Intelligent dialogue method, system, device and storage medium based on customer service recording

By mining customer service scripts from online real-world corpora, the problem of high optimization costs for intelligent customer service robot scripts was solved, resulting in richer and more diverse robot scripts and improved user experience.

CN112562678BActive Publication Date: 2024-11-29CTRIP COMP TECH SHANGHAI
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202011350237.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-26
Publication Date
2024-11-29
Estimated Expiration
2040-11-26

AI Technical Summary

Technical Problem

Optimizing the scripts of existing intelligent customer service robots is costly and requires a lot of manpower to compile real customer service scripts, leading to a development bottleneck.

Method used

By using intelligent dialogue methods based on customer service recordings, and leveraging speech recognition and text similarity matching technologies, more customer service scripts can be mined from online real-world corpora to supplement the chatbot's scripts, reducing human intervention and increasing the diversity of AI dialogues.

Benefits of technology

It reduced the cost of optimizing chatbots, enriched chatbot dialogue, improved user experience, and avoided the aversion and fatigue caused by monotonous and boring dialogue.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112562678B_ABST
    Figure CN112562678B_ABST
Patent Text Reader

Abstract

The present invention provides an intelligent dialogue method, system, device and storage medium based on customer service recordings, the method comprising: providing a customer service recording audio file and converting it into a customer service text; vectorizing the dialogue sentences and customer service sentences through different word frequency similarity algorithms, and taking the multiple customer service sentences with the highest similarity as the first set; performing word segmentation, obtaining the word vector and sentence vector after word segmentation, and taking the multiple dialogue sentences with the highest similarity as the second set; obtaining the dialogue sentences and the corresponding customer service sentences in the second set based on the semantic similarity model, and sorting and filtering the third set by similarity; supplementing the customer service sentences in the third set to the dialogue corpus; when it is necessary to express the dialogue sentence, select one of the customer service sentences associated with the dialogue sentence for output. The present invention can use the existing conversation robot configuration speech as a template, and through the text similarity matching method, dig out more customer service speech, reduce manpower investment, and improve the diversity of AI dialogue.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0002] At present, the application of intelligent customer service has been quite extensive, and many companies have their own outbound call systems or intelligent conversational robots. For each behavioral instruction of the robot, it needs to be accompanied by a corresponding text script as the robot's reply to the guest. For Ctrip's outbound task-based robots, the service objects they face are hotel customer service or guests. Because of the nature of the service, the script of the intelligent robot is different from that of the chat robot. It needs to be as rich and varied as possible and close to real people without losing formality. The current method is that the business side sorts out the real Ctrip customer service's script during the conversation with the guest as the robot's script, which requires a lot of manpower to sort out, greatly increasing the optimization cost of the conversational robot and bringing bottlenecks to the development of the conversational robot.

[0003] Therefore, the present invention provides an intelligent dialogue method, system, device and storage medium based on customer service recording. Summary of the invention

[0004] In response to the problems in the prior art, the purpose of the present invention is to provide an intelligent dialogue method, system, device and storage medium based on customer service recordings, which overcomes the difficulties of the prior art and can use the existing conversational robot configuration script as a template to mine more customer service scripts from the online real corpus through the text similarity matching method as a supplement to the robot's script. There is no need for the business side to constantly organize and summarize new and different scripts, which can not only reduce manpower investment but also improve the diversity of AI dialogues.

[0005] An embodiment of the present invention provides an intelligent dialogue method based on customer service recording, comprising the following steps:

[0006] S110, providing a customer service recording audio file, and converting the customer service recording audio file into a customer service text through voice recognition;

[0007] S120, weighting and vectorizing the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text by using at least two different word frequency similarity algorithms, and obtaining, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each of the dialogue sentences under different similarity algorithms as a first set based on the dialogue sentences;

[0008] S130, respectively segmenting the customer service sentences and the dialogue sentences in the first set to obtain word vectors and sentence vectors after segmentation, and taking the dialogue sentences in the first set with the highest similarity between the sentence vectors of the dialogue sentences and the sentence vectors of the customer service sentences as the second set based on the dialogue sentences;

[0009] S140, performing similarity calculation on the conversation sentences obtained based on the semantic similarity model and the corresponding customer service sentences in the second set, and selecting customer service sentences that meet a preset threshold as a third set based on the conversation sentences;

[0010] S160: Supplement the customer service sentences in the third set into the dialogue corpus; and

[0011] S170. When the dialogue statement needs to be expressed, one of the customer service statements associated with the dialogue statement is selected for output.

