Incoming call answering method

Abstract

The invention discloses an incoming call answering method which is used for solving the problem that an incoming call answering mode is not flexible and cannot meet user requirements. The scheme provided by the invention comprises the following steps: determining first text information based on voice content of a call of an initiator user; based on the first text information, determining at least one first knowledge answer matched with the first text information from a knowledge answer library; inputting the first text information and the at least one first knowledge answer into a generative language model to obtain second text information; and determining voice reply content based on the second text information, and sending the voice reply content to the initiator user.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method for answering incoming calls. Background Art

[0002] In the related art, the incoming call number can be identified, and if it is a specific type of phone number, such as intermediary, sales, etc., it can be automatically hung up; in the related art, automatic voice reply content can also be set, such as redial later, please leave a voice message in the meeting, etc. The above-mentioned call answering method of intercepting and hanging up or replying with fixed voice content is not flexible, cannot simulate the scene of real users communicating on the phone, and cannot meet user needs. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method for answering incoming calls, so as to solve the problem that the method for answering incoming calls is inflexible and cannot meet the needs of users.

[0004] In a first aspect, a method for answering an incoming call is provided, characterized in that it includes: determining a first text message based on the voice content of a current call by an initiating user; determining at least one first knowledge answer matching the first text message from a knowledge answer library based on the first text message; inputting the first text message and the at least one first knowledge answer into a generative language model to obtain a second text message; determining a voice reply content based on the second text message, and sending the voice reply content to the initiating user.

[0005] In a second aspect, an electronic device is provided, comprising: a first determination module, used to determine a first text message based on the voice content of a call by an initiating user; a second determination module, used to determine, based on the first text message, at least one first knowledge answer matching the first text message from a knowledge answer library; an input module, used to input the first text message and the at least one first knowledge answer into a generative language model to obtain second text information; and a third determination module, used to determine a voice reply content based on the second text information, and send the voice reply content to the initiating user.

[0006] According to a third aspect, an electronic device is provided. The electronic device comprises a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the method according to the first aspect are implemented.

[0007] According to a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method according to the first aspect are implemented.

[0008] In a fifth aspect, a computer program product is provided, the computer program product comprising a non-transitory computer-readable storage medium storing a computer program, the computer program being operable to cause a computer to execute part or all of the steps of the method of the first aspect.

[0009] In an embodiment of the present application, in a communication scenario, a knowledge answer library can be pre-set for specific users or specific types of user groups, and a generative language model can be pre-established. In this way, after obtaining the voice content of the initiator user's call, the first text information can be determined based on the voice content, and the appropriate knowledge answer can be retrieved from the knowledge answer library based on the first text information; then the first text information and the obtained knowledge answer are input into the generative language model to obtain appropriate text reply content; finally, voice reply content is generated based on the text reply content and sent to the initiator user, thereby simulating the user's real dialogue scenario, realizing a flexible automatic dialogue communication method, and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 It is a flow chart of a method for answering a call according to an embodiment of the present application.

[0011] Figure 2 It is a flow chart of a method for answering a call according to an embodiment of the present application.

[0012] Figure 3 It is a flow chart of a method for answering a call according to an embodiment of the present application.

[0013] Figure 4 It is a schematic diagram of the user interface of the answering party of a method for answering a call according to an embodiment of the present application.

[0014] Figure 5 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present application.

[0015] Figure 6 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application. The numbering of the drawings in this application is only used to distinguish the various steps in the scheme, and is not used to limit the execution order of the various steps. The specific execution order is subject to the description in the specification.

[0017] Figure 1 A flow chart of a method for answering a call provided in an embodiment of the present application is shown, and the method can be executed by an electronic device, such as a terminal device or a server device. In other words, the method can be executed by software or hardware installed in the terminal device or the server device. The server device, for example, includes an intelligent dialogue digital intelligence system, etc. Figure 1 As shown, the method may include the following steps.

