Information processing method and device based on large model, electronic equipment, readable storage medium and computer program product

Abstract

The invention provides an information processing method and device based on a large model, electronic equipment, a readable storage medium and a computer program product, and relates to the field of information processing, in particular to the field of information processing based on the large model. Obtaining a reconstructed text of the user input information; determining a detection rule corresponding to the user input information based on the reconstructed text; determining associated information corresponding to the user input information based on the reconstructed text; on the basis of the reconstructed text, the detection rule and the associated information, a template is constructed through cue words, and detection cue words corresponding to the information input by the user are determined; and based on the detection prompt word, utilizing a pre-trained content detection model to obtain a content detection result of the user input information.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of intelligent processing, in particular to the field of information processing based on large models, and specifically to a method and apparatus for information processing based on large models, an electronic device, a computer readable storage medium, and a computer program product. BACKGROUND

[0002] Artificial intelligence is a discipline that studies enabling computers to simulate some thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.) of humans, which includes both hardware and software technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing, etc.; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, and machine learning / deep learning, big data processing technology, knowledge graph technology, etc.

[0003] With the continuous development of information technology, how to determine the content to be detected from the input information has become a key link in information processing; and the rapid development of artificial intelligence technology, especially large models and neural networks, provides a better technical basis for content detection in the information processing process.

[0004] The methods described in this section can not necessarily be the methods previously conceived or adopted. Unless otherwise indicated, nothing in this section should be assumed to be prior art merely because it is included in this section. Similarly, unless otherwise indicated, issues raised in this section should not be assumed to have been recognized in any prior art. SUMMARY

[0005] The present disclosure provides a method and apparatus for information processing based on large models, an electronic device, a computer readable storage medium, and a computer program product.

[0006] According to an aspect of the present disclosure, a large model-based information processing method is provided, comprising: based on user input information, utilizing a pre-trained text reconstruction model to obtain a reconstructed text of the user input information, wherein the reconstructed text includes structured information of the user input information; based on the reconstructed text, determining a detection rule corresponding to the user input information, wherein the detection rule is used to describe the characteristics of the content in the user input information that needs to be detected; based on the reconstructed text, determining first associated information corresponding to the user input information, wherein the first associated information at least includes information with a semantic similarity greater than a preset similarity threshold with the reconstructed text, and wherein the source of the first associated information at least includes one of a pre-constructed knowledge database and an Internet platform; based on the reconstructed text, the detection rule, and the first associated information, utilizing a prompt word to construct a template to determine a detection prompt word corresponding to the user input information; and based on the detection prompt word, utilizing a pre-trained content detection model to obtain a content detection result of the user input information.

[0007] According to another aspect of the present disclosure, a training method of a text reconstruction model for information processing is provided, wherein the text reconstruction model is used to implement the information processing method, and wherein the training method of the text reconstruction model comprises: obtaining sample input information and sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by at least one of the following operations on the sample input information: variant vocabulary restoration, semantic completion, and context reconstruction; inputting the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a generative large model; based on the first reconstructed text and the sample reconstructed text, determining a first loss value of the text reconstruction model; and based on the first loss value, adjusting the parameters of the text reconstruction model.

[0008] According to another aspect of the present disclosure, a training method of a content detection model is provided, wherein the content detection model is used to implement the information processing method, and wherein the training method of the content detection model comprises: obtaining sample prompt words and sample content detection results corresponding to the sample prompt words; inputting the sample prompt words into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a generative large model; based on the first content detection result and the sample content detection result, determining a second loss value of the content detection model; and based on the second loss value, adjusting the parameters of the content detection model.

[0009] According to another aspect of the present disclosure, a large model-based information processing apparatus is provided, comprising: a text reconstruction module configured to obtain a reconstructed text of user input information based on the user input information by using a pre-trained text reconstruction model, wherein the reconstructed text includes structured information of the user input information; a detection rule acquisition module configured to determine a detection rule corresponding to the user input information based on the reconstructed text, wherein the detection rule is used to describe features of content in the user input information that needs to be detected; an associated information acquisition module configured to determine first associated information corresponding to the user input information based on the reconstructed text, wherein the first associated information at least includes information with a semantic similarity greater than a preset similarity threshold with the reconstructed text, and wherein the source of the first associated information at least includes one of a pre-constructed knowledge database and an Internet platform; a prompt word construction module configured to determine a detection prompt word corresponding to the user input information by using a prompt word construction template based on the reconstructed text, the detection rule, and the first associated information; and a content detection module configured to obtain a content detection result of the user input information by using the content detection model based on the detection prompt word.

[0010] According to another aspect of the present disclosure, a text reconstruction model training apparatus for information processing is provided, which is used to implement the text reconstruction model training method, wherein the text reconstruction model training apparatus comprises: a sample information acquisition module configured to obtain sample input information and sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by at least one of the following operations on the input sample information: variant vocabulary restoration, semantic completion, and context reconstruction; a reconstructed text generation module configured to input the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a large language model; a first loss value calculation module configured to determine a first loss value of the text reconstruction model based on the first reconstructed text and the sample reconstructed text; and a first parameter adjustment module configured to adjust parameters of the text reconstruction model based on the first loss value.

