognized based on the target probability.

[0017] Optionally, the determination unit can be configured to: determine a historical conversation text group in the above-mentioned historical conversation text set that includes the target recognition text; determine the number of texts of the historical conversation texts in the above-mentioned historical conversation text group as the first text number; determine the number of texts of the historical conversation texts in the above-mentioned historical conversation text set as the second text number; and determine the above-mentioned target probability based on the above-mentioned first text number and the above-mentioned second text number.

[0018] Optionally, the determination unit can be configured to: add the above-mentioned first text number and the above-mentioned second text number to predetermined values respectively to obtain the added first text number and the added second text number; divide the above-mentioned added first text number by the above-mentioned added second text number to obtain the divided value as the above-mentioned target probability.

[0019] Optionally, the second generating unit may be configured to: in response to determining that the target probability is greater than or equal to a first threshold, generate recognition information indicating that the target recognition text is correctly recognized.

[0020] Optionally, the second generating unit may be configured to: in response to determining that the target probability is less than a second threshold, generate recognition information characterizing a recognition error of the target recognition text, wherein the second threshold is less than or equal to the first threshold.

[0021] Optionally, the historical conversation texts in the historical conversation text set include the question and the recognition text. The second generation unit may be configured to: determine a historical conversation text group in the historical conversation text set that includes the target recognition text; filter out historical conversation texts that meet preset text conditions from the subset of the historical conversation text group as target historical conversation texts; determine a probability corresponding to a probability that the recognition text in the target historical conversation text appears in at least one historical conversation text as a historical probability, wherein the at least one historical conversation text is the historical conversation text in the historical conversation text group excluding the target historical conversation text; and generate recognition information indicating whether the target recognition text is correctly recognized based on the historical probability and the target probability.

[0022] Optionally, the second generation unit can be configured to: multiply the above historical probability by a value α to obtain a multiplied historical probability, wherein the value range of the above value α is [0, 1]; multiply the above target probability by a value 1-α to obtain a multiplied probability; add the above multiplied historical probability and the above multiplied probability to obtain an added probability; in response to determining that the above added probability is greater than or equal to a first threshold, generate recognition information representing that the above target recognition text is correctly recognized.

[0023] Optionally, the above-mentioned device also includes: using the above-mentioned recognition information to adjust the training data set of the above-mentioned speech recognition model to obtain an adjusted training data set; using the above-mentioned adjusted training data set to adjust the parameters of the above-mentioned speech recognition model to obtain an adjusted speech recognition model.

[0024] Optionally, the above-mentioned device also includes: generating semantic understanding information based on the above-mentioned recognition information and the above-mentioned target recognition text.

[0025] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0026] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner in the first aspect is implemented.

[0027] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: through the recognition information generation method of some embodiments of the present disclosure, it is possible to accurately determine whether the speech of the target dialogue user is recognized correctly. Specifically, the reason for the inaccuracy of speech recognition is that speech recognition based on automatic speech recognition technology, as the starting key node of natural language conversational interaction, often has recognition errors due to factors such as background sound, noise or multiple people talking, which in turn affects the subsequent development of the entire dialogue process. Based on this, after speech recognition, the recognition information generation method of some embodiments of the present disclosure accurately determines whether the recognized target recognition text is correct by determining the probability of the target recognition text appearing in the historical dialogue text set. By determining whether the target recognition text is correctly recognized through the above-mentioned target probability, it can largely avoid the situation where recognition errors due to factors such as background sound, noise, or multiple people talking, and improve the accuracy of determining whether the target recognition text is correctly recognized. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0029] Figure 1 is a schematic diagram of an application scenario of the identification information generation method according to some embodiments of the present disclosure;

[0030] Figure 2 is a flow chart of some embodiments of the identification information generation method according to the present disclosure;

[0031] Figure 3 is a flow chart of other embodiments of the identification information generation method according to the present disclosure;

[0032] Figure 4 is a flowchart of some further embodiments of the identification information generation method according to the present disclosure;

[0033] Figure 5 is a schematic structural diagram of some embodiments of the identification information generating device according to the present disclosure;

[0034] Figure 6 It is a structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0035] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0036] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other.

