Advertisement playing task putting method, device and equipment

Abstract

The invention discloses an advertisement playing task putting method, device and equipment, and the method comprises the steps: extracting advertisement demand information from natural language data through a natural language processing model when the natural language data for an advertisement playing task is received; generating a system operation instruction according to the advertisement demand information and a system instruction rule of the information management publishing system; optimizing the system operation instruction to obtain at least one target operation instruction; verifying each target operation instruction to obtain a verification result; and if the verification result is passed, calling a system function interface of the information management publishing system according to each target operation instruction, generating a target advertisement and putting the target advertisement to the target equipment. Therefore, the advertisement generation process is simplified in combination with the natural language processing model, so that the user can still complete advertisement generation and delivery under the condition of low familiarity of system operation steps, and the interaction efficiency and the operation portability of the information management and release system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement delivery, and in particular to a delivery method, device and equipment for an advertisement playing task. Background Art

[0002] With the rapid development of Internet technology, advertising delivery methods have gradually shifted from traditional offline delivery methods to online delivery methods, such as delivery through various applications and websites. In the increasingly fierce competition, online advertising needs to flexibly adjust advertising content according to Internet media content to meet different customer needs.

[0003] To this end, the information management and publishing system came into being, which provides enterprises with a variety of options for advertising through human-computer interactive editing. Traditional advertising solutions usually combine human-computer interaction with input devices such as keyboards and mice based on product function design, requiring users to have high requirements for the operation steps and business understanding of the advertising delivery system, and the operation process is cumbersome, making it difficult to achieve efficient advertising delivery. Summary of the invention

[0004] The present invention provides a method, device and equipment for delivering advertisement playing tasks, which solves the technical problem that the existing advertisement delivery solutions are usually designed according to product functions and combined with input devices such as keyboards and mice for human-computer interaction, and the users have high requirements for the operation steps and business understanding of the advertisement delivery system, and the operation process is cumbersome, making it difficult to achieve efficient advertisement delivery.

[0005] A first aspect of the present invention provides a method for delivering an advertisement playing task, comprising:

[0006] When natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data by using a natural language processing model;

[0007] Generate system operation instructions according to the advertising demand information and the system instruction rules of the information management and publishing system;

[0008] Optimizing the system operation instruction to obtain at least one target operation instruction;

[0009] Verifying each of the target operation instructions to obtain a verification result;

[0010] If the verification result is passed, the system function interface of the information management and publishing system is called according to each of the target operation instructions to generate a target advertisement and deliver it to the target device.

[0011] Optionally, the method further comprises:

[0012] Obtain multiple natural language text data;

[0013] Performing data cleaning and preprocessing on the various natural language text data to obtain training text data;

[0014] The training text data and the preset loss function are used to train the preset neural network model to generate an intermediate network model for identifying user intentions;

[0015] Performing model evaluation on the intermediate network model according to multiple evaluation indicators to generate evaluation results;

[0016] If the evaluation result does not meet the preset evaluation threshold, the model parameters of the intermediate network model are adjusted according to the evaluation result until the evaluation result meets the preset evaluation threshold, thereby obtaining a natural language processing model.

[0017] Optionally, when receiving the natural language data for the advertisement playing task, extracting the advertisement demand information from the natural language data by using a natural language processing model includes:

[0018] When natural language data for an advertisement playing task is received, the natural language data is input into a preset natural language processing model;

[0019] Cleaning the natural language data using the natural language processing model and dividing the natural language data into a plurality of natural language words;

[0020] Screening out stop words from the plurality of natural language words by the natural language processing model to obtain a plurality of keywords;

[0021] The keywords are classified by the natural language processing model to obtain operation targets, operation types and operation actions as advertising demand information.

[0022] Optionally, generating a system operation instruction according to the advertisement demand information and a system instruction rule of the information management and publishing system includes:

[0023] Creating an instruction template corresponding to the operation type according to the system instruction rules of the information management and publishing system;

[0024] The operation target and the operation action are respectively filled into the variable positions of the instruction template to generate a system operation instruction.

[0025] Optionally, the optimizing the system operation instruction to obtain at least one target operation instruction includes:

[0026] Determine the required resource type and resource requirement according to the operation type and operation object scale of the system operation instruction;

[0027] According to the required resource type, obtain the currently available resource amount of the corresponding type;

[0028] If the currently available resource amount is less than or equal to the resource demand amount, splitting the system operation instruction into multiple target operation instructions, or optimizing the system operation instruction into target operation instructions by using an optimization algorithm that conforms to the operation type;

[0029] If the currently available resource amount is greater than the resource requirement amount, the system operation instruction is determined as the target operation instruction.

[0030] Optionally, the verifying each of the target operation instructions to obtain a verification result includes:

[0031] Performing a security policy check on each of the target operation instructions to obtain a first check result;

[0032] Performing syntax analysis on each of the target operation instructions to obtain a second inspection result;

[0033] Performing a format compliance check on each of the target operation instructions to obtain a third check result;

[0034] If the first inspection result, the second inspection result and the third inspection result are all passed, generating a verification result of passed;

[0035] If the first inspection result, the second inspection result or the third inspection result is failed, a verification result is generated as failed.