[0012] Preferably, in step S120, the dialogue sentences in the current dialogue corpus and the customer service sentences in the customer service text are weighted and vectorized using the TF-IDF word frequency similarity algorithm and the BM25 word frequency similarity algorithm respectively.

[0013] Preferably, in step S120, a plurality of customer service sentences having the highest average similarity to each of the dialogue sentences under the two similarity algorithms are taken as the first set, and the total number of the customer service sentences meets the preset total number.

[0014] Preferably, the step S130 includes the following steps:

[0015] S131, segmenting the customer service sentences and the dialogue sentences in the first set;

[0016] S132, obtaining a word vector for each word in the sentence after word segmentation according to a preset word vector;

[0017] S133, obtaining a sentence vector by averaging the word vectors in the customer service sentence, and obtaining a sentence vector by averaging the word vectors of the dialogue sentences in the first set;

[0018] S134, obtaining the similarity between the sentence vector of the customer service sentence and the sentence vector of each of the corresponding dialogue sentences in the first set;

[0019] S135: taking a preset number of dialogue sentences with high similarity rankings in the first set as a second set.

[0020] Preferably, in step S140, a trained BERT model is used to calculate the similarity between the dialogue sentence and the corresponding customer service sentence in the second set.

[0021] Preferably, after step S140 and before step S160, the following steps are also included:

[0022] S150: Correct text errors in the customer service statements in the third set.

[0023] Preferably, in step S160, the customer service sentences in the third set are added to the dialogue corpus, and a mapping relationship between the dialogue sentences and the customer service sentences is established in the dialogue corpus according to the dialogue sentences and the third set; and

[0024] In the step S170, when the dialogue sentence needs to be expressed, one of the customer service sentences having a mapping relationship with the dialogue sentence is selected for output.

[0025] An embodiment of the present invention further provides an intelligent dialogue system based on customer service recordings, which is used to implement the above-mentioned intelligent dialogue method based on customer service recordings. The intelligent dialogue system based on customer service recordings includes:

[0026] A customer service recording module, which provides a customer service recording audio file and converts the customer service recording audio file into a customer service text through voice recognition;

[0027] A first set module, weighting and vectorizing the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text by using at least two different word frequency similarity algorithms, and obtaining, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each of the dialogue sentences under different similarity algorithms as a first set based on the dialogue sentences;

[0028] A second set module, respectively segmenting the customer service sentences and the dialogue sentences in the first set to obtain word vectors and sentence vectors after segmentation, and taking the multiple dialogue sentences in the first set with the highest similarity between the sentence vectors of the dialogue sentences and the sentence vectors of the customer service sentences as the second set based on the dialogue sentences;

[0029] A third set module, which obtains the dialogue sentences based on a semantic similarity model and calculates similarity with the corresponding customer service sentences in the second set, and selects customer service sentences that meet a preset threshold as a third set based on the dialogue sentences;

[0030] a sentence supplement module, which supplements the customer service sentences in the third set into the dialogue corpus; and

[0031] The expression output module selects and outputs one of the customer service statements associated with the dialogue statement when the dialogue statement needs to be expressed.

[0032] An embodiment of the present invention further provides an intelligent dialogue device based on customer service recording, comprising:

[0033] processor;

[0034] a memory storing executable instructions of the processor;

[0035] Among them, the processor is configured to perform the steps of the above-mentioned intelligent dialogue method based on customer service recording by executing the executable instructions.

[0036] An embodiment of the present invention also provides a computer-readable storage medium for storing a program, which, when executed, implements the steps of the above-mentioned intelligent dialogue method based on customer service recording.

[0037] The purpose of the present invention is to provide an intelligent dialogue method, system, device and storage medium based on customer service recordings, which can use the existing conversational robot configuration script as a template, and through the text similarity matching method, mine more customer service scripts from the online real corpus as a supplement to the robot's script. There is no need for the business party to constantly organize and summarize new and different scripts, which can not only reduce manpower investment but also improve the diversity of AI dialogues. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Other features, objectives and advantages of the present invention will become more apparent from a reading of the detailed description of non-limiting embodiments made with reference to the following accompanying drawings.

