Method for generating content based on agent, related device and computer program product

Abstract

The invention provides an agent-based content generation method, a related device and a computer program product, and relates to the technical field of artificial intelligence such as agents, content generation and content verification. A specific embodiment of the method comprises the following steps: identifying to-be-verified content from generation content generated by a generation agent, wherein the to-be-verified content is generated by the generation agent based on reference knowledge; generating a difference degree between the to-be-verified content and the reference knowledge; generating a first modification instruction for the target to-be-verified content in response to the fact that the target difference degree corresponding to the target to-be-verified content is greater than or equal to a difference degree threshold value; and based on the first modification indication and the reference knowledge, controlling the generation agent to refer to the reference knowledge, updating the to-be-verified content in the generated content, and generating an updated generated content. Therefore, the situation that wrong content is provided in the generated content due to the reasons that the intelligent agent understands the reference content mistakenly, model illusion and the like can be reduced and avoided, and the content generation quality is improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, specifically to artificial intelligence technologies such as intelligent agents, content generation, and content verification, and particularly to methods, apparatuses, electronic devices, computer-readable storage media, and computer program products based on intelligent agents for content generation. Background Technology

[0002] With the rapid development of artificial intelligence technology, content generation systems based on Large Language Models (LLMs) and Intelligent Agents have been widely used in various fields. For example, based on user needs and knowledge, content generation systems can efficiently and conveniently provide users with content generation services, generating text, images, videos, etc., that users expect and require.

[0003] Therefore, under such circumstances, how to further improve the quality of generated content is a matter of concern and urgent need. Summary of the Invention

[0004] This disclosure presents a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating content based on intelligent agents.

[0005] In a first aspect, embodiments of this disclosure propose a method for generating content based on an intelligent agent, comprising: identifying content to be verified from generated content generated by the generating intelligent agent, wherein the content to be verified is generated by the generating intelligent agent based on reference knowledge; generating a difference degree between the content to be verified and the reference knowledge, wherein the reference knowledge is used by the generating intelligent agent to generate the content to be verified; generating a first modification instruction for the target content to be verified in response to a target difference degree corresponding to the target content to be verified being greater than or equal to a difference degree threshold; and controlling the generating intelligent agent to update the content to be verified in the generated content by referring to the reference knowledge based on the first modification instruction and the reference knowledge, thereby generating updated generated content.

[0006] Secondly, embodiments of this disclosure propose an apparatus for content generated by an intelligent agent, comprising: a content to be verified identification unit configured to identify content to be verified from generated content generated by a generating intelligent agent, wherein the content to be verified is generated by the generating intelligent agent based on reference knowledge; a difference degree generation unit configured to generate a difference degree between the content to be verified and the reference knowledge; a first indication generation unit configured to generate a first modification indication for the target content to be verified in response to a target difference degree corresponding to the target content to be verified being greater than or equal to a difference degree threshold; and a first content update unit configured to control the generating intelligent agent to update the content to be verified in the generated content based on the first modification indication and the reference knowledge, thereby generating updated generated content.

[0007] Thirdly, embodiments of this disclosure provide an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement the agent-based content generation method as described in any implementation of the first aspect.

[0008] Fourthly, embodiments of this disclosure provide a non-transitory computer-readable storage medium storing computer instructions that enable a computer, when executed, to implement the agent-based content generation method as described in any implementation of the first aspect.

[0009] Fifthly, embodiments of this disclosure provide a computer program product including a computer program that, when executed by a processor, can implement the agent-based content generation method as described in any implementation of the first aspect.

[0010] The method, apparatus, electronic device, computer-readable storage medium, and computer program product based on agent-generated content provided in this disclosure first identify content to be verified from the generated content generated by the agent, wherein the content to be verified is generated by the agent based on reference knowledge; then, a difference degree between the content to be verified and the reference knowledge is generated; next, in response to the target difference degree corresponding to the target content to be verified being greater than or equal to a difference degree threshold, a first modification instruction for the target content to be verified is generated; finally, based on the first modification instruction and the reference knowledge, the agent is controlled to update the content to be verified in the generated content by referring to the reference knowledge, thereby generating updated generated content.

[0011] This disclosure enables the generation of content by utilizing the reference knowledge referenced by the agent during content generation, thereby reducing or avoiding the provision of erroneous content in the generated content due to the agent's misunderstanding of the reference content or model illusion, and improving the quality of content generation.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0013] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is an exemplary system architecture to which this disclosure can be applied; Figure 2 A flowchart illustrating a process for generating content based on an intelligent agent, as provided in this embodiment of the disclosure; Figure 3 A flowchart illustrating a process for generating the degree of difference between the content to be verified and reference knowledge, provided in an embodiment of this disclosure; Figure 4 A flowchart illustrating a process for generating content based on an intelligent agent, specifically implemented in an application scenario, as provided in an embodiment of this disclosure. Figure 5 A structural block diagram of an apparatus for generating content based on an intelligent agent, provided in an embodiment of this disclosure; Figure 6 This is a schematic diagram of the structure of an electronic device suitable for performing a method for generating content based on an intelligent agent, as provided in an embodiment of this disclosure. Detailed Implementation

[0014] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding; these 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 this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0015] Furthermore, the acquisition, storage, use, processing, transportation, provision, and disclosure of user personal information (such as user personal information that may be stored in trusted knowledge bases and trusted online knowledge sources mentioned later in this disclosure) in the technical solutions disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0016] Figure 1 An exemplary system architecture 100 is shown, to which embodiments of the agent-based content generation methods, apparatuses, electronic devices, and computer-readable storage media of this disclosure can be applied.