[0018] S102: Determine first text information based on the voice content of the initiator user's current call.

[0019] exist Figure 2 In the scenario shown, the executor of this embodiment can be an intelligent dialogue digital intelligence system.

[0020] In one embodiment, in this step, when the call of the initiating user is intercepted, the first text information can be determined based on the voice content of the initiating user's current call. For the initiating user, when the call of the initiating user is intercepted, the initiating user can still speak normally. At this time, the voice content of the initiating user can also be obtained, and then the subsequent steps of this embodiment are performed. The voice content can be voice inquiry content, voice question content, etc.

[0021] In other embodiments, when the incoming call of the initiating user is not intercepted, an audio or video call with the initiating user is allowed, that is, the initiating user and the receiving user can have a real audio or video call without executing the subsequent steps of this embodiment.

[0022] Regarding the triggering conditions for the above-mentioned call interception, for example, an "automatic conversation" button is displayed on the mobile phone or other electronic devices of the answering user. When the answering user is not convenient to answer the call, the button can be clicked to intercept the call of the initiating user; or, the answering user can pre-set the interception tag of the user who needs to be intercepted (such as the customer), and when the tag corresponding to the initiating user's phone number matches the above-mentioned interception tag, the call interception of the initiating user is achieved.

[0023] This embodiment first performs an interception judgment and then searches for knowledge answers through the knowledge answer database, rather than determining whether to intercept after obtaining the search results based on the knowledge answer database.

[0024] In one embodiment, this step may convert the voice of the initiating user during the call into text to obtain the first text information, that is, directly convert the voice of the initiating user into text to obtain the first text information. The first text information obtained may include, for example: "I would like to ask why the 5G message Chatbot cannot be searched". This embodiment is simple to implement and does not require excessive resource overhead.

[0025] In another embodiment, this step may convert the voice of the initiating user in this call into text to obtain a third text message, wherein the obtained third text message includes, for example, "Is it charged?"; based on the third text message and the fourth text message of one or more previous rounds of conversations (or the previous few sentences of conversations), the first text message is obtained, wherein the fourth text message includes, for example, "What is 5G message?". This step combines the third text message and the fourth text message, and the first text message obtained includes, for example, "Is 5G message charged?". This embodiment combines the information of the previous rounds of conversations, makes full use of more information, is more in line with the needs of specific application scenarios, and the obtained first text message is more accurate, which is conducive to obtaining more accurate reply content later.

[0026] S104: Based on the first text information, determine at least one first knowledge answer matching the first text information from a knowledge answer library.

[0027] This embodiment may use the inverted index and cosine similarity algorithm of the Faiss similarity search library to perform a similarity search based on the first text information and the knowledge answer library to obtain at least one most similar first knowledge answer.

[0028] This embodiment can pre-establish a knowledge answer library, which includes a large number of knowledge answers, each of which has a label attribute, a heat value attribute, etc. Each knowledge answer Di can also be pre-vectorized using a bge vector model and saved, and the vector corresponding to each knowledge answer is recorded as DVi.

[0029] The knowledge answers here can be user-related knowledge content pre-entered by the user on the receiving end, such as knowledge answers preset by the user on the receiving end for the user or user group of the initiator to which the company customer tag belongs. These knowledge answers can be relevant question and answer content related to the company customer's consultation with the receiving end.

[0030] The above-mentioned knowledge answer library may also include some user-defined reply content, and may also include some specific knowledge answers and the like.

[0031] In one embodiment, multiple specific users (such as employees in Company A) or a specific type of user group (such as credit card business salesmen, etc.) can share and use part or all of the knowledge answers in the knowledge answer library, which is conducive to reducing the cost of establishing the knowledge answer library.

