Voice call abstract method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a voice call summarization method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring voice information in a call, wherein the voice information comprises voice information of a calling party and voice information of a called party; converting the calling party voice information and the called party voice information into calling party text information and called party text information respectively; integrating the calling party text information and the called party text information into call text information; and performing summary information extraction on the call text information through a summary generation module to generate call summary information. Through the method, the user can obtain the key information of the dialogue more quickly and accurately without depending on the support of terminal equipment.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a voice call summary method, device, electronic device and storage medium. Background Art

[0002] Traditional telephone conferences, telephone interviews, and telephone surveys often face a common problem: participants need to spend a lot of time and energy to listen to the complete voice conversation to obtain key information. This inefficient communication method not only wastes valuable time, but may also make information screening difficult, affecting the accuracy and efficiency of decision-making. For example, in a telephone conference, participants often need to be online throughout the whole process, and even if most of the content is irrelevant to them, they must listen patiently to avoid missing important information. Similarly, in a telephone interview, both the interviewer and the job seeker need to invest a lot of time and repeatedly confirm the details to ensure the accurate communication of the information. In a telephone survey, the investigator needs to record the interviewee's answer word by word, which is not only time-consuming and laborious, but also prone to omissions or misunderstandings. These problems have seriously affected the efficiency and quality of communication, causing enterprises and individuals to face many inconveniences in their daily work.

[0003] At present, although there are some AI-based terminal solutions that can convert voice information into text information, this solution depends on the AI ​​program embedded in the terminal. This is not feasible for traditional terminal devices. Summary of the invention

[0004] Embodiments of the present disclosure provide a voice call summary method, device, electronic device, and storage medium.

[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0006] According to one aspect of the present disclosure, a voice call summary method is provided, comprising: acquiring voice information in a call; the voice information comprising caller voice information and called party voice information; converting the caller voice information and called party voice information into caller text information and called party text information respectively; integrating the caller text information and called party text information into call text information; and extracting summary information from the call text information through a summary generation module to generate call summary information.

[0007] In an exemplary embodiment, the method further includes: performing call category identification on the call text information through a pre-trained call classification model to obtain a target call category corresponding to the call text information.

[0008] In an exemplary embodiment, the training process of the call classification model includes: obtaining a training data set; the training data set includes at least: call text samples and category annotation information; performing data preprocessing on the call text samples to obtain preprocessed text samples; performing call category identification on the preprocessed text samples through the call classification model to be trained to obtain training call categories; obtaining a loss function based on the training call categories and category annotation information; training the call classification model to be trained based on the loss function to adjust the weight parameters in the call classification model to be trained; and obtaining the call classification model in response to the call classification model to be trained satisfying a preset convergence condition.

[0009] In an exemplary embodiment, the summary information of the call text information is extracted by a summary generation module to generate the call summary information, including: inputting the target call category into the summary generation module; according to the target call category, the summary information of the call text information is extracted by the summary generation module to generate the call summary information; the call summary information is matched with the target call category.

[0010] In an exemplary embodiment, the extracting summary information of the call text information by the summary generating module according to the target call category to generate the call summary information includes: determining at least one target prompt word matching the target call category according to the target call category; extracting at least one summary information corresponding to the target prompt word from the call text information by the summary generating module according to the at least one target prompt word; and generating the call summary information according to the at least one summary information.

[0011] In an exemplary embodiment, the calling party text information and the called party text information include: at least one set of conversation text and time stamp; the conversation text and time stamp correspond one to one.

[0012] In an exemplary embodiment, integrating the calling party text information and the called party text information into the call text information includes: integrating the calling party text information and the called party text information in time sequence based on the timestamp to generate the call text information.

[0013] In an exemplary embodiment, the method further includes: sending the call summary information to a target subscribing user.

[0014] According to another aspect of the present disclosure, a voice call summary device is provided, comprising: a voice acquisition module, configured to acquire voice information in a call; the voice information includes caller voice information and called party voice information; a text conversion module, configured to convert the caller voice information and called party voice information into caller text information and called party text information respectively; a text integration module, configured to integrate the caller text information and called party text information into call text information; a summary generation module, configured to extract summary information from the call text information through the summary generation module to generate call summary information. According to another aspect of the present disclosure, an electronic device is provided, comprising: one or more processors; a storage device, configured to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the non-ground base station switching method as described in the above-mentioned embodiment.

