Information generation method and apparatus

Abstract

The application discloses an information generation method and device, and relates to the technical field of artificial intelligence. A specific implementation manner of the method comprises the following steps: acquiring current service information; inputting the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; and inputting the current service information and the at least one keyword in the current service information into a pre-trained corpus generation model to obtain at least one simulated response information of the current service information; wherein the simulated response information comprises at least one simulated response information corresponding to each keyword. The implementation manner can improve the generation efficiency, accuracy and diversity of user response information.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an information generation method and apparatus. Background Technology

[0002] Before making official outbound calls, the service provider's outbound calling system needs to perform preparatory work such as debugging session templates and configuring the knowledge base. Some of this work requires testing by combining the service information sent by the service provider to the user with the user's response information. For example, determining whether the connection between multiple rounds of conversations is reasonable and whether the knowledge base needs to be added, deleted, or modified requires a large amount of the above data for testing. Currently, because the amount of actual user response information generated cannot meet the testing needs, user response information is generally constructed based on the experience of staff. The drawbacks are that staff cannot reflect the perspective of the general public, the accuracy and diversity of the constructed user response information are insufficient, and the construction efficiency is low. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide an information generation method and apparatus that can improve the efficiency, accuracy and diversity of user response information generation.

[0004] To achieve the above objectives, according to one aspect of the present invention, an information generation method is provided.

[0005] The information generation method of this invention includes: acquiring current service information; inputting the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; inputting the current service information and at least one keyword in the current service information into a pre-trained corpus generation model to obtain at least one simulated response information of the current service information; wherein, the simulated response information includes: at least one simulated response information corresponding to each keyword.

[0006] Optionally, the corpus generation model is trained according to the following steps: obtaining multiple historical service information entries and at least one real response entry from any historical service information entry in the service provider's historical call logs; extracting at least one keyword from any historical service information entry, and determining the real response entry containing any keyword from that historical service information entry as the real response entry corresponding to that keyword; determining any historical service information entry, the keyword in that historical service information entry, and the real response entry corresponding to each keyword as a training sample; and training the corpus generation model using the historical service information and keywords in the training sample as training input data and the real response entry corresponding to each keyword in the training sample as labels.

[0007] Optionally, the training steps of the corpus generation model further include: after determining the real response information of the historical service information containing any keyword as the real response information corresponding to the keyword, inputting the real response information corresponding to any keyword of any historical service information into a pre-trained perplexity calculation model to obtain the perplexity score of the real response information; and removing the real response information whose perplexity score meets the preset perplexity filtering conditions.

[0008] Optionally, the training steps of the corpus generation model further include: after removing real response information whose perplexity scores meet the preset perplexity filtering conditions, for multiple real response information corresponding to the same keyword of the same historical service information, arranging the multiple real response information into a queue in ascending order of perplexity scores; sequentially selecting the current real response information in the queue from the head to the tail of the queue until the real response information in the queue has been traversed or the number of selections has reached a preset number threshold; for the selected real response information, determining the similarity between the selected real response information and any subsequent real response information, and removing the subsequent real response information whose similarity is greater than the preset similarity threshold.

[0009] Optionally, the corpus generation model is a pre-trained large language model; and, inputting the current service information and at least one keyword from the current service information into the pre-trained corpus generation model includes: inputting the current service information, at least one keyword from the current service information, and interactive instructions for prompting the generation of simulated response information into the corpus generation model.

[0010] Optionally, obtaining at least one simulated response message for the current service information includes: summarizing the simulated response messages corresponding to each keyword of the current service information into simulated response messages for the current service information.

[0011] To achieve the above objectives, according to another aspect of the present invention, an information generation apparatus is provided.

[0012] The information generation apparatus of this invention includes: an information acquisition unit for acquiring current service information; a keyword extraction unit for inputting the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; and a simulated information generation unit for inputting the current service information and at least one keyword in the current service information into a pre-trained corpus generation model to obtain at least one simulated response information of the current service information; wherein the simulated response information includes at least one simulated response information corresponding to each keyword.

