Click rate prediction method and device, computer device and storage medium

Abstract

The application discloses a click rate prediction method and device, computer equipment and a storage medium, the method uses a crawler technology to crawl dynamic information in a preset time range; the preset time range is determined according to the historical statistical time; entity recognition and intent feature extraction are performed on the dynamic information to obtain entity recognition results and intent extraction results; the information industry label is determined according to the entity recognition results, and whether the information industry label and the target industry label are the same is determined; when the information industry label and the target industry label are the same, industry sentiment analysis is performed on the entity recognition results and the intent extraction results to obtain sentiment analysis results; the historical click rate wave band diagram is predicted according to the sentiment analysis results to obtain a predicted click rate wave band diagram. The application improves the accuracy of click rate prediction, and improves the feasibility and accuracy of the advertisement placement scheme recommendation.

Description

Technical Field

[0001] This invention relates to the field of click-through rate (CTR) prediction technology, and more particularly to a CTR prediction method, apparatus, computer device, and storage medium. Background Technology

[0002] With the development of science and technology, more and more new media are emerging and developing, which allows for advertising on different media to achieve information exposure and improve the efficiency of promotion such as products and events.

[0003] Current advertising technology often operates on a demand-driven basis, meaning that when there is a demand for advertising, the appropriate media platform is immediately searched for and the advertisement is placed. However, the effectiveness of the advertising on the media platform is not estimated, resulting in poor advertising results. Even if the effectiveness of the advertising on the media platform is estimated, it only assesses whether the media platform to be advertised matches the advertisement. This makes it impossible to accurately evaluate the effectiveness of the advertising on the media platform, which may lead to low advertising returns. Summary of the Invention

[0004] This invention provides a click-through rate prediction method, apparatus, computer device, and storage medium to address the problems of poor advertising performance and low advertising revenue.

[0005] A click-through rate prediction method, comprising:

[0006] Obtain a historical click-through rate (CTR) band chart; each historical CTR band chart is associated with a target industry tag and a historical statistical time period.

[0007] Web crawling technology is used to crawl dynamic information within a preset time range; the preset time range is determined based on the historical statistical time.

[0008] Entity recognition and intent feature extraction are performed on the dynamic information to obtain entity recognition results and intent extraction results;

[0009] Based on the entity recognition results, an information industry label is determined, and it is determined whether the information industry label is the same as the target industry label;

[0010] When the information industry label is the same as the target industry label, industry sentiment analysis is performed on the entity recognition result and the intent extraction result to obtain the sentiment analysis result.

[0011] Based on the sentiment analysis results, click-through rate (CTR) prediction is performed on the historical CTR band map to obtain the predicted CTR band map.

[0012] A click-through rate prediction device, comprising:

[0013] The historical data acquisition module is used to acquire historical click-through rate (CTR) band charts; each historical CTR band chart is associated with a target industry tag and a historical statistical time period.

[0014] The information crawling module is used to crawl dynamic information within a preset time range using web crawling technology; the preset time range is determined based on the historical statistical time.

[0015] The entity intent extraction module is used to perform entity recognition and intent feature extraction on the dynamic information to obtain entity recognition results and intent extraction results.

[0016] The industry tag comparison module is used to determine the information industry tag based on the entity recognition result, and to determine whether the information industry tag is the same as the target industry tag;

[0017] The industry sentiment analysis module is used to perform industry sentiment analysis on the entity recognition result and the intent extraction result when the information industry tag is the same as the target industry tag, so as to obtain the sentiment analysis result;

[0018] The click-through rate prediction module is used to predict the click-through rate of the historical click-through rate band map based on the sentiment analysis results, and obtain the predicted click-through rate band map.

[0019] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described click-through rate prediction method.

[0020] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described click-through rate prediction method.

[0021] The above-described click-through rate (CTR) prediction method, apparatus, computer equipment, and storage medium involve the following steps: The method acquires a historical CTR band map; each historical CTR band map is associated with a target industry tag and a historical statistical time period; dynamic information within a preset time range is crawled using web crawling technology; the preset time range is determined based on the historical statistical time period; entity recognition and intent feature extraction are performed on the dynamic information to obtain entity recognition results and intent extraction results; an information industry tag is determined based on the entity recognition results, and it is determined whether the information industry tag is the same as the target industry tag; when the information industry tag is the same as the target industry tag, industry sentiment analysis is performed on the entity recognition results and the intent extraction results to obtain sentiment analysis results; and CTR prediction is performed on the historical CTR band map based on the sentiment analysis results to obtain a predicted CTR band map.

