5G message content data generation method and device, equipment and storage medium

Abstract

The invention provides a content data generation method and device of a 5G message, equipment and a storage medium. The method comprises the steps of obtaining a description scheme and a service type of content data of the 5G message; generating cue words of the description scheme, and combining the description scheme and the cue words to obtain a cue word text; performing entity recognition on the cue word text to obtain a first entity word in the cue word text, and based on a sensitive word library associated with the service type, rejecting sensitive words in the first entity word to obtain compliant entity words; and based on the compliance entity words, generating content data of the description scheme by adopting a content data generation model corresponding to the description scheme. On the basis of the content data generation method, compliance detection is performed on the generated content data when the corresponding content data is generated in the 5G message service, so that the safe content data can be efficiently output through the general large model, and the safety and accuracy of the generated content data are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method, device, equipment and storage medium for generating content data of a 5G message. Background Art

[0002] At present, when generating corresponding content data for 5G message materials, the 5G message materials are usually input into the general large model used by AIGC technology to automatically generate text, pictures, videos and other content data corresponding to the 5G message materials through the large model. However, the general large model used by the existing AIGC technology cannot perform effective security assessment on the content data it outputs. If you want to improve the security of the content data output by the general large model, you need to fine-tune or train it in a targeted manner, but this is very inefficient and cannot completely solve the security problem. Therefore, how to efficiently output secure content data through a general large model is a technical problem that needs to be solved by technicians in this field. Summary of the invention

[0003] In view of this, the present disclosure proposes a content data generation method, device, equipment and storage medium for 5G messages, which can efficiently output secure content data based on a general large model.

[0004] According to one aspect of the present disclosure, a method for generating content data of a 5G message is proposed, comprising:

[0005] Obtain a description scheme and service type of content data of the 5G message;

[0006] Generate prompt words for the description scheme, and combine the description scheme and the prompt words to obtain prompt word text;

[0007] Performing entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and removing sensitive words from the first entity word based on a sensitive word library associated with the business type to obtain a compliant entity word;

[0008] Based on the compliant entity words, a content data generation model corresponding to the description scheme is adopted to generate content data of the description scheme.

[0009] In a possible implementation, based on a sensitive word library associated with the business type, sensitive words in the first entity words are removed to obtain compliant entity words, including:

[0010] Traversing each of the first entity words;

[0011] For the traversed current first entity word, determine whether there is a sensitive word similar to the current first entity word in the sensitive word library associated with the business type; if it is determined that there is a sensitive word similar to the current first entity word, remove the current first entity word as a sensitive word; if it is determined that there is no sensitive word similar to the current first entity word, traverse the next first entity word;

[0012] After the traversal is completed, the compliant entity word is obtained.

[0013] In a possible implementation manner, when it is determined that there is no sensitive word similar to the current first entity word, the method further includes:

[0014] Obtaining a synonym of the current first entity word;

[0015] Determine whether there is a sensitive word similar to the synonym of the current first entity word in the sensitive word library associated with the business type. If it is determined that there is a sensitive word similar to the synonym of the current first entity word, remove the current first entity word as a sensitive word. If it is determined that there is no sensitive word similar to the synonym of the current first entity word, perform an operation of traversing the next first entity word.

[0016] In a possible implementation, when the content data of the description scheme is generated based on the compliant entity words and using a content data generation model corresponding to the description scheme, the method includes:

[0017] Acquire a qualifier associated with the compliant entity word, wherein the qualifier is used to define a generation rule of the content data;

[0018] The compliant entity words and the qualifiers are input into a content data generation model corresponding to the description scheme to generate content data of the description scheme.

[0019] In a possible implementation manner, after generating the content data describing the scheme, the method further includes:

[0020] Calculating the compliance confidence level of the content data;

[0021] Determining whether the compliance confidence level is greater than or equal to a preset confidence level threshold;

[0022] When it is determined that the compliance confidence is greater than or equal to the confidence threshold, the content data is pushed and displayed.