[0017] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0018] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various applications for enabling information communication between the terminal devices 101, 102, and 103 and server 105 can be installed. These applications include content generation applications, content verification applications, and instant messaging applications.

[0019] Terminal devices 101, 102, and 103 and server 105 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, laptops, and desktop computers. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices, and can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here. When server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here.

[0020] Server 105 can provide various services through its built-in applications. Taking a content generation application that provides content generation services based on user needs as an example, when running this application, server 105 can achieve the following: First, it obtains the user's described generation requirements (e.g., generation requirements described in natural language) from terminal devices 101, 102, and 103 via network 104; then, server 105 calls a generation agent (e.g., utilizing a generation agent already deployed on server 105) to process the generation requirements and generate the corresponding content; then, server 105 uses the content generated by the generation agent... The system identifies content to be verified within the content, which is generated by a generative agent based on reference knowledge. Then, it generates a difference degree between the content to be verified and the reference knowledge. Next, in response to a target difference degree greater than or equal to a difference degree threshold corresponding to the target content to be verified, the server 105 generates a first modification instruction for the target content to be verified. Then, based on the first modification instruction and the reference knowledge, the server 105 controls the generative agent to update the content to be verified in the generated content, generating updated content. Finally, the server 105 can return this updated content to terminal devices 101, 102, and 103 via network 104.

[0021] It should be noted that the aforementioned generation requirements can be obtained from terminal devices 101, 102, and 103 via network 104, or they can be pre-stored locally on server 105 through various means. Therefore, when server 105 detects that this data is already stored locally (for example, when starting to process previously retained content generation tasks), it can choose to directly obtain this data from locally. In this case, the exemplary system architecture 100 may also exclude terminal devices 101, 102, and 103 and network 104.

[0022] Since deploying intelligent agents often requires significant computing resources and power, the methods for generating content based on intelligent agents provided in the subsequent embodiments of this disclosure are generally executed by a server 105 with strong computing power and abundant computing resources. Correspondingly, the device for generating content based on intelligent agents is also generally located in the server 105. However, it should also be noted that when terminal devices 101, 102, and 103 also possess sufficient computing power and resources, they can also perform the aforementioned calculations performed by the server 105 through content generation applications installed on them, thereby outputting the same results as the server 105. Especially when multiple terminal devices with different computing capabilities exist simultaneously, but the content generation application determines that the terminal device it is on has strong computing power and abundant remaining computing resources, it can allow the terminal device to perform the aforementioned calculations, thereby appropriately reducing the computing pressure on the server 105. Accordingly, the device for generating content based on intelligent agents can also be located in terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may also exclude the server 105 and the network 104.

[0023] It should be understood that Figure 1 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.

[0024] Please refer to Figure 2 , Figure 2 A flowchart of a process for generating content based on an intelligent agent, provided for an embodiment of this disclosure, includes process 200.

[0025] Process 200 specifically includes the following steps: Step 201: Identify the content to be verified from the generated content generated by the generating agent; In embodiments of this disclosure, this step is intended to be performed by the agent executing the method for generating content based on an agent (e.g., Figure 1 The server 105 shown identifies the content to be verified from the generated content generated by the generating agent.

[0026] As discussed above, the generating agent can be deployed on server 105 or on other devices independent of server 105. For ease of understanding, we will use the case where the generating agent is deployed on server 105, i.e., the execution subject described above, as an example.

[0027] In such a case, after receiving a generation request from the user (e.g., the content to be generated expressed in natural language), the executing agent can call the generating agent to process the generation request and generate the content.

[0028] It should be understood that if the generating agent is deployed on a device other than server 105, the generated content can also be generated by the generating agent in a manner similar to server 105, which will not be repeated here.

[0029] An intelligent agent that can act as a generative agent refers to a system or entity capable of perceiving its environment and autonomously making decisions and executing tasks. It can make judgments and choices based on its own goals and changes in the external environment, and take actions to achieve a specific purpose. For example, this intelligent agent can be an aggregation of one or more different models, trained using training samples corresponding to the intended use. For instance, for the aforementioned generative purpose, the generative agent can be trained using sample generation requirements and corresponding sample generation content, enabling it to, for example, process the input generation requirements, determine the semantics of the generation requirements, and utilize reference knowledge associated with those semantics to generate content that meets those requirements by summarizing the reference knowledge. For example, after determining keywords and sentences from reference knowledge, the generative agent can generate the final content based on the combination and rewriting of these keywords and sentences.

[0030] After the generative agent generates the corresponding generated content, the executing agent can obtain this generated content and identify the content to be verified generated by the generative agent based on reference knowledge (e.g., expert knowledge that can provide reference for the generative agent). For example, the executing agent can call the generative agent's generation log and generation context information, and use the generation log and generation context information to determine which content was generated by the generative agent using or leveraging reference knowledge when generating the generated content. This content can then be used as the content to be verified to check whether the generative agent made any misunderstandings or model illusions when generating this content.

[0031] In some embodiments, the executing entity can actually be configured as an intelligent agent. For example, the executing entity can be a verification intelligent agent used to verify the generated content. Furthermore, in such cases, two intelligent agents can be combined to obtain a complete integrated intelligent agent that possesses both generation and verification capabilities.

[0032] In some embodiments, the verification agent, acting as the execution subject, may have a larger scale and higher level of training than the generative agent described above. This allows for a reduction in the training and configuration costs of the generative agent by incorporating or combining verification agents, enabling the generative agent to function as a general-purpose agent for a wider range of generation tasks. For example, the generative agent can allocate more resources to understanding the generation requirements, while adjusting the accuracy of the content based on the verification and guidance results from the verification agent. Alternatively, incorporating such a verification agent can improve the capabilities of existing, already trained generative agents.