[0037] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0038] It should be noted that the modifications of "one" and "multiple" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0039] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0040] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0041] Figure 1 It is a schematic diagram of an application scenario of the identification information generation method according to some embodiments of the present disclosure.

[0042] exist Figure 1In the application scenario, the electronic device 101 can first obtain the target user voice 102. The target user voice 102 is the voice of the target dialogue user 103 in response to the question 104. In this application scenario, the target dialogue user 103 can be "Li**". The question 104 can be "Can I deliver it to you tomorrow?". Then, the electronic device 101 can generate the recognition text corresponding to the target user voice 102 as the target recognition text 105. In this application scenario, the target recognition text 105 can be "OK". Next, the electronic device 101 can determine the probability of the target recognition text 105 appearing in the historical dialogue text set 106 as the target probability 107. In this application scenario, the target probability 107 can be "70%". Finally, the electronic device 101 can generate recognition information 108 based on the target probability 107 to indicate whether the target recognition text 105 is correctly recognized.

[0043] It should be noted that the electronic device 101 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0044] It should be understood that Figure 1 The number of electronic devices in the embodiment is merely illustrative. Any number of electronic devices may be provided according to implementation requirements.

[0045] Continue to refer Figure 2 , shows a process 200 of some embodiments of the identification information generation method according to the present disclosure. The identification information generation method includes the following steps:

[0046] Step 201: Acquire the target user's voice.

[0047] In some embodiments, the execution subject of the above identification information generation method (for example Figure 1 The electronic device 101 shown can obtain the target user's voice through a wired connection or a wireless connection. The target user's voice is the voice of the target dialogue user in response to the question. The target dialogue user can be a user communicating with the intelligent customer service system. The question can be a question posed by the intelligent customer service system to the target dialogue user. For example, the question can be "Can I deliver the goods to you tomorrow?" The intelligent customer service system can be a customer service system that performs human-computer interaction with the user.

[0048] Step 202: Generate the recognition text corresponding to the target user's voice as the target recognition text.

[0049] In some embodiments, the execution entity may generate a recognition text corresponding to the target user's voice as the target recognition text. The target recognition text may be a text converted from the target user's voice. For example, the target recognition text may be "OK."

[0050] As an example, the execution entity may utilize automatic speech recognition technology to generate a recognition text corresponding to the target user's speech as the target recognition text.

[0051] Step 203: Determine the probability of the target recognition text appearing in the historical conversation text set as the target probability.

[0052] In some embodiments, the execution entity may determine a probability of the target recognized text appearing in a set of historical conversation texts as a target probability. The historical conversation texts may include the recognized text corresponding to the voice responses of historical users to the aforementioned questions. Historical users are users who have responded to the aforementioned questions.

[0053] As an example, the execution subject determines the probability of the target recognition text appearing in the historical conversation text set as the target probability, which may include the following steps:

[0054] In the first step, the execution entity may determine the number of times the target recognition text appears in each historical conversation text in the historical conversation text set.

[0055] For example, the target recognition text appears four times. The historical conversation text may include multiple target recognition texts. For example, the historical conversation text may be "OK, OK." The target recognition text is "OK." Therefore, the target recognition text "OK" appears twice in the historical conversation text "OK, OK."

[0056] In a second step, the execution entity may determine the number of historical dialogue texts in the historical dialogue text set.

[0057] For example, the number of the above historical conversation texts is 10.

[0058] The third step is to divide the above number of times by the above number to obtain the divided value as the above target probability.

[0059] For example, the number of times is 4 and the number is 10. Therefore, the probability is "40%." In response to determining that the number of times is greater than or equal to the number, the execution entity may determine the division value to be "100%" as the target probability.

[0060] In some optional implementations of some embodiments, determining the probability of the target recognition text appearing in the historical conversation text set as the target probability may include the following steps:

[0061] In a first step, the execution entity may determine a historical conversation text group including the target recognition text in the historical conversation text set.