[0039] Figure 1 It is a flow chart of the intelligent dialogue method based on customer service recording of the present invention.

[0040] Figure 2 It is a module schematic diagram of the intelligent dialogue system based on customer service recording of the present invention.

[0041] Figure 3 It is a structural schematic diagram of the intelligent dialogue device based on customer service recording of the present invention.

[0042] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention will be comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar structures, and thus their repeated description will be omitted.

[0044] Figure 1 Flowchart of the intelligent dialogue method based on customer service recording of the present invention. Figure 1 As shown, an embodiment of the present invention provides an intelligent dialogue method based on customer service recording, comprising the following steps:

[0045] S110: Provide a customer service recording audio file, and convert the customer service recording audio file into a customer service text through voice recognition.

[0046] S120. Weight and vectorize the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text using at least two different word frequency similarity algorithms, and obtain, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each dialogue sentence under different similarity algorithms as a first set based on the dialogue sentences.

[0047] S130. Segment the customer service sentences and dialogue sentences in the first set respectively to obtain word vectors and sentence vectors after segmentation, and use multiple dialogue sentences with the highest similarity between the sentence vectors of the dialogue sentences in the first set and the sentence vectors of the customer service sentences as the second set based on the dialogue sentences.

[0048] S140: Calculate similarity between the dialogue sentences obtained based on the semantic similarity model and the corresponding customer service sentences in the second set, and select customer service sentences that meet a preset threshold as a third set based on the dialogue sentences.

[0049] S150. Correct text errors in the customer service statements in the third set.

[0050] S160: Add the customer service sentences in the third set to the dialogue corpus.

[0051] S170. When a dialogue statement needs to be expressed, one of the customer service statements associated with the dialogue statement is selected for output.

[0052] The present invention uses the existing conversational robot configuration script as a template, and through the text similarity matching method, mines more customer service scripts from the online real corpus as a supplement to the robot's scripts. No manpower is required for sorting in the whole process, which greatly reduces the optimization cost of the conversational robot and contributes to the rapid iteration and experience upgrade of the conversational robot.

[0053] The purpose is to save time in manually organizing the script and enrich the robot's script.

[0054] In a preferred embodiment, in step S120, the dialogue sentences in the current dialogue corpus and the customer service sentences in the customer service text are weighted and vectorized using the TF-IDF word frequency similarity algorithm and the BM25 word frequency similarity algorithm respectively.

[0055] TF-IDF (Term Frequency-inverse Document Frequency) is a commonly used weighting technique for information retrieval and data mining, and can be used for statistical analysis methods for keywords. TF stands for term frequency, and IDF stands for inverse document frequency. TF-IDF (termfrequency-inverse document frequency) is used to evaluate the importance of a word to a document set or a corpus. The importance of a word is proportional to the number of times it appears in an article and inversely proportional to the number of times it appears in a corpus. This calculation method can effectively avoid the influence of common words on keywords and improve the relevance between keywords and articles. TF refers to the total number of times a word appears in an article. This indicator is usually normalized and defined as TF = (number of times a word appears in a document / total number of words in a document), which can prevent the results from being biased towards documents that are too long (the same word usually has a higher frequency in a long document than in a short document). IDF is the inverse document frequency. The fewer documents that contain a certain word, the larger the IDF value, indicating that the word has a strong ability to distinguish. IDF = loge (total number of documents in the corpus / number of documents containing the word + 1). The reason for +1 is to avoid the denominator being 0. TFIDF = TFxIDF. The larger the TFIDF value, the more important the feature word is to the text.

[0056] The BM25 relevance algorithm is an algorithm used to evaluate the relevance between search terms and documents. It is an algorithm based on a probabilistic retrieval model. Let's use simple words to describe the BM25 algorithm: there is a query and a batch of documents Ds. Now we need to calculate the relevance score between the query and each document D. Specifically, we first split the query to get the word $q_i$. Then the word score consists of three parts: the relevance between the word $q_i$ and D; the relevance between the word $q_i$ and D; the weight of each word. Finally, we sum the scores of each word to get the score between the query and the document.