[0036] Optionally, if the verification result is passed, calling the system function interface of the information management and publishing system according to each target operation instruction, generating a target advertisement and delivering it to a target device, includes:

[0037] If the verification result is passed, each of the target operation instructions is cached in a message queue;

[0038] Executing each of the target operation instructions in sequence from the message queue according to the cache order, so as to call the system function interface of the information management and publishing system to create a target advertisement that meets the advertisement playing task;

[0039] Calling the system function interface to deliver the target advertisement to the target device;

[0040] The target operation instructions include material acquisition instructions, material creation instructions, advertising image combination instructions, play list creation instructions, play duration creation instructions and play plan creation instructions.

[0041] Optionally, the method further comprises:

[0042] If the verification result is not passed, the error information in the verification result is returned and a re-entry prompt message is displayed.

[0043] The second aspect of the present invention further provides a device for delivering an advertisement playing task, comprising:

[0044] A model calling module, for extracting advertisement demand information from the natural language data through a natural language processing model when receiving natural language data for an advertisement playing task;

[0045] An instruction generation module, used to generate system operation instructions according to the advertising demand information and the system instruction rules of the information management and publishing system;

[0046] An instruction optimization module, used for optimizing the system operation instruction to obtain at least one target operation instruction;

[0047] An instruction verification module is used to verify each of the target operation instructions to obtain a verification result;

[0048] The instruction execution module is used to call the system function interface of the information management and publishing system according to each of the target operation instructions if the verification result is passed, generate a target advertisement and deliver it to the target device.

[0049] The third aspect of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for delivering an advertisement playback task as described in any one of the first aspect of the present invention.

[0050] It can be seen from the above technical solutions that the present invention has the following advantages:

[0051] When natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data through a natural language processing model; system operation instructions are generated according to the advertisement demand information and the system instruction rules of the information management and publishing system; the system operation instructions are optimized to obtain at least one target operation instruction; each target operation instruction is verified to obtain a verification result; if the verification result is passed, the system function interface of the information management and publishing system is called according to each target operation instruction to generate a target advertisement and deliver it to the target device. Thus, the advertisement generation process is simplified by combining the natural language processing model, so that users can still complete advertisement generation and delivery even if they are not familiar with the system operation steps, thereby improving the interactive efficiency and operational portability of the information management and publishing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 A flowchart of a method for delivering an advertisement playing task provided by an embodiment of the present invention;

[0054] Figure 2 A flowchart of the detailed steps of a method for delivering an advertisement playing task provided by an embodiment of the present invention;

[0055] Figure 3 A specific implementation flow chart of a method for delivering an advertisement playing task in an embodiment of the present invention;

[0056] Figure 4 The structure block diagram of a device for delivering advertisement playing tasks in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In the prior art, there are a wide variety of products related to information management and publishing systems. Most of these products need to be operated step by step according to the product function design to achieve the final effect of advertising to device playback: from element material production → material upload → material combination into finished advertising → create playlist → create play plan → select device → send play plan to the advertising screen, so that the specified advertisement can be played according to the preset cycle or plan. However, this human-computer interaction method mainly relies on input devices such as keyboards and mice, and the operation is relatively cumbersome, causing users to have higher requirements for the operation steps and business understanding of the information distribution system.

[0058] To this end, the embodiments of the present invention provide a method, device and equipment for delivering advertising playback tasks, which are used to solve the technical problem that the existing advertising delivery solutions are usually based on product function design combined with input devices such as keyboards and mice for human-computer interaction, which requires users to have high requirements for the operation steps and business understanding of the advertising delivery system, and the operation process is cumbersome, making it difficult to achieve efficient advertising delivery. Through a natural language human-computer dialogue method, the process from material production to advertising delivery to a specified device is solved, and the specified advertisement can be played according to a preset playback plan.

[0059] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1 , Figure 1 A flowchart of a method for delivering an advertisement playing task provided by an embodiment of the present invention.

[0061] The present invention provides a method for delivering an advertisement playing task, comprising:

[0062] Step 101, when natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data through a natural language processing model;

[0063] Natural language data refers to data input through text or voice, and in this embodiment is used to express the advertisement content, play list, play plan, and delivery device that need to be generated in the advertisement play task.

[0064] The natural language processing model refers to a computational model used to understand and analyze the natural language input by the user and generate the user's intention. The computational model can be created based on neural network models such as recurrent neural network (RNN) and long short-term memory network (LSTM).

[0065] In an embodiment of the present invention, the delivery method can be implemented by a data processor with a modular structure. When it receives natural language data, it calls a natural language processing model to perform semantic analysis on the natural language, and extracts multiple keywords or key words that match the system instruction rules of the information management and publishing system as the data basis for the subsequent generation of system operation instructions.

[0066] Step 102, generating system operation instructions according to the advertisement demand information and the system instruction rules of the information management and publishing system;

[0067] The information management and publishing system is a comprehensive platform for collecting, storing, organizing, editing, reviewing, publishing and subsequent management of various information. In this embodiment, it specifically refers to a business platform that can collect a variety of existing element materials, create element materials, synthesize advertisements and place advertisements. It has a system instruction rule dedicated to instruction generation. After obtaining the advertising demand information corresponding to the natural language data, based on the instruction content and preconditions defined by the system instruction rule, the advertising demand information is converted and formatted by template matching or rule reasoning, thereby generating system operation instructions.