[0015] According to another aspect of the present disclosure, an electronic device is provided, comprising: one or more processors; and a storage device configured to store one or more programs, so that when the one or more programs are executed by the one or more processors, the one or more processors implement the voice call summary method as described in the above embodiments.

[0016] According to another aspect 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 voice call summary method as described in the above embodiment is implemented.

[0017] The voice call summary method provided by the embodiment of the present disclosure converts the caller voice information and the called party voice information in the call into the caller text information and the called party text information respectively, and integrates them into the call text information; extracts the summary information from the call text information through the summary generation module to generate the call summary information. This method does not rely on the support of the terminal device, so that the user can obtain the key information of the conversation more quickly and accurately.

[0018] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0020] Figure 1 A schematic diagram of a network architecture of a communication system applicable to an embodiment of the present disclosure is shown;

[0021] Figure 2 A flow chart of a voice call summary method according to an embodiment of the present disclosure is shown;

[0022] Figure 3 A schematic diagram showing text information integration according to an embodiment of the present disclosure is shown;

[0023] Figure 4 A flowchart of another voice call summary method according to an embodiment of the present disclosure is shown;

[0024] Figure 5 A schematic diagram showing a voice call summary system framework according to an embodiment of the present disclosure is shown;

[0025] Figure 6 A flowchart showing a call classification model training process according to an embodiment of the present disclosure is shown;

[0026] Figure 7 A flowchart showing a process of generating call summary information according to an embodiment of the present disclosure is shown;

[0027] Figure 8 A schematic diagram showing call summary information according to an embodiment of the present disclosure;

[0028] Fig. 9 A schematic diagram showing the structure of a voice call summary device according to an embodiment of the present disclosure is shown;

[0029] Fig.10 A schematic diagram of the structure of an electronic device suitable for implementing an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0030] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0031] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0032] It should be noted that the ordinal numbers such as "first" and "second" mentioned in the embodiments of the present disclosure are used to distinguish multiple objects, and are not used to limit the order, timing, priority or importance of multiple objects, and the descriptions of "first" and "second" do not limit the objects to be different.

[0033] In order to solve the above problems, the embodiment of the present disclosure proposes a core network data storage method, which can be applied to various communication systems. For example, GSM (Global system for Mobile Communications, global mobile communications) system, CDMA (Code Division Multiple Access, code division multiple access) system, WCDMA (Wideband Code Division Multiple Access, wideband code division multiple access) system, LTE (Long Termevolution, long-term evolution) system, LTE frequency division duplex system, LTE time division duplex system, UMTS (Universal Mobiletelecommunication System, universal mobile telecommunication system), WIMAX (Worldwide Interoperability for Microwave Access, global interconnection microwave access) communication system, 5G (5th generation, fifth generation) system or communication systems applied to the future or other similar communication systems.

[0034] Figure 1 FIG. 1 is a schematic diagram showing a network architecture of a communication system applicable to an embodiment of the present disclosure. Figure 1As shown, the network architecture includes UE, RAN (Radio Access Network, wireless access network) equipment, AMF (Access and Mobility Management Function, access and mobility management function network element) network element, SMF (Session Management function, session management function) network element, UPF (User Plane Function, user plane function) network element, PCF (Policy Control Function, policy control function) network element, NSSF (Network Slice Selection Function, network slice selection function) network element, NRF (Network Repository Function, network warehouse function) network element, NWDAF (Network Data Analytics Function, network data analysis) network element, UDM (Unified Data Management, unified data management) network element, UDR (Unified Data Repository, unified data storage) network element, AUSF (Authentication Server Function, authentication server function) network element, NEF (Network Exposure Function, network open function) network element, AF (Application Function, application function) network element, and DN (Data Network, data network) connected to the operator network.

