Intelligent outbound method and device, electronic equipment and storage medium

Abstract

The invention provides an intelligent outbound method and device, electronic equipment and a storage medium, and belongs to the technical field of electronic equipment.The method comprises the steps that language recognition is conducted on first voice information input by a first user, and a first language is obtained; translating the first voice information based on a multilingual translation model to obtain semantic information; determining corresponding text reply information based on the semantic information; converting the text reply information into second voice information of the first language; and sending the second voice information to the first user. According to the intelligent outbound method and device, the electronic equipment and the storage medium provided by the invention, the problem of language limitation in an intelligent outbound system can be solved.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of electronic equipment, and more specifically, relates to an intelligent outbound calling method and device, electronic equipment, and storage medium. Background Art

[0002] The intelligent outbound call system is a system that can autonomously complete complex outbound call tasks. Through preset call processes and automated voice interaction, it can meet outbound call needs in various scenarios. For example, the intelligent outbound call system can be applied to various fields such as government and public welfare publicity and services. The intelligent outbound call system can also provide bill due reminders, customer care, financial collection reminders, etc. It is more efficient than one-way SMS notifications and users have a stronger perception. In terms of inbound calls, the intelligent outbound call system is mainly used in the telephone customer service scenarios of enterprises to answer customer questions and report the business data of customer inquiries.

[0003] Due to the different languages ​​in different regions, the intelligent outbound call system uses a single language to interact with users, which will cause difficulties for users to understand. Summary of the invention

[0004] The purpose of the present disclosure is to provide an intelligent outbound calling method and device, an electronic device, and a storage medium to solve the problem of language limitation in an intelligent outbound calling system.

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided an intelligent outbound calling method, including: Performing language recognition on first voice information input by a first user to obtain a first language; Translating the first speech information based on a multilingual translation model to obtain semantic information; Determining corresponding text reply information based on the semantic information; Converting the text reply message into a second voice message in a first language; The second voice information is sent to the first user.

[0006] A second aspect of the embodiments of the present disclosure provides an intelligent outbound calling device, including: A language recognition module, used to perform language recognition on the first voice information input by the first user to obtain a first language; A language translation module, used for translating the first speech information based on a multilingual translation model to obtain semantic information; An information reply module, used to determine corresponding text reply information based on the semantic information; An information conversion module, used for converting the text reply information into a second voice information in a first language; An information sending module is used to send the second voice information to the first user.

[0007] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the above-mentioned intelligent outbound calling method when executing the computer program.

[0008] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned intelligent outbound calling method are implemented.

[0009] The beneficial effects of the intelligent outbound calling method and device, electronic device, and storage medium provided by the embodiments of the present disclosure are: In the disclosed embodiment, the first device may be an intelligent outbound call device. When the first device receives a first voice message input by a first user, the first device first identifies the language of the first voice message, then translates the first voice message based on a multilingual translation model to obtain corresponding semantic information, and generates corresponding reply information based on the semantic information. The reply information is usually in text form. The text reply information is converted into a second voice message in the first language and sent to the first user, thereby enabling a conversation in the language used by the first user and providing a smoother and more natural interactive experience.

[0010] The disclosed embodiments can overcome language barriers, achieve effective communication and interaction with users, and help improve user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of an intelligent outbound calling method provided by an embodiment of the present disclosure; Figure 2 A structural block diagram of an intelligent outbound calling device provided in one embodiment of the present disclosure; Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0014] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0015] Please refer to Figure 1 , Figure 1 A flowchart of an intelligent outbound calling method provided by an embodiment of the present disclosure, the method comprising: S101: Perform language recognition on first voice information input by a first user to obtain a first language.