[0023] In a possible implementation, the content data includes text content data;

[0024] When calculating the compliance confidence of the text content data, it includes:

[0025] Performing entity recognition on the text content data to obtain a second entity word in the text content data;

[0026] Traversing each of the second entity words;

[0027] For the traversed current second entity word, a sensitive word similar to the current second entity word is screened out from a sensitive word library associated with the business type as a target sensitive word corresponding to the current second entity word, and a weight of the target sensitive word is obtained;

[0028] After the traversal is completed, the target sensitive words corresponding to each of the second entity words and the weights of the target sensitive words are obtained;

[0029] The compliance confidence of the text content data is calculated based on the target sensitive words corresponding to each of the second entity words and the weights of the target sensitive words.

[0030] In a possible implementation manner, when it is determined that the compliance confidence is less than the confidence threshold, prompt information indicating that the content data is non-compliant is pushed.

[0031] According to a second aspect of the present disclosure, a device for generating content data of a 5G message is provided, including:

[0032] A foreground presentation module, used to obtain a description scheme and a service type of the content data of the 5G message;

[0033] A prompt word module, used to generate a prompt word for the description scheme, and combine the description scheme and the prompt word to obtain a prompt word text;

[0034] A compliance detection module, configured to perform entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and remove sensitive words from the first entity word based on a sensitive word library associated with the business type to obtain a compliant entity word;

[0035] A model reasoning module is used to generate content data of the description scheme based on the compliant entity words and using a content data generation model corresponding to the description scheme.

[0036] According to a third aspect of the present disclosure, a device for generating content data of a 5G message is provided, including:

[0037] A processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement any one of the methods described in the first aspect of the present disclosure when executing the executable instructions.

[0038] According to a fourth aspect of the present disclosure, a non-volatile computer-readable storage medium is provided, on which computer program instructions are stored, characterized in that when the computer program instructions are executed by a processor, the method described in any one of the first aspects of the present disclosure is implemented.

[0039] The present disclosure provides a method, device, equipment and storage medium for generating content data of a 5G message, the method comprising: obtaining a description scheme and a service type of the content data of the 5G message; generating a prompt word of the description scheme, and combining the description scheme and the prompt word to obtain a prompt word text; performing entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and based on a sensitive word library associated with the service type, removing sensitive words in the first entity word to obtain a compliant entity word; based on the compliant entity word, using a content data generation model corresponding to the description scheme to generate content data of the description scheme. In the present disclosure, after obtaining the first entity word of the prompt word text, it is not directly input into the corresponding content data generation model, but firstly removes sensitive words from it to obtain compliant entity words, and then inputs the compliant entity words into the corresponding content data generation model, thereby improving the security of the content data output by the content data generation model. Through the method disclosed in the present disclosure, it is no longer necessary to fine-tune the content data generation model and can also output safe content data that does not include sensitive words, thereby improving the efficiency and security of content data generation.

[0040] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0042] Figure 1 A flow chart showing a method for generating content data of a 5G message according to an embodiment of the present disclosure is shown;

[0043] Figure 2 A schematic block diagram showing a device for generating content data of a 5G message according to an embodiment of the present disclosure is shown;

[0044] Figure 3 A schematic block diagram of a device for generating content data for a 5G message according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0045] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0046] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0047] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0048] <Method Example>

[0049] Figure 1 A flowchart of a method for generating content data of a 5G message according to an embodiment of the present disclosure is shown. Figure 1 As shown, the method includes steps S1100-S1400.

[0050] S1100, obtain the description scheme and business type of the content data of the 5G message. Among them, the description scheme is used to clarify the subject, core points and type of the content data, and the type of the content data may include at least one of text and image. The business type is used to reflect the business field to which the content data is applied. For example, when the business field to which the content data is applied is the financial field, the business type can be set to the financial field. For another example, when the business field to which the content data is applied is the cultural and tourism field, the business type can be set to the cultural and tourism field.

[0051] In one possible implementation, the system provides a front-end display interface through which the user can input the description scheme and business type of the content data. After completing the input of the description scheme and business type, click Submit, and the system can obtain the description scheme and business type of the content data.

[0052] S1200, generating prompt words for describing the scheme, and combining the description scheme and the prompt words to obtain prompt word text, wherein the prompt words for describing the scheme are used to guide the content data generation model to generate content data that meets user needs.

[0053] Specifically, when generating prompt words for describing a solution, the description solution may be directly input into a Prompt prompt word module, so that prompt words for describing the solution are automatically generated through the Prompt prompt word module.

[0054] After generating the prompt words of the description scheme, the description scheme and the prompt of the description scheme can be combined, and non-semantic characters such as punctuation marks can be removed to obtain the prompt word text.