[0035] UE can be various electronic devices, which can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; it can also be deployed on the water surface (such as ships, etc.); it can also be deployed in the air (for example, on airplanes, balloons and satellites, etc.). UE can communicate with the core network via non-RAN devices and exchange voice and / or data with RAN devices. UE can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a mobile Internet device, a wearable device, a virtual reality terminal device, an augmented reality terminal device, a wireless terminal in industrial control, a wireless terminal in unmanned driving, a wireless terminal in telemedicine, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Optionally, the clients of the applications installed in different UEs are the same, or the clients of the same type of applications based on different operating systems. Based on different terminal platforms, the specific form of the client of the application can also be different. For example, the application client can be a mobile phone client, a computer client, etc.

[0036] RAN equipment is a device in the network used to access UE to the wireless network, and may include equipment in the access network that communicates with wireless terminals through one or more sectors on the air interface. UE can access the AMF network element through RAN equipment. Specifically, when the UE accesses the AMF network element through the RAN equipment, the UE can access the AMF network element through network-side equipment such as 5G and later versions of base stations (for example: 5G NR NB), or base stations in other communication systems (for example: eNB base stations).

[0037] The AMF network element is mainly used for UE mobility management, access authentication / authorization, and is also responsible for transmitting user policies between the UE and the PCF network element.

[0038] SMF network elements are mainly used for session management, UE Internet Protocol address allocation and management, selection of endpoints for manageable user plane functions, policy control, or charging function interfaces, and downlink data notification.

[0039] The UPF network element can be used for packet routing and forwarding, or QoS processing of user plane data, etc. User data can be accessed to the DN through this network element.

[0040] The PCF network element is a unified policy framework used to guide network behavior and provide policy rule information for control plane functional network elements (such as AMF network elements and SMF network elements).

[0041] The NSSF network element is mainly used to select appropriate network slices for UE services.

[0042] NRF network elements are mainly used to provide registration and discovery functions for network elements or services provided by network elements.

[0043] NWDAF network elements can collect data from various network functions and perform analysis and prediction.

[0044] The UDM network element is mainly used to manage the UE's contract information. For example, during the authentication process, it performs authentication vector calculation, key deduction, user identity decryption, etc. In the resynchronization process, it verifies AUTS according to the corresponding algorithm and initiates a re-authentication process.

[0045] The UDR network element is mainly used to store structured data information, including contract information, policy information, and network data or business data defined in a standard format.

[0046] The AUSF network element is mainly used to perform security authentication on terminal devices.

[0047] NEF network element is located between 5G core network and external third-party application function body, and may also have some internal application functions. It is responsible for managing the network data open to the outside world. All external applications that want to access the internal data of 5G core network must go through NEF network element. NEF network element provides corresponding security guarantee to ensure the security of external application to the network, and provides functions such as external application QoS customization capability opening, mobility status event subscription, and application function request distribution.

[0048] AF network elements are mainly used to convey the requirements of the application side to the network side, such as QoS requirements and user status event subscriptions. AF network elements can be third-party functional entities or application services deployed by operators. For application functional entities of third-party applications, they can also be authorized through NEF network elements when interacting with the core network. For example, a third-party application function directly sends a request to the NEF network element, and the NEF network element determines whether the AF network element is allowed to send the request. If the verification is successful, the request will be forwarded to the corresponding PCF network element or UDM network element.

[0049] It should be understood that the above-mentioned network elements or functions can be network elements in hardware devices, software functions running on dedicated hardware, or virtualized functions instantiated on a platform (e.g., a cloud platform).

[0050] Figure 2 A flow chart of a voice call summary method according to an embodiment of the present disclosure is shown. The voice call summary method can be applied to a cloud host, a virtual machine, an operator server or a proxy server. Figure 2 As shown, the voice call summary method may include the following steps.

[0051] In step S210, voice information in the call is obtained; the voice information includes the calling party's voice information and the called party's voice information.

