Text proofreading method, device, equipment, medium and program product

Abstract

The present application provides a text proofreading method, device, equipment, medium and program product, which relates to the field of artificial intelligence technology, wherein the method comprises: obtaining the speech recognition result of the current query node, obtaining the target pinyin corresponding to the text to be proofread according to the pinyin of each character in the text to be proofread, and proofreading the text to be proofread according to the pre-configured pinyin proofreading set and the target pinyin to obtain the proofread text. In this technical solution, by configuring a pinyin proofreading set for the text that is easily misrecognized in the current query node, and configuring a pinyin proofreading set for the text that is easily misrecognized in the business process where the current query node is located, the text range involved in the proofreading can be accurately controlled, so that the text obtained by the speech recognition technology is more accurate.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a text proofreading method, apparatus, device, medium, and program product. Background Art

[0002] With the development of artificial intelligence technology, intelligent customer service systems have been applied. Intelligent customer service systems obtain users' spoken voice information through voice communication with users, and analyze the users' true intentions from the voice information, so as to select corresponding words to answer users according to the users' true intentions, which can reduce labor costs.

[0003] In the existing technology, automatic speech recognition technology is mainly used to recognize the voice information spoken by the user and convert it into text information, and then keywords are extracted from the text to obtain the user's true intention.

[0004] However, the existing automatic speech recognition technology is prone to text confusion during the speech recognition process, resulting in inaccurate recognized text information. Summary of the Invention

[0005] The present application provides a text proofreading method, apparatus, device, medium and program product for solving the problem of inaccurate text recognized by automatic speech recognition technology.

[0006] In a first aspect, an embodiment of the present application provides a text proofreading method, comprising:

[0007] Obtaining a speech recognition result for the current query node, wherein the speech recognition result includes text to be proofread obtained by recognizing the speech input at the current query node;

[0008] According to the pinyin of each character in the text to be proofread, obtaining the target pinyin corresponding to the text to be proofread;

[0009] According to the pre-configured pinyin proofreading set and the target pinyin, the text to be proofread is proofread to obtain the proofread text; wherein, the pinyin proofreading set includes a first pinyin proofreading set or a second pinyin proofreading set, the first pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the business process where the current query node is located, the second pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the current query node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

[0010] In a possible design of the first aspect, proofreading the text to be proofread based on a preconfigured pinyin proofreading set and the target pinyin to obtain the proofread text includes:

[0011] Obtaining the business process where the current query node is located, and determining whether the business process is configured with the first pinyin proofreading set;

[0012] If the business process is configured with the first pinyin proofreading set, proofreading the text to be proofread according to the first pinyin proofreading set and the target pinyin;

[0013] If the business process is not configured with the first pinyin proofreading set, obtaining the second pinyin proofreading set configured at the current query node;

[0014] The text to be proofread is proofread according to the second pinyin proofreading set and the target pinyin to obtain the proofread text.

[0015] In another possible design of the first aspect, proofreading the text to be proofread based on the first pinyin proofreading set and the target pinyin includes:

[0016] According to the target pinyin and the corrected pinyins in each first pinyin proofreading set, determining a target first pinyin proofreading set that matches the target pinyin in each first pinyin proofreading set;

[0017] Correct the text to be proofread into the corrected text in the target first pinyin proofreading set.

[0018] In another possible design of the first aspect, determining a target first pinyin proofreading set that matches the target pinyin in each first pinyin proofreading set based on the target pinyin and the corrected pinyin in each first pinyin proofreading set includes:

[0019] Obtaining the confused pinyin corresponding to the target pinyin, wherein the similarity value between the confused pinyin and the target pinyin is higher than a preset threshold;

[0020] The target pinyin and the confused pinyin are compared with the corrected pinyins in each first pinyin proofreading set to determine a target first pinyin proofreading set that matches the target pinyin.

[0021] In another possible design of the first aspect, proofreading the text to be proofread based on the second pinyin proofreading set and the target pinyin includes:

[0022] According to the target pinyin and the corrected pinyins in each second pinyin proofreading set, determining a target second pinyin proofreading set that matches the target pinyin in each second pinyin proofreading set;

[0023] Correct the text to be proofread into the corrected text in the target second pinyin proofreading set.

[0024] In another possible design of the first aspect, determining a target second pinyin proofreading set that matches the target pinyin in each second pinyin proofreading set based on the target pinyin and the corrected pinyin in each second pinyin proofreading set includes:

[0025] Obtaining a confused pinyin corresponding to the target pinyin, wherein a similarity value between the confused pinyin and the target pinyin is higher than a preset threshold;

[0026] The target pinyin and the confused pinyin are compared with the corrected pinyin in each second pinyin proofreading set to determine a target second pinyin proofreading set that matches the target pinyin.

[0027] In another possible design of the first aspect, obtaining the business process in which the current query node is located includes:

[0028] Determine the previous node and next node of the current query node according to a preset flow chart;

[0029] The business process in which the current query node is located is determined according to the previous node and the next node of the current query node.

[0030] In another possible design of the first aspect, obtaining the target pinyin corresponding to the text to be proofread based on the pinyin of each character in the text to be proofread includes:

[0031] Delete the phonetic symbols of each character's pinyin to obtain the corresponding pinyin without phonetic symbols for each character;

[0032] The non-phonetic pinyin of each character is separated according to a preset separator to obtain the target pinyin corresponding to the character to be proofread.