[0016] In this embodiment, the first device is an intelligent outbound call device, and the first device can automatically communicate with the first user to realize the intelligent outbound call function. First, the first voice information input by the first user can be obtained through an audio acquisition device such as a microphone, and then the first voice information is feature extracted to obtain feature parameters in the first voice information. The feature parameters in the first voice information are input into the language recognition model to obtain the corresponding first language. Among them, the language recognition model can be trained based on machine learning or deep learning models. During the training process, voice samples of different languages ​​are used as input, and feature representations of different languages ​​are learned through the model. S102: Translate the first voice information based on a multilingual translation model to obtain semantic information.

[0017] In this embodiment, the first voice information can be first converted into text information by voice recognition technology, and then the first voice information of the first user can be translated into machine-understandable semantic information by using a multilingual translation model. The multilingual translation model is based on neural machine translation (NMT) technology, and uses a deep learning algorithm to train a large number of bilingual or multilingual parallel corpora to learn the mapping relationship between different languages. The multilingual translation model will consider multiple aspects such as the contextual information of the text, the grammatical structure of the first language, and the semantic meaning to generate accurate and fluent translation results. For some complex language phenomena, such as polysemy of a word, differences in grammatical structure, etc., the model will use its own knowledge base and reasoning ability to make appropriate adjustments and optimizations.

[0018] S103: Determine corresponding text reply information based on the semantic information.

[0019] In this embodiment, the first device uses natural language processing (NLP) to perform in-depth analysis on the semantic information obtained by translation. The analysis process includes tasks such as entity recognition, relationship extraction, and intent recognition. Through semantic analysis, key information in the first voice information, such as names, place names, time, events, etc., as well as the intentions or needs of the first user, can be identified. After understanding the semantic information of the first user, an internal knowledge base or an external data source can be queried to obtain information or answers that match the intentions or needs of the first user. The knowledge base usually contains predefined reply templates, FAQs, domain knowledge graphs, etc. By querying the knowledge base, the first device can quickly find content related to the intentions or needs of the first user, and provide materials for generating reply information. According to the predefined reply template and the results of semantic analysis, filling the corresponding slot can generate text reply information for the semantic information (that is, the intentions or needs of the first user).

[0020] S104: Convert the text reply information into a second voice information in the first language.

[0021] In this embodiment, after the text reply information is determined, the text reply information can be converted into a second voice information output in the first language through speech synthesis technology. In the speech synthesis process, it is necessary to configure the parameters of speech synthesis, and set the language setting parameter to the first language to generate the second voice information matching the language input by the first user.

[0022] S105: Send a second voice message to the first user.

[0023] In this embodiment, the first device sends the converted second voice information to the first user, and the first user can receive a reply from the first device through voice, thus completing a complete automatic voice interaction process.

[0024] From the above, it can be concluded that in this embodiment, when the first device receives the first voice information input by the first user, it first identifies the language of the first voice information, and then translates the first voice information based on a multilingual translation model to obtain corresponding semantic information, and generates corresponding reply information based on the semantic information. The reply information is usually a text reply information, and the text reply information is converted into a second voice information in the first language and sent to the first user, thereby realizing a conversation according to the language used by the first user and providing a smoother and more natural interactive experience.

[0025] The disclosed embodiments can overcome language barriers, achieve effective communication and interaction with users, and help improve user experience.

[0026] In one embodiment of the present disclosure, the intelligent outbound calling method further includes: In response to the number of parallel corpora corresponding to the plurality of language pairs in the first sample set being greater than a first threshold, a multilingual translation model is trained based on the first sample set.

[0027] In response to the number of parallel corpora corresponding to the first language pair in the first sample set being less than or equal to a first threshold, the parallel corpora corresponding to the first language pair are expanded to obtain a second sample set, and a multilingual translation model is trained based on the second sample set; the first language pair is any language pair in the first sample set.

[0028] In this embodiment, a specific implementation method for training a multilingual translation model is provided. During the training process, it is necessary to collect parallel corpora corresponding to multiple language pairs as the first sample set. Parallel corpora refer to text pairs with the same or similar content between two or more languages, which are the basic data for training the translation model. For example, to achieve automatic translation between Chinese and Spanish, it is necessary to collect parallel corpora of the Chinese-Spanish language pair in advance to train the multilingual translation model.