[0022] This invention performs entity recognition and intent feature extraction on dynamic information within a preset time range. Then, assuming the dynamic information and historical click-through rate (CTR) waveforms belong to the same industry, it performs sentiment analysis on the entity recognition and intent extraction results. Based on the sentiment analysis results, it predicts the historical CTR waveforms. This combines real-time dynamic information with CTR, providing data support for CTR prediction and improving its accuracy. Furthermore, after predicting CTR, the predicted CTR waveforms allow for rapid adjustments to advertising timing or strategy recommendations, improving the feasibility and accuracy of advertising recommendations and ultimately increasing advertising revenue. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of an application environment for the click-through rate prediction method in one embodiment of the present invention;

[0025] Figure 2 This is a flowchart of a click-through rate prediction method according to an embodiment of the present invention;

[0026] Figure 3 This is a block diagram of a click-through rate prediction device according to an embodiment of the present invention;

[0027] Figure 4 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The click-through rate prediction method provided in this embodiment of the invention can be applied to, for example... Figure 1 The application environment shown. Specifically, this click-through rate (CTR) prediction method is applied in a click-through rate (CTR) prediction system, which includes, as shown in the example. Figure 1The diagram illustrates a client and server. The client and server communicate over a network to address issues of poor advertising performance and low advertising revenue. The client, also known as the user terminal, is the program that provides local services to the client, corresponding to the server. The client can be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0030] In one embodiment, such as Figure 2 As shown, a click-through rate prediction method is provided, which is then applied to... Figure 1 Taking the server in the example, the following steps are included:

[0031] S10: Obtain a historical click-through rate (CTR) band chart; each CTR band chart is associated with a target industry tag and a historical statistical time period.

[0032] Understandably, a historical click-through rate (CTR) chart represents the CTR trend of an ad over a period of time after it has been placed on a media platform. This historical CTR chart can be plotted by statistically analyzing the daily click-through rate (CTR) of an ad after it has been placed on the media platform. The target industry tag refers to the industry associated with this historical CTR chart. For example, the target industry tag could be the education industry, e-commerce industry, gaming industry, etc. That is, in this embodiment, the historical CTR chart is industry-related; one target industry tag corresponds to one historical CTR chart. Figure 1 One-to-one correspondence; furthermore, in this embodiment, the historical click-through rate (CTR) band chart corresponding to each target industry tag can be drawn by collecting the click counts of different advertising campaigns under that target industry tag. The historical statistical time refers to the statistical time period of the click counts in the historical CTR band chart. For example, the historical statistical time can be one week, one month, or one quarter, etc.

[0033] S20: Use web crawling technology to crawl dynamic information within a preset time range; the preset time range is determined based on the historical statistical time.

[0034] In essence, web scraping is the technique of retrieving information from sources such as web pages using programs or scripts. The preset time range is determined based on historical statistical periods. For example, if the historical statistical period is from July 21st to July 27th, 2021, then the last point in time is July 27th. Therefore, the preset time range can be determined as seven days after that last point in time (e.g., from July 28th to August 3rd, 2021). Dynamic information refers to things like policy information and event information.

[0035] S30: Perform entity recognition and intent feature extraction on the dynamic information to obtain entity recognition results and intent extraction results.

[0036] Entity recognition is the process of extracting entities from dynamic information to determine the industry to which the dynamic information belongs. Intent feature extraction is the process of extracting the intent of actions from dynamic information to perform sentiment analysis based on the intent extraction results and entity recognition results.

[0037] S40: Determine the information industry label based on the entity recognition result, and determine whether the information industry label is the same as the target industry label.

[0038] Understandably, information industry tags are used to represent industry information in dynamic information. For example, information industry tags can be game industry tags, education industry tags, e-commerce industry tags, etc.

[0039] Specifically, after performing entity recognition and intent feature extraction on the dynamic information, and obtaining the entity recognition results and intent extraction results, the corresponding information industry tag can be determined based on the entity recognition results. For example, if a dynamic information is "Double Eleven is coming soon, and the e-commerce war is about to begin," the entity that can be identified is "e-commerce," and thus the information industry tag is determined to be the e-commerce industry tag. After determining the information industry tag, the information industry tag is compared with the target industry tag to determine whether the information industry tag and the target industry tag are the same.

[0040] S50: When the information industry label is the same as the target industry label, perform industry sentiment analysis on the entity recognition result and the intent extraction result to obtain the sentiment analysis result.

[0041] Specifically, after determining the information industry label based on the entity recognition result and determining whether the information industry label is the same as the target industry label, if the information industry label is the same as the target industry label, industry sentiment analysis is performed on the entity recognition result and the intent extraction result. That is, the impact of the intent extraction result on the entity recognition result is determined, and the sentiment analysis result is obtained. In this embodiment, the sentiment analysis result mainly includes positive sentiment and negative sentiment.