[0052] In the disclosed embodiment, the voice call summary method can be applied to an operator server. When the two parties establish a call connection, the voice call summary service is started to obtain the voice information during the call. The voice information includes: the voice information of the caller and the voice information of the called party. Since the service is deployed on the operator server, the voice of the caller and the voice of the called party can be distinguished directly based on the call link. The voice stream files of the calling and called directions are obtained based on different call links.

[0053] In an exemplary embodiment, the operator server may detect whether a user has subscribed to the voice call summary service. If a user has subscribed to the voice call summary service, the voice call summary service is enabled based on the established call connection. Otherwise, the voice call summary service is not enabled. The subscribed user may be the caller, the called party, or other third party.

[0054] In step S220, the calling party voice information and the called party voice information are converted into calling party text information and called party text information respectively.

[0055] In the disclosed embodiment, the voice information is converted into text information based on the deployed speech-to-text model. As mentioned above, the voice information includes the voice information of the caller and the voice information of the called party. Therefore, the speech-to-text model also converts the voice information of the caller and the voice information of the called party respectively to generate corresponding text information of the caller and the text information of the called party. Through this conversion process, it can be ensured that the voice information can enter the subsequent processing stage in a structured form. Since the speech-to-text technology is relatively mature, the specific conversion algorithm adopted is not introduced here.

[0056] In an exemplary embodiment, if Figure 3 As shown, the caller text information and the called party text information respectively include multiple dialogue texts. The dialogue text is used to represent a speech paragraph expressed by one party during the call. The speech-to-text model also adds a corresponding timestamp to each dialogue text during the corresponding language-to-text conversion process. The timestamp is used to represent the time information of the dialogue. The time information can be the actual time or the process time during the call. Based on this, the caller text information and the called party text information include: at least one set of dialogue text and timestamp. The dialogue text and timestamp correspond one to one.

[0057] like Figure 3 As shown in the figure, the content of the caller's text message is: Hello, Mr. Li, I am Ms. Zhang from a market research company. We are conducting a study on consumer behavior. Do you have time to answer a few questions now? Thank you very much. First of all, how much time do you spend shopping online each week? I see. What types of products do you usually buy?

[0058] like Figure 3 As shown, the content of the text message of the called party is: Hello, Ms. Zhang, yes, I am very happy to do so. I think I will spend about two to three hours a week. Mainly daily necessities, such as food and cleaning supplies. I will also buy some electronic products or books.

[0059] In step S230, the calling party text information and the called party text information are integrated into a call text message.

[0060] In the disclosed embodiment, the calling party text information and the called party text information are text-integrated to form call text information based on the conversation between the calling party and the called party.

[0061] In an exemplary embodiment, if Figure 3 As shown, based on the timestamps of the respective conversation texts in the calling party text information and the called party text information, the calling party text information and the called party text information are integrated in chronological order to generate the conversation text information.

[0062] In an exemplary embodiment, if Figure 3 As shown, according to whether the conversation text comes from the caller text information or the called party text information, the speaker's identity information can be added to the call text information. The identity information is used to identify the speaker's identity. For example, "A" and "B"; "caller" and "called party".

[0063] like Figure 3 As shown, the content of the call text message:

[0064] A: Hello, Mr. Li, I am Miss Zhang from a market research company. We are conducting a study on consumer behavior. Do you have time to answer a few questions now?

[0065] B: Hello, Miss Zhang. Yes, I'd love to.

[0066] A: Thank you very much. First of all, how much time do you spend on online shopping every week?

[0067] B: I think about two to three hours a week.

[0068] A: I see. What types of products do you usually buy?

[0069] B: Mainly daily necessities, such as food, cleaning supplies, etc. I also buy some electronic products or books.

[0070] It can be seen that based on the timestamps corresponding to the various conversation texts, the caller text information and the called party text information can be integrated into logical conversation text information.

[0071] In an exemplary embodiment, the caller text information and the called party text information may be preprocessed before the text integration to organize and standardize the related text information. The preprocessing may include illegal character processing. For example, character encoding unification: converting the text into a unified encoding format, such as UTF-8; special character replacement: replacing multiple consecutive spaces with a single space. Replace or delete unprintable characters, such as bell (\a), backspace (\b), formfeed (\f), etc.; illegal string cleaning: removing strings that may cause processing errors, such as HTML tags, XML tags, JavaScript codes, etc.; emoticon and other symbol processing: as needed, you can choose to retain, delete or replace emoticons and other special symbols. Through the above preprocessing, the text data can be cleaned and standardized to provide clean and orderly text information input for subsequent processing processes.