[0033] In another possible design of the first aspect, the method further includes:

[0034] Obtaining a first user intent that matches the text to be proofread and a matching value corresponding to the first user intent;

[0035] Obtaining a second user intent that matches the proofread text and a matching value corresponding to the second user intent;

[0036] determining, according to a matching value corresponding to the first user intent and a matching value corresponding to the second user intent, the first user intent or the second user intent as the user's true intent;

[0037] According to the real user intention, answer information corresponding to the voice input at the current query node is obtained.

[0038] In another possible design of the first aspect, the method further includes:

[0039] The text to be proofread and the proofread text are written into a preset interaction record.

[0040] In another possible design of the first aspect, the method further includes:

[0041] A proofreading report is generated according to the text to be proofread, the pre-configured pinyin proofreading set and the proofread text.

[0042] In a second aspect, an embodiment of the present application provides a text proofreading device, comprising:

[0043] A result acquisition module, configured to acquire a speech recognition result of a current query node, wherein the speech recognition result includes text to be proofread obtained by recognizing the speech input at the current query node;

[0044] A pinyin conversion module, configured to obtain a target pinyin corresponding to the text to be proofread based on the pinyin of each character in the text to be proofread;

[0045] A text proofreading module is used to proofread the text to be proofread according to a pre-configured pinyin proofreading set and the target pinyin to obtain the proofread text; wherein, the pinyin proofreading set includes a first pinyin proofreading set or a second pinyin proofreading set, the first pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the business process where the current query node is located, the second pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the current query node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

[0046] In a third aspect, an embodiment of the present application provides a computer device, comprising: a processor, and a memory communicatively connected to the processor;

[0047] The memory stores computer-executable instructions;

[0048] The processor executes the computer-executable instructions stored in the memory to implement the above method.

[0049] In a fourth aspect, an embodiment of the present application provides a readable storage medium, in which computer instructions are stored. When the computer instructions are executed by a processor, they are used to implement the above method.

[0050] In a fifth aspect, an embodiment of the present application provides a program product, comprising computer instructions, which implement the above method when executed by a processor.

[0051] The text proofreading method, apparatus, equipment, medium and program product provided in the embodiments of the present application can accurately control the scope of text involved in proofreading by configuring a pinyin proofreading set for text that is easily misrecognized in the current query node, and configuring a pinyin proofreading set for text that is easily misrecognized in the business process where the current query node is located, so that the text obtained by speech recognition technology is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application;

[0053] Figure 1 A schematic diagram of a text proofreading method scenario provided in an embodiment of the present application;

[0054] Figure 2 A flowchart of a first embodiment of a text proofreading method according to an embodiment of the present application;

[0055] Figure 3 A schematic diagram of the business process structure provided in the embodiment of the present application;

[0056] Figure 4 A flowchart of a second embodiment of the text proofreading method provided in the present application;

[0057] Figure 5 A schematic diagram of data processing interaction of the intelligent customer service system provided in an embodiment of the present application;

[0058] Figure 6 A schematic diagram of the structure of a text proofreading device provided in an embodiment of the present application;

[0059] Figure 7 A schematic diagram of the structure of a computer device provided in an embodiment of the present application.

[0060] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0061] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0062] First, let’s explain the terms involved in this application:

[0063] Media Resource Server:

[0064] Under the control of control devices (soft switches, application servers), media resource servers provide the media resource functions required to implement various services on IP networks, including service audio provision, conferencing, interactive voice response (IVR), notifications, and high-level language services.

[0065] Media Resource Control Protocol:

[0066] The Media Resource Control Protocol (MRCP) is a computer network application layer communication protocol used by voice servers to provide various voice services (such as speech recognition, speech synthesis, and voiceprint recognition) to clients.

[0067] Automatic speech recognition technology:

[0068] Automatic Speech Recognition (ASR) is a technology that converts human speech into text.

[0069] Natural Language Processing:

[0070] Natural Language Processing (NLP) is a discipline that integrates linguistics, mathematics, and computer science. Its core goal is to convert human natural language into computer-readable instructions. In other words, it allows machines to understand human language.

[0071] Figure 1 Schematic diagram of the text proofreading method provided in the embodiment of this application. Figure 1 As shown, the intelligent customer service system 10 can proactively dial the user's phone number. After the user answers the call on the mobile terminal 11, the intelligent customer service system 10 can inquire the user, such as asking the user whether he needs to apply for installment services, and the user can answer the inquiry questions. The intelligent customer service system 10 needs to feedback the corresponding information to the user based on the user's reply, and ultimately realize intelligent question and answer, reducing manual input. In actual life applications, in order to improve the interactive effect of question and answer, the intelligent customer service system 10 needs to feedback accurate words to the user based on the user's reply. This process mainly involves recognizing the user's voice, extracting text and identifying the user's true intention based on the text, and finally finding the corresponding words based on the user's true intention and feeding them back to the user.

[0072] The intelligent customer service system 10 mainly includes four modules: media resource server, intelligent voice product, NLP and application system. The data processing involved in a basic interaction process of the intelligent customer service system 10 mainly includes the following aspects: (1) The user's voice is transmitted to the intelligent voice product through the media resource server. The ASR model in the intelligent voice product converts the user's voice into text and returns it to the media resource server; (2) The media resource server sends the recognition result to the application system, and the application system calls the NLP model to understand the user's intention and then selects the reply words according to the user's intention. Then, the media resource server is driven to send a speech-to-speech request to the intelligent voice product; (3) The intelligent voice product converts the words into speech, transmits it to the media resource server, and plays it to the user. In this data processing process, due to the presence of homophones, a large number of confusing sounds, and the limitations of the ASR model, it is easy to cause the text results converted by the ASR model to be incorrect, which is likely to affect the entire intelligent customer service system and cause the question-answering interaction effect to be unsatisfactory.