[0055] S1300, performing named entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and based on a sensitive word library associated with the business type, removing sensitive words from the first entity word to obtain a compliant entity word. Specifically, the prompt word text may be subjected to named entity recognition using a BERT model to obtain all first entity words in the prompt word text.

[0056] After all the first entity words are obtained, sensitive words in the first entity words can be removed based on the sensitive word library associated with the business type, so as to obtain compliant entity words.

[0057] First of all, it should be noted that the system has created a corresponding sensitive word library for each business type in advance, and established an association relationship between the business type and the corresponding sensitive word library. In this way, after obtaining the business type, the sensitive word library associated with the business type can be determined based on the association relationship, and then sensitive words can be removed based on the sensitive word library associated with the business type. Among them, the sensitive word library of each business type includes sensitive words that may appear in each business scenario.

[0058] In a possible implementation, when sensitive words in the first entity word are removed based on a sensitive word library associated with the business type to obtain a compliant entity word, the following steps may be included:

[0059] First, traverse each first entity word.

[0060] Second, for the current first entity word traversed, determine whether there is a sensitive word similar to the current first entity word in the sensitive word library associated with the business type. If it is determined that there is a sensitive word similar to the current first entity word, the current first entity word is eliminated as a sensitive word. If it is determined that there is no sensitive word similar to the current first entity word, traverse the next first entity word.

[0061] Third, after the traversal is completed, the remaining first entity words are the compliant entity words.

[0062] In a possible implementation, when it is determined that there is no sensitive word similar to the current first entity word, the following steps may also be included:

[0063] Obtain a synonym for the current first entity word; determine whether there is a sensitive word similar to the synonym of the current first entity word in the sensitive word library associated with the business type; if it is determined that there is a sensitive word similar to the synonym of the current first entity word, remove the current first entity word as a sensitive word; if it is determined that there is no sensitive word similar to the synonym of the current first entity word, perform an operation of traversing the next first entity word; after the traversal is completed, take all the remaining first entity words as compliant entity words.

[0064] In another possible implementation, when there is a sensitive word similar to the current first entity word or a synonym of the first entity word in the sensitive word library associated with the business type, the operation of removing the current first entity word is not performed, but the current first entity word is replaced with a specified character according to a preset replacement rule, thereby realizing the hiding of sensitive information. The specified character can be a set fixed character such as "*", or a manually configured word set according to the user's needs, which will not be repeated here.

[0065] S1400, based on the compliant entity words, using a content data generation model corresponding to the description scheme to generate content data for the description scheme.

[0066] It should be noted that the system is provided with different content data generation models for different types of content data. For example, when the types of content data include text and image, the system will be provided with a content data generation model for generating text content and a content data generation model for generating image content. Among them, the content data generation model for generating text content can be a chatglm3 model, and the content data generation model for generating image content can be an sdxl model.

[0067] After obtaining the compliant entity words corresponding to the description scheme, the type of content data included in the description scheme will be parsed, and the content data generation model corresponding to the description scheme will be called according to the type of content data. The compliant entity words corresponding to the description scheme will be input into the corresponding content data generation model, thereby generating the content data of the description scheme through the corresponding content data generation model.

[0068] For example, if the type of the content data in the parsed description scheme is text, the chatglm3 model can be called, and the compliant entity words corresponding to the description scheme can be input into the chatglm3 model to generate the text content corresponding to the description scheme through the chatglm3 model.

[0069] For example, if the type of content data in the parsed description scheme is image, the sdxl model can be called, and the compliant entity words corresponding to the description scheme can be input into the sdxl model to generate the image content corresponding to the description scheme through the sdxl model.

[0070] In a possible implementation, in order to improve the accuracy of content data generation, when generating content data of the description scheme based on the compliant entity words and using a content data generation model corresponding to the description scheme, the following is further included:

[0071] First, obtain the qualifiers associated with the compliant entity words. The qualifiers are used to limit the generation rules of the content data. For example, to generate a poster with a national flag, the qualifiers can be "the national flag cannot be damaged" or "the national flag cannot have stains".