[0033] For example, if the executing agent determines, based on the generation log, that the explanation “XXX” about concept A generated by the generating agent (e.g., “concept A can be understood as XXX” recorded in the generated content) was generated using “A is XXX” in reference knowledge A, then the executing agent can identify “concept A can be understood as XXX” as the content to be verified.

[0034] Furthermore, the sources of reference knowledge can be the local memory units of the generating agent, the knowledge base configured in association with the generating agent, or online knowledge sources that can be called and used by the generating agent, etc.

[0035] In some embodiments, considering that the generating agent often uses reference knowledge to generate some explanatory and factual content (or provides conclusive content), and that this content actually has a significant impact on the user's reading and generation quality, in order to more comprehensively and accurately identify these contents to be verified and balance recognition efficiency, the executing agent can choose to directly search for and locate the contents to be verified based on the nature of the content, so as to ensure and improve the reading and generation quality without excessively affecting the efficiency of the final output of the generated content.

[0036] For example, the executing entity can first identify the content type of each part of the generated content (for example, identify the content type of each unit based on sentences).

[0037] Then, the implementing entity can extract the content that is a factual statement.

[0038] For example, the executing entity can analyze and determine whether a sentence is a declarative sentence based on its sentence structure and pattern. If it is, the executing entity can extract the content type of "factual statement" from the "declarative sentence".

[0039] Next, the executing agent can identify whether these factual statements were generated by the generative agent based on reference knowledge. For example, by calling the generation log, it can determine the generation process of the "statement" and whether reference knowledge was used in it.

[0040] In some embodiments, after determining the "statement," the executing agent can also use prompts to inquire with the generating agent about the basis for generating the "statement" and whether any reference knowledge was consulted. Thus, in this way, the executing agent can determine the content to be verified without needing to access the generating agent's historical actions or have access to the generating agent's generation logs. This allows the executing agent to be configured more flexibly and decoupled from the generating agent, improving the deployment flexibility of the generating agent.

[0041] Accordingly, if the "statement" was indeed generated by the generating agent with reference to existing knowledge, the executing agent can use it as content to be verified. Thus, only factual content that significantly impacts reading quality and generation quality is selected for verification.

[0042] If not, the implementing entity can make further decisions based on the content related to knowledge bases and knowledge sources discussed below. The relevant content can be found in the content discussed below, which will not be explained here for the time being.

[0043] Step 202: Generate the degree of difference between the content to be verified and the reference knowledge; In the embodiments of this disclosure, based on step 201 above, this step aims to have the executing entity independently compare the content to be verified with its corresponding reference knowledge, relative to the generating agent, and generate the degree of difference between the two.

[0044] For example, the executing entity can break down the reference knowledge into multiple sentences and then generate reference semantic features for each sentence.

[0045] Then, the executing entity compares the reference semantic features of these sentences with the semantic features of the content to be verified, and after determining the similarity of each reference semantic feature, uses the highest similarity to generate the degree of difference between the content to be verified and the reference knowledge. For example, the executing entity can determine the degree of difference in the form of (1 - highest similarity).

[0046] Next, if there is a target content to be verified with a difference degree greater than or equal to the difference degree threshold (this difference degree threshold can be preset based on the assumption that the semantic difference between the two is too large and that the two cannot be considered to point to the same concept or content), or in other words, if the target difference degree corresponding to the target content to be verified is greater than or equal to the difference degree threshold, then the executing entity can "consider" that the target content to be verified is inaccurate (i.e., its difference from the corresponding reference knowledge is too large).

[0047] Accordingly, the executing entity can respond to this, that is, in response to the target difference degree corresponding to the target content to be verified being greater than or equal to the difference degree threshold, by selecting to execute step 203 for the target content to be verified.

[0048] Step 203: Generate the first modification instruction for the target content to be verified; In the embodiments of this disclosure, based on step 202, this step aims to have the aforementioned execution entity generate a first modification instruction for the target content to be verified when the target content to be verified exists. This first modification instruction can be used to instruct the generating agent to re-reference the reference knowledge corresponding to the target content to be verified, and modify or correct the target content to be verified so that it can meet the requirements corresponding to the aforementioned difference threshold.

[0049] Then, the executing entity can use the first modification instruction to request the generated agent to perform modification actions by executing the next step 204.

[0050] Step 204: Based on the first modification instruction and reference knowledge, control the generating agent to refer to the reference knowledge, update the content to be verified in the generated content, and generate updated generated content.

[0051] In the embodiments of this disclosure, as discussed above, based on step 203, this step aims to have the executing entity, based on and utilizing the first modification instruction and reference knowledge, control the generating agent to reference the reference knowledge, update the content to be verified in the generated content, and generate updated generated content.

[0052] In some embodiments, if the executing agent, as a verification agent, may be superior to the generating agent in terms of model size and capabilities (e.g., understanding of reference knowledge), it may further indicate, when providing reference knowledge, the "sentence" in the reference knowledge that has the highest similarity to the target content to be verified, in order to assist and "inform" the generating agent that the most relevant content at present may be that "sentence," thereby helping the generating agent to more quickly and granularly locate the specific content in the reference knowledge that can actually be referenced.