[0032] In one embodiment, this step may determine at least one knowledge answer with the highest score from the knowledge answer library; wherein the score is obtained based on at least one of the following: 1) The similarity (or relevance) between the first text information and the knowledge answer, specifically, the similarity between the vector obtained by processing the first text information and the knowledge answer vector, and the corresponding similarity scores are F1~F n Generally, the higher the similarity, the higher the score obtained.

[0033] 2) The matching degree between the label corresponding to the knowledge answer and the label corresponding to the initiating user. Generally, the higher the label matching degree, the higher the score obtained.

[0034] Each knowledge answer may correspond to multiple tags, and the initiator user also has multiple tags. This embodiment can determine the number of knowledge answers that match the initiator user's tags. The value range of this number is, for example, M1~M n The labels include, for example, company customers, company colleagues, superiors, leaders, teachers, students, family members, etc.

[0035] 3) The heat information of the knowledge answer. Usually, each knowledge answer has an attribute heat value, which can be the frequency of occurrence in the conversation records in the recent period of time, recorded as R1~R n .

[0036] In a specific embodiment, the score of each knowledge answer can be obtained based on the following formula:

[0037] This formula combines the similarity between the first text information and the knowledge answer, the label matching degree, and the popularity information of the knowledge answer, which is conducive to calculating the best score value and obtaining more accurate reply content later.

[0038] In one embodiment, before S104, the method further includes: based on the tag corresponding to the initiating user, determining at least one second knowledge answer matching the tag from the knowledge answer library; then S104 may be determining at least one first knowledge answer matching the first text information from the at least one second knowledge answer.

[0039] Since the knowledge answer library usually contains a lot of knowledge, this embodiment determines the second knowledge answer from the knowledge answer library in advance based on the label corresponding to the initiating user, and then determines at least one first knowledge answer that matches the first text information from the second knowledge answer. This not only helps to reduce the amount of data processing, but also helps to improve the accuracy of the first knowledge answer obtained by making full use of the label corresponding to the initiating user, which is conducive to obtaining more accurate reply content later.

[0040] S106: Input the first text information and the at least one first knowledge answer into a generative language model to obtain second text information.

[0041] The generative language model may be a dialogue-type generative language model, which is used to generate natural language text based on existing knowledge and simulate real dialogue scenarios. For example, key information such as subject, predicate, and object is added to the above-mentioned first knowledge answer to obtain second text information.

[0042] In one embodiment, S106 may be inputting the first text information and the at least one first knowledge answer into a generative language model to add at least one of a subject, a predicate and an object to the first knowledge answer to obtain the second text information.

[0043] In one embodiment, the input of the generative language model further includes at least one of the following: 1) Summary information of the most recent call or calls with the initiating user.

[0044] 2) A fourth text message of one or more previous rounds of conversations with the initiating user.

[0045] 3) The label corresponding to the initiating user.

[0046] This embodiment is conducive to obtaining more accurate reply content by inputting more information.

[0047] S108: Determine voice reply content based on the second text information, and send the voice reply content to the initiating user.

[0048] This embodiment can send the voice reply content to the communication layer corresponding to the initiator user, and the communication layer will specifically send the voice reply content to the electronic device of the initiator user, such as a mobile phone.

[0049] In one embodiment, multiple second text information is obtained based on the generative language model. In this step, determining the voice reply content based on the second text information includes: obtaining user input information, and the user input information is used to determine the selected second text information from multiple second text information; converting the selected second text information into voice reply content.

[0050] In this embodiment, the answering user can select the best reply content from multiple reply contents, so that the best reply content is sent to the initiating user, thereby improving the user experience.

[0051] In one embodiment, converting the selected second text information into voice reply content includes: determining acoustic features based on the selected second text information; performing parameter perturbation processing on the acoustic features to obtain perturbed acoustic features; inputting the perturbed acoustic features into a vocoder model for reverse synthesis to generate a voice waveform of the voice reply content.

[0052] This embodiment can simulate the timbre of the answering user and play the voice reply content to the initiating user, which is helpful to simulate the real audio and video communication scene and improve the user experience.