[0013] Optionally, the corpus generation model is trained according to the following steps: obtaining multiple historical service information entries and at least one real response entry from any historical service information entry in the service provider's historical call logs; extracting at least one keyword from any historical service information entry, and determining the real response entry containing any keyword from that historical service information entry as the real response entry corresponding to that keyword; determining any historical service information entry, the keyword in that historical service information entry, and the real response entry corresponding to each keyword as a training sample; and training the corpus generation model using the historical service information and keywords in the training sample as training input data and the real response entry corresponding to each keyword in the training sample as labels.

[0014] Optionally, the training steps of the corpus generation model further include: after determining the real response information of the historical service information containing any keyword as the real response information corresponding to the keyword, inputting the real response information corresponding to any keyword of any historical service information into a pre-trained perplexity calculation model to obtain the perplexity score of the real response information; and removing the real response information whose perplexity score meets the preset perplexity filtering conditions.

[0015] Optionally, the training steps of the corpus generation model further include: after removing real response information whose perplexity scores meet the preset perplexity filtering conditions, for multiple real response information corresponding to the same keyword of the same historical service information, arranging the multiple real response information into a queue in ascending order of perplexity scores; sequentially selecting the current real response information in the queue from the head to the tail of the queue until the real response information in the queue has been traversed or the number of selections has reached a preset number threshold; for the selected real response information, determining the similarity between the selected real response information and any subsequent real response information, and removing the subsequent real response information whose similarity is greater than the preset similarity threshold.

[0016] Optionally, the corpus generation model is a pre-trained large language model; and the information generation unit is further used to: input the current service information, at least one keyword in the current service information, and interactive instructions for prompting the generation of simulated response information into the corpus generation model.

[0017] Optionally, the information generation unit is further configured to: summarize the simulated response information corresponding to each keyword of the current service information into simulated response information of the current service information.

[0018] To achieve the above objectives, according to another aspect of the present invention, an electronic device is provided.

[0019] An electronic device according to the present invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the information generation method provided by the present invention.

[0020] To achieve the above objectives, according to another aspect of the present invention, a computer-readable storage medium is provided.

[0021] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the information generation method provided by the present invention.

[0022] To achieve the above objectives, according to another aspect of the present invention, a computer program product is provided.

[0023] One computer program product of the present invention includes a computer program that, when executed by a processor, implements the information generation method provided by the present invention.

[0024] According to the technical solution of the present invention, the embodiments described above have the following advantages or beneficial effects: After obtaining the current service information, the server uses a keyword extraction model to extract keywords from the information and a corpus generation model to process the information and its keywords, thereby generating simulated response information corresponding to each keyword. Through the training and use of artificial intelligence models such as the keyword extraction model and the corpus generation model, automatic simulation generation of user response information is achieved. Because the artificial intelligence model can learn the conversational characteristics from a general audience's perspective during training, the generated simulated response information is highly accurate and diverse, and the generation efficiency is high, thus solving the problem of insufficient data volume encountered in existing outbound call system testing technologies.

[0025] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0026] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein: Figure 1 This is a schematic diagram of the main steps of the information generation method in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the specific execution steps of the information generation method in this embodiment of the invention; Figure 3 This is a schematic diagram of the components of the information generation device in an embodiment of the present invention; Figure 4 This is an exemplary system architecture diagram that can be applied thereto according to embodiments of the present invention; Figure 5This is a schematic diagram of the electronic device structure used to implement the information generation method in the embodiments of the present invention. Detailed Implementation

[0027] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0028] It should be noted that, unless otherwise specified, the embodiments of the present invention and the technical features thereof can be combined with each other.

[0029] Figure 1 This is a schematic diagram of the main steps of the information generation method in an embodiment of the present invention.