[0053] The method for generating content based on an intelligent agent provided in this disclosure first identifies content to be verified from the generated content generated by the intelligent agent, wherein the content to be verified is generated by the intelligent agent based on reference knowledge; then, a difference degree between the content to be verified and the reference knowledge is generated; next, in response to the target difference degree corresponding to the target content to be verified being greater than or equal to a difference degree threshold, a first modification instruction for the target content to be verified is generated; finally, based on the first modification instruction and the reference knowledge, the intelligent agent is controlled to refer to the reference knowledge to update the content to be verified in the generated content, thereby generating updated generated content. Thus, after generating content using an intelligent agent, the generated content can be verified using the reference knowledge referenced during content generation, reducing or avoiding the provision of erroneous content in the generated content due to the intelligent agent's misunderstanding of the reference content, model illusions, etc., thereby improving the quality of content generation.

[0054] In some embodiments, if the degree of difference corresponding to each content to be verified is less than the aforementioned degree of difference threshold, the executing entity can consider that each content to be verified is in accordance with the reference knowledge. In response, the executing entity can choose to output the generated content, for example, by providing the generated content to the target device used by the user based on a preset communication path.

[0055] Therefore, this method can avoid generating or providing incorrect content due to the generating agent's misunderstanding of the reference knowledge.

[0056] In some embodiments, if the executing entity detects target content to be verified, after generating updated content as discussed in process 200 above, it can identify new content to be verified again and determine whether new target content to be verified will be included, or whether the old target content to be verified has actually been modified to meet the requirements, by re-executing process 200.

[0057] Accordingly, if the modification is included, the executing agent can similarly generate the first modification instruction again to instruct the generating agent to make the corresponding modifications until the modification is no longer included, at which point the final result is output. For ease of understanding, this method can also be called cyclic verification or multi-round verification.

[0058] Therefore, by using this cyclical and multi-round verification method, the generated content is checked repeatedly until the verification results of each "content to be verified" included therein meet the requirements.

[0059] In some embodiments, even if the content to be verified can "match" the reference knowledge used for reference, the quality of the generated content may still be affected because the reference knowledge itself is unreliable. Therefore, in order to identify whether the reference knowledge is reliable and to determine whether the reference knowledge is trustworthy and reliable, a trustworthy knowledge base can be maintained in advance using reference knowledge whose content quality has been confirmed to meet the requirements. This way, when it is determined that the reference knowledge comes from the trustworthy knowledge base, the reference knowledge can be considered trustworthy and reliable.

[0060] Correspondingly, this trusted knowledge base can also be used and accessed by generative agents, enabling them to utilize this high-quality, trusted reference knowledge to generate content.

[0061] Similarly, in some embodiments, trusting online (external) knowledge sources can be chosen to eliminate the need for detailed, granular maintenance of every piece of reference knowledge, while still achieving the effect of identifying the reliability of reference knowledge and enriching and expanding the sources of credible reference knowledge. For example, the websites of authoritative publishers can be trusted, and reference knowledge originating from these trusted online knowledge sources can be considered reliable and trustworthy.

[0062] Accordingly, in such cases, when performing step 202 as described above, the executing entity may, as an alternative, first determine the source of the reference knowledge. If the reference knowledge comes from a pre-determined trusted knowledge base or a trusted online knowledge source, the executing entity will then generate a difference score between the content to be verified and the reference knowledge. This avoids the problem of low-quality content to be verified due to unreliable or untrustworthy reference knowledge used by the generating agent.

[0063] In some embodiments, since even if the reference knowledge comes from an untrusted online knowledge source (i.e., the reference knowledge is neither from a trusted knowledge base nor from a trusted knowledge source), it does not necessarily mean that the reference knowledge is unreliable, in such cases, the executing entity can actually decide whether the reference knowledge is actually unreliable or untrustworthy by using the content in trusted knowledge bases and trusted knowledge sources to assist in the judgment.

[0064] This is to avoid limiting the ability to generate intelligent agents to the size of the trusted knowledge base and the trusted scale of online knowledge sources, and to avoid reducing the overall ability to generate content due to untimely updates of the trusted knowledge base or untimely trust of online knowledge sources.

[0065] For easier understanding, you can also refer to... Figure 3 . Figure 3This disclosure provides a flowchart of a process for generating a difference between content to be verified and reference knowledge, including process 300. For example, process 300 can be used as an alternative or alternative implementation of step 202 in process 200 when the reference knowledge comes from an untrusted online knowledge source.

[0066] Process 300 specifically includes the following steps: Step 301: Generate the semantic similarity between the reference knowledge and the trusted reference knowledge in the pre-determined trusted knowledge base; Specifically, as discussed above, if the reference knowledge comes from an untrusted online knowledge source (for example, the executing entity can determine whether the reference knowledge comes from an untrusted online knowledge source by generating the access path for the reference knowledge), then the executing entity can respond by generating the semantic similarity between the reference knowledge and the trusted reference knowledge in a pre-determined trusted knowledge base.

[0067] For example, the executing entity can generate semantic features corresponding to the reference knowledge, and then compare these semantic features with the semantic features of the trusted reference knowledge in the trusted knowledge base to determine and generate the semantic similarity of the reference knowledge to each trusted reference knowledge in the trusted knowledge base.

[0068] Next, if there is a target credible reference knowledge with a corresponding semantic similarity greater than or equal to the first similarity threshold (usually, the standard for determining whether the reference knowledge and the credible reference knowledge are semantically close and consistent can be considered), the executing entity can respond to this and choose to continue executing step 302.

[0069] Step 302: Generate the degree of difference between the content to be verified and the reference knowledge.