[0053] In one embodiment, the method further includes: when no suitable reply content is obtained based on the generative language model, outputting prompt information, for example, reminding the answering user to reply to the initiating user when he is free.

[0054] The call answering method provided in the embodiment of the present application can pre-set a knowledge answer library and pre-establish a generative language model for specific users or specific types of user groups in a communication scenario. In this way, after obtaining the voice content of the call of the initiating user, the first text information can be determined based on the voice content, and the appropriate knowledge answer can be retrieved from the knowledge answer library based on the first text information; the first text information and the obtained knowledge answer are then input into the generative language model to obtain appropriate text reply content; finally, voice reply content is generated based on the text reply content and sent to the initiating user, thereby simulating the user's real dialogue scenario, realizing a flexible automatic dialogue communication method, and improving the user experience.

[0055] The call answering method provided in the embodiment of the present application, during a call, for a specific initiator user or user group, based on the information of the communicating parties and the knowledge information of the knowledge answer library, calculates the relevance between the communication content and the knowledge answer information during the communication process, and after the most relevant knowledge answer content is inferred by the generative language model, uses text-to-speech technology and accesses a random algorithm to generate specific voice content for intelligent communication, which can simulate real user dialogue scenarios, realize flexible dialogue communication methods, and enhance user experience.

[0056] The embodiments of the present application have a lot of room for expansion under future intelligent dialogues, and can achieve rich application effects. For enterprises, it can replace the basic intelligent voice customer service system, and realize the intelligent dialogue method with the initiator based on the enterprise's preset data, so as to efficiently and intelligently complete the customer service business of the enterprise. For individual users, it can meet various scenarios. For example, when it is inconvenient to answer the phone, business consultations from company customers, work understanding of company colleagues, calls from family members, etc. can all be answered and communicated intelligently according to specific knowledge data, efficiently share personal busy official duties, quickly reply to urgent business, intelligently handle private or family affairs, etc. In summary, the embodiments of the present application have a wide range of applications and high commercial value.

[0057] To explain in detail the incoming call answering method provided in the embodiment of the present application, a specific embodiment will be described below.

[0058] In this embodiment, in a communication scenario, for a specific user or a specific type of user group, a call answering method for intelligent dialogue communication is implemented based on communication big data such as information of the communication initiator user, information of the answering user, system customized information, etc., combined with text relevance algorithms, generative language models, text-to-speech models, random perturbation algorithms, etc.

[0059] When communicating with the initiating user, the intelligent dialogue digital intelligence system recognizes the voice input by the sender and converts it into text, and then calculates the text relevance, label relevance, and heat value of the communication big data knowledge answer library to comprehensively calculate the most appropriate reply content. The reply content is integrated with the historical call content summary, the current call content information, and the user's current question into a prompt, and then input into the generative language model for reasoning to obtain the reply text. The text-to-speech technology is combined with the random perturbation algorithm to generate a voice reply content with a specific tone, and the voice reply content is communicated to the initiating user. For specific users or specific types of user groups, the technology of intelligent voice reply dialogue based on specific big data content is realized.

[0060] like Figure 2 As shown, Figure 2 The schematic flow of this embodiment is shown, including the following steps: Step 1: The initiator user initiates the communication. Figure 2 Phone access in.

[0061] Step 2: The communication layer gives the phone number to the number detection and interception system.

[0062] Step 3: The number detection and interception system outputs the detection and interception information back to the communication layer.

[0063] Step 4: The communication layer transfers the communication to the intelligent dialogue digital intelligence system.

[0064] Step 5: The intelligent dialogue digital intelligence system receives the communication input stream, parses it, and obtains the communication output stream content, which is then returned to the communication layer.

[0065] Step 6: The communication layer sends the content of the communication output stream to the intercepted user.