[0030] like Figure 1 As shown, the information generation method of this embodiment of the invention can be executed by a server (hereinafter referred to as the server) for generating simulated response information. The specific execution steps are as follows: Step S101: Obtain Current Service Information. In this step, the server obtains the current service information that the service provider has called the user. Calls made by the service provider to the user can include situations such as: the service provider has already called the user, the service provider will call the user, or the service provider plans to call the user. "Calling" refers to sending information such as text, images, voice, and video through communication devices. The service information can be any information related to the service provider's business, such as reminders, recommendations, or marketing information. The information format can be text, images, voice, or video. If the service information is in image, voice, or video format, it can be converted to text format for subsequent processing. The following explanation uses voice service information as an example.

[0031] Step S102: Input the current service information into the pre-trained keyword extraction model to obtain at least one keyword from the current service information. In this step, the server inputs the acquired current service information into the trained keyword extraction model to obtain at least one keyword from the current service information. The keyword extraction model can be a pre-trained large language model, which can improve the accuracy of keyword extraction.

[0032] Step S103: Input the current service information and at least one keyword from the current service information into a pre-trained corpus generation model to obtain at least one simulated response message for the current service information. Specifically, the simulated response information, in contrast to the user's actual response information, refers to information automatically generated by an artificial intelligence model to simulate the user's actual response information. The simulated response information for the current service information may include at least one simulated response message corresponding to each keyword, where the information corresponding to each keyword refers to information containing that keyword. The corpus generation model can be a pre-trained large language model. Through the above steps, the server can automatically and efficiently generate simulated response information corresponding to the current service information based on the current service information, its keywords, and the corpus generation model, resulting in information with high accuracy and richness.

[0033] Figure 2 This is a schematic diagram illustrating the specific execution steps of the information generation method in this embodiment of the invention. See [link / reference]. Figure 2 .

[0034] Step S201: Obtain historical service information (i.e., service information from historical periods) and actual response information (i.e., actual user response information from historical periods) from historical call logs. In this step, historical service information and corresponding actual response information are first obtained from the service provider's historical call logs, and then training samples are constructed based on the historical service information and actual response information.

[0035] Step S202: Perform keyword-based relevance filtering on the actual response information. In this step, the previously obtained actual response information is filtered for relevance, and actual response information with low relevance to historical service information is removed. Specifically, after obtaining multiple historical service information entries and at least one actual response entry for any historical service information entry from the service provider's historical call logs, a known keyword extraction tool (such as jieba) can be used to extract at least one keyword from any historical service information entry. The actual response information entry containing any keyword is identified as the actual response information corresponding to that keyword, and the actual response information entry not containing any keyword is removed.

[0036] In other words, for a given historical service information and its multiple real response messages, the keywords are first extracted from the historical service information. Then, it is determined whether each real response message contains a keyword. If a real response message contains any keyword, it indicates that the real response message is highly relevant to the historical service information, so the real response message is retained and identified as the real response message corresponding to that keyword. If a real response message does not contain any keyword, it indicates that the real response message is less relevant to the historical service information, so the real response message is removed to improve the data quality of the training samples.

[0037] Step S203: Perform perplexity filtering on the real response information. In this step, the real response information corresponding to any keyword of any historical service information can be input into a pre-trained perplexity calculation model to obtain the perplexity score of the real response information. Then, real response information whose perplexity scores meet the preset perplexity filtering conditions is removed. The perplexity score is used to represent the degree of semantic certainty. For example, it can take values ​​between [0,1], with higher semantic certainty resulting in a lower perplexity score. In one embodiment, perplexity can be manually labeled to annotate the training data to train the perplexity calculation model. For example, the information "I am eating" with a high degree of semantic certainty can be labeled with a perplexity score of 0.1, and the information "I am very tall" with ambiguous semantics can be labeled with a perplexity score of 0.8. After the perplexity calculation model is trained, the perplexity score of the input information can be calculated and output.