[0029] During the training process, if the number of parallel corpora corresponding to multiple language pairs in the first sample set is large enough (i.e., greater than the first threshold), the multilingual translation model can be trained directly based on the first sample set. On the contrary, if the number of parallel corpora corresponding to a certain language pair (e.g., the first language pair) in the first sample set is small (i.e., less than or equal to the first threshold), in order to ensure the performance and generalization ability of the multilingual translation model, the parallel corpora corresponding to the first language pair need to be expanded, and then the multilingual translation model is trained based on the second sample set obtained after the expansion.

[0030] It can be concluded from the above that this embodiment expands the corpus by evaluating the number of parallel corpora, which is conducive to building a multilingual translation model with better performance and stronger generalization ability.

[0031] In one embodiment of the present disclosure, each language pair includes a second language and a third language. Expand the parallel corpus corresponding to the first language pair, including: If the number of parallel corpora corresponding to the second language pair and the third language pair are both greater than a second threshold, multiple fourth language pairs are obtained based on the second language pair and the corresponding third language pair; each fourth language pair includes the second language in the second language pair and the third language in the third language pair, the second language pair and the corresponding third language pair have the same fourth language, the second language pair includes the second language and the fourth language, and the third language pair includes the fourth language and the third language; and the parallel corpora corresponding to the first language pair are expanded based on the fourth language pairs.

[0032] If the amount of the corpus in the second language is greater than a fourth threshold, the second corpus corresponding to the second language is translated into a third corpus, and the parallel corpus corresponding to the first language is expanded based on the third corpus.

[0033] If the number of parallel corpora corresponding to the fifth language pair is greater than a fifth threshold, and the similarity between the third language and the fifth language is greater than the similarity threshold, the parallel corpora corresponding to the first language pair are expanded based on the parallel corpora corresponding to the fifth language pair.

[0034] In this embodiment, a specific implementation method for expanding the parallel corpus corresponding to the first language pair is provided.

[0035] If the first language pair includes the second language and the corresponding third language, when the parallel corpus corresponding to the first language pair is relatively small, if there is a language (for example, the fourth language) such that the number of second language pairs consisting of the second language and the fourth language, and the number of third language pairs consisting of the fourth language and the third language are both greater than the second threshold, then the fourth language is used as an intermediate bridge, a fourth language pair of the second language and the third language is newly created, and the fourth language pair is used as a newly added sample to expand the parallel corpus corresponding to the first language pair.

[0036] For example, if the second language is Chinese, the third language is Spanish, and the fourth language is English, and the first language pair is Chinese-Spanish, if there are fewer parallel corpora for Chinese-Spanish, but there are more parallel corpora for Chinese-English and English-Spanish, then English can be used as an intermediate bridge to create a new fourth language pair of Chinese-Spanish. For example, the English word for apple is "apple" and the Spanish word is "manzana". The second language pair of Chinese-English can be "apple-apple", and the third language pair of English-Spanish can be "apple-manzana". Based on this, the fourth language pair of Chinese-Spanish can be "apple-manzana".

[0037] In addition to the above situation, if the parallel corpus corresponding to the first language pair is small, but the amount of corpus data of the second language in the first language pair (including the second language and the third language) is greater than the fourth threshold, the second corpus corresponding to the second language can be translated into the third corpus, and the second corpus and the third corpus can be used as new parallel corpora to expand the parallel corpus corresponding to the first language pair.

[0038] For example, the second language is Chinese, the third language is French, and the first language pair is Chinese-French. If there are fewer parallel corpora corresponding to Chinese-Spanish, but more Chinese corpora, the existing translation model can be used to translate Chinese into French. In this way, new Chinese-French parallel corpora can be generated to expand the parallel corpora corresponding to the first language pair.