[0073] In the prior art, there are two main approaches to converting user speech into text using ASR models. One approach is to divide ASR models into basic models and specific models. Basic models are suitable for less specialized fields, and due to their wide scope of coverage, their recognition accuracy is slightly lower. Specific models, on the other hand, are specially trained for highly specialized fields, such as finance and insurance, and have the advantage of high recognition rates in specific fields. However, specific models require a lot of tuning and training, such as collecting corpus training sets and adjusting algorithm training parameters. The entire process is time-consuming, labor-intensive, and costly. Furthermore, specific models have a narrow scope of application, and when new businesses emerge, they cannot be quickly adjusted and adapted. Another approach is to utilize a third-party self-learning training platform to add popular keywords, customized language models, acoustic models, etc., and allow users to independently train and tune to solve some simple recognition problems. However, this method incurs high costs for using the self-learning training platform. Furthermore, this type of optimization (e.g., for popular keywords) can easily lead to overcorrection, mistaking the correct recognition results of the old model for configured popular keywords with similar pronunciations. Furthermore, this type of optimization is coarse-grained and cannot control the scope of application (e.g., processes and process nodes), resulting in a wide range of impact.

[0074] In response to the above problems, the text proofreading method, device, equipment, medium and program product provided in the embodiments of the present application further correct the ASR recognition results by adding a pinyin proofreading set based on pinyin recognition and error correction based on specific business processes and specific process nodes, and ultimately improve the accuracy of speech-to-text conversion, prepare for subsequent NLP preprocessing, and improve the interactive effect of the entire intelligent customer service system.

[0075] The technical solution of the present application is described in detail below through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0076] Figure 2 This is a flow chart of the text proofreading method provided in the embodiment of the present application. This method can be applied to the intelligent customer service system, where the intelligent voice product converts the voice into speech and then further proofreads the text in the speech. Figure 2 As shown, the method may specifically include the following steps:

[0077] S201: Obtain the speech recognition result of the current query node.

[0078] The speech recognition result includes the text to be proofread, obtained by recognizing the speech input at the current query node. The speech can be the user's spoken speech, such as the speech generated by the user answering the intelligent customer service system's questions after the intelligent customer service system calls the user's phone.

[0079] For example, the speech recognition result can be obtained by recognizing the user's speech through the ASR model, which includes text. For example, if the intelligent customer service system asks the user whether he needs installment payment, and the user says "need installment payment", the text obtained by the ASR model from the user's speech recognition can be "need installment payment" or "need divergence payment". Among them, the latter is the text recognition result generated by the ASR's incorrect recognition of the user's speech because divergence and installment payment are homophones. In this case, the speech recognition result of "need divergence payment" needs to be proofread to the former using the speech recognition proofreading method of this application.

[0080] In this embodiment, the business processes and corresponding query nodes vary depending on the field involved in the intelligent customer service system. For example, if the intelligent customer service system is involved in the financial field, the business processes may include loan business, repayment business, etc. For example, the query nodes included in the loan business include the node for inquiring whether to inquire about installments, the node for inquiring about the number of installments, the node for inquiring about the end of the loan, etc.

[0081] For example, Figure 3 A schematic diagram of the business process provided in this application embodiment, such as Figure 3 As shown, it includes multiple nodes, such as whether to apply for installment payments, the number of installments, and the end node. Each node has a corresponding dialogue and a set of possible user intents. For example, for the installment payment node, the dialogue might ask the user, "Do you want to apply for installment payments?" The intent set includes "affirmative intent" and "negative intent." A business process is a structure composed of nodes, connected by arrows.

[0082] In some embodiments, Figure 3 The process structure of the business process in Figure 3 can be used as a preset process table, thereby determining the previous node and the next node of the current inquiry node, and thus determining the business process in which the current inquiry node is located.

[0083] Exemplarily, the business process may include an installment business handling process (i.e., the business process in Figure 3 ) and a user complaint handling business process. Different business processes are configured with different first pinyin proofreading sets. Figure 3

[0084] S202. According to the pinyin of each character in the text to be proofread, obtain the target pinyin corresponding to the text to be proofread.

[0085] In this embodiment, the text to be proofread includes at least two Chinese characters, such as "需要分歧". Each Chinese character has a corresponding pinyin. After splicing the pinyin of each Chinese character, the target pinyin is obtained, such as "xuyaofenqi".

[0086] S203. According to the pre-configured pinyin proofreading set and the target pinyin, proofread the text to be proofread to obtain the proofread text.

[0087] Among them, the pinyin proofreading set includes a first pinyin proofreading set or a second pinyin proofreading set. The first pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text included in the business process in which the current inquiry node is located. The second pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text included in the current inquiry node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

[0088] In this embodiment, the pinyin proofreading set includes pinyin and the text corresponding to the pinyin. Taking the pinyin as "fenqi" and the corresponding text as "分期" as an example. If the text to be proofread is "需要分歧", the target pinyin is "xuyaofenqi". At this time, the text to be proofread will be proofread to "分期" as the proofread text.