[0038] The above perplexity filtering conditions can be customized according to actual needs. For example, it could be to remove information with a perplexity score greater than a preset perplexity threshold, or to sort all true response information in ascending order of perplexity score (from smallest to largest), retain the first preset number of true response information, and remove the remaining true response information. Through these steps, the certainty of true response information in the training samples can be improved, thereby improving the quality of the generated data of the corpus generation model.

[0039] Step S204: Perform similarity filtering on the real response information. In this step, information with high similarity can be removed from multiple real response messages of the same historical service information, thereby improving the diversity of real response information and avoiding data duplication. In this embodiment of the invention, similarity filtering can be performed in two ways.

[0040] In the first approach, for multiple genuine response messages corresponding to the same keyword in the same historical service information, the messages are first arranged into a queue in ascending order of perplexity score (from smallest to largest). Then, genuine response messages are selected sequentially from the front to the back of the queue until all genuine response messages in the queue have been traversed or a preset selection threshold has been reached (i.e., a cutoff condition has been met). For each selected genuine response message, its similarity to any subsequent genuine response message is determined. Response messages with a similarity score greater than the preset similarity threshold are removed. It can be understood that subsequent genuine response messages have a higher perplexity score than the selected ones.

[0041] For example, for multiple genuine response messages corresponding to the same keyword in the same historical service information, the multiple genuine response messages are first arranged in ascending order of perplexity score into queues a1, a2, a3, a4, and a5. Then, the current genuine response message in the queue is selected sequentially from the head to the tail, i.e., a1 is selected first, and the similarity between a1 and any subsequent genuine response message a2, a3, a4, or a5 is calculated. If the similarity between a1 and a2 is 0.3, the similarity between a1 and a3 is 0.9, the similarity between a1 and a4 is 0.4, and the similarity between a1 and a5 is 0.1, then the subsequent message a3, which has a similarity greater than the similarity threshold (0.8), is removed, and the current messages in the queue are a1, a2, a4, and a5. Next, a2 is selected in order from the front to the back of the queue. The similarity between a2 and a4 is calculated to be 0.85, and the similarity between a2 and a5 is 0.1. Therefore, the information following a4 with a similarity greater than the similarity threshold is removed. The current information in the queue is a1, a2, and a5. Since there is no other information after a5, the traversal of the queue is complete, and the process ends.

[0042] In the second approach, the scope of the similarity filtering is expanded to include all genuine response information from the same historical service. Specifically, for multiple genuine response messages from the same historical service, they are first arranged into a queue in ascending order of perplexity score. Then, genuine response messages are selected sequentially from the front to the back of the queue until all genuine response messages have been traversed or a preset selection threshold has been reached. For each selected genuine response message, its similarity to any subsequent genuine response message is determined, and subsequent genuine response messages with similarity scores greater than the preset similarity threshold are removed. This step removes information with high perplexity scores from similar information within a certain range, thereby improving the diversity of training data, avoiding data duplication, and ensuring a certain degree of determinism in the data, which helps improve model performance.

[0043] Step S205: Form training samples to train the corpus generation model. In this step, any historical service information, the keywords in that historical service information, and the real response information corresponding to each keyword are determined as a training sample. The real response information can be the real response information obtained after processing in steps S202, S203, and S204. Thus, by matching the real response information based on keywords with historical service information, a relevance screening is formed for the real response information; by detecting and filtering the perplexity of the real response information, semantically ambiguous real response information can be removed; for multiple real response information corresponding to the same keyword in the same historical service information, the similarity between the real response information is detected in ascending order of perplexity score, and the information with relatively high perplexity scores among the real response information with high similarity can be removed, thereby ensuring the diversity of response information in the training samples.

[0044] It should be noted that steps S202, S203, and S204 are not mandatory steps. In practical applications, any one or any two of these steps can be selected for execution, or any step can be selected to directly determine any historical service information, the keywords in the historical service information, and the real response information of the historical service information as training samples to train the corpus generation model.