[0070] This allows the executing entity to first verify the quality of content from a trusted knowledge base when the reference knowledge comes from an untrusted online knowledge source, rather than simply "rejecting" it outright. This enables the generating agent to have a wider range of reference knowledge to choose from and use.

[0071] In some optional implementations of this embodiment, if the semantic similarity between each trusted reference knowledge in the determined trusted knowledge base and the reference knowledge is less than the first similarity threshold, the executing entity may further choose to try to use the trusted online knowledge source to verify the quality of the reference knowledge.

[0072] Accordingly, in the absence of the aforementioned reliable reference knowledge of the target, the executing entity may also respond by choosing to execute step 303.

[0073] Step 303: Retrieve supplementary reference knowledge related to the reference knowledge from the online knowledge source for credit granting; Specifically, in this step, the executing entity can choose to use reference knowledge as search information to retrieve supplementary reference knowledge related to that reference knowledge from online knowledge sources. For example, the executing entity can construct a search question using the reference knowledge and use the search question to search through various online knowledge sources to attempt to retrieve supplementary reference knowledge related to that reference knowledge.

[0074] If the retrieved supplementary reference knowledge contains target supplementary reference knowledge whose semantic similarity to the reference knowledge is greater than or equal to the second similarity threshold (this second similarity threshold can be the same as or different from the first similarity threshold mentioned above. For example, the second similarity threshold can be greater than the first similarity threshold mentioned above), then the executing entity can choose to jump to step 302 above and select to generate the difference between the content to be verified and the reference knowledge.

[0075] Thus, in this way, the executing entity can use trusted online knowledge sources to verify reference knowledge from untrusted knowledge sources, thereby enabling the generating agent to have a wider range of reference knowledge selection and usage capabilities.

[0076] In some embodiments, if multiple supplementary reference knowledge exists, the executing entity may also choose to integrate each of the multiple supplementary reference knowledge, compare the integrated result with the reference knowledge, generate a corresponding semantic similarity, and compare the semantic similarity with a second similarity threshold to determine whether the reference knowledge is available. This avoids misidentification due to different trusted online knowledge sources only partially recording reference knowledge.

[0077] In some optional implementations of this embodiment, if the semantic similarity between the retrieved supplementary reference knowledge and the reference knowledge is less than the second similarity threshold (or, each supplementary reference knowledge is less than the second similarity threshold, and the above-mentioned "integration result" is also less than the second similarity threshold), that is, there is no "target supplementary reference knowledge", it means that the supplementary reference knowledge cannot be used to prove or verify whether the reference knowledge is credible or reliable. In such a case, the executing entity may choose not to generate the difference between the content to be verified and the reference knowledge, but instead choose to delete the content to be verified, so as to avoid outputting unverified and potentially unreliable content to be verified.

[0078] For example, the executing agent can similarly instruct the generating agent to delete the content to be verified in the generated content by issuing a modification instruction.

[0079] In some embodiments, to prevent the generating agent from generating unreliable content by referencing the unreliable reference knowledge in other content besides the content to be verified, the modification instructions sent by the executing agent can also be used to instruct the generating agent to prohibit referencing the reference knowledge during the generation of the generated content, so as to comprehensively correct the potential errors in the generated content caused by the use of the unreliable reference knowledge.

[0080] For example, the above process 300 may also include step 304, in which the subject being executed may choose to execute when the semantic similarity between the supplementary reference knowledge retrieved by the executing subject and the reference knowledge is less than the second similarity threshold, that is, when there is no target supplementary reference knowledge.

[0081] Step 304: Generate a second modification instruction for the generated content; Specifically, as discussed above, this second modification can be used to instruct the generating agent to prohibit referencing reference knowledge during the generation of generated content.

[0082] Step 305: Based on the second modification instruction and reference knowledge, control the generating agent to update the unreliable content generated by the reference knowledge in the generated content, and generate updated generated content.

[0083] Specifically, similar to step 204 above, the executing entity can send the second modification instruction and reference knowledge to the generating agent, and control the generating agent to update the unreliable content generated by the reference knowledge in the generated content, and generate updated generated content.

[0084] It should be understood that, in the process of verification through loop verification and multi-round verification as discussed above, the executing entity can also perform the action of "prohibiting reference to reference knowledge" in each round, which will not be repeated here.

[0085] Furthermore, as discussed above, if the "statement" was not generated by the generative agent with reference to reference knowledge, the executing agent can also choose to extract it and similarly verify its accuracy by searching for trusted reference knowledge in a trusted knowledge base that meets the semantic similarity requirements (e.g., greater than or equal to the first similarity threshold) and by searching for verifiable reference knowledge from trusted online knowledge sources. If it is inaccurate, the executing agent can provide this reference knowledge to the generative agent and similarly instruct the executing agent to refer to the provided reference knowledge to modify and adjust the "statement" to improve the overall quality of the generated content.

[0086] Based on any of the above embodiments, where the generated content and the updated generated content can be provided to the user, that is, where the "target content to be verified" is no longer included, the executing entity can also add or present knowledge acquisition links for obtaining reference knowledge in the provided generated content and the updated generated content.

[0087] For example, knowledge acquisition links can be added at the end of the generated content in the content delivery interface. This allows users to understand which references were used in generating the content and facilitates access to more detailed references when needed, thus improving the efficiency and experience of knowledge acquisition.

[0088] In some optional implementations of this embodiment, in order to facilitate users to access the reference knowledge and generated content (and update the generated content) pointed to by the knowledge acquisition link and to obtain more specific content (e.g., the content to be verified mentioned above), the executing entity may also choose to adjust the presentation result (e.g., text result) corresponding to the content to be verified to a different style of "control" that can be used by the user.