[0068] Since information management and publishing systems usually have a variety of different settings such as architectures, functions and operation execution rules, in order to facilitate the subsequent generation of target advertisements for various information management and publishing systems, after obtaining the advertising demand information, the information is matched with the system instruction rules of the information management and publishing system. For example, the advertising demand information is matched with the instruction rules of "operation target + operation type + operation action" to generate system operation instructions corresponding to each step.

[0069] The system operation instruction may be single or multiple, and the specific number is determined according to the advertising demand information generated by the natural language data processing model.

[0070] Step 103, optimizing the system operation instruction to obtain at least one target operation instruction;

[0071] Since the information publishing management system may not only perform a single task, and the resources of the server where the system is running are not unlimited, before generating the target advertisement, the system operation instructions can be simplified or optimized according to the resource usage of the information publishing management system to improve the execution efficiency of subsequent instructions and obtain at least one target operation instruction.

[0072] Step 104, verify each target operation instruction to obtain a verification result;

[0073] After obtaining the target operation instructions, in order to improve the stability and execution reliability of the information release management system, each target operation instruction is verified through syntax analysis, rule matching, etc. to check the grammatical correctness, format standardization and compliance with the system security policy of the target operation instructions, and generate verification results based on the comprehensive results of various checks.

[0074] Step 105: If the verification result is passed, the system function interface of the information management and publishing system is called according to each target operation instruction, and the target advertisement is generated and delivered to the target device.

[0075] If the verification result is passed, it means that the target operation instruction can be executed at this time. Each target operation instruction can be sent in turn to call the system function interface of the information management and publishing system to perform the corresponding operation through the information management and publishing system, generate the target advertisement corresponding to the natural language data and deliver it to the target device.

[0076] The target advertisement includes but is not limited to at least one picture and video, and is delivered to the target device according to the play list and play plan set by the natural language data, so as to perform the planned play of the target advertisement on the target device.

[0077] In an embodiment of the present invention, when natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data through a natural language processing model; system operation instructions are generated according to the advertisement demand information and the system instruction rules of the information management and publishing system; the system operation instructions are optimized to obtain at least one target operation instruction; each target operation instruction is verified to obtain a verification result; if the verification result is passed, the system function interface of the information management and publishing system is called according to each target operation instruction to generate a target advertisement and deliver it to the target device. Thus, the advertisement generation process is simplified by combining the natural language processing model, so that users can still complete advertisement generation and delivery even if they are not familiar with the system operation steps, thereby improving the interactive efficiency and operation portability of the information management and publishing system.

[0078] See also Figure 2 , Figure 2 This is a flowchart of detailed steps of a method for delivering an advertisement playing task in an embodiment of the present invention, wherein step 103 is detailed.

[0079] An embodiment of the present invention provides a method for delivering an advertisement playing task, comprising:

[0080] Step 201, when natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data through a natural language processing model;

[0081] In an example of the present invention, step 201 may include the following sub-steps:

[0082] When receiving natural language data for an advertisement playing task, inputting the natural language data into a preset natural language processing model;

[0083] The natural language data is cleaned and divided into multiple natural language words through the natural language processing model;

[0084] Using a natural language processing model, stop words are removed from multiple natural language words to obtain multiple keywords;

[0085] Classify keywords through a natural language processing model to obtain the operation target, operation type, and operation action as advertisement demand information.

[0086] In this embodiment, an input dialog box or a voice dialog box can be displayed on the visualization interface. When natural language data for an advertisement playback task is received from the user or an external source in the dialog box, the natural language data is input into the natural language processing model for analysis and understanding. The natural language processing model performs data cleaning on the natural language data to remove noise information such as extra spaces, line breaks, and special characters in the natural language data, and then uses a word segmentation tool to divide the cleaned natural language data into multiple natural language words. Stop words such as "de", "shi", "zai", "le", etc. are screened out from the multiple natural language words to reduce the data volume and highlight the key information of the natural language data, obtaining multiple keywords, and the keywords can be sorted and output according to the original word order.

[0087] After the extraction of keywords is completed, the natural language processing model classifies the keywords, performs language analysis on the keywords according to the internal structure and parameters to extract the user's intention, and finally obtains information such as the operation target, operation type, and operation action as advertisement demand information.

[0088] For example, the received natural language data is "I want to create a playback plan to play xxx advertisement from A year B month C day to D year E month F day". After keyword extraction by the natural language processing model, "create", "playback plan", and "play xxx advertisement from A year B month C day to D year E month F day" are obtained. At this time, the operation target is the playback plan, the operation type is create, and the operation action is to play the advertisement according to a specific time.

[0089] It should be noted that multiple types of keywords can be preset to judge the task type of the natural language data, such as "advertisement", "create", "promote", "launch", "play", "picture", and "material", etc. When the natural language data is received in the dialog box, first judge whether the input is completed. If the input is completed, retrieve whether the above types of keywords exist in the input natural language data. If they exist, it indicates that it meets the advertisement playback task, and the natural language processing model can be called for further semantic analysis.