[0072] It should be noted here that the system has a pre-mapped relationship between various compliant entity words and qualifier sets. In this way, after obtaining the compliant entity words, the qualifier set corresponding to each compliant word can be found based on the mapping relationship between it and the qualifier set. The qualifier set includes multiple qualifiers set for the compliant words. When the qualifier set corresponding to the compliant entity word is an empty set, no qualifiers are added to the current compliant entity word; when the qualifier set corresponding to the compliant entity word is a non-empty set, all qualifiers in the corresponding qualifier set are added to the current compliant entity word.

[0073] Second, the compliance entity words and qualifiers are input into the content data generation model associated with the business type to generate content data describing the solution. The specific steps are as mentioned above and will not be described in detail here.

[0074] In a possible implementation, in order to further improve the security of generated content data, after generating the content data describing the scheme, the following steps are further included:

[0075] First, calculate the compliance confidence of the content data. The specific steps are as follows:

[0076] First, entity recognition is performed on the text content data to obtain the second entity word in the text content data. The specific process is referred to the step of performing named entity recognition on the prompt word text, which will not be repeated here.

[0077] Secondly, traverse each second entity word. For the current second entity word traversed, select sensitive words similar to the current second entity word from the sensitive word library associated with the business type as the target sensitive words corresponding to the current second entity word, and obtain the weight of the target sensitive word. After the traversal is completed, the target sensitive words corresponding to each second entity word and the weight of the target sensitive word are obtained.

[0078] Specifically, for each second entity word, the similarity between the second entity word and each sensitive word in the sensitive word library associated with the business type is calculated, and the sensitive word with a similarity greater than a threshold and a maximum similarity is selected as the target sensitive word of the second entity word. Each sensitive word in the sensitive word library is pre-set with a corresponding sensitive word weight, so that after the target sensitive word is determined, the weight of the target sensitive word can be obtained at the same time. Referring to this step, the target sensitive word corresponding to each second entity word and the weight of the target sensitive word can be obtained.

[0079] Finally, based on the target sensitive words corresponding to each second entity word and the weight of the target sensitive words, the compliance confidence of the text content data is calculated. The specific calculation formula is as follows:

[0080]

[0081] In the formula, X is the combination of the feature vectors of each second entity word, and Y is the combination of the feature vectors of each target sensitive word. i is the feature vector of the second entity word of the i-th entity in X, y i is the feature vector of the i-th target sensitive word in Y, where the i-th entity word in X corresponds to the i-th target sensitive word in Y. For example, the target sensitive word corresponding to the second entity word in X is the second one in Y. S(x i ,y i ) is the similarity between the i-th second entity word and the i-th target sensitive word, W(y i ) is the weight of the i-th target sensitive word in Y. It should be noted that the weight of each sensitive word can be configured according to the specific situation, and its value range can be [0~0.2].

[0082] The compliance confidence of the text content data can be calculated using the above calculation formula.

[0083] For image content data, when calculating its compliance confidence, it is necessary to first perform multi-target detection on the image through the YOLOv8 model, generate text labels for each detection target, perform entity recognition on the text labels of each detection target, obtain the entity words corresponding to each text label, and use the entity words corresponding to each text label as the second entity words corresponding to the image content data. Furthermore, after obtaining the second entity words corresponding to the image content data, the corresponding target sensitive words and the weights of the target sensitive words can be obtained by referring to the above text, and the compliance confidence of the image content data can be calculated.

[0084] Second, determine whether the calculated compliance confidence is greater than or equal to the preset confidence threshold. When it is determined that the compliance confidence is greater than or equal to the confidence threshold, the model-generated content data is pushed for display. The confidence threshold can be set according to the specific situation. For example, the confidence threshold can be set to 0.98. When it is determined that the compliance confidence is less than the confidence threshold, a prompt message indicating that the content data is non-compliant is pushed. The prompt message indicating that the content data is non-compliant can be set according to specific needs. For example, it can be set to "The content data does not meet the business security requirements. Please correct the description plan."

[0085] In a possible implementation, sensitive words eliminated during the content generation process will also be stored, and when a prompt message indicating that the content data is not compliant is pushed, the proposed sensitive words will be pushed and displayed to facilitate the user to make subsequent adjustments to the description plan.

[0086] It should be noted that when conducting compliance testing, it is necessary to manually enter the detection rules in advance, including the sensitive word library, sensitive word detection rules, extended prompt words, and association strategy rules used in the above-mentioned compliance testing.