[0072] In step S240, a summary generation module extracts summary information from the call text information to generate call summary information.

[0073] In the disclosed embodiment, the summary information of the call text information is extracted by a pre-trained summary generation module. Since the call text information has been integrated into text information with dialogue logic, the summary generation module constructed based on the large language model can extract the target summary information through the semantic relationship between the dialogue texts, and then generate the call summary information.

[0074] In an exemplary embodiment, as described above, the voice call summary method can be applied to an operator server and trigger the voice call summary service based on user subscription. Based on this, after the call summary information is generated, the call summary information can be sent to the target subscribed user to send the summary information of the call to the target subscribed user. Since the voice call summary method is deployed on the operator side, it does not rely on the terminal to have relevant AI analysis capabilities, and can directly provide users with relevant call summary information based on the relevant services subscribed by the user.

[0075] The voice call summary method provided by the embodiment of the present disclosure converts the caller voice information and the called party voice information in the call into the caller text information and the called party text information respectively, and integrates them into the call text information; extracts the summary information from the call text information through the summary generation module to generate the call summary information. This method does not rely on the support of the terminal device, so that the user can obtain the key information of the conversation more quickly and accurately.

[0076] Figure 4 A flowchart of another voice call summary method according to an embodiment of the present disclosure is shown. Figure 5A schematic diagram showing a voice call summary system framework according to an embodiment of the present disclosure is shown.

[0077] In the disclosed embodiment, Figure 4 Steps S410, S420, S430 and S450 in the voice call summary method are respectively Figure 2 Steps S210, S220, S230 and S240 in the voice call summary method are corresponding and will not be repeated here.

[0078] In the embodiment of the present disclosure, Figure 2 Based on the voice call summary method shown, Figure 4 The voice call summary method may further include the following steps.

[0079] In step S440, the call category of the call text information is identified by using a pre-trained call classification model to obtain a target call category corresponding to the call text information.

[0080] In the disclosed embodiment, the call summary content required in different application scenarios is also different. Therefore, in order to more accurately extract effective summary information from the call text information, the call category of the call text information is also identified through a pre-trained call classification model to obtain a target call category corresponding to the call text information. The call category is a number of pre-set call categories. For example, telephone conference, telephone interview, questionnaire, small talk, etc. The call classification model adopts a classification model. The call classification model is used to determine the probability value between the call text information and each call category based on semantic analysis of the call text information. According to the probability value, the call category matched by the call text information is determined.

[0081] In the disclosed embodiment, the target call category determined by the call classification model can be used to adjust the summary generation module so that the summary generation module can extract summary information suitable for the application scenario from the call text information for the call category scenario.

[0082] Figure 6 FIG. 4 is a flowchart showing the call classification model training process of an embodiment of the present disclosure. Figure 6 As shown, the training process of the call classification model may include the following steps.

[0083] In step S610, a training data set is obtained; the training data set at least includes: call text samples and category labeling information.

[0084] In the disclosed embodiment, relevant training data sets are collected from data sources (such as internal enterprise databases, online forums, conversation records, etc.). The training data set includes a large amount of sample data. Each sample data includes at least: a call text sample and category annotation information. The call text sample is the call text information of the sample data. The category annotation information is used to identify the call category corresponding to the sample data. The call category is a number of pre-set call categories. For example, telephone conference, telephone interview, questionnaire, small talk, etc.

[0085] In step S620, data preprocessing is performed on the call text sample to obtain a preprocessed text sample.

[0086] In the embodiment of the present disclosure, data preprocessing is performed on the call text sample, and the data preprocessing is used to organize and standardize the relevant call text samples.

[0087] In an exemplary embodiment, the data preprocessing may convert the text into lowercase, remove punctuation marks, special characters, etc.