[0062] As an example, a set of historical conversation texts includes: a first historical conversation text, a second historical conversation text, and a third historical conversation text. The target recognition text is "OK." The first historical conversation text is "OK." The second historical conversation text is "OK." The third historical conversation text is "OK, OK." Therefore, the historical conversation text group includes: the second historical conversation text and the third historical conversation text.

[0063] In a second step, the execution entity may determine the number of texts of the historical conversation texts in the historical conversation text group as the first text number.

[0064] As an example, the historical conversation text group includes: a second historical conversation text and a third historical conversation text. Then, the number of historical conversation texts in the historical conversation text group is "2".

[0065] In a third step, the execution entity may determine the number of historical conversation texts in the historical conversation text set as the second number of texts.

[0066] As an example, the historical conversation text set includes: a first historical conversation text, a second historical conversation text, and a third historical conversation text. Then the number of historical conversation texts in the historical conversation text set is "3".

[0067] In the fourth step, the execution entity determines the target probability according to the number of the first texts and the number of the second texts.

[0068] In response to the number of the first texts being 0, the target probability may be directly determined as a target probability value, for example, "100%".

[0069] As an example, the execution entity may directly divide the number of the first texts by the number of the second texts to obtain the divided value as the target probability. For example, the number of the first texts is "2" and the number of the second texts is "3". The target probability is "66.6%".

[0070] Optionally, determining the target probability according to the number of the first texts and the number of the second texts may include the following steps:

[0071] In the first step, the execution entity may add the number of first texts and the number of second texts to a predetermined value, respectively, to obtain the added number of first texts and the added number of second texts. For example, the predetermined value may be "1." The number of first texts is "2." The number of second texts is "3." The added number of first texts is "3." The added number of second texts is "4."

[0072] In a second step, the execution entity may divide the added number of first texts by the added number of second texts to obtain a division value as the target probability.

[0073] As an example, the number of the first text after the addition is "3", the number of the second text after the addition is "4", and the target probability is "75%".

[0074] Step 204: Based on the target probability, generate recognition information indicating whether the target recognition text is correctly recognized.

[0075] In some embodiments, the execution entity may generate recognition information indicating whether the target recognition text is correctly recognized based on the target probability.

[0076] As an example, the execution entity may first add a preset probability value to the target probability. For example, the preset probability value is "1%." Then, in response to determining that the added probability value is greater than or equal to a first threshold, recognition information is generated indicating that the target text is correctly recognized. For example, the first threshold is "70%." In response to determining that the added probability value is less than the first threshold, recognition information is generated indicating that the target text is incorrectly recognized.

[0077] In some optional implementations of some embodiments, in response to determining that the target probability is greater than or equal to a first threshold, the execution entity may generate recognition information indicating that the target recognition text is correctly recognized. For example, the first threshold may be "70%."

[0078] Optionally, in response to determining that the target probability is less than a second threshold, the execution entity may generate recognition information indicating a recognition error in the target recognition text. The second threshold is less than or equal to the first threshold. For example, the second threshold may be 50% or 70%.

[0079] In some optional implementations of some embodiments, the execution entity may generate semantic understanding information based on the recognition information and the target recognition text. The semantic understanding information may be the output of a semantic understanding module. The semantic understanding module may include at least one of the following: a natural language understanding (NLU) module, a dialogue management (DM) module, and a natural language generation (NLG) module.

[0080] As an example, the execution entity may feed the recognition information and the target recognition text back to the semantic understanding model to generate semantic understanding information. For example, the semantic understanding model may be an intent classification model (i.e., a model in the natural language understanding module that implements intent classification). The execution entity may feed the recognition information and the target recognition text back to the intent classification model to generate more accurate semantic understanding information. Specifically, the intent classification model may generate a proportional probability that the target recognition text belongs to the target category. The execution entity may determine whether the proportional probability has reference value based on the recognition information. For example, if the recognition information indicates that the target recognition text is correctly recognized, the proportional probability has reference value. If the recognition information indicates that the target recognition text is incorrectly recognized, the proportional probability has no reference value. In addition, the intent classification model may generate an initial proportional probability that the target recognition text belongs to the target category. The execution entity may multiply the proportional probability by the probability corresponding to the recognition information, and use the multiplication result as the final proportional probability that the target recognition text belongs to the target category. In this way, intent classification is achieved through the final proportional probability.