[0066] like Figure 3 As shown, Figure 3 The specific technical solution implementation instructions are shown, including the following steps: First, answer the call from the other party and identify the label corresponding to the other party's user number. Based on the label, extract the knowledge answer belonging to the label type (corresponding to the second knowledge answer in the previous text), which is recorded as the {Di} set. Each knowledge answer Di has been pre-vectorized using the bge vector model and saved, and the vector corresponding to each knowledge answer Di is recorded as DVi. The extracted knowledge answer is used to generate the content of the intelligent answer in the next step. See steps 1 to 7 below.

[0067] Step 1: The initiating user initiates a telephone call to the receiving user, and the request reaches the communication service system. The system can be the existing communication system layer of the operator or the self-built communication service system of the receiving user's mobile terminal.

[0068] Step 2: The communication system receives an incoming call, parses and identifies the number of the initiator of the incoming call, and passes the number to the call interception detection system to determine whether it needs to be intercepted.

[0069] Step 3: The call interception detection system receives the incoming number, matches the number with the number label library that has been entered into the system and the number label type that the receiving user needs to intercept, and returns the interception result and the number label information to the communication system. The label here can be company customers, company colleagues, superiors, leaders, teachers, students, family members, etc.

[0070] Step 4: The communication system determines whether the communication request needs to be intercepted based on the result of the previous step. If it needs to be intercepted, proceed to the next step.

[0071] Step 5: The communication system returns a call answering response signal to the initiating user.

[0072] Step 6: The communication system synchronizes the communication detection information to the intelligent dialogue digital intelligence system. The detection information includes the initiator's user number, user tag and other data. The tags here can be company customers, company colleagues, superiors, leaders, teachers, students, family members, etc. The user tags here are recorded as T1~Te.

[0073] Step 7: Based on the label information in the previous step, the intelligent dialogue digital intelligence system extracts the knowledge answers that belong to the label type (corresponding to the second knowledge answer in the previous article), which are recorded as the set {Di}. Each knowledge answer has label attributes, heat value attributes, etc. Each knowledge answer Di can also be pre-vectorized using the bge vector model and saved, and the vector corresponding to each knowledge answer is recorded as DVi. Here, the knowledge answer can be the user-related knowledge content pre-entered by the user on the receiving end, such as the knowledge answers preset by the user on the receiving end for the user or user group of the initiator to which the company customer label belongs. These knowledge answers can be related questions and answers related to the company customer's consultation with the receiving end.

[0074] In this embodiment, the intelligent conversation digital intelligence system can also record the summary of each call content to the database, extract the latest 3 call content summaries with the initiator user from the database, and use them as the pre-made premise for the current call. The summary content is recorded as: P1, P2, P3. The summary record of each call is generated in a specific summary prompt, the call content is integrated into the large language model, and the output content of the large model is recorded as the summary Pi of the call and saved.

[0075] The following is an example of a summary prompt, where content is the original text of the call:

[0076] Example of large model output content Pi:

[0077] After executing steps 1 to 7, the call information of the initiating user is received, and the voice is converted into text W0. The text is combined with the content of the previous round of conversations, and a specific optimized prompt is used based on the large model to supplement the appropriate subject, predicate and object to form a complete question text content (corresponding to the first text information in the previous text), improve the accuracy of knowledge matching, and record the text content as W1. W1 is input into the bge vector model, and the output is used as the converted text vector V1.

[0078] Then, the inverted index and cosine similarity algorithm of the Faiss similarity search library are used to search the similarity between V1 and the {Di} set, and the top n most similar knowledge answers are obtained. The similarity score corresponding to each knowledge answer is Fi. The number of matches between the knowledge answer and the tag of the initiating user is Mi, and the knowledge has an attribute heat value. Finally, the formula is used to calculate the Si value, that is, to calculate the score of the knowledge answer. See steps 8 to 12 below.

[0079] In one embodiment, the optimized prompt is:

[0080] For example:

[0081] The output W1 after input into the large model is:

[0082] Step 8: After the initiator's phone is connected, voice communication is initiated and the voice communication content is input into the communication system.