[0045] During model training, historical service information and keywords from the training samples can be used as training input data, and the real response information corresponding to each keyword in the training samples can be used as labels to train the corpus generation model. The trained corpus generation model can generate response information corresponding to each keyword based on the input service information and keywords. The above describes the model training process; the following describes the model usage process.

[0046] Step S206: The server obtains current service information. In this step, the server obtains current service information that the service provider has already sent to the user, will send to the user, or plans to send to the user.

[0047] Step S207: The server inputs the current service information into the keyword extraction model to obtain keywords. In this step, the server can input the current service information into a pre-trained keyword extraction model to obtain the keywords output by the keyword extraction model. The keyword extraction model can be a pre-trained large language model, which can be trained separately from the corpus generation model or jointly with the corpus generation model.

[0048] Step S208: The server obtains the trained corpus generation model. In this step, the server determines the corpus generation model that has undergone the previous training process.

[0049] Step S209: The server inputs the current service information and its keywords into the corpus generation model to obtain the simulated response information corresponding to each keyword. In this step, the server inputs the current service information, at least one keyword from the current service information, and an interactive instruction for prompting the generation of simulated response information into the corpus generation model, thereby obtaining the simulated response information corresponding to each keyword output by the corpus generation model.

[0050] Step S210: The server aggregates the simulated response information corresponding to each keyword to obtain the simulated response information for the current service information. In this step, the server can aggregate the simulated response information corresponding to each keyword of the current service information into the simulated response information for the current service information. For example, if the current service information is: "Hello, we cordially invite you to participate in Brand A's event. Tires are as low as 300 yuan per tire, and you can also get a free canvas bag and other exquisite gifts. Please come and check it out in our mall." The extracted keywords are: brand, tire, 300, canvas bag. The generated simulated response information is: Hello, what brand is this (corresponding to brand); Hello, what brand day is it? I don't have time (corresponding to brand); Hmm, I just changed my tires (corresponding to tire); Hello, I'm not buying tires for now (corresponding to tire); What product is 300 yuan (corresponding to 300); Hmm, never mind, I just bought a canvas bag (corresponding to canvas bag).

[0051] In the technical solution of this invention embodiment, after obtaining the current service information, the server uses a keyword extraction model to extract keywords from the current service information and a corpus generation model to process the current service information and its keywords, thereby generating simulated response information corresponding to each keyword. Through the training and use of artificial intelligence models such as the keyword extraction model and the corpus generation model, automatic simulation generation of user response information is achieved. Because the artificial intelligence model can learn the conversational characteristics from a general public perspective through the training process, the generated simulated response information has high accuracy and diversity, and high generation efficiency, thus solving the problem of insufficient data volume encountered in the testing of outbound call systems in the prior art.

[0052] It should be noted that the technical solutions of this invention, including the collection, updating, analysis, processing, use, transmission, and storage of user personal information, all comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security.

[0053] For the foregoing method embodiments, they are described as a series of actions for ease of description. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, and some steps may actually be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential for implementing the present invention.

[0054] To facilitate better implementation of the above-described solutions of the embodiments of the present invention, related apparatus for implementing the above-described solutions is also provided below.

[0055] Please see Figure 3 As shown, the information generation device 300 provided in this embodiment of the invention may include: an information acquisition unit 301, a keyword extraction unit 302, and a simulated information generation unit 303.

[0056] The information acquisition unit 301 is used to acquire current service information; the keyword extraction unit 302 is used to input the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; the simulated information generation unit 303 is used to input the current service information and at least one keyword in the current service information into a pre-trained corpus generation model to obtain at least one simulated response information of the current service information; wherein, the simulated response information includes: at least one simulated response information corresponding to each keyword.

[0057] In this embodiment of the invention, the corpus generation model is trained according to the following steps: obtaining multiple historical service information entries and at least one real response information entry for any historical service information entry from the service provider's historical call logs; extracting at least one keyword from any historical service information entry and determining the real response information entry containing any keyword from that historical service information entry as the real response information corresponding to that keyword; determining any historical service information entry, the keyword in that historical service information entry, and the real response information corresponding to each keyword as a training sample; and training the corpus generation model using the historical service information and keywords in the training sample as training input data and the real response information corresponding to each keyword in the training sample as labels.