[0081] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: through the recognition information generation method of some embodiments of the present disclosure, it is possible to accurately determine whether the speech of the target dialogue user is recognized correctly. Specifically, the reason for the inaccuracy of speech recognition is that speech recognition based on automatic speech recognition technology, as the starting key node of natural language conversational interaction, often has recognition errors due to factors such as background sound, noise or multiple people talking, which in turn affects the subsequent development of the entire dialogue process. Based on this, after speech recognition, the recognition information generation method of some embodiments of the present disclosure accurately determines whether the recognized target recognition text is correct by determining the probability of the target recognition text appearing in the historical dialogue text set. By determining whether the target recognition text is correctly recognized through the above-mentioned target probability, it can largely avoid the situation where recognition errors due to factors such as background sound, noise, or multiple people talking, and improve the accuracy of determining whether the target recognition text is correctly recognized.

[0082] Further references Figure 3 , shows a process 300 of another embodiment of the identification information generation method according to the present disclosure. The identification information generation method includes the following steps:

[0083] Step 301: Acquire the target user's voice.

[0084] Step 302: Generate the recognition text corresponding to the target user's voice as the target recognition text.

[0085] Step 303: Determine the probability of the target recognition text appearing in the historical conversation text set as the target probability.

[0086] In some embodiments, the specific implementation of steps 301-303 and the technical effects thereof can be referred to in Figure 2 Steps 201-203 in the corresponding embodiment will not be repeated here.

[0087] Step 304: Determine a historical conversation text group in the historical conversation text set that includes the target recognition text.

[0088] In some embodiments, the execution entity (e.g. Figure 1The electronic device 101 shown can determine a group of historical conversation texts in the historical conversation text set that includes the target recognition text. The historical conversation texts in the historical conversation text set include the question and the recognition text. Here, the historical conversation text group refers to the historical conversation texts in the historical conversation text set. The conversation text generation time corresponding to the historical conversation texts in the historical conversation text group is earlier than the conversation text generation time corresponding to the target conversation text. Optionally, the historical conversation texts can be text pairs. For example, a historical conversation text may be "question, recognition text." In addition, each historical conversation text has a conversation text generation time. The conversation text generation time is the time when the historical user corresponding to the historical conversation text responded to the question. For example, the historical conversation text set includes a first historical conversation text corresponding to a first historical user, a second historical conversation text corresponding to a second historical user, and a third historical conversation text corresponding to a third historical user. The target recognition text may be "OK." The first historical conversation text corresponding to the first historical user may be "Can I deliver it to you tomorrow? Yes." The second historical conversation text corresponding to the second historical user may be "Can I deliver it to you tomorrow? Yes." The first historical conversation text corresponding to the third historical user may be "Can I deliver it to you tomorrow? Yes." The conversation text generation time corresponding to the first historical conversation text may be "November 1, 2012." The conversation text generation time corresponding to the second historical conversation text may be "November 2, 2012." The conversation text generation time corresponding to the third historical conversation text may be "November 3, 2012." The historical conversation text group then includes: the second historical conversation text corresponding to the second historical user and the third historical conversation text corresponding to the third historical user.

[0089] Step 305 , selecting historical conversation texts that meet preset text conditions from the above historical conversation text group as target historical conversation texts.

[0090] In some embodiments, the execution entity may select a historical conversation text that meets a preset text condition from the historical conversation text group as the target historical conversation text. The preset text condition may be that the generation time of the corresponding conversation text of the historical conversation text is the same as the generation time of the latest conversation text.

[0091] For example, a historical conversation text group includes a second historical conversation text corresponding to a second historical user and a third historical conversation text corresponding to a third historical user. The conversation text generation time corresponding to the second historical conversation text may be "November 2, 2012." The conversation text generation time corresponding to the third historical conversation text may be "November 3, 2012." The target historical conversation text is then the third historical conversation text.