[0057] In a preferred embodiment, in step S120, a plurality of customer service sentences having the highest average similarity to each dialogue sentence under two similarity algorithms are taken as the first set, and the total number of customer service sentences meets the preset total number.

[0058] In a preferred embodiment, step S130 includes the following steps:

[0059] S131. Segment the customer service sentences and dialogue sentences in the first set.

[0060] S132. Obtain a word vector for each word in the sentence after word segmentation according to a preset word vector.

[0061] S133. Obtain a sentence vector by averaging the word vectors in the customer service sentences, and obtain a sentence vector by averaging the word vectors of the dialogue sentences in the first set.

[0062] S134. Obtain the similarity between the sentence vector of the customer service sentence and the sentence vector of each corresponding dialogue sentence in the first set.

[0063] S135. Take a preset number of dialogue sentences with high similarity rankings in the first set as the second set.

[0064] In a preferred embodiment, in step S140, a trained BERT model is used to calculate the similarity between the dialogue sentence and the corresponding customer service sentence in the second set. The full name of the BERT model is: Bidirectional Encoder Representations from Transformer. As can be seen from the name, the goal of the BERT model is to use large-scale unlabeled corpus training to obtain a representation of the text containing rich semantic information, that is, the semantic representation of the text, and then fine-tune the semantic representation of the text in a specific NLP task, and finally apply it to the NLP task. For example, the process of training the semantic representation of the text with the BERT model is like learning basic subjects such as Chinese, mathematics, English, physics, chemistry, and biology in high school to consolidate the basic knowledge. The fine-tuning of the model's parameters in a specific NLP task is equivalent to further strengthening the existing basic knowledge and the selected major during college, so as to obtain professional skills that can be applied to actual scenarios.

[0065] In a preferred embodiment, in step S160, the customer service sentences in the third set are added to the dialogue corpus, and a mapping relationship between the dialogue sentences and the customer service sentences is established in the dialogue corpus according to the dialogue sentences and the third set.

[0066] In step S170, when a dialogue sentence needs to be expressed, one of the customer service sentences that have a mapping relationship with the dialogue sentence is selected for output. For example, when the dialogue sentence A that needs to be expressed is "Take a plane to Beijing tomorrow", the dialogue sentence A corresponds to (according to the calculation process in the previous dialogue corpus, the sentences that have a mapping relationship with A) customer service sentence B "Take a plane to Beijing tomorrow", customer service sentence C "Tomorrow's plane ticket, to Beijing", and customer service sentence D "Fly to Beijing, tomorrow's ticket", then one of customer service sentence B, customer service sentence C, and customer service sentence D is selected for output. Since the customer service sentence itself is based on colloquialism and humanization, it helps to enrich the emotionality of the conversation robot's dialogue, making the user feel more like communicating with a real customer service, thereby improving the user experience.

[0067] In a variation, in step S170, when a dialogue sentence needs to be expressed, one is selected from the dialogue sentence and the set of customer service sentences with a mapping relationship between the dialogue sentences for output. For example, when the dialogue sentence A to be expressed is "take a plane to Beijing tomorrow", the dialogue sentence A corresponds to (based on the previous operation process in the dialogue corpus, has a mapping relationship with A) customer service sentence B "take a plane to Beijing tomorrow", customer service sentence C "tomorrow's plane ticket, to Beijing", customer service sentence D "fly to Beijing, tomorrow's ticket", then when outputting, the dialogue sentence A, customer service sentence B, customer service sentence C and customer service sentence D form a set, and a sentence (dialogue sentence or customer service sentence) is selected from this set for output.

[0068] The present invention is mainly implemented by the following method, which mainly comprises the following steps:

[0069] Initial text screening: The first step is the corpus recall step. According to the two frequency-based methods of TF-IDF and BM25, the robot's speech after word segmentation and the online corpus are scored respectively, and the sentences are vectorized. Then the similarity between the robot's speech and the online corpus is calculated, and the 200 online corpora with the highest similarity between the two algorithms are taken as the results of the initial text screening. Because the complexity of the TF-IDF and BM25 algorithms is lower than that of the model, the prediction speed of the two methods is relatively fast in the face of a large online corpus. As a preliminary screening stage with not particularly high accuracy requirements, it is a better choice.