[0083] Step 9: The communication system synchronizes the communication content of the voice input to the intelligent dialogue digital intelligence system.

[0084] Step 10: The intelligent dialogue digital intelligence system analyzes the voice content of the initiating user, converts the voice content into text W0 through ASR speech recognition technology, and combines W0 with the previous dialogue content through a large model to complete key information such as subject, predicate and object to form a complete question. The text of the complete question is recorded as W1 (corresponding to the first text information in the previous text).

[0085] For example, W1 is:

[0086] Step 11: The intelligent dialogue digital intelligence system inputs W1 into the bge vector model and outputs it as the converted text vector V1.

[0087] Step 12: The intelligent dialogue digital intelligence system uses the inverted index of the Faiss similarity search library and the cosine similarity algorithm to perform a similarity search between V1 and the {Di} set to obtain the top n most similar knowledge answers. The scores corresponding to the n knowledge answers are obtained based on at least one of the following: 1) The similarity (or relevance) between the first text information and the knowledge answer, which can be the similarity between the vector obtained by processing the first text information and the knowledge answer vector. The corresponding similarity scores are F1~F n Generally, the higher the similarity, the higher the score obtained.

[0088] 2) The matching degree between the label corresponding to the knowledge answer and the label corresponding to the initiating user. Generally, the higher the label matching degree, the higher the score obtained.

[0089] The knowledge answer may correspond to multiple tags, and the initiator user may also have multiple tags. This embodiment can determine the number of knowledge answers that match the initiator user's tags. The value range of this number is, for example, M1~M n The labels include, for example, company customers, company colleagues, superiors, leaders, teachers, students, family members, etc.

[0090] 3) The heat information of the knowledge answer. Usually, each knowledge answer has an attribute heat value, which can be the frequency of occurrence in the conversation records in the recent period of time, recorded as R1~R n .

[0091] In a specific embodiment, the score of each knowledge answer can be obtained based on the following formula:

[0092] This formula combines the similarity between the first text information and the knowledge answer, the label matching degree, and the popularity information of the knowledge answer, which is conducive to calculating the best score value and obtaining more accurate reply content later.

[0093] Step 13: The intelligent conversation digital intelligence system integrates the summary information P1, P2, P3 of the last three calls; the conversation content C of the previous N rounds i~n (corresponding to the fourth text information in the previous text); the user's question, the first K knowledge answers Ai~Ak obtained according to the score S, and the user tags T1~Te and W1 are integrated into prompt, and the input is a GPT-like generative language model to generate a suitable reply content FA (second text information). And identify the output result FA, determine whether there is a suitable reply result, and if not, call up the system's memo to record the phone information and conversation content, and remind the user to reply when free.

[0094] The prompt integration rules are:

[0095] The following examples illustrate:

[0096] Among them, the first line below the call content 4 is the content of W1, the known conditions 1~4 are Ai~An, and prompt is a question sentence integrated based on W1 and Ai~An.

[0097]

[0098] Among them, FA is the output result of the generative language model, which can be found in Figure 4 The interface diagram shown.

[0099] Step 14: The intelligent dialogue digital intelligence system identifies the FA text content (second text message). If there is a suitable reply, it will use text-to-speech technology to convert it into a voice reply. If there is no suitable reply, the caller number, call time, and call question record will be called up through deeplink to the system's memo application and recorded in the memo record.

[0100] The text-to-speech technology here can use the Tacotron 2 model and combine it with a random perturbation algorithm to generate speech content with a specific human voice. The specific steps are as follows: Feature generation: The text to be synthesized is input into the model, and the Tacotron2 model generates corresponding acoustic features, such as spectrograms.

[0101] Parameter perturbation: Randomly perturb the speaking rate of the acoustic features generated by the Tacotron2 model by increasing or decreasing the duration of each syllable.