[0089] Then, if the executing entity detects an interaction with the presentation result of the content to be verified, it can respond by presenting target access links for reference knowledge associated with the content to be verified, in connection with the presentation result. For example, these target access links can be presented in the form of a "list," hovering above and near the presentation result.

[0090] Therefore, this not only allows users to use "controls" to understand which content is generated based on reference knowledge, but also enables them to conveniently and efficiently obtain the necessary links through interaction with these controls, thus improving the user's interactive experience.

[0091] In some embodiments, the execution entity may be allowed or required to provide a verification process in certain scenarios. For example, in scenarios where the verification capabilities of the execution entity are adjusted, the execution entity can similarly utilize an interface to visually provide feedback on the verification process. For instance, the generated content can be displayed on the interface. Then, after marking the content to be verified in the generated content, a window can be used to display the reference knowledge corresponding to the content to be verified (or, a link to obtain the reference knowledge), as well as the execution entity's verification result for the reference knowledge (i.e., if the target difference is greater than or equal to the difference threshold, "verification failed" can be displayed).

[0092] This allows for a visual and dynamic view of the verification process for users who wish to understand it.

[0093] Similarly, in some scenarios where users need to know which content has been verified, the executing entity can also present indicators or markers in the interface in association with the aforementioned "controls," and use these indicators or markers (e.g., "verified") to indicate that the corresponding content has been verified.

[0094] In some embodiments, after providing or outputting the generated content and updated generated content to the user (e.g., sending it to the target device via a preset communication path), the user can also provide feedback to the executing entity regarding any parts of the reference knowledge they believe are inappropriate. For example, the user can send a knowledge association instruction to the executing entity regarding the target generated content in the generated content and updated generated content, instructing the executing entity to read the feedback reference knowledge provided in the knowledge association instruction to modify or adjust the target generated content (i.e., adjust the target generated content using the feedback reference knowledge).

[0095] Accordingly, if the executing agent receives a knowledge association instruction from the target device for updating the target generated content, it can respond by reading the feedback reference knowledge provided by the association instruction. Then, the executing agent generates a third modification instruction for the target generated content, that is, an instruction to instruct the generating agent to update the target generated content with reference to the feedback reference knowledge.

[0096] It should be understood that if the target generated content is the previous "content to be verified," that is, generated by the executing agent based on reference knowledge, the executing agent can also ask the user whether the reference knowledge needs to be replaced through feedback and interaction, i.e., replace it with the feedback reference knowledge. If the user indicates that a replacement is needed, the third modification instruction can also be similar to the "prohibition" situation discussed above, instructing the generating agent to check and modify all parts of the generated content and updated generated content that reference the reference knowledge (i.e., comprehensively adjust the parts that previously referenced the reference knowledge by referring to the "feedback reference knowledge").

[0097] Then, the executing entity can control the generating agent to update the already output generated content, update the generated content, and update it again, based on the third modification instruction and feedback reference knowledge.

[0098] For ease of understanding, the process of updating the target generated content in the updated generated content can be understood as "another update". Correspondingly, the result of updating the updated generated content can be described as updating the generated content again.

[0099] Then, after the execution entity completes the generation, it can output the updated result to the target device again, for example, the content to be updated again.

[0100] This allows users to conveniently and personally utilize generative agents to complete generation tasks according to their needs by adjusting the way reference knowledge is used.

[0101] In some embodiments, in a multi-round verification scenario, if a certain content to be verified persists in multiple rounds (for example, the content to be verified is considered the target content to be verified in a preset number of rounds), then in such a case, the executing agent may choose to instruct the generating agent to delete that part of the content, so as to avoid the verification loop from being unable to break out due to the inability of the executing agent and the generating agent to reach an agreement.

[0102] To enhance understanding, this disclosure also provides a specific implementation scheme based on a particular application scenario. For easier comprehension, please refer to [link / reference needed]. Figure 4 . Figure 4 The flowchart illustrates a process for generating content based on an intelligent agent, specifically implemented in an application scenario according to an embodiment of this disclosure, including process 400.

[0103] In process 400, agent 410 and agent 415 may be included, wherein agent 410 may be executor of the generating agent described above, and agent 415 may be executor of the "executing agent" of this disclosure, i.e., the verification agent.

[0104] In process 400, agent 410 can first generate content 420 by executing S401. For example, the generated content 420 can be generated based on the content generation requirements provided by the user (not shown in the figure) (e.g., the generation requirement could be exemplarily "generate a speech about introducing and explaining C"). For example, agent 410 can generate content 420 based on reference knowledge 430.

[0105] After agent 410 generates generated content 420, agent 415 can identify the content to be verified from the generated content 420 generated by agent 410 by executing S402. For example, it can identify the content to be verified as "factual statement" 421.

[0106] Next, agent 415 can generate a difference 422 between the content to be verified 421 and the reference knowledge 430 by executing S403.

[0107] For example, if the (target) difference 422 corresponding to the content to be verified 421 is greater than or equal to the difference threshold mentioned above, in such a case, the agent 415 may choose to continue executing S404 to generate a modification instruction 423 for the content to be verified 421.

[0108] Then, agent 415 can execute S405 to control agent 410 to refer to reference knowledge 430 based on modification instruction 423 and reference knowledge 430, update the content to be verified 421 in the generated content 420, and generate updated generated content 425.

[0109] Accordingly, agent 410 can be controlled by agent 415, referencing reference knowledge 430, to update the content to be verified 421 in the generated content 420, and generate updated generated content 425 (for example, correcting the description or statement of "C" from "XXX" to "YYY").