[0039] In addition, if there is a language (for example, the fifth language) that satisfies the similarity between the third language and the fifth language (the similarity between the third language and the fifth language is greater than the similarity threshold), and the number of parallel corpora corresponding to the fifth language pair consisting of the second language and the fifth language is greater than the fifth threshold, then the fifth language pair can be used as the first language pair, and the parallel corpora corresponding to the first language pair can be expanded. Among them, if two languages ​​are similar in grammar, vocabulary or semantics, they can be considered similar.

[0040] For example, the second language is Chinese, the third language is Spanish, and the first language pair is Chinese-Spanish. If there are fewer parallel corpora corresponding to Chinese-Spanish, but more parallel corpora corresponding to Chinese-Portuguese, then the parallel corpora corresponding to Chinese-Portuguese can be used as the parallel corpora for Chinese-Spanish.

[0041] It can be concluded from the above that this embodiment selects corresponding expansion methods to expand the parallel corpus corresponding to the first language pair based on different situations of the first language pair, which is conducive to improving the accuracy of the expanded data.

[0042] In one embodiment of the present disclosure, the intelligent outbound calling method further includes: The grammatical similarity, lexical similarity and semantic similarity between the third language and the fifth language are calculated respectively.

[0043] The similarity between the third language and the fifth language is obtained by weighted summing up the grammatical similarity, lexical similarity and semantic similarity.

[0044] In this embodiment, when calculating the grammatical similarity between the third language and the fifth language, a feature vector may be generated based on indicators of corresponding grammatical features (e.g., the number of types of parts of speech, the number of change rules of different parts of speech, the position change of sentence components, the types of special sentence patterns, etc.), and then the distance between the two feature vectors is calculated to obtain the grammatical similarity between the third language and the fifth language. The distance between the two feature vectors may be calculated using existing Euclidean distance and cosine similarity methods, which are not limited here.

[0045] When calculating the vocabulary similarity between the third language and the fifth language, you can use a pre-trained word vector model (such as Word2Vec, GloVe, etc.) to convert the vocabulary in the two languages ​​into vector representations. For each word in the third language, find the most similar word in the vocabulary vector space of the fifth language (which can be determined by calculating the cosine similarity between the vectors). Count the number of similar vocabulary pairs, and calculate the average cosine similarity of all similar vocabulary pairs as the vocabulary similarity of the two languages.

[0046] When calculating the semantic similarity between the third language and the fifth language, a large-scale parallel corpus can be used, that is, a corpus containing corresponding texts in two languages. The proportion of sentence pairs in the two languages ​​with completely corresponding semantics to the total number of sentence pairs in the parallel corpus can be counted. For example, if there are 1,000 parallel sentence pairs, of which 700 sentence pairs are very close in semantics, then the semantic similarity between the two languages ​​can be considered to be 0.7.

[0047] On the basis of obtaining the grammatical similarity, lexical similarity and semantic similarity, the grammatical similarity, lexical similarity and semantic similarity may be weightedly summed to obtain the similarity between the third language and the fifth language.

[0048] In one embodiment of the present disclosure, the intelligent outbound calling method further includes: An interaction topic and a first interaction stage of the interaction topic are determined based on historical interaction data of the first user.

[0049] The interactive content corresponding to the second interactive stage of the interactive theme is determined based on the first interactive stage of the interactive theme; the second interactive stage is the next stage of the first interactive stage.

[0050] Interact with the first user based on the interactive content.

[0051] In this embodiment, it is considered that in some application scenarios, multiple interactions are required between the first device and the first user, each interaction has a different focus, and the content of the multiple interactions is coherent. Therefore, the first device of this embodiment will first perform an in-depth analysis based on the historical interaction data each time it interacts, and identify the main areas and topics of concern to the first user through methods such as text classification and keyword extraction. For example, if the first user mentions words such as fitness and health preservation many times in historical interactions, it can be determined that health management is a potential interaction topic. At the same time, according to the feedback content of the first user at different stages, the interaction stage (that is, the first interaction stage) of the interaction topic is determined. For example, if the first user mainly asks about basic information about health management when interacting with the first device, the interaction stage can be determined as the product information understanding stage.