[0089] In this embodiment, the number of the first pinyin proofreading set and the second pinyin proofreading set can be multiple. For example, there is a node where the recognition error rate of the ASR model at this node is very high, and many of the texts obtained by user speech recognition at this node are incorrect. For example, the ASR model recognizes "三期" said by the user as "三七" and "九期" as "酒器". At this time, two second pinyin proofreading sets are required for this node to proofread "三七" and "酒器" respectively.

[0090] Among them, in addition to the high recognition error rate of the ASR model at a certain node, the ASR model may also have a high recognition error rate for certain voices in the entire business process. At this time, multiple first pinyin proofreading sets are required to proofread those voices in the business process that are prone to misrecognition.

[0091] In this embodiment, the first pinyin proofreading set is for proofreading speech that is easily misrecognized in the entire business process, while the second pinyin proofreading set is for proofreading speech that is easily misrecognized in a single node.

[0092] For example, the pinyin proofreading sets configured for business processes and nodes are shown in Table 1 below:

[0093]

[0094] Table 1

[0095] The embodiment of the present application configures a corresponding pinyin proofreading set at the current query node or the business process where the current query node is located, so that the text that is easily misrecognized in the current query node or business process can be accurately proofread, so that the text obtained after ASR recognition can more accurately describe the voice input by the user.

[0096] In some embodiments, the above step S203 can be implemented by the following steps:

[0097] Obtain the business process where the current query node is located, and determine whether the business process is configured with the first pinyin proofreading set;

[0098] If the business process is configured with a first pinyin proofreading set, the text to be proofread is proofread according to the first pinyin proofreading set and the target pinyin;

[0099] If the business process is not configured with the first pinyin proofreading set, then obtain the second pinyin proofreading set configured at the current query node;

[0100] Proofread the text to be proofread according to the second pinyin proofreading set and the target pinyin.

[0101] In this embodiment, the first and second pinyin proofreading sets have different proofreading targets. The pinyin and text included in the first pinyin proofreading set are mainly for Chinese characters that may be misrecognized in the entire business process. The pinyin and text included in the second pinyin proofreading set are mainly for Chinese characters that may be misrecognized in specific nodes.

[0102] Among them, multiple first pinyin proofreading sets or multiple second pinyin proofreading sets can be configured at the current inquiry node. Exemplarily, taking Table 1 above as an example, three second pinyin proofreading sets <sanqi, three phases>, <xinlong, Xinlong>, and <xinglong, Xinlong> can be set at the current inquiry node. If the target pinyin is sanqi, the text to be proofread will be proofread to the proofread text "three phases".

[0103] In this embodiment, the purpose of configuring the first pinyin proofreading set or the second pinyin proofreading set is to avoid overcorrection. For example, if the text recognized by the ASR model from the user's speech is "disagreement", and if the current inquiry node is in the business process of handling user complaints at this time, it should not be corrected to "installment", but it should be considered that the recognition result of the ASR model is correct.

[0104] By configuring the first pinyin proofreading set or the second pinyin proofreading set at the current inquiry node in the embodiment of the present application, words or characters within different ranges can be proofread, the proofreading granularity can be accurately controlled, and for the limited range of a specific business and the limited response content that users may have, a targeted pinyin proofreading set is configured to avoid overcorrection.

[0105] Based on the above embodiment, in some embodiments, the above "proofread the text to be proofread according to the first pinyin proofreading set and the target pinyin" can be specifically implemented through the following steps:

[0106] Determine a target first pinyin proofreading set that matches the target pinyin in each first pinyin proofreading set according to the target pinyin and the corrected pinyin in each first pinyin proofreading set;

[0107] Correct the text to be proofread to the corrected text in the target first pinyin proofreading set.

[0108] In this embodiment, the corrected pinyin and the corrected text in the first pinyin proofreading set can be in the form of key-value pairs, that is, <key, value>, where key represents the corrected pinyin and value represents the corrected text. Exemplarily, the number of first pinyin proofreading sets can be one, for example, <fenqi, installment>. If the target pinyin contains fenqi or the target pinyin is fenqi, the text to be proofread will be proofread to "installment".

[0109] By setting the proofreading pinyin and the proofreading text in the embodiment of the present application, the misrecognized text output by the ASR model can be proofread to obtain a more accurate text, which provides strong support for the subsequent work of the intelligent customer service system and improves the interaction effect between the intelligent customer service system and users.

[0110] Further, in some embodiments, the above step of "determining a target first pinyin proofreading set that matches the target pinyin from each first pinyin proofreading set according to the target pinyin and the corrected pinyins in each first pinyin proofreading set" can be specifically implemented through the following steps:

[0111] Obtain the confusing pinyin corresponding to the target pinyin;

[0112] Compare the target pinyin and the confusing pinyin with the corrected pinyins in each first pinyin proofreading set to determine a target first pinyin proofreading set that matches the target pinyin.

[0113] Among them, the similarity value between the confusing pinyin and the target pinyin is higher than a preset threshold. Exemplarily, taking the target pinyin as fenqi, the confusing pinyin can be the postnasal sound that is easily confused with the target pinyin. For example, the confusing pinyin can be fengqi.