[0102] Parameter perturbation: Randomly perturb the pitch of the acoustic features generated by the Tacotron2 model by modifying the sampling frequency of the signal, extending or shortening the sampling interval, etc.

[0103] Parameter perturbation: Randomly perturb the volume of the acoustic features generated by the Tacotron2 model to increase or decrease the energy of the signal and change the loudness of the sound.

[0104] Parameter anti-synthesis: The perturbed acoustic features are input into the vocoder model for anti-synthesis to generate the corresponding speech waveform.

[0105] Step 15: The intelligent dialogue digital intelligence system sends the voice response content obtained in step 14 as the dialogue content to the communication system.

[0106] Step 16: The communication system transmits the voice reply content to the device of the sender user for playback communication.

[0107] The call answering method provided in the embodiment of the present application combines the voice content of the communication initiator user in the communication scenario, a multi-dimensional relevance matching algorithm based on a preset knowledge answer library, integrates historical call record summaries based on a generative language model, and combines the current call content, the user's questions and the call answering technology retrieved from the relevant knowledge base to complete the process of intelligent dialogue voice communication.

[0108] The call answering method provided in the embodiment of the present application adopts the inverted index and cosine similarity algorithm of the Faiss search library in the communication scenario, matches the specific knowledge big data of a specific user group, and combines the intelligent dialogue method of the generative AI model and the TTS model to fill the gap that the existing technology cannot have intelligent dialogue in this communication field.

[0109] The call answering method provided in the embodiment of the present application combines appropriate prompts according to specific conditions in the communication field and inputs them into a generative large language AI model, and combines the technology of synthesizing speech with a text-to-speech TTS model and a random perturbation algorithm to achieve a realistic anthropomorphic dialogue scenario. This makes up for the effect of realistic and fluent dialogue that cannot be achieved by the existing technology in the communication field.

[0110] Figure 5 A schematic structural diagram of an electronic device 500 provided in an embodiment of the present application is shown, and the electronic device 500 includes the following modules.

[0111] The first determining module 502 is used to determine the first text information based on the voice content of the initiator user's current call.

[0112] The second determination module 504 is configured to determine, based on the first text information, at least one first knowledge answer matching the first text information from a knowledge answer library.

[0113] The input module 506 is used to input the first text information and the at least one first knowledge answer into the generative language model to obtain second text information.

[0114] The third determination module 508 is used to determine the voice reply content based on the second text information, and send the voice reply content to the initiating user.

[0115] The electronic device provided by the embodiment of the present application can pre-set a knowledge answer library for a specific user or a specific type of user group in a communication scenario, and pre-establish a generative language model. In this way, after obtaining the voice content of the initiator user's current call, the first text information can be determined based on the voice content, and a suitable knowledge answer can be retrieved from the knowledge answer library based on the first text information; then the first text information and the obtained knowledge answer are input into the generative language model to obtain a suitable text reply content; finally, a voice reply content is generated based on the text reply content and sent to the initiator user, thereby simulating the user's real dialogue scenario, realizing a flexible automatic dialogue communication method, and improving the user experience. .

[0116] The electronic device 500 provided in the embodiment of the present application can execute any embodiment described in the aforementioned method embodiment and realize the functions and beneficial effects of any embodiment described in the aforementioned method embodiment, which will not be described in detail here.

[0117] Among them, the above modules in the electronic device provided by the embodiment of the present application can also implement the method steps provided by the above method embodiment. Alternatively, the electronic device provided by the embodiment of the present application can also include other modules in addition to the above modules to implement the method steps provided by the above method embodiment. And the electronic device provided by the embodiment of the present application can achieve the technical effects that can be achieved by the above method embodiment.

[0118] An embodiment of the present application also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned call answering embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0119] Figure 6 The hardware structure diagram of the electronic device provided by the embodiment of the present application is shown. Referring to the figure, at the hardware level, the electronic device includes a processor, and may also include an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (Random-Access Memory, RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage, etc. Of course, the electronic device may also include hardware required for other services.