[0087] In order to clearly illustrate the method of the present invention, the method of the present invention is further illustrated below in conjunction with a specific example. In this example, the step of generating content data includes:

[0088] Step 1: Manually enter the detection rules in advance, including:

[0089] Sensitive word library: establish multiple sensitive word libraries based on business type association. Each sensitive word library contains sensitive words that may appear in various business scenarios and sets corresponding weights for each sensitive word.

[0090] The association strategy, the mapping relationship between each compliant entity word and the qualifier set, can be expressed as:

[0091] MAP(X)=Y

[0092] Among them, X is the set of all compliant entity words, and Y is the set of qualifiers. In other words, a compliant entity word in set X can find a corresponding set of qualifiers. When Y is an empty set Φ, no qualifier processing is performed; when Y is a non-empty set, all qualifiers in the corresponding set Y need to be added to the compliant entity word.

[0093] Detection rules are the processing rules after sensitive words are detected, including:

[0094] Rule 1: Tolerance rule, sets the tolerance threshold for detecting negative sensitive words, and the tolerance threshold is set to 0.98.

[0095] Rule 2: Elimination rule: when sensitive words are detected in the generated content data, sensitive words are eliminated.

[0096] Rule 3: Replacement rule. When sensitive words are detected in the generated content data, the sensitive words are replaced with "*" or manually configured words.

[0097] Step 2: The user enters the description scheme and business type of the content data through the front-end display interface provided by the system and clicks Submit. The description scheme and business type of the content data received by the system are taken as parameter 1. Parameter 1 is passed to the Prompt prompt word module, which generates a prompt word for the description scheme, and combines parameter 1 and the prompt word to remove punctuation marks and other non-semantic characters as parameter 2, which is then transmitted to the compliance detection module for compliance detection.

[0098] Step 3: The compliance detection module receives parameter 2, performs named entity recognition on parameter 2, and obtains all first entity words. Based on the sensitive word library associated with the business type, it searches for sensitive words similar to the first entity words, removes the entity words in the first entity words, and the remaining entity words are compliant entity words. The compliant entity words are obtained as parameter 3, and the removed first entity words are used as parameter 4.

[0099] When removing sensitive words from the first entity words in the sensitive word library associated with the business type to obtain compliant entity words, the first entity word can be searched in the corresponding sensitive word libraries according to the business type, and the first entity words corresponding to the found sensitive words can be removed to obtain compliant entity words.

[0100] The first entity word can also be searched for corresponding sensitive word libraries according to the business type, and synonyms of the first entity word that cannot be found in the corresponding sensitive word library are identified. After synonym identification, the corresponding sensitive word library is searched again, and the first entity word that has been found to be a corresponding sensitive word is removed to obtain a compliant entity word.

[0101] After searching the corresponding sensitive word library for the first entity word, the sensitive words found are not removed, but the first entity word is replaced with specified characters according to a preset replacement rule, thereby hiding the sensitive information.

[0102] It should be noted that the method for obtaining the compliant entity word parameter 3 is not limited.

[0103] Step 4: Parameter 3 can be directly input into the model inference module. After receiving parameter 3, the model inference module parses the type of content data included in the content data description scheme of parameter 1, and calls different models for inference generation according to the type of content data. Text data calls the chatglm3 model to generate text content data, and image data calls the sdxl model to generate image content data. The generated content data is used as parameter 5 to further test the security of parameter 5.

[0104] You can also add associated qualifiers to parameter 3 and input them into the model reasoning module, then parse the type of content data contained in the content data description scheme in parameter 1, and call different models to generate content data according to the type of content data. The specific content data generation method is the same as above, so I will not go into details here.

[0105] Step 5: Calculate compliance confidence based on the generated data type.