[0114] Exemplarily, there can be two first pinyin proofreading sets. For example, <fengqi, installment> and <fenqi, installment>. That is, the target first pinyin proofreading set that matches the target pinyin is <fenqi, installment>, and the target first pinyin proofreading set that matches the confusing pinyin is <fengqi, installment>. Through these two target first pinyin proofreading sets, the finally proofread text is "installment".

[0115] In this embodiment, by obtaining the confusing sound of the target pinyin and configuring a pinyin proofreading set that matches the confusing sound in the first pinyin proofreading set, it is possible to proofread the text misrecognized by the ASR model due to inaccurate pronunciation into the correct text, improving the recognition accuracy.

[0116] In some embodiments, the above step of "proofreading the text to be proofread according to the second pinyin proofreading set and the target pinyin" can be specifically implemented through the following steps:

[0117] Determine a target second pinyin proofreading set that matches the target pinyin from each second pinyin proofreading set according to the target pinyin and the corrected pinyins in each second pinyin proofreading set;

[0118] Correct the text to be proofread into the corrected text in the target second pinyin proofreading set.

[0119] Further, in some embodiments, the above "determining a target second pinyin proofreading set that matches the target pinyin from each second pinyin proofreading set according to the target pinyin and the corrected pinyins in each second pinyin proofreading set" can be specifically implemented through the following steps:

[0120] Obtain the confusing pinyin corresponding to the target pinyin;

[0121] Compare the target pinyin and the confused pinyin with the corrected pinyin in each second pinyin proofreading set to determine the target second pinyin proofreading set that matches the target pinyin.

[0122] Among them, the similarity value between the confused pinyin and the target pinyin is higher than the preset threshold.

[0123] Exemplarily, the second pinyin proofreading set can include three, such as <sanqi, three periods>, <xinlong, Xinlong>, <xinglong, Xinlong>. When the target pinyin is xinlon, the corresponding confused pinyins can be xinlong and xinglong. At this time, the target second pinyin proofreading sets corresponding to the confused pinyins are <xinlong, Xinlong> and <xinglong, Xinlong>, and the finally proofread text is Xinlong.

[0124] In this embodiment, by obtaining the confused sounds of the target pinyin and configuring the pinyin proofreading set that matches the confused sounds in the second pinyin proofreading set, it is possible to proofread the text misrecognized by the ASR model due to inaccurate pronunciation into the correct text, improving the recognition accuracy.

[0125] In some embodiments, the above step S202 can be specifically implemented through the following steps:

[0126] Delete the phonetic symbols of the pinyin of each character to obtain the non-phonetic pinyin corresponding to each character;

[0127] According to the preset delimiter, separate the non-phonetic pinyins of each character to obtain the target pinyin corresponding to the text to be proofread.

[0128] In this embodiment, the text recognized by the ASR model has tones, which are reflected by the phonetic symbols of the pinyin. For example, the phonetic symbols of "period" and "歧" are different. However, when some users speak, the pronunciation of some characters may be confused. For example, a user may pronounce a character that should be pronounced in the fourth tone as the second tone, causing misrecognition by the ASR model. Therefore, the pinyin of each character can be removed. For example, both "period" and "歧" will be converted to the non-phonetic pinyin qi.

[0129] Exemplarily, the preset delimiter can be a slash, a horizontal bar, a punctuation mark, etc. For example, the target pinyin corresponding to "installment" is fen-qi.

[0130] In the embodiment of the present application, by converting each character into non-phonetic pinyin, it is possible to be compatible with the problem of different pronunciation tones of different users. At the same time, setting delimiters for separation can avoid unclear boundaries of the pinyins of multiple characters, and can effectively improve the recognition effect and compatibility.

[0131] In some embodiments, the above method may further include the following steps:

[0132] Obtain a first user intention that matches the text to be proofread and a matching value corresponding to the first user intention;

[0133] Obtain a second user intention that matches the proofread text and a matching value corresponding to the second user intention;

[0134] Determine the first user intention or the second user intention as the true user intention according to the matching value corresponding to the first user intention and the matching value corresponding to the second user intention;

[0135] Obtain a reply message corresponding to the voice input at the current inquiry node according to the true user intention.

[0136] Exemplarily, taking "Sanqi" as the text to be proofread and "Sanqi" as the proofread text, in the intelligent customer service system, there is a mapping relationship between the speech recognition result and the intention. The first user intention mapped and matched by "Sanqi" may be a traditional Chinese medicine in the intelligent customer service system, and the matching value corresponding to this first user intention may be 0.1. The second user intention mapped and matched by "Sanqi" is that the user needs to divide the loan into three installments in the intelligent customer service system. The matching value corresponding to this second user intention is 0.8. Since the matching value corresponding to the second user intention is larger, the second user intention is selected as the true user intention, and then the corresponding conversation is matched according to the true user intention and used as the reply message to reply to the user.