[0070] Model 1 is used to calculate text similarity: After the initial screening of the text, a relatively complex and more accurate method is required to screen the speech. Model 1 uses the word vector averaging method. Specifically, first, the robot speech and the online corpus screened in step 1 are segmented, and then the vector corresponding to each word is obtained from the previously pre-trained word vector. Finally, the word vectors in the sentence are averaged to obtain the final vector of the sentence. Then the similarity between the robot speech and the online corpus is calculated, and it is determined whether it is greater than the set threshold. If it is greater than the threshold, continue with the following steps, otherwise it is deemed that there is no qualified corpus with high similarity, and the process ends.

[0071] Model 2 is used to calculate text similarity: Since the standards for selecting available online customer service scripts are very strict, the design of the entire process includes multiple layers of different screening methods. Model 2 is a deep semantic model. The pre-trained BERT model is fine-tuned with Ctrip data and the retrained semantic matching model is used to calculate the similarity between two texts from a semantic perspective. Similarly, the robot scripts and the online predictions left after the screening in step 2 are calculated for similarity, and it is determined whether it is greater than the set threshold. If it is greater than the threshold, continue with the following steps, otherwise it is deemed that there is no qualified corpus with high similarity and the process ends.

[0072] Text normalization: After the first three steps of screening, the remaining online corpus can be used as the new mining scripts for configuration. However, because the screened text is obtained by transcribing the real online recordings, there may be transcription errors and overly colloquial expressions, so text normalization is also required, including:

[0073] The original sentence is: "Then I will confirm the order for you, okay?" The repeated words need to be removed, and the corrected result is: "Then I will confirm the order for you, okay?"

[0074] The original sentence is: "Sorry, there is no relevant information at the moment. Please restore it in the background. Thank you and goodbye." The incorrectly transcribed keywords need to be corrected. The corrected result is: "Sorry, there is no relevant information at the moment. Please reply in the background. Thank you and goodbye."

[0075] Finally, the standardized text is used as the newly discovered excellent customer service words, configured and put online as a robot

[0076] The supplement of existing speech makes the intelligent robot more rich and human from the perspective of speech, avoiding the disgust and fatigue of customers due to monotonous and boring speech.

[0077] The intelligent dialogue method based on customer service recordings of the present invention can use the existing conversational robot configuration words as a template, and through the text similarity matching method, it can mine more customer service words from the online real corpus as a supplement to the robot's words. There is no need for the business party to constantly organize and summarize new and different words, which can not only reduce manpower investment but also improve the diversity of AI dialogues.

[0078] Figure 2 Schematic diagram of the module of the intelligent dialogue system based on customer service recording of the present invention. Figure 2 As shown, the embodiment of the present invention also provides an intelligent dialogue system based on customer service recordings, which is used to implement the above-mentioned intelligent dialogue method based on customer service recordings. The intelligent dialogue system based on customer service recordings 5 ​​includes

[0079] The customer service recording module 51 provides a customer service recording audio file and converts the customer service recording audio file into a customer service text through voice recognition.

[0080] The first set module 52 weights and vectorizes the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text by using at least two different word frequency similarity algorithms, and obtains, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each dialogue sentence under different similarity algorithms as a first set based on the dialogue sentences.

[0081] The second set module 53 performs word segmentation on the customer service sentences and dialogue sentences in the first set respectively to obtain word vectors and sentence vectors after word segmentation, and takes multiple dialogue sentences in the first set with the highest similarity between the sentence vectors of the dialogue sentences and the sentence vectors of the customer service sentences as the second set based on the dialogue sentences.

[0082] The third set module 54 obtains the dialogue sentences based on the semantic similarity model and performs similarity calculation on the corresponding customer service sentences in the second set, and selects the customer service sentences that meet the preset threshold as the third set based on the dialogue sentences.

[0083] The text correction module 55 corrects text errors of the customer service sentences in the third set.

[0084] The sentence supplement module 56 supplements the customer service sentences in the third set to the dialogue corpus. And

[0085] The expression output module 57 selects and outputs one of the customer service sentences associated with the dialogue sentence when a dialogue sentence needs to be expressed.