[0088] In an exemplary embodiment, the data preprocessing may perform word segmentation, stem extraction, and stop word removal on the data.

[0089] In an exemplary embodiment, the data preprocessing may be divided. For example, the data may be divided into a training set, a validation set, and a test set. For example, 70% of the data may be divided into a training set, 15% of the data may be divided into a validation set, and 15% of the data may be divided into a test set.

[0090] In step S630, the call classification model to be trained is used to perform call classification on the pre-processed text sample to obtain a training call category.

[0091] In the disclosed embodiment, a call classification model to be trained is constructed. The call classification model to be trained can adopt a classification model architecture constructed by transfer learning. The call classification model to be trained can be used to identify the call category of the preprocessed text sample to obtain the training call category. The processing process is similar to the aforementioned step S440 and will not be repeated here.

[0092] In an exemplary embodiment, the output layer of the call classification model to be trained can be replaced with a fully connected layer suitable for text classification tasks. The number of output nodes of the fully connected layer is set to 4, corresponding to four categories: call conference, telephone interview, questionnaire, and small talk.

[0093] In step S640, a loss function is obtained according to the training call category and the category labeling information.

[0094] In the disclosed embodiment, a corresponding loss function is obtained based on the training call category obtained by the call classification model to be trained and the category labeling information corresponding to the call text sample. The loss function can be constructed based on the cross entropy loss function.

[0095] In step S650, the call classification model to be trained is trained according to the loss function to adjust the weight parameters in the call classification model to be trained.

[0096] In the disclosed embodiment, the call classification model to be trained is trained according to the loss function to adjust the weight parameters in the call classification model to be trained, so that the classification result of the call classification model to be trained is close to the category labeling information.

[0097] In an exemplary embodiment, during the training process, the optimizer Adam and the learning rate 0.00001 are used. The training batch size is 16, which can be adjusted according to the GPU resources and data size. The number of training cycles is determined based on the performance on the validation set, usually starting from 10 epochs until the performance on the validation set stops improving or starts to decrease (overfitting). The model is trained using the training set and validated with the validation set after each epoch.

[0098] In step S660, in response to the call classification model to be trained satisfying a preset convergence condition, the call classification model is obtained.

[0099] In the disclosed embodiment, in response to the call classification model to be trained satisfying a preset convergence condition, a trained call classification model is obtained.

[0100] The voice call summary method provided by the embodiment of the present disclosure can adjust the aforementioned summary generation module by determining the target call category through the call classification model, so that the summary generation module can extract summary information suitable for the application scenario from the call text information for the call category scenario.

[0101] Figure 7 FIG. 1 is a flowchart showing a process of generating call summary information in an embodiment of the present disclosure. Figure 4 Based on the voice call summary method shown, Figure 7 The voice call summary method may further include the following steps.

[0102] In step S710, the target call category is input into the summary generation module.

[0103] In the disclosed embodiment, the target call category obtained by the call classification model according to the call text information is input into the summary generation module, so that the summary generation module can extract the summary according to the target call category.

[0104] In step S720, according to the target call category, the summary generation module extracts summary information from the call text information to generate the call summary information; the call summary information matches the target call category.

[0105] In the disclosed embodiment, according to the target call category, the summary information of the call text information is extracted by the summary generation module, so that the summary generation module can extract the summary information suitable for the application scenario from the call text information for the call category scenario. The summary generation module generates the call summary information according to the extracted summary information. The call summary information matches the target call category.

[0106] In an exemplary embodiment, a plurality of target prompt words matching the call categories are pre-set. The summary generation module can extract corresponding summary information from the call text information according to the plurality of target prompt words. Specifically, the following steps may be included.

[0107] According to the target call category, determining at least one target prompt word matching the target call category;

[0108] According to the at least one target prompt word, extracting at least one summary information corresponding to the target prompt word from the call text information by the summary generation module;

[0109] The call summary information is generated according to the at least one summary information.