[0137] In the embodiment of the present application, by separately sending the text before proofreading and the text after proofreading to determine the corresponding user intention, and taking the user intention with the largest matching value as the true user intention, it can provide a fallback for speech proofreading, avoid overcorrection during the speech proofreading process, and improve the interaction effect of the intelligent customer service system. [[ID=1']]

[0138] Exemplarily, Figure 4 is a schematic flowchart of the second embodiment of the text proofreading method provided by the embodiment of the present application. As Figure 4 shown, the method may specifically include the following steps:

[0139] S401. Ask the user;

[0140] S402. Obtain the customer's response voice;

[0141] S403. Identify the speech recognition result through the intelligent voice product ASR model;

[0142] S404. Convert the speech recognition result into the target pinyin;

[0143] S405. Have a business process pinyin proofreading set;

[0144] S4051. Have the next business process pinyin proofreading set;

[0145] S4052, obtaining a corrected pinyin and corrected text of a business process pinyin proofreading set;

[0146] S4053, matching the corrected pinyin with the target pinyin;

[0147] S4054, replacing the speech recognition result with the correction text;

[0148] S4055, traverse the next business process pinyin proofreading set;

[0149] S406, the current query node is configured with a pinyin proofreading set;

[0150] S4061, there is a next node pinyin proofreading set;

[0151] S4062, obtaining the corrected pinyin and corrected text of a pinyin proofreading set of a node;

[0152] S4063, matching the corrected pinyin with the target pinyin;

[0153] S4064, replacing the speech recognition result with the correction text;

[0154] S4065, traverse the next node process pinyin proofreading set;

[0155] S407, obtaining the user intention corresponding to the speech recognition result and the user intention corresponding to the correction text;

[0156] S408: Determine the real user intention.

[0157] In some embodiments, the above method further comprises the following steps:

[0158] The text to be proofread and the proofread text are written into the preset interaction record.

[0159] The embodiment of the present application establishes a post-tracking mechanism to write both the text to be proofread and the proofread text into the interaction record, which can facilitate the managers of the intelligent customer service system to compare the recordings afterwards and verify whether the proofreading effect meets expectations.

[0160] In some embodiments, the above method further comprises the following steps:

[0161] Generate a proofreading report based on the text to be proofread, the pre-configured pinyin proofreading set, and the proofread text.

[0162] In this embodiment, a proofreading report can be issued for subsequent analysis by the management personnel of the intelligent customer service system to determine whether the proofreading effect meets expectations.

[0163] Figure 5This is a diagram of the data processing interaction of the intelligent customer service system provided in this embodiment. The intelligent customer service system mainly involves four major modules: resource media server, intelligent voice product, NLP, and application system, which involve the following specific processes:

[0164] S501, start the outbound call task;

[0165] S502, dial the user's phone number;

[0166] S503: The user answers the call;

[0167] S504, request dialogue;

[0168] S505, return to the speech;

[0169] S506: Request text-to-speech conversion;

[0170] S507, return the synthesized speech;

[0171] S508, play the synthesized speech;

[0172] S509: The user answers the question;

[0173] S510: Send a speech recognition request;

[0174] S511. Return the speech recognition result;

[0175] S512: Send the speech recognition result;

[0176] S513, performing pinyin proofreading and error correction on the recognition result;

[0177] S514, sending proofreading results;

[0178] S515: Return user intention;

[0179] S516: Mapping reply words according to user intent;

[0180] S517. Return to the reply script.

[0181] Among them, the main contents of steps S501-S503 are that the application system starts the outbound call task, triggers the media resource server to dial the customer's phone, and the customer answers the call; the main contents of steps S504-S505 are that the media resource server informs the application system that the call has been connected, the application system starts the process, and returns the opening words to the media resource server; the main contents of steps S506-S508 are that the media resource server sends the opening words to the intelligent voice server, the intelligent voice server returns the voice synthesized by the TTS module, and the media resource server plays it for the customer to listen to; the main contents of steps S509-S512 are that the customer answers the question, the media resource server sends the customer's voice to the intelligent voice product, the ASR module of the intelligent voice product converts the customer's voice into text, and returns it to the media resource server; the media resource server sends it to the application system, requesting a reply word; the main contents of steps S513-S515 are that the application system receives the reply word, checks whether the business process and the process node are configured with a pinyin proofreading set (the set is<K,V> (Key-value pairs, K is pinyin, V is the correct text.) If configured, the system iterates through the pinyin in the configuration set and compares it with the ASR recognition results. If the comparison is successful, the corresponding Chinese character in the ASR recognition result is replaced with the Chinese character corresponding to the pinyin in the configuration set. The verified ASR recognition result is then sent to the NLP, which returns the customer intent. Steps S516-S517 focus on the application system mapping the corresponding response based on the customer intent and returning it to the media resource server. This completes a closed loop of customer interaction.

[0182] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0183] Figure 6 This is a structural diagram of the text proofreading device provided in the embodiment of this application. The device can be integrated into the intelligent customer service system, or it can be independent of the intelligent customer service system and work together with the intelligent customer service system to implement this technical solution. Figure 6 As shown, the text proofreading device 60 includes: a result acquisition module 61, a pinyin conversion module 62 and a text proofreading module 63.

[0184] The result acquisition module 61 is used to obtain the speech recognition result of the current query node. The pinyin conversion module 62 is used to obtain the target pinyin corresponding to the text to be proofread based on the pinyin of each character in the text to be proofread. The text proofreading module 63 is used to proofread the text to be proofread based on a pre-configured pinyin proofreading set and the target pinyin to obtain the proofread text.

[0185] Among them, the speech recognition result includes the text to be proofread obtained by recognizing the speech input at the current query node, the pinyin proofreading set includes the first pinyin proofreading set or the second pinyin proofreading set, the first pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the business process where the current query node is located, the second pinyin proofreading set includes the corrected pinyin and corrected text for proofreading the text contained in the current query node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

[0186] In some embodiments, the text proofreading module may be specifically used to:

[0187] Obtain the business process where the current query node is located, and determine whether the business process is configured with the first pinyin proofreading set;

[0188] If the business process is configured with a first pinyin proofreading set, the text to be proofread is proofread according to the first pinyin proofreading set and the target pinyin;

[0189] If the business process is not configured with the first pinyin proofreading set, then obtain the second pinyin proofreading set configured at the current query node;

[0190] The text to be proofread is proofread according to the second pinyin proofreading set and the target pinyin to obtain the proofread text.