[0110] After receiving the updated generated content 425, agent 415 can similarly read the new content to be verified from it and continue to check whether agent 410 needs to make further modifications in the same way as described above. Since this process is actually equivalent to the process described above, for ease of understanding, we will directly use the case where no further modifications are needed as an example.

[0111] In cases where no further modification to the updated content 425 is required, the agent 415 may choose to execute S406 to output the updated content 425 to, for example, a target device (not shown in the figure), so that a user of the target device can obtain and use the updated content 425.

[0112] Accordingly, in the updated content 425 output, the intelligent agent 415 can also adjust the text content 440 into a control that can be triggered by the user, so that the user can trigger the control through interaction with the text content 440 to obtain a link (not shown in the figure) for obtaining reference knowledge 430. Accordingly, the user can also understand that the text 440 is associated with a "link" through its visual style.

[0113] Furthermore, when updating the generated content 425, the agent 415 can also add an indicator 445 to help users intuitively understand that the content has been verified.

[0114] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a device for generating content based on an intelligent agent. This device embodiment is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0115] like Figure 5As shown, the device 500 for content generation based on an intelligent agent in this embodiment may include: a content to be verified identification unit 501, a difference degree generation unit 502, a first instruction generation unit 503, and a first content update unit 504. The content to be verified identification unit 501 is configured to identify the content to be verified from the generated content generated by the generating intelligent agent, wherein the content to be verified is generated by the generating intelligent agent based on reference knowledge; the difference degree generation unit 502 is configured to generate the difference degree between the content to be verified and the reference knowledge; the first instruction generation unit 503 is configured to generate a first modification instruction for the target content to be verified in response to a target difference degree corresponding to the target content to be verified being greater than or equal to a difference degree threshold; the first content update unit 504 is configured to control the generating intelligent agent to update the content to be verified in the generated content based on the first modification instruction and the reference knowledge, thereby generating updated generated content.

[0116] In this embodiment, the specific processing and technical effects of the following components in the agent-based content generation device 500—namely, the content recognition unit 501, the difference generation unit 502, the first indication generation unit 503, and the first content update unit 504—can be found in reference [reference needed]. Figure 2 The relevant descriptions of steps 201-204 in the corresponding embodiments will not be repeated here.

[0117] In some optional implementations of this embodiment, the difference generation unit 502 is further configured to generate a difference between the content to be verified and the reference knowledge in response to the reference knowledge being derived from a predetermined trusted knowledge base or a trusted online knowledge source.

[0118] In some optional implementations of this embodiment, the difference generation unit 502 is further configured to generate a difference between the content to be verified and the reference knowledge in response to the reference knowledge coming from an untrusted online knowledge source and the semantic similarity between the reference knowledge and the target trusted reference knowledge in a pre-determined trusted knowledge base being greater than or equal to a first similarity threshold.

[0119] In some optional implementations of this embodiment, the difference generation unit 502 includes: a supplementary knowledge retrieval subunit, configured to retrieve supplementary reference knowledge associated with the reference knowledge from the trusted online knowledge source in response to the reference knowledge being from an untrusted online knowledge source and the semantic similarity between each trusted reference knowledge in a pre-determined trusted knowledge base and the reference knowledge being less than a first similarity threshold; and a difference generation subunit, configured to generate the difference between the content to be verified and the reference knowledge in response to the retrieval of target supplementary reference knowledge with a semantic similarity greater than or equal to a second similarity threshold.

[0120] In some optional implementations of this embodiment, the apparatus 500 further includes: a second instruction generation unit configured to generate a second modification instruction for the generated content in response to the fact that the semantic similarity between the retrieved supplementary reference knowledge and the reference knowledge is less than a second similarity threshold, wherein the second modification instruction generating agent prohibits referencing the reference knowledge during the generation of the generated content; and a second content updating unit configured to control the generating agent to update unreliable content generated by referencing the reference knowledge in the generated content based on the second modification instruction and the reference knowledge, thereby generating updated generated content.

[0121] In some optional implementations of this embodiment, the apparatus 500 further includes: a link acquisition unit, configured to add knowledge acquisition links for acquiring reference knowledge in the updated generated content.

[0122] In some optional implementations of this embodiment, the apparatus 500 further includes: a link acquisition and presentation unit, configured to, in response to detecting an interactive action related to the presentation result of the content to be verified, present a target acquisition link of reference knowledge associated with the content to be verified in association with the presentation result.

[0123] In some optional implementations of this embodiment, the content to be verified identification unit 501 is further configured to identify content to be verified that is a factual statement from the generated content generated by the generating agent.

[0124] In some optional implementations of this embodiment, the device 500 further includes: a first content output unit configured to output updated generated content to a target device; a feedback knowledge reading unit configured to read feedback reference knowledge provided by the instruction association in response to receiving a knowledge association instruction from the target device for the target generated content in the updated generated content; a third instruction generation unit configured to generate a third modification instruction for the target generated content; a third content update unit configured to control the generating agent to update the target generated content in the updated generated content based on the third modification instruction and the feedback reference knowledge, and generate updated generated content again; and a second content output unit configured to output updated generated content again to the target device.

[0125] In some optional implementations of this embodiment, the apparatus 500 further includes: a third content output unit, configured to output generated content in response to each content to be verified having a difference degree less than a difference degree threshold.

[0126] This embodiment exists as a device embodiment corresponding to the above method embodiment. The device for generating content based on an intelligent agent provided in this embodiment can verify the generated content by using the reference knowledge referenced when generating the content after the intelligent agent generates the content, so as to reduce or avoid the situation that the intelligent agent provides erroneous content in the generated content due to misunderstanding of the reference content, model illusion, etc., and improve the quality of content generation.