[0110] In the disclosed embodiment, a plurality of target prompt words that match the target call category are determined. Based on the plurality of target prompt words, corresponding summary generation prompt questions can be generated. The summary generation prompt question is proposed through a large language model, so as to extract at least one summary information corresponding to the target prompt word from the call text information. A summary information template that matches the call category is pre-set in advance. The extracted plurality of summary information is filled into the template according to the format of the summary information template, so as to generate the call summary information.

[0111] like Figure 8 As shown, for Figure 3The call text information shown is determined by the call classification model to be a questionnaire type of application scenario. According to the pre-set prompt words that match the questionnaire type, including: questionnaire topic, object, questionnaire questions and answers. Based on the relevant prompt words, a summary generation prompt question is generated: Please summarize the conversation content based on this conversation and output a questionnaire template. The template includes the questionnaire topic, object, questionnaire questions and answers. By inputting the summary generation prompt question into the summary generation module, the summary generation module extracts summary information related to "questionnaire topic, object, questionnaire questions and answers" from the call text information. The extracted relevant summary information is organized in the form of a preset template to generate the following. Figure 8 The call summary information is shown.

[0112] The voice call summary method provided by the embodiment of the present disclosure determines a plurality of target prompt words that match the target call category, thereby extracting at least one summary information corresponding to the target prompt word from the call text information, and then filling the extracted plurality of summary information into the template according to the format of the summary information template, thereby generating the call summary information. By setting the prompt words corresponding to the call category in advance, the summary information can be extracted more accurately, thereby generating the call summary information with a standardized format.

[0113] Based on the same inventive concept, the embodiment of the present disclosure provides a voice call summary device, as described in the following embodiment. Since the principle of solving the problem in the device embodiment is similar to that in the above method embodiment, the implementation of the device embodiment can refer to the implementation of the above method embodiment, and the repeated parts will not be repeated.

[0114] Fig. 9 FIG. 1 is a schematic diagram showing the structure of a voice call summary device according to an embodiment of the present disclosure. Fig. 9 As shown, the voice call summary device 900 may include: a voice acquisition module 910 , a text conversion module 920 , a text integration module 930 and a summary generation module 940 .

[0115] The voice acquisition module 910 is configured to acquire voice information in a call; the voice information includes the voice information of the calling party and the voice information of the called party.

[0116] The text conversion module 920 is configured to convert the calling party voice information and the called party voice information into calling party text information and the called party text information respectively.

[0117] The text integration module 930 is configured to integrate the calling party text information and the called party text information into a call text message.

[0118] The summary generation module 940 is configured to extract summary information from the call text information through a summary generation module to generate call summary information.

[0119] In an exemplary embodiment, a call classification module is further included; the call classification module is configured to identify the call category of the call text information through a pre-trained call classification model to obtain a target call category corresponding to the call text information.

[0120] In an exemplary embodiment, the training process of the call classification model includes: obtaining a training data set; the training data set includes at least: call text samples and category annotation information; performing data preprocessing on the call text samples to obtain preprocessed text samples; performing call category identification on the preprocessed text samples through the call classification model to be trained to obtain training call categories; obtaining a loss function based on the training call categories and category annotation information; training the call classification model to be trained based on the loss function to adjust the weight parameters in the call classification model to be trained; and obtaining the call classification model in response to the call classification model to be trained satisfying a preset convergence condition.

[0121] In an exemplary embodiment, the summary generation module 940 is further configured to input the target call category into the summary generation module; based on the target call category, the summary generation module extracts summary information from the call text information to generate the call summary information; and the call summary information matches the target call category.

[0122] In an exemplary embodiment, the summary generation module 940 is further configured to determine, based on the target call category, at least one target prompt word that matches the target call category; based on the at least one target prompt word, extract at least one summary information corresponding to the target prompt word from the call text information through the summary generation module; and generate the call summary information based on the at least one summary information.

[0123] In an exemplary embodiment, the calling party text information and the called party text information include: at least one set of conversation text and time stamp; the conversation text and time stamp correspond one to one.

[0124] In an exemplary embodiment, the text integration module 930 is further configured to integrate the calling party text information and the called party text information in chronological order based on the timestamp to generate the call text information.