[0052] Based on the determination of the interaction theme and the first interaction stage of the interaction theme, combined with the interaction development pattern of similar users under the same theme in historical data, the corresponding interaction content of the next stage (i.e., the second interaction stage) can be determined. For example, by analyzing the interaction history of a large number of users under the theme of health management, it is found that after understanding the basic services, users will further pay attention to the price and effect of the services. The next interaction based on the interaction content of the second interaction stage is easy to arouse the interest of the first user and improve the interaction effect.

[0053] It can be concluded from the above that this embodiment gradually promotes the development of interactive topics based on the historical interaction data of the first user, and provides users with a more personalized, coherent and valuable interactive experience.

[0054] In one embodiment of the present disclosure, determining an interaction topic and a first interaction stage of the interaction topic based on historical interaction data of a first user includes: Perform keyword extraction on historical interaction data.

[0055] The interaction topics and the first interaction stage of the interaction topics are determined based on the keywords.

[0056] In this embodiment, a specific implementation method for determining the interaction theme and the first interaction stage of the interaction theme based on the historical interaction data of the first user is provided. First, a method based on word frequency statistics can be used to count the frequency of each word in the historical interaction data, and select words with higher frequency as keywords. For example, if the words "product", "function", "price" and so on appear more times in the historical interaction data, then these words may be selected as keywords. Then, cluster analysis is performed on the extracted keywords, and the related keywords are classified into one category, and each category represents a potential interaction theme. For example, keywords such as "mobile phone", "screen" and "camera" are classified into one category, representing the interaction theme of "mobile phone products"; keywords such as "price", "discount" and "cost-effectiveness" are classified into one category, representing the interaction theme of "product price". On this basis, the main interaction theme is determined according to the importance and frequency of each keyword. For example, if the frequency of keywords related to "mobile phone products" in the historical interaction data is much higher than that of keywords of other topics, then "mobile phone products" can be determined as the main interaction theme.

[0057] After obtaining the interaction theme, the user's current status and needs under the interaction theme can be inferred in combination with the context information to determine the first interaction stage. For example, if the keywords mainly revolve around the basic information of the product, such as "mobile phone model", "screen size", "processor performance", etc., then the first interaction stage can be determined as the "product information understanding stage"; if the keywords involve product price comparisons and preferential activities, such as "price", "discount", "promotion", etc., then the first interaction stage can be determined as the "price comparison stage".

[0058] It can be concluded from the above that this embodiment determines the interaction theme and the first interaction stage of the interaction theme based on the historical interaction data of the first user, providing a basis for subsequent interaction content generation and personalized services.

[0059] In one embodiment of the present disclosure, determining the interaction content corresponding to the second interaction stage of the interaction theme based on the first interaction stage of the interaction theme includes: An interaction template corresponding to the second interaction stage is selected from the set template library.

[0060] The interaction template is determined as the interaction content corresponding to the second interaction stage.

[0061] In this embodiment, a specific implementation method for determining the interactive content corresponding to the second interactive stage is provided. A series of interactive templates can be pre-designed as a template library according to different interactive themes and interactive stages. These templates can cover various common business scenarios and user needs. For example, for an interactive theme of e-commerce product recommendation, templates for different interactive stages can be designed, such as the template for the product information understanding stage can include [product name], [product features 1], [product features 2] and other contents; the template for the price comparison stage can include [product name], [specific price], [promotion 1], [promotion 2] and other contents.

[0062] In the actual interaction process, a corresponding interaction template may be selected from a template library based on the interaction theme and the interaction stage in which the interaction theme is located, and the content of the interaction template may be determined as the corresponding interaction content.