[0011] According to another aspect of the present disclosure, there is provided a content detection model training apparatus for implementing the training method of the content detection model, wherein the content detection model training apparatus comprises: a sample prompt word obtaining module configured to obtain a sample prompt word and a sample content detection result corresponding to the sample prompt word; a detection result generating module configured to input the sample prompt word into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a large language model; a second loss value calculating module configured to determine a second loss value of the content detection model based on the first content detection result and the sample content detection result; and a second parameter adjusting module configured to adjust parameters of the content detection model based on the second loss value.

[0012] According to another aspect of the present disclosure, there is provided 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, and the instructions are executed by the at least one processor to enable the at least one processor to perform the information processing method and the model training method.

[0013] According to another aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, causes the processor to implement the information processing method and the model training method as described above.

[0014] According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the information processing method and the model training method as described above.

[0015] According to one or more embodiments of the present disclosure, the user input information is reconstructed by a pre-trained text reconstruction model to improve the understandability of the text, and the detection rule of the user input information is determined according to the reconstructed text, and at least first associated information with a semantic similarity greater than a preset similarity threshold to the reconstructed text is determined from a pre-constructed knowledge database or an Internet platform; then, the detection prompt word of the user input information is determined by constructing a template using a prompt word according to the reconstructed text, the detection rule, and the first associated information, the reasoning process of the content detection model is guided and optimized, and the content detection is more accurate and efficient.

[0016] It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings illustrate exemplary embodiments and constitute a part of the specification. Together with the description of the specification, they serve to explain exemplary implementation of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals designate similar, but not necessarily identical, elements.

[0018] Figure 1 A schematic diagram illustrating an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure; Figure 2 A flowchart of an information processing method based on a large model according to an embodiment of the present disclosure is shown; Figure 3 A flowchart of a method for training a text reconstruction model for information processing according to an embodiment of the present disclosure is shown; Figure 4 A flow chart of a method for training a content detection model according to an embodiment of the present disclosure is shown; Figure 5 A block diagram of an information processing device based on a large model according to an embodiment of the present disclosure is shown; Figure 6 A block diagram of a text reconstruction model training device for information processing according to an embodiment of the present disclosure is shown; Figure 7 A block diagram of a content detection model training apparatus according to an embodiment of the present disclosure is shown; Figure 8 A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0019] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0020] In this disclosure, unless otherwise specified, the use of terms such as "first" and "second" to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, while in some cases, based on the context of the description, they may also refer to different instances.

[0021] The terminology used in the description of the various described examples in the present disclosure is for the purpose of describing particular examples only and is not intended to be limiting. As used in this description, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. Furthermore, the use of the terms "a" and "an" and "the" and "at least one" are intended to include the meaning of "one or more." Additionally, the use of the term "including," as well as other forms such as "includes" and "included," is not limiting. Also, terms such as "example" and "exemplary" are utilized merely for illustrative purposes and do not reflect a partial or full scope of the subject matter.

[0022] Embodiments of the present disclosure will be described in detail with reference to the drawings, wherein:

[0023] Figure 1 A schematic diagram of an example system 100 in which various methods and apparatus described herein can be implemented in accordance with embodiments of the present disclosure is shown. Referring to Figure 1 , the system 100 includes one or more client devices 101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 that couple the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105, and 106 can be configured to execute one or more application programs.

[0024] In embodiments of the present disclosure, the server 120 can run one or more services or software applications that enable execution of information processing methods and model training methods.

[0025] In certain embodiments, the server 120 can also provide other services or software applications, which can include non-virtual and virtual environments. In certain embodiments, these services can be provided as web-based services or cloud services, such as to users of the client devices 101, 102, 103, 104, 105, and / or 106 under a software as a service (SaaS) model.

[0026] In Figure 1 In the illustrated configuration, the server 120 can include one or more components that implement the functionality performed by the server 120. These components can include software components that are executable by one or more processors, hardware components, or combinations thereof. Users operating the client devices 101, 102, 103, 104, 105, and / or 106 can in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by the components. It should be understood that various different system configurations are possible, which can differ from the system 100. Thus, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0027] A user can use the client devices 101, 102, 103, 104, 105, and / or 106 to implement the recommendation of content. The client devices can provide an interface that enables a user of the client device to interact with the client device. The client devices can also output information to the user via the interface. Although Figure 1 Only six client devices are depicted, but those skilled in the art will appreciate that the present disclosure can support any number of client devices.

[0028] The client devices 101, 102, 103, 104, 105, and / or 106 can include various types of computer devices, such as portable handheld devices, general purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service kiosk devices, service robots, gaming systems, thin clients, various messaging devices, sensors or other sensing devices, and the like. These computer devices can run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices can include cellular telephones, smartphones, tablet computers, personal digital assistants (PDAs), and the like. Wearable devices can include head-mounted displays (such as smart glasses) and other devices. Gaming systems can include various handheld gaming devices, Internet-enabled gaming devices, and the like. The client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and can use various communication protocols.