[0086] The intelligent dialogue system based on customer service recordings of the present invention can use the existing conversational robot configuration script as a template, and through the text similarity matching method, it can mine more customer service scripts from the online real corpus as a supplement to the robot's script. There is no need for the business party to constantly organize and summarize new and different scripts, which can not only reduce manpower investment but also improve the diversity of AI dialogues.

[0087] The embodiment of the present invention further provides an intelligent dialogue device based on customer service recording, comprising a processor and a memory, wherein executable instructions of the processor are stored. The processor is configured to execute the steps of the intelligent dialogue method based on customer service recording by executing the executable instructions.

[0088] As shown above, the intelligent dialogue system based on customer service recordings of this embodiment of the present invention can use the existing conversational robot configuration words as a template, and through the text similarity matching method, it can mine more customer service words from the online real corpus as a supplement to the robot's words. There is no need for the business party to constantly organize and summarize new and different words, which can not only reduce manpower investment but also improve the diversity of AI dialogues.

[0089] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as systems, methods or program products. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits", "modules" or "platforms".

[0090] Figure 3 This is a schematic diagram of the structure of the intelligent dialogue device based on customer service recording of the present invention. Figure 3 The electronic device 600 according to this embodiment of the present invention is described. Figure 3 The electronic device 600 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0091] like Figure 3 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0092] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 executes the steps of various exemplary embodiments of the present invention described in the electronic prescription circulation processing method section of this specification. For example, the processing unit 610 can execute the following steps: Figure 1Follow the steps shown in .

[0093] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0094] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0095] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0096] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms, etc.

[0097] The embodiment of the present invention also provides a computer-readable storage medium for storing a program, and the steps of the intelligent dialogue method based on customer service recording are implemented when the program is executed. In some possible implementations, various aspects of the present invention can also be implemented in the form of a program product, which includes a program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps of various exemplary embodiments of the present invention described in the above-mentioned electronic prescription circulation processing method section of this specification.

[0098] As shown above, the intelligent dialogue system based on customer service recordings of this embodiment of the present invention can use the existing conversational robot configuration words as a template, and through the text similarity matching method, it can mine more customer service words from the online real corpus as a supplement to the robot's words. There is no need for the business party to constantly organize and summarize new and different words, which can not only reduce manpower investment but also improve the diversity of AI dialogues.

[0099] Figure 4 Schematic diagram of the structure of the computer-readable storage medium of the present invention. Figure 4 As shown, a program product 800 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0100] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The 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 readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable 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.

[0101] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program codes contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0102] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0103] In summary, the purpose of the present invention is to provide an intelligent dialogue method, system, device and storage medium based on customer service recordings. The intelligent dialogue system based on customer service recordings of the present invention can use the existing conversational robot configuration words as a template, and through the text similarity matching method, it can mine more customer service words from the online real corpus as a supplement to the robot's words. There is no need for the business party to constantly organize and summarize new and different words, which can not only reduce manpower investment, but also improve the diversity of AI dialogues.

[0104] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. An intelligent dialogue method based on customer service recordings, It is characterized in that The following steps are involved: S110, providing a customer service recording audio file, and converting the customer service recording audio file into a customer service text through voice recognition; S120, weighting and vectorizing the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text by using at least two different word frequency similarity algorithms, and obtaining, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each of the dialogue sentences under different similarity algorithms as a first set based on the dialogue sentences; S130, respectively segmenting the conversation sentence and the corresponding customer service sentence in the first set, and obtaining a word vector for each word in the segmented sentence according to a preset word vector; A sentence vector is obtained by averaging the word vectors in the customer service sentences, and a sentence vector is obtained by averaging the word vectors of the dialogue sentences in the first set, and a plurality of the customer service sentences in the first set having the highest similarity between the sentence vectors in the customer service sentences and the sentence vectors of the dialogue sentences are used as a second set based on the dialogue sentences; S140, performing similarity calculation on the conversation sentences obtained based on the semantic similarity model and the corresponding customer service sentences in the second set, and selecting customer service sentences that meet a preset threshold as a third set based on the conversation sentences; S160: Supplement the customer service sentences in the third set into the dialogue corpus; as well as S170. When the dialogue statement needs to be expressed, one of the customer service statements associated with the dialogue statement is selected for output.