[0191] In some embodiments, the text proofreading module may be specifically used to:

[0192] According to the target pinyin and the corrected pinyin in each first pinyin proofreading set, a target first pinyin proofreading set matching the target pinyin is determined in each first pinyin proofreading set;

[0193] Correct the text to be proofread into the corrected text in the target first pinyin proofreading set.

[0194] In some embodiments, the text proofreading module may be specifically used to:

[0195] Get the confused pinyin corresponding to the target pinyin;

[0196] The target pinyin and the confused pinyin are compared with the corrected pinyin in each first pinyin proofreading set to determine a target first pinyin proofreading set that matches the target pinyin.

[0197] Among them, the similarity values ​​between the confused pinyin and the target pinyin are both higher than the preset threshold.

[0198] In some embodiments, the text proofreading module may be specifically used to:

[0199] According to the target pinyin and the corrected pinyin in each second pinyin proofreading set, a target second pinyin proofreading set matching the target pinyin is determined in each second pinyin proofreading set;

[0200] Correct the text to be proofread into the corrected text in the target second pinyin proofreading set.

[0201] In some embodiments, the text proofreading module may be specifically used to:

[0202] Get the confused pinyin corresponding to the target pinyin;

[0203] The target pinyin and the confused pinyin are compared with the corrected pinyin in each second pinyin proofreading set to determine a target second pinyin proofreading set that matches the target pinyin.

[0204] Among them, the similarity value between the confused pinyin and the target pinyin is higher than a preset threshold.

[0205] In some embodiments, the text proofreading module may be specifically used to:

[0206] According to the preset flow chart, determine the previous node and next node of the current query node;

[0207] The business process of the current query node is determined based on the previous node and the next node of the current query node.

[0208] In some embodiments, an intention acquisition module is further included, which is used to:

[0209] Obtaining a first user intent that matches the text to be proofread and a matching value corresponding to the first user intent;

[0210] Obtaining a second user intent that matches the proofread text and a matching value corresponding to the second user intent;

[0211] Determining the first user intention or the second user intention as the user's true intention based on the matching value corresponding to the first user intention and the matching value corresponding to the second user intention;

[0212] According to the real user intention, the answer information corresponding to the voice input at the current query node is obtained.

[0213] In some embodiments, a writing module is also included for writing the text to be proofread and the proofread text into a preset interaction record.

[0214] In some embodiments, a generation module is also included for generating a proofreading report based on the text to be proofread, a pre-configured pinyin proofreading set and the proofread text.

[0215] The device provided in the embodiments of the present application can be used to execute the method in the above embodiments. Its implementation principle and technical effects are similar and will not be repeated here.

[0216] It should be noted that it should be understood that the division of the various modules of the above device is merely a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. Moreover, these modules can all be implemented in the form of software called by a processing element; they can also all be implemented in the form of hardware; some modules can also be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, the result acquisition module can be a separately established processing element, or it can be integrated into a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device to perform the functions of the above result acquisition module. The implementation of other modules is similar. In addition, these modules can all or partly be integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0217] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code on a processing element, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0218] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When software is used for implementation, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrations. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)).

[0219] Figure 7 This is a schematic diagram of the structure of the computer device provided in the embodiment of the present application. Figure 7 As shown, the computer device 70 includes at least one processor 71 , a memory 72 , a bus 73 and a communication interface 74 .

[0220] The processor 71 , the communication interface 74 and the memory 72 communicate with each other via the bus 73 .

[0221] The communication interface 74 is used to communicate with other devices and includes a communication interface for data transmission and a display interface or operation interface for human-computer interaction.

[0222] The processor 71 is used to execute computer-executable instructions stored in the memory 72, and specifically can execute relevant steps in the method described in the above embodiment.

[0223] The processor may be a central processing unit, an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in a computer device may be processors of the same type, such as one or more CPUs, or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0224] The memory 72 is used to store computer-executable instructions. The memory 72 may include a high-speed RAM memory, or may also include a non-volatile memory, such as at least one disk memory.

[0225] This embodiment further provides a readable storage medium, in which computer instructions are stored. When at least one processor of a computer device executes the computer instructions, the computer device executes the text proofreading methods provided by the various aforementioned embodiments.

[0226] This embodiment also provides a program product, comprising computer instructions stored in a readable storage medium. At least one processor of a computer device can read the computer instructions from the readable storage medium and execute the computer instructions to cause the computer device to implement the text proofreading methods provided in the various embodiments described above.