[0106] When the generated content data includes text data, parameter 5 is subjected to named entity recognition to obtain each second entity word. For each second entity word, the similarity between the second entity word and each sensitive word in the sensitive word library associated with the business type is calculated, and the sensitive words with a similarity greater than a threshold and a maximum similarity are screened out as the target sensitive words of the second entity word, and the weight of the target sensitive word is obtained. Finally, the target sensitive words corresponding to each second entity word and the weight of the target sensitive word are obtained. The confidence calculation formula is used to calculate the compliance confidence of the generated content data. The specific calculation formula is as follows:

[0107]

[0108] In the formula, X is the combination of the feature vectors of each second entity word, and Y is the combination of the feature vectors of each target sensitive word. i is the feature vector of the second entity word of the i-th entity in X, y i is the feature vector of the i-th target sensitive word in Y, where the i-th entity word in X corresponds to the i-th target sensitive word in Y. For example, the target sensitive word corresponding to the second entity word in X is the second one in Y, S(x i ,y i ) is the similarity between the i-th second entity word and the i-th target sensitive word, W(y i ) is the weight of the i-th target sensitive word in Y. The weight of each sensitive word can be configured according to the specific situation, and its value range can be [0~0.2].

[0109] When the generated content data includes image data, parameter 5 uses the YOLOv8 model to perform multi-target detection on the image to generate text labels, generate text labels for each detection target, perform entity recognition on the text labels of each detection target, obtain the entity words corresponding to each text label, and use the entity words corresponding to each text label as the second entity words corresponding to the image content data. Obtain the second entity words. Refer to the above to obtain the corresponding target sensitive words and the weights of the target sensitive words for each second entity word, and calculate the compliance confidence of the image content data.

[0110] Step 6: If the calculated confidence level is greater than or equal to 0.98, parameter 5 is output, and the content data of parameter 5 is pushed and displayed.

[0111] Step 7: When the calculated compliance confidence is less than 0.98, push a prompt message indicating that the content data is not compliant. It should be noted that the prompts related to content data non-compliance can be set according to actual needs, for example, it can be set to "the content data does not meet the business security requirements, please correct the description plan."

[0112] When pushing and displaying, the sensitive words removed in parameter 4 can be pushed and displayed to remind the user to adjust the description plan.

[0113] Therefore, the present disclosure provides a method for generating content data of a 5G message, the method comprising: obtaining a description scheme and a service type of the content data of the 5G message; generating a prompt word of the description scheme, and combining the description scheme and the prompt word to obtain a prompt word text; performing entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and based on a sensitive word library associated with the service type, removing sensitive words in the first entity word to obtain a compliant entity word; based on the compliant entity word, using a content data generation model corresponding to the description scheme to generate content data of the description scheme. Based on the above content data generation method, when generating corresponding content data for the 5G message service, compliance detection is performed on the generated content data, so that secure content data can be efficiently output through a general large model, thereby improving the security and accuracy of the generated content data, and at the same time, there is no need to fine-tune the general large model for training, saving computing resources for data set preprocessing and training.

[0114] <Device Example>

[0115] The present disclosure provides a 5G message content data generation device 100, referring to Figure 2 , including: a foreground display module 110, a prompt module 120, a compliance detection module 130, and a model reasoning module 140. The foreground display module 110 is used to obtain the description scheme and business type of the content data of the 5G message; the prompt word module 120 is used to generate prompt words for the description scheme, and combine the description scheme and the prompt words to obtain prompt word text; the compliance detection module 130 is used to perform entity recognition on the prompt word text to obtain the first entity word in the prompt word text, and based on the sensitive word library associated with the business type, remove the sensitive words in the first entity word to obtain the compliant entity word; the model reasoning module 140 is used to generate the content data of the description scheme based on the compliant entity word and adopt the content data generation model corresponding to the description scheme.

[0116] <Equipment Embodiment>

[0117] Furthermore, according to another aspect of the present disclosure, a 5G message content data generating device 200 is also provided. Figure 3The content data generating device 200 of the 5G message of the embodiment of the present disclosure includes a processor 210 and a memory 220 for storing executable instructions of the processor 210. The processor 210 is configured to implement any of the above-mentioned methods for generating content data of the 5G message when executing the executable instructions.

[0118] Here, it should be noted that the number of processors 210 may be one or more. At the same time, in the 5G message content data generation device 200 of the embodiment of the present disclosure, an input device 230 and an output device 240 may also be included. Among them, the processor 210, the memory 220, the input device 230 and the output device 240 may be connected through a bus or in other ways, which are not specifically limited here.

[0119] The memory 220, as a computer-readable storage medium, can be used to store software programs, computer executable programs, and various modules, such as: programs or modules corresponding to the content data generation method of the 5G message of the embodiment of the present disclosure. The processor 210 executes various functional applications and data processing of the content data generation device 200 of the 5G message by running the software programs or modules stored in the memory 220.