[0063] From the above, it can be concluded that this embodiment determines the interactive content corresponding to the second interactive stage of the interactive theme based on the template library. When the interactive content needs to be determined, it can be quickly retrieved and selected from the template library without generating content from scratch, which greatly shortens the response time and enables timely interaction with users. Corresponding to the intelligent outbound calling method in the above embodiment, Figure 2 This is a block diagram of the structure of an intelligent outbound calling device provided by an embodiment of the present disclosure. For ease of explanation, only the parts related to the embodiment of the present disclosure are shown. Figure 2 The intelligent outbound calling device 20 includes: a language recognition module 21, a language translation module 22, an information reply module 23, an information conversion module 24 and an information sending module 25.

[0064] The language recognition module 21 is used to perform language recognition on the first voice information input by the first user to obtain the first language.

[0065] The language translation module 22 is used to translate the first speech information based on a multilingual translation model to obtain semantic information.

[0066] The information reply module 23 is used to determine the corresponding text reply information based on the semantic information.

[0067] The information conversion module 24 is used to convert the text reply information into a second voice information in the first language.

[0068] The information sending module 25 is used to send the second voice information to the first user.

[0069] In one embodiment of the present disclosure, the language translation module 22 is specifically used for: In response to the number of parallel corpora corresponding to the plurality of language pairs in the first sample set being greater than a first threshold, a multilingual translation model is trained based on the first sample set.

[0070] In response to the number of parallel corpora corresponding to the first language pair in the first sample set being less than or equal to a first threshold, the parallel corpora corresponding to the first language pair are expanded to obtain a second sample set, and a multilingual translation model is trained based on the second sample set; the first language pair is any language pair in the first sample set.

[0071] In one embodiment of the present disclosure, each language pair includes a second language and a third language, and the language translation module 22 is further configured to: If the number of parallel corpora corresponding to the second language pair and the third language pair are both greater than a second threshold, multiple fourth language pairs are obtained based on the second language pair and the corresponding third language pair; each fourth language pair includes the second language in the second language pair and the third language in the third language pair, the second language pair and the corresponding third language pair have the same fourth language, the second language pair includes the second language and the fourth language, and the third language pair includes the fourth language and the third language; and the parallel corpora corresponding to the first language pair are expanded based on the fourth language pairs.

[0072] If the amount of the corpus in the second language is greater than a fourth threshold, the second corpus corresponding to the second language is translated into a third corpus, and the parallel corpus corresponding to the first language is expanded based on the third corpus.

[0073] If the number of parallel corpora corresponding to the fifth language pair is greater than a fifth threshold, and the similarity between the third language and the fifth language is greater than the similarity threshold, the parallel corpora corresponding to the first language pair are expanded based on the parallel corpora corresponding to the fifth language pair.

[0074] In one embodiment of the present disclosure, the language translation module 22 is further used for: The grammatical similarity, lexical similarity and semantic similarity between the third language and the fifth language are calculated respectively.

[0075] The similarity between the third language and the fifth language is obtained by weighted summing up the grammatical similarity, lexical similarity and semantic similarity.

[0076] In one embodiment of the present disclosure, the information reply module 23 is specifically used to: An interaction topic and a first interaction stage of the interaction topic are determined based on historical interaction data of the first user.

[0077] The interactive content corresponding to the second interactive stage of the interactive theme is determined based on the first interactive stage of the interactive theme; the second interactive stage is the next stage of the first interactive stage.

[0078] Interact with the first user based on the interactive content.

[0079] In one embodiment of the present disclosure, the information reply module 23 is further configured to: Perform keyword extraction on historical interaction data.

[0080] The interaction topics and the first interaction stage of the interaction topics are determined based on the keywords.

[0081] In one embodiment of the present disclosure, the information reply module 23 is further configured to: An interaction template corresponding to the second interaction stage is selected from the set template library.

[0082] The interaction template is determined as the interaction content corresponding to the second interaction stage.

[0083] See also Figure 3 , Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 3 The electronic device 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of each module / unit in the above-mentioned device embodiments, such as Figure 2 The functions of modules 21 to 25 are shown.

[0084] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0085] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.

[0086] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0087] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure can execute the implementation methods described in the first and second embodiments of the intelligent outbound call method provided in the embodiments of the present disclosure, and can also execute the implementation methods of the electronic device described in the embodiments of the present disclosure, which will not be repeated here.