[0227] In this application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship; in the formula, the character " / " indicates that the previous and next associated objects are in a "division" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0228] It is understood that the various numerical numbers involved in the embodiments of the present application are only for the convenience of description and are not intended to limit the scope of the embodiments of the present application. In the embodiments of the present application, the order of the sequence numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0229] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A text proofreading method, characterized in that: include: Obtaining a speech recognition result of the current query node, wherein the speech recognition result includes the text to be proofread obtained by recognizing the speech input at the current query node; According to the pinyin of each character in the text to be proofread, obtaining the target pinyin corresponding to the text to be proofread; According to the pre-configured pinyin proofreading set and the target pinyin, the text to be proofread is proofread to obtain the proofread text; wherein the pinyin proofreading set includes a first pinyin proofreading set or a second pinyin proofreading set, the first pinyin proofreading set includes the corrected pinyin and the corrected text for proofreading the text contained in the business process where the current inquiry node is located, the second pinyin proofreading set includes the corrected pinyin and the corrected text for proofreading the text contained in the current inquiry node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

2. The method according to claim 1, characterized in that The step of proofreading the text to be proofread according to the pre-configured pinyin proofreading set and the target pinyin to obtain the proofread text includes: Obtaining the business process where the current query node is located, and determining whether the business process is configured with the first pinyin proofreading set; If the business process is configured with the first pinyin proofreading set, proofreading the text to be proofread according to the first pinyin proofreading set and the target pinyin; If the business process is not configured with the first pinyin proofreading set, obtaining the second pinyin proofreading set configured at the current query node; The text to be proofread is proofread according to the second pinyin proofreading set and the target pinyin to obtain the proofread text.

3. The method according to claim 2, characterized in that Proofreading the text to be proofread according to the first pinyin proofreading set and the target pinyin includes: According to the target pinyin and the corrected pinyins in each first pinyin proofreading set, a target first pinyin proofreading set matching the target pinyin is determined in each first pinyin proofreading set; Correct the text to be proofread into the corrected text in the target first pinyin proofreading set.

4. The method according to claim 3, characterized in that The step of determining a target first pinyin proofreading set matching the target pinyin from each first pinyin proofreading set according to the target pinyin and the corrected pinyin in each first pinyin proofreading set comprises: Obtaining a confused pinyin corresponding to the target pinyin, wherein the similarity value between the confused pinyin and the target pinyin is higher than a preset threshold; The target pinyin and the confused pinyin are compared with the corrected pinyin in each first pinyin proofreading set to determine a target first pinyin proofreading set that matches the target pinyin.

5. The method according to claim 2, characterized in that: Proofreading the text to be proofread according to the second pinyin proofreading set and the target pinyin includes: According to the target pinyin and the corrected pinyins in each second pinyin proofreading set, a target second pinyin proofreading set matching the target pinyin is determined in each second pinyin proofreading set; Correct the text to be proofread into the corrected text in the target second pinyin proofreading set.

6. The method according to claim 5, characterized in that The step of determining a target second pinyin proofreading set matching the target pinyin from each second pinyin proofreading set according to the target pinyin and the corrected pinyin in each second pinyin proofreading set comprises: Obtaining a confused pinyin corresponding to the target pinyin, wherein a similarity value between the confused pinyin and the target pinyin is higher than a preset threshold; The target pinyin and the confused pinyin are compared with the corrected pinyin in each second pinyin proofreading set to determine a target second pinyin proofreading set that matches the target pinyin.

7. The method according to claim 2, characterized in that: The obtaining of the business process of the current query node includes: According to a preset flow chart, determine the previous node and the next node of the current query node; The business process in which the current query node is located is determined according to the previous node and the next node of the current query node.

8. The method according to claim 1, characterized in that: The step of obtaining the target pinyin corresponding to the text to be proofread according to the pinyin of each text in the text to be proofread includes: Delete the phonetic symbol of each character's pinyin to obtain the corresponding pinyin without phonetic symbol of each character; The non-phonetic pinyin of each character is separated according to the preset separator to obtain the target pinyin corresponding to the character to be proofread.

9. The method according to claim 1, characterized in that: The method further comprises: Acquire a first user intention matching the text to be proofread and a matching value corresponding to the first user intention; Acquire a second user intent that matches the proofread text and a matching value corresponding to the second user intent; determining, according to a matching value corresponding to the first user intention and a matching value corresponding to the second user intention, the first user intention or the second user intention as a true user intention; According to the real user intention, the answer information corresponding to the voice input at the current query node is obtained.

10. The method according to claim 1, characterized in that The method further comprises: The text to be proofread and the proofread text are written into a preset interaction record.

11. The method according to claim 10, characterized in that The method further comprises: A proofreading report is generated according to the text to be proofread, the pre-configured pinyin proofreading set and the proofread text.

12. A text proofreading device, characterized in that: include: A result acquisition module, used to acquire the speech recognition result of the current query node, wherein the speech recognition result includes the text to be proofread obtained by recognizing the speech input at the current query node; A pinyin conversion module, used for obtaining the target pinyin corresponding to the text to be proofread according to the pinyin of each character in the text to be proofread; A text proofreading module is used to proofread the text to be proofread according to a pre-configured pinyin proofreading set and the target pinyin to obtain the proofread text; wherein the pinyin proofreading set includes a first pinyin proofreading set or a second pinyin proofreading set, the first pinyin proofreading set includes a corrected pinyin and a corrected text for proofreading the text contained in the business process where the current inquiry node is located, the second pinyin proofreading set includes a corrected pinyin and a corrected text for proofreading the text contained in the current inquiry node, and the corrected text in the first pinyin proofreading set is different from the corrected text in the second pinyin proofreading set.

13. A computer device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 11.

14. A readable storage medium, characterized in that: The readable storage medium stores computer instructions, which are used to implement the method according to any one of claims 1 to 11 when executed by a processor.

15. A program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images