[0029] The networks 110 can be any type of networks known to those skilled in the art that can support data communications using any of a plurality of available protocols, including without limitation TCP / IP, SNA, IPX, etc. As examples only, one or more of the networks 110 can be a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0030] The server 120 can include one or more general purpose computers, special purpose server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other appropriate arrangement and / or combination. The server 120 can include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that can be virtualized to maintain virtual storage devices for the servers). In various embodiments, the server 120 can run one or more services or software applications that provide the functionality described below.

[0031] The computing units in the server 120 can run one or more operating systems including any of the operating systems described above, as well as any commercially available server operating systems. The server 120 can also run any of a variety of additional server applications and / or mid-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0032] In some embodiments, the server 120 can include one or more applications to analyze and consolidate data feeds and / or event updates from users of the client devices 101, 102, 103, 104, 105, and / or 106. The server 120 can also include one or more applications to display the data feeds and / or real-time events via one or more display devices of the client devices 101, 102, 103, 104, 105, and / or 106.

[0033] In some embodiments, the server 120 can be a server of a distributed system, or a server combined with a blockchain. The server 120 can also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. The cloud server is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and virtual private server (VPS, Virtual Private Server) services.

[0034] The system 100 can also include one or more databases 130. In certain embodiments, these databases can be used to store data and other information. For example, one or more of the databases 130 can be used to store information such as audio files and video files. The databases 130 can reside in various locations. For example, databases used by the server 120 can be local to the server 120 or can be remote from the server 120 and can communicate with the server 120 via a network-based or dedicated connection. The databases 130 can be of different types. In certain embodiments, databases used by the server 120 can be, for example, relational databases. One or more of these databases can store, update, and retrieve data to and from the databases in response to commands.

[0035] In certain embodiments, one or more of the databases 130 can also be used by applications to store application data. Databases used by applications can be databases of different types, such as key-value stores, object stores, or regular stores supported by file systems.

[0036] Figure 1 The system 100 can be configured and operated in various ways to enable the various methods and apparatuses described according to the present disclosure to be applied.

[0037] Figure 2 is a flowchart illustrating a large model-based information processing method 200 according to an embodiment of the present disclosure, as Figure 2 As shown, the information processing method can include: a step S202, obtaining a reconstructed text of user input information based on the user input information, by using a pre-trained text reconstruction model, wherein the reconstructed text includes structured information of the user input information; a step S204, determining a detection rule corresponding to the user input information based on the reconstructed text, wherein the detection rule is used to describe features of content in the user input information that needs to be detected; a step S206, determining first associated information corresponding to the user input information based on the reconstructed text, wherein the first associated information at least includes information with a semantic similarity greater than a preset similarity threshold with the reconstructed text, and wherein the source of the first associated information at least includes one of a pre-constructed knowledge database and an Internet platform; a step S208, constructing a template by using a prompt word based on the reconstructed text, the detection rule, and the first associated information, to determine a detection prompt word corresponding to the user input information; and a step S210, obtaining a content detection result of the user input information by using a pre-trained content detection model based on the detection prompt word.

[0038] In this embodiment, the user input information is reconstructed by a pre-trained text reconstruction model to improve the understandability of the text; and a detection rule of the user input information is determined according to the reconstructed text, and at least first associated information with a semantic similarity greater than a preset similarity threshold to the reconstructed text is determined from a pre-constructed knowledge database or an Internet platform; and a detection prompt word of the user input information is determined by constructing a template using a prompt word according to the reconstructed text, the detection rule and the first associated information, so as to guide and optimize the reasoning process of the content detection model, so that the content detection is more accurate and efficient.

[0039] In some embodiments, the detection rule database can be a relational database, a NoSQL database, an in-memory database structure, and other databases that can realize the correspondence between the detection rule and the industry category, and the core design is to take the industry category as the key index or query condition.

[0040] For example, when the user input information is: "5 people, go to Xi'an, history lovers, public transportation is convenient, which three scenic spots to play", the reconstructed text of the user input information obtained by using the text reconstruction model at least includes: key entities: number of people (5 people), destination (Xi'an), interest label (history lovers) and traffic preference (public transportation is convenient); standardized expression: convert colloquial description into database queryable fields, such as "history lovers" mapped to "interest label" and "public transportation is convenient" mapped to "traffic convenience attribute"; add context: based on the tourism field, supplement implicit information (such as scenic spot type, location).

[0041] Correspondingly, taking "tourism field" as the key index or query condition, the detection rule corresponding to the user input information in the detection rule is determined, and the query logic based on the detection rule gives priority to: destination filtering: scenic spot location in Xi'an; interest matching: scenic spot label contains "history", "culture" or "heritage"; traffic convenience: distance from subway station or bus stop is less than 500 meters; number of people: 5-person group does not affect core recommendation (scenic spot capacity is usually sufficient); recommended number: 3.