[0092] Step 306: Determine the probability that the recognized text in the target historical conversation text appears in at least one historical conversation text as the historical probability.

[0093] In some embodiments, the execution entity may determine a probability that the recognized text in the target historical conversation text appears in at least one historical conversation text, as a historical probability, for example, the historical probability is "40%."

[0094] Optionally, the above historical probability can be generated in the same manner as the target probability. Specifically, the historical probability can be generated by the following steps:

[0095] The first step is to obtain a set of historical conversation texts for the target historical conversation text as a candidate historical conversation text set, wherein the conversation text generation time of the candidate historical conversation texts in the candidate historical conversation text set is earlier than the conversation text generation time of the target historical conversation text.

[0096] The second step is to determine the candidate historical conversation text group that includes the target recognition text in the above candidate historical conversation text set.

[0097] The third step is to determine the number of candidate historical conversation texts in the candidate historical conversation text group as the third text number.

[0098] The fourth step is to determine the number of candidate historical dialogue texts in the candidate historical dialogue text set as the fourth text number.

[0099] The fifth step is to determine the historical probability based on the third text number and the fourth text number.

[0100] Step 307: Generate recognition information indicating whether the target recognition text is correctly recognized based on the historical probability and the target probability.

[0101] In some embodiments, the execution entity may generate recognition information indicating whether the target recognition text is correctly recognized based on the historical probability and the target probability.

[0102] As an example, first, the execution subject may determine the difference in probability value between the historical probability and the target probability. Secondly, in response to determining that the probability difference is a negative value or 0, and the target probability is greater than or equal to a first threshold, the execution subject may generate recognition information representing that the target recognition text is correctly recognized. In addition, in response to determining that the probability difference is a positive value, and the target probability is less than the first threshold, the execution subject may generate recognition information representing that the target recognition text is incorrectly recognized. In addition, in response to determining that the probability difference is a negative value or 0, and the target probability is less than the first threshold, the execution subject may generate recognition information representing that the target recognition text is correctly recognized. Next, in response to determining that the probability difference is a positive value, and the target probability is greater than the first threshold, the execution subject may generate recognition information representing that the target recognition text is incorrectly recognized.

[0103] In some optional implementations of some embodiments, generating recognition information indicating whether the target recognition text is correctly recognized based on the historical probability and the target probability may include the following steps:

[0104] In the first step, the execution entity may multiply the historical probability by a value α to obtain the multiplied historical probability, wherein the value α has a range of [0, 1].

[0105] For example, the value α can be "0.4", the historical probability is "40%", and the multiplied historical probability is "16%".

[0106] In the second step, the execution entity may multiply the target probability by the value 1-α to obtain the multiplied probability.

[0107] For example, if the target probability is "60%", the value 1-α is "60%", and the probability after multiplication is "36%".

[0108] In a third step, the execution entity may add the multiplied historical probability and the multiplied probability to obtain the added probability.

[0109] For example, if the historical probability after multiplication is "16%", the probability after multiplication is "36%", and the probability after addition is "52%".

[0110] In a fourth step, in response to determining that the added probability is greater than or equal to the first threshold, the execution entity may generate recognition information indicating that the target recognition text is correctly recognized.

[0111] For example, the first threshold is "50%".

[0112] Optionally, in response to determining that the added probability is less than a second threshold, the execution entity may generate recognition information indicating a recognition error in the target recognition text.

[0113] The second threshold is less than or equal to the first threshold, for example, the second threshold is 40%.

[0114] from Figure 3 It can be seen that Figure 2 Compared with the description of some corresponding embodiments, Figure 3 The identification information generation method process 300 in some corresponding embodiments further highlights the specific steps for generating identification information based on historical probabilities and target probabilities. Thus, the solutions described in these embodiments utilize the historical probabilities corresponding to target historical conversation texts in the historical conversation text set and the aforementioned probabilities to generate more accurate and valuable identification information.

[0115] Further references Figure 4 , shows a process 400 of some further embodiments of the identification information generation method according to the present disclosure. The identification information generation method includes the following steps:

[0116] Step 401: Acquire the target user's voice.