[0090] In an example of the present invention, a neural network model can be used as the basic architecture, and combined with the collected natural language text data to train and generate a natural language processing model. The method further includes the following steps S11 - S15:

[0091] S11. Obtain various natural language text data;

[0092] The types of natural language text data include, but are not limited to, operation step keywords of the information management and publishing system, technical documents, user manuals, operation manuals, FAQ materials, etc. By acquiring a variety of natural language text data, a diverse training data foundation is provided for the natural language processing model.

[0093] S12, performing data cleaning and preprocessing on various natural language text data to obtain training text data;

[0094] In this embodiment, after obtaining a variety of natural language text data, since there may be many special characters in the natural language text data, such as punctuation marks (some punctuation marks may have little effect on semantic understanding, and whether to remove them can be determined according to the specific situation), mathematical symbols, tabs, etc., these special characters may belong to noise information in some cases. Data cleaning and preprocessing are performed on various types of natural language text data to remove the noise information therein, and preprocessing is performed to standardize the natural language text data, such as date format standardization and number format standardization, so as to obtain training text data.

[0095] S13, training a preset neural network model using training text data and a preset loss function to generate an intermediate network model for identifying user intent;

[0096] After obtaining a large amount of training text data, the training text data can be used to train preset neural network models such as recurrent neural networks (RNNs), long short-term memory networks (LSTMs), etc. During the training process, supervised learning or unsupervised learning methods can be used to select appropriate loss functions and optimization algorithms based on specific circumstances, optimize the parameters or structure of the neural network model, and generate an intermediate network model for identifying user intent and extracting advertising demand information.

[0097] S14, performing model evaluation on the intermediate network model according to multiple evaluation indicators to generate evaluation results;

[0098] S15. If the evaluation result does not meet the preset evaluation threshold, the model parameters of the intermediate network model are adjusted according to the evaluation result until the evaluation result meets the preset evaluation threshold, thereby obtaining a natural language processing model.

[0099] After completing the construction of the intermediate network model, the model is evaluated using indicators such as accuracy, recall, and F1 value to obtain evaluation results to determine whether the model is overfitting or underfitting. If the evaluation results do not meet the preset evaluation threshold, that is, the accuracy and F1 value continue to increase in the training set, but begin to decrease in the validation set, it indicates that overfitting has occurred; if the accuracy and F1 value are at a low level in both the training set and the validation set and there is no obvious improvement trend, it indicates that the model is underfitting.

[0100] At this time, the parameters and structure of the model can be adjusted according to the evaluation results, such as the learning rate of the intermediate network model, the regularization parameters, or the number of hidden layers, the number of neurons, the number of attention heads, etc. in the neural network to improve the performance of the model until the evaluation results meet the preset evaluation threshold and obtain the natural language processing model. Specifically, taking the learning rate as an example, if the model is overfitting, you can try to reduce the learning rate so that the model parameters are updated more slowly, thereby possibly avoiding overfitting the training set data. Conversely, if underfitting occurs, you can try to increase the learning rate and speed up the update of the model parameters so as to learn the patterns in the data more quickly.

[0101] In addition to adjusting the model parameters of the intermediate network model, the model structure can also be adjusted. If overfitting occurs, the model structure can be simplified, such as reducing the number of hidden layers of the neural network, removing some unnecessary features, etc. On the contrary, if the model is underfitting, the model structure can be complicated, such as increasing the number of hidden layers, adding new features, etc. If the performance of the model is still not ideal after multiple parameter adjustments, the model type can be changed.

[0102] Step 202, generating system operation instructions according to the advertisement demand information and the system instruction rules of the information management and publishing system;

[0103] In this embodiment, the system operation instructions can be generated in two ways: template matching and rule reasoning, so as to match the system instruction rules of different information relationship publishing systems.

[0104] In one example of the present invention, step 202 may include the following sub-steps:

[0105] Create an instruction template corresponding to the operation type according to the system instruction rules of the information management and publishing system;

[0106] Fill the operation target and operation action into the variable position of the instruction template respectively to generate the system operation instruction.

[0107] In this embodiment, after obtaining the advertisement demand information, an instruction template corresponding to the operation type is created according to the system instruction rules of the information management and publishing system. Specifically, in the information publishing and management system, the operation instruction may include the production of materials, the creation of a playlist, the setting of a play plan, and the sending of the play plan to the advertisement playback device after selecting the device, so the instruction rule obtained may be: "operation target + operation type + operation action".

[0108] Taking "play plan" as an example, there are operation functions and rules related to "play plan" in the information release system. :

[0109] a) Create a play schedule: The operation instruction format is: CREATE_SCHEDULE ['schedule_name']['play_list']['start_time']['end_time'].