[0042] Based on Xi'an real scenic spot data and detection rule matching: Terracotta Warriors and Horses: historical relics, consistent with "history lovers" interest, can be reached by bus and subway from Xi'an urban area; Xi'an City Wall: ancient city walls, high historical value; multiple entrances close to subway and bus coverage, suitable for group tour; Shaanxi History Museum: historical museum, rich in Zhou, Qin and Tang Dynasty cultural relics, located in the urban area, close to the subway station.

[0043] Based on the reconstructed text and the determination of the detection rule, the accuracy of detecting key information in the user input information and the accuracy of processing the user input information are improved.

[0044] It should be noted that in the technical solutions of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information comply with relevant laws and regulations and do not violate public order and good customs.

[0045] In some embodiments, the pre-constructed knowledge database is a collection of information in a private / authorized environment that is quickly accessible before responding to a query, which has been sorted and structured (or semi-structured); wherein the information is processed by cleaning, sorting, classifying, indexing, etc. to facilitate efficient retrieval. And wherein the knowledge database can be in the form of a relational database, a document database, a knowledge graph, a vector database, etc.; usually the knowledge database is stored locally or near-end, providing authoritative and verified factual information.

[0046] In some embodiments, the internet platform refers to accessing public web resources in real time or near real time to obtain information; involving web crawling, API calling or search engine integration technology; through the internet platform, the latest and ongoing information (news, social media updates, stock prices, weather, latest events, forum discussions, etc.) can be obtained. The internet platform can provide the latest and real-time information and trends to supplement the information outside the coverage of the knowledge base, especially in emerging, niche or rapidly changing fields.

[0047] In some embodiments, a dual-source collaborative work of the pre-constructed knowledge database and the internet platform is also included, specifically, according to the user's intention and strategy, it is decided whether to query the knowledge base first (for factual and stable problems) or to query the internet directly (for timeliness and news type problems), or to query both at the same time, to overcome the limitations of a single source, thereby improving the robustness of the information processing method.

[0048] Corresponding to the above-mentioned embodiments in the tourism field, the pre-constructed knowledge database at least includes the basic attributes of the scenic spots, such as scenic spot name, geographic location, belonging city, basic ticket price, opening time, etc.; through the internet platform, at least real-time news, comments, flow information corresponding to the scenic spots are obtained.

[0049] In some embodiments, in step S202, based on the user input information, a pre-trained text reconstruction model is used to obtain the reconstructed text of the user input information, including at least one of the following operations on the user input information: variant vocabulary restoration, semantic completion and context reconstruction.

[0050] In the embodiment, the text reconstruction model improves the clarity and standardization of the semantic of the user input information by at least one of variant vocabulary reduction, semantic completion, and context reconstruction.

[0051] In some embodiments, the variant vocabulary reduction includes identifying non-standard, colloquial, abbreviated, misspelled, dialectal, or internet slang, etc. “variant” vocabulary in the user input information, and reducing or replacing it with standard, normative, written vocabulary or phrases to improve the standardization of the input information, making the description of key entities, attributes, actions in the input information more clear and accurate.

[0052] In some embodiments, the semantic completion includes making explicit supplement and complete expression in the reconstructed text based on the information implied in the user input, not explicitly stated, or dependent on context / common sense to understand; the text reconstruction model uses its pre-trained language understanding ability to infer the missing but necessary semantic components, making the input information expression more complete, the logic more clear, eliminating ambiguity, and ensuring that the extracted structured information is accurate and comprehensive.

[0053] In some embodiments, the context reconstruction includes logical reorganization, sentence conversion, emphasis, or perspective conversion of the user input information without changing the original semantics of the user input information, so that it is more consistent with a specific expression purpose (such as report, summary, database entry) or more clearly highlights the structured information; to optimize the logicality and readability of information presentation, and make the core structured elements (entities, attributes, relationships, events) more prominent and easily identifiable. The context reconstruction includes adjusting the time sequence, causal relationship, primary and secondary relationship, etc.

[0054] In some embodiments, determining the detection rule corresponding to the user input information according to the reconstructed text includes determining the industry category corresponding to the user input information based on the reconstructed text; and in response to determining that the detection rule database stores the industry category, obtaining the detection rule corresponding to the industry category from the detection rule database.

[0055] In the embodiment, the determination of the industry category corresponding to the user input information based on the reconstructed text accurately reflects the intention or theme of the user input information, and when the detection rule database already has a rule for the industry category, the rule is directly reused, saving computing resources and time cost.

[0056] In some embodiments, the information processing method further includes, in response to determining that the detection rule database does not store the industry category corresponding to the user input information, determining a first detection rule for the industry category based on the industry category and a preset detection rule template.

[0057] In the embodiment, in response to determining that the industry category corresponding to the user input information does not exist in the rule detection database, a first detection rule of the industry category is generated according to a preset detection rule template, so as to improve the robustness of the information processing method.