[0117] Step 402: Generate the recognition text corresponding to the target user's voice as the target recognition text.

[0118] Step 403: Determine the probability of the target recognition text appearing in the historical conversation text set as the target probability.

[0119] Step 404: generating recognition information indicating whether the target recognition text is correctly recognized based on the target probability.

[0120] In some embodiments, the specific implementation of steps 401-404 and the technical effects thereof can be referred to in Figure 2 Steps 201-204 in the corresponding embodiment will not be repeated here.

[0121] Step 405 : Using the recognition information, adjust the training data set of the speech recognition model to obtain an adjusted training data set.

[0122] In some embodiments, the execution entity (e.g. Figure 1The electronic device 101 shown can use the above-mentioned recognition information to adjust the training data set of the above-mentioned speech recognition model to obtain an adjusted training data set. The above-mentioned speech recognition model may include: an acoustic model and a language model. The above-mentioned acoustic model can be, but is not limited to, one of the following: an end-to-end acoustic model, a Gaussian Mixed Model (GMM) + a Hidden Markov Model (HMM), or a Deep Neural Network (DNN) + HMM. The language model can be, but is not limited to, one of the following: an N-Gram language model.

[0123] As an example, the above-mentioned execution entity can assign corresponding data weights to each training data in the training data set of the above-mentioned speech recognition model according to the target probability corresponding to the recognition information, so as to obtain the assigned training data set as the above-mentioned adjusted training data set.

[0124] Step 406: Use the adjusted training data set to adjust the parameters of the speech recognition model to obtain an adjusted speech recognition model.

[0125] In some embodiments, the execution entity may use the adjusted training data set to adjust parameters of the speech recognition model to obtain an adjusted speech recognition model.

[0126] As an example, the above-mentioned execution entity can use the above-mentioned adjusted training data set to retrain the speech recognition model through back propagation to update the model parameters and obtain the retrained speech recognition model as the above-mentioned adjusted speech recognition model.

[0127] from Figure 4 It can be seen that Figure 2 Compared with the description of some corresponding embodiments, Figure 4 The process 400 of the recognition information generation method in some corresponding embodiments further highlights the specific steps of updating the parameters of the speech recognition model. Therefore, the solutions described in these embodiments use the target probability corresponding to the recognition information to train a speech recognition model with more accurate recognition.

[0128] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an identification information generating device. These device embodiments are similar to Figure 2 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0129] like Figure 5As shown, a recognition information generating device 500 includes: an acquisition unit 501, a first generation unit 502, a determination unit 503, and a second generation unit 504. The acquisition unit 501 is configured to acquire a target user's voice, wherein the target user's voice is a response from a target conversation user to a question; the first generation unit 502 is configured to generate a recognition text corresponding to the target user's voice as a target recognition text; the determination unit 503 is configured to determine the probability of the target recognition text appearing in a historical conversation text set as a target probability; and the second generation unit 504 is configured to generate recognition information indicating whether the target recognition text is correctly recognized based on the target probability.

[0130] In some optional implementations of some embodiments, the determination unit 503 in the above-mentioned device 500 can be further configured to: determine a historical conversation text group including the target recognition text in the above-mentioned historical conversation text set; determine the number of texts of the historical conversation texts in the above-mentioned historical conversation text group as the first text number; determine the number of texts of the historical conversation texts in the above-mentioned historical conversation text set as the second text number; and determine the above-mentioned target probability based on the above-mentioned first text number and the above-mentioned second text number.

[0131] In some optional implementations of some embodiments, the determination unit 503 in the above-mentioned device 500 can be further configured to: add the above-mentioned first text number and the above-mentioned second text number to predetermined values respectively to obtain the added first text number and the added second text number; divide the above-mentioned added first text number by the above-mentioned added second text number to obtain the divided value as the above-mentioned target probability.

[0132] In some optional implementations of some embodiments, the second generation unit 504 in the above-mentioned device 500 can be further configured to: in response to determining that the above-mentioned target probability is greater than or equal to a first threshold, generate recognition information representing that the above-mentioned target recognition text is correctly recognized.