[0110] Schedule_name: Playback schedule name

[0111] Play_list: Play advertisement list

[0112] Start_time: Playback start time

[0113] End_time: playback end time

[0114] b) Delete the play schedule: The operation instruction format is: DELETE_SCHEDULE['schedule_id']

[0115] Schedule_id: Playback schedule id

[0116] Parse the user's intention and extract key information and conditions. For example, if the user inputs "I want to create a play plan to play xxx advertisement from A year B month C day to D year E month F day", the system extracts the advertising demand information as operation target: play plan, operation type: create, operation action: play advertisement according to a specific time period, and the specific time period is A year B month C day to D year E month F day, and the advertisement to be played is xxx advertisement. Fill the advertising demand information into the above instruction template, and the generated system operation instruction is as follows:

[0117] CREATE_SCHEDULE ['Schedule 1']['xxx Advertising']['AB-C']['DE-F']

[0118] In the above example, through natural language processing model analysis, the user's intention to create a broadcast plan was accurately understood, and the corresponding advertising demand information was extracted. Then, according to the operation instruction format of the information management and publishing system, the extracted operation target, type, and action were filled into the corresponding positions, thereby generating system operation instructions that can be executed in this system.

[0119] In addition, step 202 may include the following sub-steps:

[0120] The current status of the test information management and publishing system;

[0121] Determine whether the information management and publishing system meets the execution conditions of the operation type in the current state;

[0122] If the conditions are met, the system instruction rules of the information management and publishing system are matched by the operation action and the operation target to generate a system operation instruction in the form of a logical expression or a conditional statement;

[0123] If not met, a preset error message is output.

[0124] In this embodiment, the system operation instruction may also be generated by rule reasoning, and the rule reasoning relies on a series of pre-established rules, which describe the logical relationship between the user intention and the operation instruction.

[0125] For example, the user intention is parsed to extract key information and conditions. For example, for the user input "I want to create a play plan to play xxx advertisements from A year B month C day to D year E month F day", the system extracts the advertising demand information as operation target: play plan, operation type: create, operation action: play advertisements according to a specific time period, and the specific time period is A year B month C day to D year E month F day, and the advertisement played is xxx advertisement. The parsed information is matched and inferred with the rules. In the inference process, some uncertainty and ambiguity may need to be handled. For example, if the user inputs "I want to create a play plan to play advertisements" and does not clearly give the specific advertisement content and play time, the system may need to further interact with the user, or determine the specific play time and content based on context information (such as the most recent play time and advertisement content), and then perform rule inference. Further judge whether the current state of the current information management and publishing system meets the execution conditions of the operation type, such as whether the target device has played advertisements during the period. If not, it is determined to be satisfied. At this time, the operation action and operation target can be used to match the system instruction rules of the information management and publishing system to generate system operation instructions in the form of logical expressions or conditional statements.

[0126] Step 203, determining the required resource type and resource requirement according to the operation type and operation object scale of the system operation instruction;

[0127] In this embodiment, after obtaining the system operation instruction, it is necessary to analyze the consumed resources to determine whether the information management and publishing system can execute the instruction. Specifically, according to the operation type and operation object scale of the system operation instruction, the system resources (such as memory, CPU time, disk I / O, etc.) required when executing the instruction can be determined, and the required resource type and resource demand can be obtained, which can be statistically calculated using the existing resource calculation algorithm.

[0128] For example, for advertising image combination instructions, which involve the acquisition, creation and combination of multiple materials, a large amount of disk I / O resources may be required. At this time, according to the operation type (generating advertising images) and the scale of the operation object (multiple materials), for data processing operations, if it involves caching of a large amount of data, storage of intermediate results, etc., a larger memory capacity may be required. For example, for data analysis of a large data set, it may be necessary to reserve hundreds of MB or even several GB of memory for storing data samples and intermediate results. For file system operations, when reading or writing large files, sufficient memory may be required to cache file contents to avoid frequent disk I / O. Generally, large file reading may require tens to hundreds of MB of memory cache. For network communication operations, when receiving a large number of network requests, sufficient memory may be required to cache the received request data. The specific capacity depends on the amount of request data and the number of requests, and may range from a few MB to tens of MB. Similarly, in data processing operations, complex data analysis, multi-table joint queries, etc. usually take a long time to complete, which may range from a few seconds to a few minutes or even longer, depending on the amount of data and the complexity of the algorithm. File system operations generally require relatively little CPU time, unless they involve complex parsing or processing of file contents, such as format conversion of large files, which may require several minutes of CPU time. Network communication operations may require more CPU time in high-concurrency situations, such as queuing and retrying a large number of network requests, which generally ranges from a few seconds to tens of seconds, depending on the number of requests and processing complexity.

[0129] Step 204, obtaining the currently available resource amount of the corresponding type according to the required resource type;

[0130] At the same time, the current available resource amount of the information management and publishing system is obtained according to the required resource type. For example, if it is a disk I / O resource, the currently allocatable disk I / O resource is obtained; if it is CPU time, the current CPU utilization is obtained.

[0131] Step 205: if the current available resource amount is less than or equal to the resource demand amount, the system operation instruction is split into multiple target operation instructions, or the system operation instruction is optimized into the target operation instruction by using an optimization algorithm that conforms to the operation type;

[0132] For example, if it is estimated that executing an instruction requires a large amount of disk I / O resources, and the system's current disk I / O is busy, then you can choose an optimization strategy such as batch processing of data, and optimize the original instruction to back up a large number of files at once to an instruction to back up smaller batches of files in batches, so as to reduce the instantaneous impact on disk I / O and improve the overall performance of the system. Alternatively, if it is found that executing a certain instruction requires a lot of CPU time, and the current CPU utilization is high, then you can consider optimizing the instruction to use a more efficient algorithm or data structure to perform the operation, such as optimizing a common sorting operation instruction to an instruction that uses a quick sorting algorithm, thereby optimizing and adjusting the system operation instructions to improve the execution efficiency of the subsequent system.