[0058] In some embodiments, the detection rule template specifies the core logical structure and key components of the detection rule, but the specific details (such as matching content, threshold, parameter) therein are placeholders or configurable items to be filled; when it is confirmed that the industry category corresponding to the user input information does not exist in the rule detection database, the detection rule template provides a standardized starting point to ensure that the generated rule follows a consistent structure and logical paradigm, so as to facilitate the execution and maintenance of the detection rule.

[0059] For example, the detection rule template includes a template ID / name, a target detection object description, an applicable scenario, a version number, and creation / modification information, etc.

[0060] The template ID / name is used to uniquely identify the detection rule; the target detection object description is used to describe what type of information the detection rule is applicable to detect; the applicable scenario is used to describe which type of scenario (such as text, structured data, specific source, etc.) the detection rule is most suitable for processing; the version number facilitates iterative updating of the detection rule; and the creation / modification information is used to record the creator of the template and the modification history.

[0061] In some embodiments, the method further includes adding the first detection rule as the detection rule of the industry category to the detection rule database.

[0062] In the embodiment, the detection rule database is dynamically expanded according to the generated first detection rule, which improves the expansibility of the information processing method.

[0063] In some embodiments, the adding of the first detection rule as the detection rule of the industry category to the detection rule database supports multiple rule types (keyword, regular, logical expression, script, model reference); for example, a JSON field is used to store structured rule configuration, or a special rule description language (DSL) or other rule adding methods are used.

[0064] In some embodiments, the method further includes storing historical versions of the detection rule to facilitate rollback optimization; for example, a detection rule version record field and a version priority record field can be designed.

[0065] In some embodiments, the first association information further includes information of at least one same keyword as the reconstructed text.

[0066] In the embodiment, the first associated information is determined based on the keyword, the associated range is limited to information directly related to the core theme of the user input, the interference of irrelevant information is reduced, and the relevance of the detection result is improved.

[0067] In some embodiments, the keyword includes a word or phrase in the user input information that can represent the core content, theme or important concept of the user input information, such as a professional term or a specific entity (such as a person's name, place name, organization name, product name, etc.).

[0068] In some embodiments, the method further comprises: determining, based on the reconstructed text, second associated information corresponding to the user input information using a first retrieval technique of a plurality of predetermined retrieval techniques; and determining, based on the reconstructed text and the second associated information, the first associated information corresponding to the user input information using a second retrieval technique different from the first retrieval technique of the plurality of retrieval techniques, wherein the plurality of retrieval techniques at least includes semantic similarity retrieval and keyword matching retrieval.

[0069] In the embodiment, the plurality of retrieval techniques is used to overcome the limitations of a single retrieval technique at the expense of a certain retrieval efficiency, and the matching degree of the first associated information and the user input information is improved.

[0070] In some embodiments, a first retrieval technique (such as semantic similarity retrieval) of a plurality of predefined retrieval techniques can be used to convert the reconstructed text into a text vector, find an item (such as a document, a paragraph, a knowledge graph node, a product description segment, etc.) most similar to the reconstructed text embedding vector in the semantic space, return a set of candidate results sorted by similarity, i.e. the second associated information, and the second associated information is most similar in meaning to the user input information. Wherein the similarity can be measured by cosine similarity or dot product or other methods.

[0071] Further, the core entity (such as a person, a place, a product, an event) in the reconstructed text is determined, and the core entity is used as a query anchor point; the anchor point retrieval is used as the second retrieval technique to determine information existing the query anchor point from the second associated information as the first associated information corresponding to the user input information.

[0072] In some embodiments, the first associated information corresponding to the user input information can also be determined from the second associated information based on user historical behavior data using behavior matching retrieval as the second retrieval technique.

[0073] In some embodiments, the keyword matching retrieval can be used as the first retrieval technique based on the reconstructed text to determine the second associated information corresponding to the user input information, and the semantic similarity retrieval can be used as the second retrieval technique to determine, from the second associated information, information having a similarity greater than a second similarity threshold with the reconstructed text as the first associated information corresponding to the user input information. The retrieval techniques described above for the second associated information and the first associated information corresponding to the user input information are not limited to the retrieval techniques listed above, and combinations of different retrieval techniques can also be adjusted according to the application scenarios of the information processing method.

[0074] In some embodiments, the method further includes: in response to determining that the confidence of the content detection model is less than a confidence threshold, manually detecting the user input information; and adjusting the detection rule corresponding to the user input information in the detection rule database using the result of the manual detection.

[0075] In the present embodiment, the user input information corresponding to the discriminant result with low model confidence is manually detected, and the detection rule of the detection rule database is dynamically adjusted according to the result of the manual detection, so as to improve the matching degree of the detection rule and the accuracy of the content detection.

[0076] In some embodiments, the confidence output by the model is compared with a preset confidence threshold, and when it is determined that the confidence of the content detection model is less than the confidence threshold, the user input information marked as "low confidence" is put into a manual review queue. A content reviewer (or a domain expert) logs in to a dedicated review platform to view the user input information and its context, and based on professional knowledge, platform rules, laws and regulations, etc., independently and carefully reviews the information, marks the content to be detected or the features of the content to be detected, and stores the final determination result of the manual reviewer in association with the user input information, the model determination result and the low confidence record that initially triggered the manual review.