[0058] As a preferred embodiment, the training steps of the corpus generation model further include: after determining the real response information of the historical service information containing any keyword as the real response information corresponding to the keyword, inputting the real response information corresponding to any keyword of any historical service information into a pre-trained perplexity calculation model to obtain the perplexity score of the real response information; and removing the real response information whose perplexity scores meet the preset perplexity filtering conditions.

[0059] Preferably, the training steps of the corpus generation model further include: after removing real response information whose perplexity scores meet the preset perplexity filtering conditions, for multiple real response information corresponding to the same keyword of the same historical service information, arranging the multiple real response information into a queue in ascending order of perplexity scores; sequentially selecting the current real response information in the queue from the head to the tail of the queue until the real response information in the queue has been traversed or the number of selections has reached a preset number threshold; for the selected real response information, determining the similarity between the selected real response information and any subsequent real response information, and removing the subsequent real response information whose similarity is greater than the preset similarity threshold.

[0060] In one embodiment, the corpus generation model is a pre-trained large language model; and the information generation unit 303 is further configured to: input current service information, at least one keyword in the current service information, and interactive instructions for prompting the generation of simulated response information into the corpus generation model.

[0061] Furthermore, in this embodiment of the invention, the information generation unit 303 is further configured to: summarize the simulated response information corresponding to each keyword of the current service information into simulated response information of the current service information.

[0062] According to the technical solution of this invention, after obtaining current service information, the server uses a keyword extraction model to extract keywords from the current service information and a corpus generation model to process the current service information and its keywords, thereby generating simulated response information corresponding to each keyword. Through the training and use of artificial intelligence models such as the keyword extraction model and the corpus generation model, automatic simulation generation of user response information is achieved. Because the artificial intelligence model can learn the conversational characteristics from a general public perspective through the training process, the generated simulated response information has high accuracy and diversity, and high generation efficiency, thus solving the problem of insufficient data volume encountered in the testing of outbound call systems in the prior art.

[0063] Figure 4 An exemplary system architecture 400 is shown that can be applied to the information generation method or information generation apparatus of the present invention.

[0064] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, and 403, network 404, and server 405 (this architecture is merely an example; the components included in a specific architecture may be adjusted according to the specific application). Network 404 serves as the medium for providing a communication link between terminal devices 401, 402, and 403 and server 405. Network 404 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0065] Users can use terminal devices 401, 402, and 403 to interact with server 405 via network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, and 403, such as simulated message generation applications (for example only).

[0066] Terminal devices 401, 402, and 403 can be various electronic devices with displays that support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0067] Server 405 can be a server that provides various services, such as a backend server that supports the simulation information generation application operated by the user using terminal devices 401, 402, and 403 (for example only). The backend server can process received simulation information generation requests and feed back the processing results (such as generated simulation response information - for example only) to terminal devices 401, 402, and 403.

[0068] It should be noted that the information generation method provided in the embodiments of the present invention is generally executed by server 405, and correspondingly, the information generation device is generally set in server 405.

[0069] It should be understood that Figure 4 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0070] The present invention also provides an electronic device. The electronic device according to an embodiment of the present invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the information generation method provided by the present invention.

[0071] The following is for reference. Figure 5 It shows a schematic diagram of the structure of a computer system 500 suitable for implementing an electronic device according to embodiments of the present invention. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0072] like Figure 5As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 502 or programs loaded from storage section 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the computer system 500. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0073] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.

[0074] In particular, according to the embodiments disclosed in this invention, the processes described in the above main step diagrams can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the main step diagrams. In the above embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit 501, it performs the functions defined in the system of this invention.