[0133] Step 206, if the current available resource amount is greater than the resource demand amount, the system operation instruction is determined as the target operation instruction;

[0134] Step 207, verify each target operation instruction to obtain a verification result;

[0135] In an example of the present invention, step 207 may include the following sub-steps:

[0136] Performing a security policy check on each target operation instruction to obtain a first check result;

[0137] Performing syntax analysis on each target operation instruction to obtain a second inspection result;

[0138] Performing format compliance check on each target operation instruction to obtain a third check result;

[0139] If the first check result, the second check result and the third check result are all passed, then the generated verification result is passed;

[0140] If the first check result, the second check result or the third check result is failed, then the generated verification result is failed.

[0141] In this embodiment, in order to improve the stability of the system and the reliability of execution, each target operation instruction may be further verified after obtaining the target operation instruction, and the instructions may be checked by using methods such as syntax analysis and rule matching.

[0142] Perform security policy checks on target operation instructions to determine whether the target operation instructions have corresponding execution permissions, whether there are security risk contents, etc. This can be achieved by detecting the permission range and regular expression of the login information to which the natural language data belongs.

[0143] At the same time, a grammatical analysis is performed on each target operation instruction to determine whether the target operation instruction complies with the lexical rules and syntactic rules, which can be achieved through a lexical analyzer and a syntactic analyzer.

[0144] At the same time, the format of each target operation instruction is checked for compliance. After clarifying the system's various parameter requirements, sorting requirements, and structure requirements for various instructions and parameters, the corresponding target operation instructions are matched one by one for format checking. For example, for date parameters, the format of "YYYY-MM-DD" must be used.

[0145] After the corresponding inspection results are generated respectively, if the inspection results are all passed, it is determined that the verification result of the target operation instruction is passed; if any inspection result is not passed, it is determined that the inspection result is not passed.

[0146] Step 208: If the verification result is passed, the system function interface of the information management and publishing system is called according to each target operation instruction to generate a target advertisement and deliver it to the target device.

[0147] In one example of the present invention, step 208 may include the following sub-steps:

[0148] If the verification result is passed, each target operation instruction is cached in the message queue;

[0149] Execute each target operation instruction from the message queue in sequence according to the cache order, so as to call the system function interface of the information management and publishing system to create a target advertisement that meets the advertisement playing task;

[0150] Call the system function interface to deliver the target advertisement to the target device;

[0151] Among them, the target operation instructions include material acquisition instructions, material creation instructions, advertising image combination instructions, play list creation instructions, play duration creation instructions and play plan creation instructions.

[0152] In the embodiment of the present invention, if the verification results of each target operation instruction are all passed, they can be sent to the information management and publishing system through a message queue such as redis, RabbitMQ, AWS sqs, so as to call the system function interface of the information management and publishing system such as API in turn according to each target operation instruction to perform corresponding operations to create a target advertisement that meets the advertisement playback task. After the target advertisement is generated, the target advertisement is delivered to the target device for playback through the information management and publishing system.

[0153] The material acquisition instruction is used to call the relevant system function interface to obtain relevant image materials from the material library. The material creation instruction is used to call the relevant system function interface to create relevant image materials using a preset image generation model. The advertising image combination instruction is used to call the relevant system function interface to combine or superimpose the obtained image materials in a certain order and add relevant special effects to obtain advertising images. The playlist creation instruction is used to call the relevant system function interface to create a playlist and insert the specified advertising image into the playlist. The playback duration creation instruction is used to call the relevant system function interface to set the playback duration of each advertising image and the execution date of the playback plan to obtain the final playlist. The playback plan creation instruction is used to call the relevant system function interface to obtain the playlist and related advertising images as target advertisements, and set the target devices to be delivered.

[0154] In addition, for the process of calling the system function interface to perform operations, the operation results can be fed back and displayed in real time on the visual interface so that users can know the progress of the operation in real time, such as displaying pictures generated by text on the interface, or displaying the set playback plan time period.

[0155] In another example of the present invention, the method further includes:

[0156] If the verification result fails, the error message in the verification result is returned and a re-enter prompt message is displayed.

[0157] The error message refers to the specific check type and specific target operation instruction that caused the verification result to fail, such as security policy check failure, syntax analysis failure and / or format compliance check failure. At the same time, a re-entry prompt message is displayed to inform the user how to re-enter the natural language data.

[0158] See also Figure 3 , Figure 3 The present invention is a flowchart of a method for delivering an advertisement playing task according to an embodiment of the present invention.

[0159] For ease of understanding, the following explanation is given using an example from material production, advertisement production, play plan creation to device playback.