[0077] The detection rule of the detection rule database is dynamically adjusted according to the result of the manual detection, which aims to analyze the loopholes or deficiencies of the current detection rule library (such as the existing rules being too broad, resulting in a large number of ambiguous safe contents being misjudged as suspicious; the existing rules being too strict or outdated, missing some rule variants that should be identified), to adjust the threshold, range and logic condition of the existing rules, so as to make them more accurate.

[0078] In some embodiments, adjusting the detection rules of the detection rule database further comprises: removing rules that are invalid, outdated or generate a large number of false positives; adjusting the order, weight or logical relationship of rule execution. The adjusted rule library will be re-applied to subsequent automated detection, and when a new input information similar to the previous low confidence case is encountered next time, since the detection rule database has been updated, the detection rule itself makes a more accurate judgment, thereby reducing the probability of triggering manual review again.

[0079] It should be noted that the confidence threshold can be finely tuned according to the application scenario of the information processing method (trade-off between accuracy and recall rate), manual review capability, risk tolerance and other factors to improve the adaptability of the information processing method.

[0080] Figure 3 is a training method of a text reconstruction model for information processing according to an embodiment of the present disclosure, wherein the text reconstruction model is used to implement the information processing method, and wherein the training method of the text reconstruction model comprises: step S302, obtaining sample input information and sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by at least one of the following operations on the input sample information: variant vocabulary restoration, semantic completion and context reconstruction; step S304, inputting the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a generative large model; step S306, determining a first loss value of the text reconstruction model based on the first reconstructed text and the sample reconstructed text; and step S308, adjusting parameters of the text reconstruction model based on the first loss value.

[0081] In some embodiments, adjusting the parameters of the text reconstruction model based on the first loss value comprises: in response to determining that the first loss value is greater than a first loss threshold, ending the training of the text reconstruction model, otherwise adjusting the parameters of the text reconstruction model based on the first loss value, executing the obtaining of the sample input information and the sample reconstructed text corresponding to the sample input information and the subsequent steps until the first loss value is greater than the first loss threshold.

[0082] Figure 4 is a training method of a content detection model according to an embodiment of the present disclosure, as Figure 4As shown, the content detection model is used to implement the information processing method, and a training method of the content detection model includes: step S402, obtaining a sample prompt word and a sample content detection result corresponding to the sample prompt word; step S404, inputting the sample prompt word into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a generative large model; step S406, determining a second loss value of the content detection model based on the first content detection result and the sample content detection result; and step S408, adjusting parameters of the content detection model based on the second loss value.

[0083] In some embodiments, the adjusting parameters of the content detection model based on the second loss value includes, in response to determining that the second loss value is greater than a second loss threshold, ending the training of the content detection model, otherwise adjusting parameters of the content detection model based on the second loss value, executing the obtaining of the sample prompt word and the sample content detection result corresponding to the sample prompt word and subsequent steps until the second loss value is greater than the second loss threshold.

[0084] Figure 5 is a block diagram illustrating an information processing apparatus 500 according to an embodiment of the present disclosure, as shown in Figure 5 As shown, the information processing apparatus includes: a text reconstruction module 502 configured to obtain a reconstructed text of user input information based on the user input information by using a pre-trained text reconstruction model, wherein the reconstructed text includes structured information of the user input information; a detection rule obtaining module 504 configured to determine a detection rule corresponding to the user input information based on the reconstructed text, wherein the detection rule is used to describe features of content in the user input information that needs to be detected; an associated information obtaining module 506 configured to determine first associated information corresponding to the user input information based on the reconstructed text, wherein the first associated information at least includes information with a semantic similarity greater than a preset similarity threshold with the reconstructed text, and wherein the source of the first associated information at least includes one of a pre-constructed knowledge database and an Internet platform; a prompt word construction module 508 configured to determine a detection prompt word corresponding to the user input information by using a prompt word construction template based on the reconstructed text, the detection rule, and the first associated information; and a content detection module 510 configured to obtain a content detection result of the user input information by using the content detection model based on the detection prompt word.

[0085] Figure 6 is a block diagram illustrating a text reconstruction model training apparatus 600 for information processing according to an embodiment of the present disclosure, as shown in Figure 6As shown, the text reconstruction model training apparatus includes: a sample information acquisition module 602, configured to acquire sample input information and sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by at least one of the following operations on the input sample information: variant vocabulary restoration, semantic completion, and context reconstruction; a reconstructed text generation module 604, configured to input the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a large language model; a first loss value calculation module 606, configured to determine a first loss value of the text reconstruction model based on the first reconstructed text and the sample reconstructed text; and a first parameter adjustment module 608, based on the first loss value, to adjust the parameters of the text reconstruction model.