[0133] In some optional implementations of some embodiments, the second generation unit 504 in the above-mentioned device 500 can be further configured to: in response to determining that the above-mentioned target probability is less than a second threshold, generate recognition information characterizing the above-mentioned target recognition text recognition error, wherein the above-mentioned second threshold is less than or equal to the above-mentioned first threshold.

[0134] In some optional implementations of some embodiments, the historical conversation texts in the historical conversation text set include the question and the recognition text. Furthermore, the second generation unit 504 in the apparatus 500 may be further configured to: determine a historical conversation text group in the historical conversation text set that includes the target recognition text; filter out historical conversation texts that meet preset text conditions from the subset of the historical conversation text group as target historical conversation texts; determine a probability corresponding to a probability that the recognition text in the target historical conversation text appears in at least one historical conversation text as a historical probability, wherein the at least one historical conversation text is the historical conversation text in the historical conversation text group excluding the target historical conversation text; and generate recognition information indicating whether the target recognition text is correctly recognized based on the historical probability and the target probability.

[0135] In some optional implementations of some embodiments, the second generation unit 504 in the above-mentioned device 500 can be further configured to: multiply the above-mentioned historical probability by a value α to obtain a multiplied historical probability, wherein the value range of the above-mentioned value α is [0, 1]; multiply the above-mentioned target probability by a value 1-α to obtain a multiplied probability; add the above-mentioned multiplied historical probability and the above-mentioned multiplied probability to obtain an added probability; in response to determining that the above-mentioned added probability is greater than or equal to a first threshold, generate recognition information representing that the above-mentioned target recognition text is correctly recognized.

[0136] In some optional implementations of some embodiments, the apparatus 500 further includes a data adjustment unit and a parameter adjustment unit (not shown). The data adjustment unit may be configured to adjust the training data set of the speech recognition model using the recognition information to obtain an adjusted training data set. The parameter adjustment unit may be configured to adjust the parameters of the speech recognition model using the adjusted training data set to obtain an adjusted speech recognition model.

[0137] In some optional implementations of some embodiments, the apparatus 500 further includes a third generating unit (not shown in the figure). The third generating unit may be configured to generate semantic understanding information based on the recognition information and the target recognition text.

[0138] It is understood that the units described in the device 500 are similar to those in the reference Figure 2 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 500 and the units included therein, and will not be repeated here.

[0139] Reference below Figure 6, which shows an electronic device (eg, Figure 1 Schematic diagram of the structure of the electronic device 101)600. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0140] like Figure 6 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the electronic device 600 are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0141] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead. Figure 6 Each block shown in the figure may represent one device, or may represent multiple devices as needed.

[0142] In particular, according to some embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network via the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of some embodiments of the present disclosure are performed.

[0143] It should be noted that in some embodiments of the present disclosure, the computer-readable medium mentioned above may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0144] In some embodiments, the client and server can communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0145] The computer-readable medium may be included in the electronic device, or may exist independently and not incorporated into the electronic device. The computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: obtain a target user voice, wherein the target user voice is a response from a target conversation user to a system question; generate recognized text corresponding to the target user voice as target recognized text; determine a probability of the target recognized text appearing in a historical conversation text set as a target probability; and, based on the target probability, generate recognition information indicating whether the target recognized text was correctly recognized.

[0146] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0147] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0148] The units described in some embodiments of the present disclosure may be implemented in software or hardware. The units described may also be provided in a processor. For example, they may be described as follows: a processor comprising: an acquisition unit, a first generation unit, a determination unit, and a second generation unit. The names of these units do not, in some cases, constitute limitations on the units themselves. For example, the acquisition unit may also be described as a "unit for acquiring the target user's voice."