2. The intelligent dialogue method based on customer service recording according to claim 1, It is characterized in that In the step S120, the dialogue sentences in the current dialogue corpus and the customer service sentences in the customer service text are weighted and vectorized using the TF-IDF word frequency similarity algorithm and the BM25 word frequency similarity algorithm respectively.

3. The intelligent dialogue method based on customer service recording according to claim 2, It is characterized in that In the step S120, a plurality of customer service sentences having the highest average similarity to each of the dialogue sentences under the two similarity algorithms are taken as a first set, and the total number of the customer service sentences meets a preset total number.

4. The intelligent dialogue method based on customer service recording according to claim 1, It is characterized in that The step S130 includes the following steps: S131, performing word segmentation on the dialogue sentence and the corresponding customer service sentence in the first set; S132, obtaining a word vector for each word in the sentence after word segmentation according to a preset word vector; S133, obtaining a sentence vector by averaging the word vectors in the dialogue sentence, and obtaining a sentence vector by averaging the word vectors of the customer service sentences in the first set; S134, obtaining the similarity between the sentence vector of the dialogue sentence and the sentence vector of each of the customer service sentences in the first set; S135. Take a preset number of customer service statements with high similarity rankings in the first set as a second set.

5. The intelligent dialogue method based on customer service recording according to claim 1, It is characterized in that In the step S140, a trained BERT model is used to calculate the similarity between the dialogue sentence and the corresponding customer service sentence in the second set.

6. The intelligent dialogue method based on customer service recording according to claim 1, It is characterized in that The steps after step S140 and before step S160 also include: S150: Correct text errors in the customer service statements in the third set.

7. The intelligent dialogue method based on customer service recording according to claim 1, It is characterized in that In the step S160, the customer service sentences in the third set are added to the dialogue corpus, and a mapping relationship between the dialogue sentences and the customer service sentences is established in the dialogue corpus according to the dialogue sentences and the third set; as well as In the step S170, when the dialogue sentence needs to be expressed, one of the customer service sentences having a mapping relationship with the dialogue sentence is selected for output.

8. An intelligent dialogue system based on customer service recordings, used to implement the intelligent dialogue method based on customer service recordings as claimed in claim 1, It is characterized in that include: A customer service recording module, which provides a customer service recording audio file and converts the customer service recording audio file into a customer service text through voice recognition; A first set module, weighting and vectorizing the dialogue sentences in the dialogue corpus and the customer service sentences in the customer service text by using at least two different word frequency similarity algorithms, and obtaining, based on the dialogue sentences, a plurality of customer service sentences with the highest similarity to each of the dialogue sentences under different similarity algorithms as a first set based on the dialogue sentences; The second set module performs word segmentation on the dialogue sentences and the corresponding customer service sentences in the first set, respectively, and obtains a word vector of each word in the sentence after word segmentation according to a preset word vector; obtains a sentence vector by averaging the word vectors in the customer service sentences, and obtains a sentence vector by averaging the word vectors of the dialogue sentences in the first set, and uses the plurality of customer service sentences in the first set whose sentence vectors have the highest similarity to the sentence vectors of the dialogue sentences as the second set based on the dialogue sentences; A third set module, which obtains the dialogue sentences based on a semantic similarity model and calculates similarity with the corresponding customer service sentences in the second set, and selects customer service sentences that meet a preset threshold as a third set based on the dialogue sentences; A sentence supplement module, which supplements the customer service sentences in the third set into the dialogue corpus; as well as The expression output module selects and outputs one of the customer service statements associated with the dialogue statement when the dialogue statement needs to be expressed.

9. An intelligent conversation device based on customer service recordings, It is characterized in that include: processor; a memory storing executable instructions of the processor; Wherein, the processor is configured to execute the steps of the intelligent dialogue method based on customer service recording as described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium for storing a program, It is characterized in that When the program is executed, the steps of the intelligent dialogue method based on customer service recording described in any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Method and system for processing outbound conversation

    CN109618067A

  • Dialogue generation method, dialogue generation device, electronic device and storage medium

    CN109977207A