[0160] When the user inputs natural language for the advertising task in the visual interface: "Please draw a picture of tomatoes for use as an advertising picture, and then add a text description of the price of 2.9 yuan per pound to this picture", it is judged whether the input has been completed. If not, it continues to respond to the user's input "Put the drawn picture into the material library and combine it with another picture named: vegetable basket to form an advertising picture", "Create a play list, put the advertising picture generated above into the play list, and set the playback time of this advertising picture to 10 seconds" and "Create a playback plan, put the play list created above into this playback plan, set it to start playing at 8 am every day and end at 6 pm. And put this playback plan into store A". The model call module analyzes and understands the natural language input by the user through the natural language processing model to identify the user's intention. The instruction generation module generates corresponding operation instructions based on the user's intention and the operation requirements and rules of the information release system. The default system can directly execute the operation instructions without optimizing the operation instructions. The instruction verification module verifies the generated operation instructions, checks the syntax correctness, format standardization and compliance with the system security policy. If the instruction verification passes, the operation instruction is sent to the instruction execution module. The instruction execution module executes the operation instruction, and according to the natural language input by the user, it will eventually deliver the generated advertisement to the device in the designated store according to the user's needs, and feedback the operation results to the user. The user can see the generated pictures, playlists, play plans, and delivery devices on the interface.

[0161] In an embodiment of the present invention, when natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data through a natural language processing model; system operation instructions are generated according to the advertisement demand information and the system instruction rules of the information management and publishing system; the system operation instructions are optimized to obtain at least one target operation instruction; each target operation instruction is verified to obtain a verification result; if the verification result is passed, the system function interface of the information management and publishing system is called according to each target operation instruction to generate a target advertisement and deliver it to the target device. Thus, the advertisement generation process is simplified by combining the natural language processing model, so that users can interact with the information management and publishing system through natural language without having to learn complex operation instructions and interfaces, which greatly reduces the user's usage threshold and improves user experience and usage efficiency.

[0162] See also Figure 4 , Figure 4 A structural block diagram of a device for delivering an advertisement playing task in an embodiment of the present invention is shown.

[0163] An embodiment of the present invention provides a device for delivering an advertisement playing task, comprising:

[0164] The model calling module 401 is used to extract advertisement demand information from the natural language data through a natural language processing model when natural language data for an advertisement playing task is received;

[0165] Instruction generation module 402, used to generate system operation instructions according to the advertisement demand information and the system instruction rules of the information management and publishing system;

[0166] The instruction optimization module 403 is used to optimize the system operation instruction to obtain at least one target operation instruction;

[0167] The instruction verification module 404 is used to verify each target operation instruction and obtain a verification result;

[0168] The instruction execution module 405 is used to call the system function interface of the information management and publishing system according to each target operation instruction if the verification result is passed, generate the target advertisement and deliver it to the target device.

[0169] Optionally, the device further includes a model training module, which is specifically used to:

[0170] Obtain multiple natural language text data;

[0171] Perform data cleaning and preprocessing on various natural language text data to obtain training text data;

[0172] The preset neural network model is trained using training text data and a preset loss function to generate an intermediate network model for identifying user intent;

[0173] Perform model evaluation on the intermediate network model according to multiple evaluation indicators and generate evaluation results;

[0174] If the evaluation result does not meet the preset evaluation threshold, the model parameters of the intermediate network model are adjusted according to the evaluation result until the evaluation result meets the preset evaluation threshold, thereby obtaining a natural language processing model.

[0175] Optionally, the model calling module 401 is specifically used for:

[0176] When receiving natural language data for an advertisement playing task, inputting the natural language data into a preset natural language processing model;

[0177] The natural language data is cleaned and divided into multiple natural language words through the natural language processing model;

[0178] Using a natural language processing model, stop words are removed from multiple natural language words to obtain multiple keywords;

[0179] Keywords are classified through a natural language processing model to obtain operation targets, operation types, and operation actions as advertising demand information.

[0180] Optionally, the instruction generation module 402 is specifically used for:

[0181] Create an instruction template corresponding to the operation type according to the system instruction rules of the information management and publishing system;

[0182] Fill the operation target and operation action into the variable position of the instruction template respectively to generate the system operation instruction.

[0183] Optionally, the instruction optimization module 403 is specifically used for:

[0184] Determine the required resource type and resource demand according to the operation type and operation object scale of the system operation instruction;

[0185] Obtain the current available resources of the corresponding type according to the required resource type;

[0186] If the current available resource amount is less than or equal to the resource demand amount, the system operation instruction is split into multiple target operation instructions, or the system operation instruction is optimized into the target operation instruction by using an optimization algorithm that conforms to the operation type;

[0187] If the current available resource amount is greater than the resource demand amount, the system operation instruction is determined as the target operation instruction.

[0188] Optionally, the instruction checking module 404 is specifically used for:

[0189] Performing a security policy check on each target operation instruction to obtain a first check result;

[0190] Performing syntax analysis on each target operation instruction to obtain a second inspection result;

[0191] Performing format compliance check on each target operation instruction to obtain a third check result;

[0192] If the first check result, the second check result and the third check result are all passed, then the generated verification result is passed;

[0193] If the first check result, the second check result or the third check result is failed, then the generated verification result is failed.