[0086] Figure 7 is a block diagram illustrating a content detection model training apparatus 700 according to an embodiment of the present disclosure, as shown in Figure 7 The training apparatus includes: a sample prompt word acquisition module 702, configured to acquire a sample prompt word and a sample content detection result corresponding to the sample prompt word; a detection result generation module 704, configured to input the sample prompt word into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a large language model; a second loss value calculation module 706, configured to determine a second loss value of the content detection model based on the first content detection result and the sample content detection result; and a second parameter adjustment module 708, configured to adjust the parameters of the content detection model based on the second loss value.

[0087] It should be noted that Figure 5 The various modules of the apparatus 500 shown in Figure 2 Correspond to the various steps of the method 200 described with reference to the foregoing method 200. Therefore, the operations, features and advantages described above for the method 200 also apply to the apparatus 500 and the modules and units included therein. Figure 6 The various modules of the apparatus 600 shown in Figure 3 Correspond to the various steps of the method 300 described with reference to the foregoing method 300. Therefore, the operations, features and advantages described above for the method 300 also apply to the apparatus 600 and the modules and units included therein. Figure 7 The various modules of the apparatus 700 shown in Figure 4 Correspond to the various steps of the method 400 described with reference to the foregoing method 400. Therefore, the operations, features and advantages described above for the method 400 also apply to the apparatus 700 and the modules and units included therein. For the sake of brevity, certain operations, features and advantages are not described here.

[0088] According to embodiments of the present disclosure, an electronic device, a readable storage medium, and a computer program product are also provided.

[0089] Reference will now be made to Figure 8 A block diagram of an electronic device 800 that can be a server or a client of the present disclosure, which is an example of a hardware device that can be applied to aspects of the present disclosure, will now be described. The electronic device is intended to represent various forms of digital electronic computing devices such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computing devices. The electronic device can also represent various forms of mobile devices such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present disclosure described and / or claimed in this document.

[0090] As shown in Figure 8 The electronic device 800 includes a computing unit 801 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 8808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0091] A plurality of components in the electronic device 800 are connected to the I / O interface 805, including an input unit 806, an output unit 807, the storage unit 808, and a communication unit 809. The input unit 806 can be any type of device that can input information to the electronic device 800, can receive inputted digital or character information, and generate key signal inputs related to user settings and / or function controls of the electronic device, and can include, but is not limited to, a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and / or a remote controller. The output unit 807 can be any type of device that can present information, and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 808 can include, but is not limited to, a magnetic disk, an optical disk. The communication unit 809 allows the electronic device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0092] The computing unit 801 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, and the like. The computing unit 701 performs various methods and processes described above, such as the method 200, the method 300, and the method 400. For example, in some embodiments, the method 200, the method 300, and the method 400 can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded onto the RAM 803 and executed by the computing unit 801, one or more steps of the method 200, the method 300, and the method 400 described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the method 200, the method 300, and the method 400 by other any appropriate means, such as by means of firmware.

[0093] Various implementations of the systems and techniques described above herein can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip system (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0094] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package, or entirely on a remote machine or server.

[0095] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0096] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; 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 acoustic, speech, or tactile input.

[0097] The systems and techniques described herein can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described herein), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0098] The computer system can include clients and servers. The clients and the servers are generally remote from each other and typically interact through a communication network. The relationship of client and server is one of communication and distribution, with the server receiving requests from the client and transmitting data to the client. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0099] It should be understood that various forms of flow shown above can be used with orders of steps being re-sequenced, added, or deleted. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure are achieved, which are not limited herein.

[0100] While embodiments or examples of the present disclosure have been described with reference to the figures, it is understood that the methods, systems, and apparatuses described above are merely exemplary embodiments or examples and the scope of the present disclosure is not limited by these embodiments or examples. Various elements of the embodiments or examples can be omitted or replaced by equivalent elements. In addition, the steps can be performed in an order different from that described in the present disclosure. Further, the various elements of the embodiments or examples can be combined in various ways. It is important that many of the elements described herein can be replaced by equivalent elements that appear after the present disclosure as technology evolves.

Claims

1. A large model-based information processing method, comprising: Based on the user input information, using a pre-trained text reconstruction model, obtain a reconstructed text of the user input information, wherein the reconstructed text includes structured information of the user input information; Determining, based on the reconstructed text, a detection rule corresponding to the user input information, wherein the detection rule is used to describe features of content to be detected in the user input information; Based on the reconstructed text, determining first associated information corresponding to the user input information, wherein the first associated information at least includes information having a semantic similarity with the reconstructed text greater than a preset similarity threshold, and wherein a source of the first associated information includes at least one of the following: a pre-constructed knowledge database and an Internet platform; Based on the reconstructed text, the detection rule, and the first association information, constructing a template using prompt words to determine a detection prompt word corresponding to the user input information; and Based on the detection prompt word, a pre-trained content detection model is used to obtain a content detection result of the user input information.