[0125] In an exemplary embodiment, a summary sending module is also included; the summary sending module is configured to send the call summary information to a target subscribing user.

[0126] Fig.10 The structure diagram of the electronic device suitable for implementing the exemplary embodiment of the present disclosure is shown. Fig.10 The electronic device 1000 according to this embodiment of the present invention will be described. Fig.10 The electronic device 1000 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0127] like Fig.10 As shown, the electronic device 1000 is in the form of a general computing device. The components of the electronic device 1000 may include but are not limited to: the at least one processing unit 1010, the at least one storage unit 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.

[0128] The storage unit stores program codes, which can be executed by the processing unit 1010, so that the processing unit 1010 executes the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0129] The storage unit 1020 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 10201 and / or a cache storage unit 10202 , and may further include a read-only storage unit (ROM) 10203 .

[0130] The storage unit 1020 may also include a program / utility 10204 having a set (at least one) of program modules 10205, such program modules 10205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0131] Bus 1030 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0132] The electronic device 1000 may also communicate with one or more external devices 1070 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1000, and / or communicate with any device that enables the electronic device 1000 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1050. Furthermore, the electronic device 1000 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the electronic device 1000 via a bus 1030. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0133] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present invention can also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.

[0134] The program product for implementing the above method according to the embodiment of the present invention may adopt a portable compact disk read-only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus, or device.

[0135] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0136] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0137] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0138] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0139] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0140] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0141] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0142] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

[0143] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A voice call summary method, characterized in that: The method comprises: Acquire voice information during a call; the voice information includes the calling party's voice information and the called party's voice information; Converting the calling party voice information and the called party voice information into calling party text information and called party text information respectively; Integrating the calling party text information and the called party text information into a call text message; The summary information is extracted from the call text information by a summary generation module to generate call summary information.

2. The method according to claim 1, characterized in that The method further comprises: The call category of the call text information is identified by using a pre-trained call classification model to obtain a target call category corresponding to the call text information.

3. The method according to claim 2, characterized in that The training process of the call classification model includes: Obtaining a training data set; the training data set at least includes: call text samples and category annotation information; Performing data preprocessing on the call text sample to obtain a preprocessed text sample; Performing call category identification on the preprocessed text sample using a call classification model to be trained to obtain a training call category; Obtaining a loss function according to the training call category and the category labeling information; Training the call classification model to be trained according to the loss function to adjust weight parameters in the call classification model to be trained; In response to the call classification model to be trained satisfying a preset convergence condition, the call classification model is obtained.

4. The method according to claim 2, characterized in that: The extracting summary information from the call text information by the summary generating module to generate the call summary information includes: inputting the target call category into the summary generation module; According to the target call category, the summary information of the call text information is extracted by the summary generation module to generate the call summary information; the call summary information matches the target call category.

5. The method according to claim 4, characterized in that The extracting summary information from the call text information by the summary generating module according to the target call category to generate the call summary information includes: According to the target call category, determining at least one target prompt word matching the target call category; According to the at least one target prompt word, extracting at least one summary information corresponding to the target prompt word from the call text information by the summary generation module; The call summary information is generated according to the at least one summary information.

6. The method according to claim 1, characterized in that The calling party text information and the called party text information include: at least one set of conversation text and time stamp; the conversation text and time stamp correspond one to one.

7. The method according to claim 6, characterized in that The step of integrating the calling party text information and the called party text information into the call text information includes: Based on the timestamp, the calling party text information and the called party text information are integrated in time sequence to generate the call text information.

8. The method according to claim 1, characterized in that The method further comprises: The call summary information is sent to the target subscribed user.

9. A voice call summary device, characterized in that: include: A voice acquisition module is configured to acquire voice information in a call; the voice information includes the calling party's voice information and the called party's voice information; A text conversion module, configured to convert the calling party voice information and the called party voice information into calling party text information and called party text information respectively; A text integration module, configured to integrate the calling party text information and the called party text information into a call text message; The summary generation module is configured to extract summary information from the call text information through the summary generation module to generate call summary information.

10. An electronic device, characterized in that: include: one or more processors; A storage device configured to store one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images