[0194] Optionally, the instruction execution module 405 is specifically used to:

[0195] If the verification result is passed, each target operation instruction is cached in the message queue;

[0196] Execute each target operation instruction from the message queue in sequence according to the cache order, so as to call the system function interface of the information management and publishing system to create a target advertisement that meets the advertisement playing task;

[0197] Call the system function interface to deliver the target advertisement to the target device;

[0198] Among them, the target operation instructions include material acquisition instructions, material creation instructions, advertising image combination instructions, play list creation instructions, play duration creation instructions and play plan creation instructions.

[0199] Optionally, the instruction execution module 405 is further configured to:

[0200] If the verification result fails, the error message in the verification result is returned and a re-enter prompt message is displayed.

[0201] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for delivering an advertisement playback task as described in any embodiment of the present invention.

[0202] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0203] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0204] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0205] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0206] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for delivering an advertisement playing task, characterized in that: include: When natural language data for an advertisement playing task is received, advertisement demand information is extracted from the natural language data by using a natural language processing model; Generate system operation instructions according to the advertising demand information and the system instruction rules of the information management and publishing system; Optimizing the system operation instruction to obtain at least one target operation instruction; Verifying each of the target operation instructions to obtain a verification result; If the verification result is passed, the system function interface of the information management and publishing system is called according to each of the target operation instructions to generate a target advertisement and deliver it to the target device.

2. The method according to claim 1, characterized in that The method further comprises: Obtain text data in multiple natural languages; Performing data cleaning and preprocessing on the various natural language text data to obtain training text data; The training text data and the preset loss function are used to train the preset neural network model to generate an intermediate network model for identifying user intentions; Performing model evaluation on the intermediate network model according to multiple evaluation indicators to generate evaluation results; If the evaluation result does not meet the preset evaluation threshold, the model parameters of the intermediate network model are adjusted according to the evaluation result until the evaluation result meets the preset evaluation threshold, thereby obtaining a natural language processing model.

3. The method according to claim 1, characterized in that When receiving the natural language data for the advertisement playing task, extracting the advertisement demand information from the natural language data by using the natural language processing model includes: When natural language data for an advertisement playing task is received, the natural language data is input into a preset natural language processing model; Cleaning the natural language data using the natural language processing model and dividing the natural language data into a plurality of natural language words; Screening out stop words from the plurality of natural language words by the natural language processing model to obtain a plurality of keywords; The keywords are classified by the natural language processing model to obtain operation targets, operation types and operation actions as advertising demand information.

4. The method according to claim 3, characterized in that The generating of system operation instructions according to the advertisement demand information and the system instruction rules of the information management and publishing system includes: Creating an instruction template corresponding to the operation type according to the system instruction rules of the information management and publishing system; The operation target and the operation action are respectively filled into the variable positions of the instruction template to generate a system operation instruction.

5. The method according to claim 1, characterized in that The optimizing the system operation instruction to obtain at least one target operation instruction includes: Determine the required resource type and resource requirement according to the operation type and operation object scale of the system operation instruction; According to the required resource type, obtain the currently available resource amount of the corresponding type; If the currently available resource amount is less than or equal to the resource demand amount, splitting the system operation instruction into multiple target operation instructions, or optimizing the system operation instruction into target operation instructions by using an optimization algorithm that conforms to the operation type; If the currently available resource amount is greater than the resource requirement amount, the system operation instruction is determined as the target operation instruction.

6. The method according to claim 1, characterized in that The verifying each of the target operation instructions to obtain a verification result includes: Performing a security policy check on each of the target operation instructions to obtain a first check result; Performing syntax analysis on each of the target operation instructions to obtain a second inspection result; Performing a format compliance check on each of the target operation instructions to obtain a third check result; If the first inspection result, the second inspection result and the third inspection result are all passed, generating a verification result of passed; If the first inspection result, the second inspection result or the third inspection result is failed, a verification result is generated as failed.

7. The method according to claim 1, characterized in that If the verification result is passed, calling the system function interface of the information management and publishing system according to each target operation instruction, generating a target advertisement and delivering it to the target device, including: If the verification result is passed, each of the target operation instructions is cached in a message queue; Executing each of the target operation instructions in sequence from the message queue according to the cache order, so as to call the system function interface of the information management and publishing system to create a target advertisement that meets the advertisement playing task; Calling the system function interface to deliver the target advertisement to the target device; The target operation instructions include material acquisition instructions, material creation instructions, advertising image combination instructions, play list creation instructions, play duration creation instructions and play plan creation instructions.

8. The method according to any one of claims 1 to 7, characterized in that: Also includes: If the verification result is not passed, the error information in the verification result is returned and a re-entry prompt message is displayed.

9. A device for delivering an advertisement playing task, characterized in that: include: A model calling module, for extracting advertisement demand information from the natural language data through a natural language processing model when receiving natural language data for an advertisement playing task; An instruction generation module, used to generate system operation instructions according to the advertising demand information and the system instruction rules of the information management and publishing system; An instruction optimization module, used for optimizing the system operation instruction to obtain at least one target operation instruction; An instruction verification module is used to verify each of the target operation instructions to obtain a verification result; The instruction execution module is used to call the system function interface of the information management and publishing system according to each of the target operation instructions if the verification result is passed, generate a target advertisement and deliver it to the target device.

10. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for delivering the advertisement playing task as described in any one of claims 1 to 8.

Images