[0149] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0150] The above description is only an illustration of some preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for generating identification information, comprising: Acquire the target user's voice, wherein the target user's voice is the voice of the target dialogue user in response to the question; Generating a recognition text corresponding to the target user's voice as a target recognition text; Determining a probability of the target recognition text appearing in a historical conversation text set as a target probability, wherein the historical conversation text includes: recognition text corresponding to a reply voice of a historical user in response to the question; Generating recognition information indicating whether the target recognition text is correctly recognized based on the target probability, including: determining a historical conversation text group including the target recognition text in the historical conversation text set; Filtering out a historical conversation text that meets a preset text condition from the historical conversation text group as a target historical conversation text, wherein the preset text condition is that the generation time of the corresponding conversation text of the historical conversation text is the same as the generation time of the latest conversation text; Acquire a historical conversation text set for the target historical conversation text as a candidate historical conversation text set; Determining a probability that the recognized text in the target historical conversation text appears in the candidate historical conversation text set as a historical probability; According to the historical probability and the target probability, recognition information is generated to indicate whether the target recognition text is correctly recognized.

2. The method according to claim 1, wherein Determining the probability of the target recognition text appearing in the historical conversation text set as the target probability includes: Determining a historical conversation text group including a target recognition text in the historical conversation text set; Determining the number of texts of the historical conversation texts in the historical conversation text group as a first text number; Determining the number of texts of the historical conversation texts in the historical conversation text set as a second number of texts; The target probability is determined according to the first number of texts and the second number of texts.

3. The method according to claim 2, wherein: The determining the target probability according to the first number of texts and the second number of texts includes: Adding the first text number and the second text number to a predetermined value respectively to obtain the added first text number and the added second text number; The number of the first texts after addition is divided by the number of the second texts after addition to obtain a divided value as the target probability.

4. The method according to claim 1, wherein Generating recognition information indicating whether the target recognition text is correctly recognized according to the target probability includes: In response to determining that the target probability is greater than or equal to a first threshold, recognition information indicating that the target recognition text is correctly recognized is generated.

5. The method according to claim 4, wherein Generating recognition information indicating whether the target recognition text is correctly recognized according to the target probability includes: In response to determining that the target probability is less than a second threshold, recognition information characterizing a recognition error of the target recognition text is generated, wherein the second threshold is less than or equal to the first threshold.

6. The method according to claim 1, wherein The historical dialogue texts in the historical dialogue text set include: the questions and the recognition texts.

7. The method according to claim 6, wherein: The generating, based on the historical probability and the target probability, recognition information indicating whether the target recognition text is correctly recognized includes: Multiplying the historical probability by a value α to obtain a multiplied historical probability, wherein the value α is in the range of [0, 1]; Multiply the target probability by the value 1-α to obtain the multiplied probability; Adding the multiplied historical probability and the multiplied probability to obtain a summed probability; In response to determining that the added probability is greater than or equal to a first threshold, recognition information indicating that the target recognition text is correctly recognized is generated.

8. The method according to claim 1, wherein The method further comprises: Using the recognition information, adjusting a training data set of a speech recognition model to obtain an adjusted training data set; The adjusted training data set is used to adjust the parameters of the speech recognition model to obtain an adjusted speech recognition model.

9. The method according to claim 1, wherein The method further comprises: Semantic understanding information is generated based on the recognition information and the target recognition text.

10. An identification information generating device, comprising: an acquisition unit configured to acquire a target user voice, wherein the target user voice is a voice reply of the target dialogue user to the question; A first generating unit is configured to generate a recognition text corresponding to the target user voice as a target recognition text; a determination unit configured to determine a probability of the target recognition text appearing in a historical conversation text set as a target probability, wherein the historical conversation text includes: recognition text corresponding to a reply voice of a historical user in response to the question; The second generating unit is configured to generate recognition information representing whether the target recognition text is correctly recognized based on the target probability, including: determining a historical conversation text group including the target recognition text in the historical conversation text set; screening out historical conversation texts that meet a preset text condition from the historical conversation text group as target historical conversation texts, wherein the preset text condition is that the generation time of the conversation text corresponding to the historical conversation text is the same as the generation time of the latest conversation text; obtaining a historical conversation text set for the target historical conversation text as a candidate historical conversation text set; determining the probability of the recognition text in the target historical conversation text appearing in the candidate historical conversation text set as a historical probability; and generating recognition information representing whether the target recognition text is correctly recognized based on the historical probability and the target probability.

11. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 9.

12. A computer-readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images