[0088] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0089] The computer-readable storage medium may be an internal storage unit of the electronic device of any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0090] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0092] In the several embodiments provided in the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present disclosure.

[0094] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0095] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. An intelligent outbound calling method, characterized in that: include: Performing language recognition on first voice information input by a first user to obtain a first language; Translating the first speech information based on a multilingual translation model to obtain semantic information; Determining corresponding text reply information based on the semantic information; Converting the text reply message into a second voice message in a first language; The second voice information is sent to the first user.

2. The intelligent outbound calling method according to claim 1, characterized in that: Also includes: In response to the number of parallel corpora corresponding to the plurality of language pairs in the first sample set being greater than a first threshold, training the multilingual translation model based on the first sample set; In response to the number of parallel corpora corresponding to the first language pair in the first sample set being less than or equal to a first threshold, the parallel corpora corresponding to the first language pair are expanded to obtain a second sample set, and the multilingual translation model is trained based on the second sample set; the first language pair is any language pair in the first sample set.

3. The intelligent outbound calling method according to claim 2, characterized in that: Each of the language pairs includes a second language and a third language, The expanding the parallel corpus corresponding to the first language pair includes: If the number of parallel corpora corresponding to the second language pair and the third language pair is greater than a second threshold, a plurality of fourth language pairs are obtained based on the second language pair and the corresponding third language pair; each of the fourth language pairs includes the second language in the second language pair and the third language in the third language pair, the second language pair and the corresponding third language pair have the same fourth language, the second language pair includes the second language and the fourth language, and the third language pair includes the fourth language and the third language; and the parallel corpora corresponding to the first language pair are expanded based on the fourth language pairs; If the amount of the corpus in the second language is greater than a fourth threshold, translating the second corpus corresponding to the second language into a third corpus, and expanding the parallel corpus corresponding to the first language pair based on the third corpus; If the number of parallel corpora corresponding to the fifth language pair is greater than a fifth threshold, and the similarity between the third language and the fifth language is greater than a similarity threshold, the parallel corpora corresponding to the first language pair are expanded based on the parallel corpora corresponding to the fifth language pair.

4. The intelligent outbound calling method according to claim 3, characterized in that: Also includes: Calculate the grammatical similarity, lexical similarity and semantic similarity between the third language and the fifth language respectively; The grammatical similarity, the lexical similarity and the semantic similarity are weightedly summed to obtain the similarity between the third language and the fifth language.

5. The intelligent outbound calling method according to claim 1, characterized in that: The determining corresponding text reply information based on the semantic information includes: Determine an interaction topic and a first interaction stage of the interaction topic based on historical interaction data of the first user; Determining the interactive content corresponding to the second interactive stage of the interactive theme based on the first interactive stage of the interactive theme; the second interactive stage is the next stage of the first interactive stage; Interact with the first user based on the interactive content.

6. The intelligent outbound calling method according to claim 5, characterized in that: The determining of the interaction topic and the first interaction stage of the interaction topic based on the historical interaction data of the first user includes: Performing keyword extraction on the historical interaction data; An interaction topic and a first interaction stage of the interaction topic are determined based on the keyword.

7. The intelligent outbound calling method according to claim 5, characterized in that: The determining, based on the first interaction stage of the interaction theme, the interaction content corresponding to the second interaction stage of the interaction theme includes: Selecting an interaction template corresponding to the second interaction stage from a set template library; The interaction template is determined as the interaction content corresponding to the second interaction stage.

8. An intelligent outbound calling device, characterized in that: include: A language recognition module, used to perform language recognition on the first voice information input by the first user to obtain a first language; A language translation module, used for translating the first speech information based on a multilingual translation model to obtain semantic information; An information reply module, used to determine corresponding text reply information based on the semantic information; An information conversion module, used for converting the text reply information into a second voice information in a first language; An information sending module is used to send the second voice information to the first user.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images