[0120] The processor, network interface and memory can be interconnected through an internal bus, which can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0121] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0122] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a device for locating the target user at the logical level. The processor executes the program stored in the memory and is specifically used to execute: Figure 1-4 The method disclosed in the illustrated embodiment realizes the functions and beneficial effects of each method described in the foregoing method embodiments, which will not be described in detail here.

[0123] The above application Figure 1-4The method disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0124] The electronic device can also execute any embodiment described in the aforementioned method embodiment, and realize the functions and beneficial effects of any embodiment described in the aforementioned method embodiment, which will not be repeated here.

[0125] Of course, in addition to software implementation methods, the electronic device of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0126] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned call answering method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0127] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. The computer program can be operated to enable a computer to execute part or all of the steps of the above-mentioned call answering method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0128] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0132] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0133] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0134] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0135] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0137] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for answering a call, characterized in that: include: Determine the first text information based on the voice content of the initiator user's current call; Based on the first text information, determining at least one first knowledge answer matching the first text information from a knowledge answer library; Inputting the first text information and the at least one first knowledge answer into a generative language model to obtain second text information; The voice reply content is determined based on the second text information, and the voice reply content is sent to the initiating user.

2. The method according to claim 1, characterized in that Determining at least one knowledge answer matching the first text information from the knowledge answer library includes: determining at least one knowledge answer with the highest score from the knowledge answer library; wherein the score is obtained based on at least one of the following: The similarity between the first text information and the knowledge answer; The matching degree between the label corresponding to the knowledge answer and the label corresponding to the initiating user; The popularity information of the knowledge answer.

3. The method according to claim 1, characterized in that The determining of the first text information based on the voice content of the current call by the initiating user includes: Converting the voice of the initiating user during the call into text to obtain third text information; The first text information is obtained based on the third text information and fourth text information of one or more previous rounds of conversations.

4. The method according to claim 1, characterized in that: The determining of the first text information based on the voice content of the current call by the initiating user includes: In the case that the incoming call of the initiating user is intercepted, the first text information is determined based on the voice content of the current call of the initiating user.

5. The method according to claim 4, characterized in that In the case that the incoming call of the initiating user is not blocked, an audio or video call with the initiating user is allowed.

6. The method according to claim 1, characterized in that The method further comprises: Based on the tag corresponding to the initiating user, determining at least one second knowledge answer matching the tag from the knowledge answer library; Wherein, determining at least one first knowledge answer matching the first text information from the knowledge answer library includes: determining at least one first knowledge answer matching the first text information from the at least one second knowledge answer.

7. The method according to claim 1, characterized in that The input of the generative language model also includes at least one of the following: Summary information of the most recent call or calls with the initiating user; a fourth text message of one or more previous rounds of conversations with the initiating user; The label corresponding to the initiator user.

8. The method according to any one of claims 1 to 7, characterized in that: The second text information obtained based on the generative language model is a plurality of pieces, and the determining of the voice reply content based on the second text information includes: Acquire user input information, where the user input information is used to determine the selected second text information from a plurality of second text information; The selected second text message is converted into voice reply content.

9. The method according to claim 8, characterized in that The step of converting the selected second text message into voice reply content includes: determining an acoustic feature based on the selected second text information; Performing parameter perturbation processing on the acoustic feature to obtain a perturbed acoustic feature; The disturbed acoustic features are input into the vocoder model for anti-synthesis to generate a speech waveform of the speech reply content.

10. The method according to claim 1, characterized in that The step of inputting the first text information and the at least one first knowledge answer into a generative language model to obtain the second text information comprises: The first text information and the at least one first knowledge answer are input into a generative language model to add at least one of a subject, a predicate and an object to the first knowledge answer to obtain second text information.

Images