2. The information processing method according to claim 1, wherein: The determining, based on the reconstructed text, a detection rule corresponding to the user input information includes: Determining the industry category corresponding to the user input information based on the reconstructed text; and In response to determining that the industry category is stored in the detection rule database, a detection rule corresponding to the industry category is acquired from the detection rule database.

3. The information processing method according to claim 2 further includes, in response to determining that the industry category corresponding to the user input information does not exist in the detection rule database, determining a first detection rule for the industry category based on the industry category and a preset detection rule template. 4 . The information processing method according to claim 3 , further comprising adding the first detection rule as a detection rule for the industry category to the detection rule database.

5. The information processing method according to any one of claims 1 to 4, wherein: The first association information also includes information containing at least one keyword identical to that of the reconstructed text.

6. The information processing method according to any one of claims 1 to 5, further comprising: Based on the reconstructed text, using a first retrieval technique among a plurality of predetermined retrieval techniques, determining second associated information corresponding to the user input information; as well as Based on the reconstructed text and the second associated information, a second retrieval technology different from the first retrieval technology among multiple retrieval technologies is used to determine the first associated information corresponding to the user input information, wherein the multiple retrieval technologies include at least semantic similarity retrieval and keyword matching retrieval.

7. The information processing method according to any one of claims 1 to 6, wherein: The obtaining of the reconstructed text of the user input information by using a pre-trained text reconstruction model based on the user input information includes performing at least one of the following operations on the user input information: variant vocabulary restoration, semantic completion, and context reconstruction.

8. The information processing method according to any one of claims 2 to 7, further comprising: In response to determining that the confidence of the content detection model is less than a confidence threshold, manually detecting the user input information; as well as The detection rule corresponding to the user input information in the detection rule database is adjusted using the result of the manual detection.

9. A method for training a text reconstruction model for information processing, wherein: The text reconstruction model is used to implement the information processing method according to any one of claims 1 to 8, and wherein the training method of the text reconstruction model includes: Obtaining sample input information and a sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by performing at least one of the following operations on the input sample information: variant vocabulary restoration, semantic completion, and context reconstruction; Inputting the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a generative large model; Determining a first loss value of the text reconstruction model based on the first reconstructed text and the sample reconstruction; and Based on the first loss value, parameters of the text reconstruction model are adjusted.

10. A method for training a content detection model, wherein: The content detection model is used to implement the information processing method according to any one of claims 1 to 8, and wherein the training method of the content detection model includes: Obtaining a sample prompt word and a sample content detection result corresponding to the sample prompt word; Inputting the sample prompt word into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a generative large model; determining a second loss value of the content detection model based on the first content detection result and the sample content detection result; and Based on the second loss value, parameters of the content detection model are adjusted.

11. An information processing device based on a large model, comprising: A text reconstruction module is configured to obtain a reconstructed text of the user input information using a pre-trained text reconstruction model based on the user input information, wherein the reconstructed text category includes structured information of the user input information; a detection rule acquisition module configured to determine a detection rule corresponding to the user input information based on the reconstructed text, wherein the detection rule is used to describe features of content to be detected in the user input information; a related information acquisition module configured to determine, based on the reconstructed text, first related information corresponding to the user input information, wherein the first related information at least includes information having a semantic similarity with the reconstructed text greater than a preset similarity threshold, and wherein a source of the first related information includes at least one of the following: a pre-constructed knowledge database and an Internet platform; a prompt word construction module configured to determine a detection prompt word corresponding to the user input information by using a prompt word construction template based on the reconstructed text, the detection rule, and the first association information; and The content detection module is configured to obtain a content detection result of the user input information based on the detection prompt word and using the content detection model.

12. A text reconstruction model training device for information processing, the text reconstruction model training device is used to implement the text reconstruction model training method according to claim 9, wherein: The text reconstruction model training device comprises: a sample information acquisition module configured to acquire sample input information and a sample reconstructed text corresponding to the sample input information, wherein the sample reconstructed text is obtained by performing at least one of the following operations on the input sample information: variant vocabulary restoration, semantic completion, and context reconstruction; a reconstructed text generation module configured to input the sample input information into the text reconstruction model to obtain a first reconstructed text of the sample input information, wherein the text reconstruction model is constructed based on a large language model; A first loss value calculation module is configured to determine a first loss value of the text reconstruction model based on the first reconstructed text and the sample reconstruction; and A first parameter adjustment module adjusts parameters of the text reconstruction model based on the first loss value.

13. A content detection model training device, the content detection model training device being used to implement the content detection model training method according to claim 10, wherein: The content detection model training device comprises: a sample prompt word acquisition module configured to acquire sample prompt words and sample content detection results corresponding to the sample prompt words; a detection result generating module configured to input the sample prompt word into the content detection model to obtain a first content detection result, wherein the content detection model is constructed based on a large language model; a second loss value calculation module configured to determine a second loss value of the content detection model based on the first content detection result and the sample content detection result; and The second parameter adjustment module is configured to adjust parameters of the content detection model based on the second loss value.

14. An electronic device comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

15. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-10.

16. A computer program product comprising a computer program, wherein When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images