[0127] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0128] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0129] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded into random access memory (RAM) 603 from storage unit 608. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0130] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0131] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the agent-based content generation method. For example, in some embodiments, the agent-based content generation method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the agent-based content generation method described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the agent-based content generation method by any other suitable means (e.g., by means of firmware).

[0132] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0133] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0134] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, 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 devices, magnetic storage devices, or any suitable combination of the foregoing.

[0135] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0136] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0137] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, also known as cloud computing servers or cloud hosts, which are hosting products within the cloud computing service ecosystem to address the management difficulties and weak business scalability inherent in traditional physical hosts and Virtual Private Servers (VPS) services. Servers can also be categorized as distributed system servers or servers incorporating blockchain technology.

[0138] According to the technical solution of the present disclosure, after the intelligent agent generates content, the reference knowledge referenced when generating the content can be used to verify the generated content, so as to reduce or avoid the situation that the intelligent agent provides erroneous content in the generated content due to misunderstanding of the reference content, model illusion, etc., and improve the quality of content generation.

[0139] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution provided in this disclosure can be achieved, and this is not limited herein.

[0140] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating content based on intelligent agents, comprising: Identify the content to be verified from the generated content generated by the generating agent, wherein the content to be verified is generated by the generating agent based on reference knowledge; Generate the degree of difference between the content to be verified and the reference knowledge; In response to the target difference degree corresponding to the target content to be verified being greater than or equal to the difference threshold, a first modification instruction is generated for the target content to be verified. Based on the first modification instruction and the reference knowledge, the generating agent is controlled to update the content to be verified in the generated content by referring to the reference knowledge, and to generate updated generated content.

2. The method according to claim 1, wherein, The generation of the difference between the content to be verified and the reference knowledge includes: In response to the fact that the reference knowledge comes from a pre-determined trusted knowledge base or a trusted online knowledge source, a difference degree between the content to be verified and the reference knowledge is generated.

3. The method according to claim 1, wherein, The generation of the difference between the content to be verified and the reference knowledge includes: In response to the fact that the reference knowledge comes from an untrusted online knowledge source, and the semantic similarity between the reference knowledge and the target trusted reference knowledge in a pre-determined trusted knowledge base is greater than or equal to a first similarity threshold, a difference degree between the content to be verified and the reference knowledge is generated.

4. The method according to claim 1, wherein, The generation of the difference between the content to be verified and the reference knowledge includes: In response to the fact that the reference knowledge comes from an untrusted online knowledge source, and the semantic similarity between each trusted reference knowledge in the pre-determined trusted knowledge base and the reference knowledge is less than a first similarity threshold, supplementary reference knowledge associated with the reference knowledge is retrieved from the trusted online knowledge source. In response to the retrieval of target supplementary reference knowledge with a semantic similarity greater than or equal to a second similarity threshold with the reference knowledge, a difference degree between the content to be verified and the reference knowledge is generated.

5. The method according to claim 4, further comprising: In response to the fact that the semantic similarity between the retrieved supplementary reference knowledge and the reference knowledge is less than the second similarity threshold, a second modification instruction is generated for the generated content, wherein the second modification instruction prohibits the generating agent from referring to the reference knowledge during the process of generating the generated content; Based on the second modification instruction and the reference knowledge, the generating agent is controlled to update the unreliable content generated in the generated content with reference to the reference knowledge, and to generate updated generated content.

6. The method according to claim 1, further comprising: Add a knowledge acquisition link to the updated content to obtain the reference knowledge.

7. The method according to claim 6, further comprising: In response to detecting an interaction with the presentation result of the content to be verified, a target acquisition link of the reference knowledge associated with the content to be verified is presented in association with the presentation result.

8. The method according to claim 1, wherein, The step of identifying the content to be verified from the generated content generated by the generating agent, wherein the content to be verified is generated by the generating agent based on reference knowledge, includes: Identify content to be verified that is a factual statement from the content generated by the generative agent.

9. The method according to claim 1, further comprising: The updated content is output to the target device; In response to receiving a knowledge association instruction from the target device for the target generated content in the updated generated content, the feedback reference knowledge provided by the instruction association is read. Generate a third modification instruction for the content generated for the target; Based on the third modification instruction and the feedback reference knowledge, the generating agent is controlled to update the target generated content in the updated generated content by referring to the feedback reference knowledge, and generate updated generated content again; The updated content is output to the target device.

10. The method according to any one of claims 1-9, further comprising: In response to the fact that the difference degree corresponding to each of the contents to be verified is less than the difference degree threshold, the generated content is output.

11. An apparatus for generating content based on intelligent agents, comprising: The content to be verified identification unit is configured to identify the content to be verified from the generated content generated by the generating agent, wherein the content to be verified is generated by the generating agent based on reference knowledge; The difference generation unit is configured to generate the difference between the content to be verified and the reference knowledge; The first instruction generation unit is configured to generate a first modification instruction for the target content to be verified in response to a target difference degree greater than or equal to a difference degree threshold. The first content update unit is configured to control the generating agent to update the content to be verified in the generated content based on the first modification instruction and the reference knowledge, thereby generating updated generated content.

12. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the agent-based content generation method according to any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the agent-based content generation method according to any one of claims 1-10.

14. A computer program product comprising a computer program that, when executed by a processor, implements the method for generating content based on an intelligent agent according to any one of claims 1-10.

Images