[0120] The input device 230 may be used to receive input numbers or signals. The signals may be key signals related to user settings and function control of the device / terminal / server. The output device 240 may include display devices such as display screens.

[0121] <Storage Medium Embodiment>

[0122] According to another aspect of the present disclosure, a non-volatile computer-readable storage medium is also provided, on which computer program instructions are stored. When the computer program instructions are executed by the processor 210, any of the above-mentioned methods for generating content data of 5G messages is implemented.

[0123] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for generating content data of a 5G message, characterized in that: include: Obtain a description scheme and service type of content data of the 5G message; Generate prompt words for the description scheme, and combine the description scheme and the prompt words to obtain prompt word text; Performing entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and removing sensitive words from the first entity word based on a sensitive word library associated with the business type to obtain a compliant entity word; Based on the compliant entity words, a content data generation model corresponding to the description scheme is adopted to generate content data of the description scheme.

2. The method according to claim 1, characterized in that When the sensitive words in the first entity words are removed based on the sensitive word library associated with the business type to obtain the compliant entity words, the method includes: Traversing each of the first entity words; For the traversed current first entity word, determine whether there is a sensitive word similar to the current first entity word in the sensitive word library associated with the business type; if it is determined that there is a sensitive word similar to the current first entity word, remove the current first entity word as a sensitive word; if it is determined that there is no sensitive word similar to the current first entity word, traverse the next first entity word; After the traversal is completed, the compliant entity word is obtained.

3. The method according to claim 2, characterized in that In the case where it is determined that there is no sensitive word similar to the current first entity word, the method further includes: Obtaining a synonym of the current first entity word; Determine whether there is a sensitive word similar to the synonym of the current first entity word in the sensitive word library associated with the business type. If it is determined that there is a sensitive word similar to the synonym of the current first entity word, remove the current first entity word as a sensitive word. If it is determined that there is no sensitive word similar to the synonym of the current first entity word, perform an operation of traversing the next first entity word.

4. The method according to claim 1, characterized in that: When the content data of the description scheme is generated based on the compliant entity words and using a content data generation model corresponding to the description scheme, the method includes: Acquire a qualifier associated with the compliant entity word, wherein the qualifier is used to define a generation rule of the content data; The compliant entity words and the qualifiers are input into a content data generation model corresponding to the description scheme to generate content data of the description scheme.

5. The method according to claim 1, characterized in that: After generating the content data describing the scheme, the method further comprises: Calculating the compliance confidence level of the content data; Determining whether the compliance confidence level is greater than or equal to a preset confidence level threshold; When it is determined that the compliance confidence is greater than or equal to the confidence threshold, the content data is pushed and displayed.

6. The method according to claim 5, characterized in that: When the content data includes text content data and the compliance confidence of the text content data is calculated, the method includes: Performing entity recognition on the text content data to obtain a second entity word in the text content data; Traversing each of the second entity words; For the traversed current second entity word, a sensitive word similar to the current second entity word is screened out from a sensitive word library associated with the business type as a target sensitive word corresponding to the current second entity word, and a weight of the target sensitive word is obtained; After the traversal is completed, the target sensitive words corresponding to each of the second entity words and the weights of the target sensitive words are obtained; The compliance confidence of the text content data is calculated based on the target sensitive words corresponding to each of the second entity words and the weights of the target sensitive words.

7. The method according to claim 5, characterized in that: When it is determined that the compliance confidence is less than the confidence threshold, prompt information indicating that the content data is non-compliant is pushed.

8. A device for generating content data of a 5G message, characterized in that: include: A foreground presentation module, used to obtain a description scheme and a service type of the content data of the 5G message; A prompt word module, used to generate a prompt word for the description scheme, and combine the description scheme and the prompt word to obtain a prompt word text; A compliance detection module, configured to perform entity recognition on the prompt word text to obtain a first entity word in the prompt word text, and remove sensitive words from the first entity word based on a sensitive word library associated with the business type to obtain a compliant entity word; A model reasoning module is used to generate content data of the description scheme based on the compliant entity words and using a content data generation model corresponding to the description scheme.

9. A device for generating content data of a 5G message, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the method described in any one of claims 1 to 7 when executing the executable instructions.

10. A non-volatile computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images