[0075] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0076] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0077] The units described in the embodiments of the present invention can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor can be described as including: an information acquisition unit, a keyword extraction unit, and a simulated information generation unit. The names of these units do not necessarily limit the specific unit; for example, the information acquisition unit can also be described as "a unit that provides current service information to the keyword extraction unit."

[0078] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, and when the device executes the one or more programs, the steps performed by the device include: acquiring current service information; inputting the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; inputting the current service information and at least one keyword in the current service information into a pre-trained corpus generation model to obtain at least one simulated response information for the current service information; wherein the simulated response information includes at least one simulated response information corresponding to each keyword.

[0079] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the information generation method provided by the present invention.

[0080] In the technical solution of this invention embodiment, after obtaining the current service information, the server uses a keyword extraction model to extract keywords from the current service information and a corpus generation model to process the current service information and its keywords, thereby generating simulated response information corresponding to each keyword. Through the training and use of artificial intelligence models such as the keyword extraction model and the corpus generation model, automatic simulation generation of user response information is achieved. Because the artificial intelligence model can learn the conversational characteristics from a general public perspective through the training process, the generated simulated response information has high accuracy and diversity, and high generation efficiency, thus solving the problem of insufficient data volume encountered in the testing of outbound call systems in the prior art.

[0081] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An information generation method, characterized in that, include: Get current service information; The current service information is input into a pre-trained keyword extraction model to obtain at least one keyword from the current service information; The current service information and at least one keyword from the current service information are input into a pre-trained corpus generation model to obtain at least one simulated response information for the current service information; wherein, the simulated response information includes at least one simulated response information corresponding to each keyword.

2. The method according to claim 1, characterized in that, The corpus generation model is trained according to the following steps: Obtain multiple historical service information entries from the service provider's historical call logs, as well as at least one genuine response entry for any historical service information entry; Extract at least one keyword from any historical service information, and determine the real response information of the historical service information that contains any keyword as the real response information corresponding to that keyword; Each historical service information, the keywords in that historical service information, and the real response information corresponding to each keyword are determined as a training sample. The corpus generation model is trained using historical service information and keywords from the training samples as training input data and real response information corresponding to each keyword in the training samples as labels.

3. The method according to claim 2, characterized in that, The training steps of the corpus generation model further include: After determining the real response information of the historical service information containing any keyword as the real response information corresponding to the keyword, the real response information corresponding to any keyword of any historical service information is input into the pre-trained perplexity calculation model to obtain the perplexity score of the real response information. Remove genuine response information that matches the preset perplexity filtering criteria based on the perplexity score.

4. The method according to claim 3, characterized in that, The training steps of the corpus generation model further include: After removing the real response information that meets the preset confusion filtering conditions, for multiple real response information corresponding to the same keyword of the same historical service information, the multiple real response information are arranged into a queue in ascending order of confusion score; The current real response information in the queue is selected sequentially from the front to the back of the queue until the real response information in the queue has been traversed or the number of selections reaches a preset threshold. For each selected real response information, the similarity between the selected real response information and any subsequent real response information is determined, and subsequent real response information with a similarity greater than a preset similarity threshold is removed.

5. The method according to claim 1, characterized in that, The corpus generation model is a pre-trained large language model; and the step of inputting the current service information and at least one keyword from the current service information into the pre-trained corpus generation model includes: The current service information, at least one keyword in the current service information, and the interactive instructions for prompting the generation of the simulated response information are input into the corpus generation model.

6. The method according to claim 1, characterized in that, The at least one simulated response message for obtaining the current service information includes: The simulated response information corresponding to each keyword of the current service information is summarized into the simulated response information of the current service information.

7. An information generation device, characterized in that, include: The information acquisition unit is used to acquire current service information; The keyword extraction unit is used to input the current service information into a pre-trained keyword extraction model to obtain at least one keyword in the current service information; The simulated information generation unit is used to input the current service information and at least one keyword from the current service information into a pre-trained corpus generation model to obtain at least one simulated response information for the current service information; wherein, the simulated response information includes at least one simulated response information corresponding to each keyword.

8. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images