[0042] In one embodiment, after step S40, that is, after determining whether the information industry label is the same as the target industry label, the method further includes:

[0043] When the information industry label is different from the target industry label, the historical click-through rate band chart is recorded as the predicted click-through rate band chart.

[0044] Understandably, after determining the information industry label based on the entity recognition result and determining whether the information industry label is the same as the target industry label, if the information industry label is different from the target industry label, it indicates that the dynamic information may not have an impact on the historical click-through rate (CTR) band chart. Therefore, the historical CTR band chart can be directly recorded as the predicted CTR band chart. Furthermore, directly recording the historical CTR band chart as the predicted CTR band chart in this embodiment is merely an example, meaning it only indicates that the trend of clicks on future advertising campaigns within that industry will not change significantly. Therefore, the predicted CTR band chart only needs to be similar to the trend of the historical CTR band chart, allowing for a small deviation (e.g., a deviation of 0.1 or 0.2).

[0045] S60: Based on the sentiment analysis results, perform click-through rate prediction on the historical click-through rate band map to obtain the predicted click-through rate band map.

[0046] Understandably, the click-through rate prediction in this embodiment mainly involves predicting the changes in the historical click-through rate band map over a future period of time, thereby obtaining the predicted click-through rate band map.

[0047] In one embodiment, step S60, namely, predicting the click-through rate (CTR) of the historical CTR band map based on the sentiment analysis results to obtain a predicted CTR band map, includes:

[0048] When the sentiment analysis results represent positive sentiment, an upward prediction is made on the historical click-through rate band map to obtain an upward prediction band map.

[0049] When the sentiment analysis results represent reverse sentiment, a downward prediction is made on the historical click-through rate band map to obtain a downward prediction band map.

[0050] Understandably, the results of sentiment analysis include both positive and negative sentiment; positive sentiment indicates that the dynamic information has a positive meaning, while negative sentiment indicates that the dynamic information has a negative meaning.

[0051] Specifically, after performing industry sentiment analysis on the entity recognition results and the intent extraction results to obtain the sentiment analysis results, if the sentiment analysis results represent positive sentiment, an upward prediction is made on the historical click-through rate (CTR) band chart to obtain an upward prediction band chart. Understandably, compared to the historical CTR band chart, the upward prediction band chart shows a higher trend in click counts, meaning that the CTR of the upward prediction band chart is higher than that of the historical CTR band chart over a certain period. For example, suppose a dynamic news item says "Double Eleven is coming soon, the e-commerce war is about to begin." This dynamic news item mentions the e-commerce industry, and for the e-commerce industry, "Double Eleven" is an event that will have a significant positive impact on the e-commerce industry. Therefore, the CTR of advertising in the e-commerce industry will show a certain upward change. Thus, an upward prediction of the CTR can be made on the historical CTR band chart to obtain the upward prediction band chart.

[0052] Furthermore, industry sentiment analysis is performed on the entity recognition results and the intent extraction results to obtain sentiment analysis results. If the sentiment analysis results represent negative sentiment, a decline prediction is made on the historical click-through rate (CTR) band map to obtain a decline prediction band map. Understandably, compared to the historical CTR band map, the trend of clicks in the rise prediction band map is higher, meaning that the CTR of the rise prediction band map is higher than that of the historical CTR band map over a period of time. For example, suppose a dynamic news item states, "Due to the epidemic, various transportation routes need to be closed, and it is recommended not to travel unless absolutely necessary to avoid the spread of the epidemic." The entity recognition results for "transportation routes" and "tourism" belong to the tourism industry. The intent feature for "transportation routes" is "closed," and the intent feature for "tourism" is "not to travel." This is information with a negative impact on the tourism industry, therefore, the CTR for tourism industry advertisements will show a certain decline. Therefore, a decline prediction can be made on the historical CTR band map to obtain the decline prediction band map.

[0053] Furthermore, in addition to the aforementioned positive and negative sentiments, neutral sentiment may also exist. When representing neutral sentiment in sentiment analysis results, historical click-through rate (CTR) band charts can be recorded as predicted CTR band charts. For example, suppose a dynamic news item states "The e-sports industry should maintain its healthy development at this stage." In this dynamic news item, "e-sports" refers to the gaming industry, and the corresponding intent extraction result is "maintain." Therefore, the CTR for advertising in the gaming industry will not show significant changes compared to historical CTR band charts, and thus, historical CTR band charts can be recorded as predicted CTR band charts.

[0054] In this embodiment, by performing entity recognition and intention feature extraction on the dynamic information within a preset time range, and on the premise that the dynamic information and the historical click-through rate band chart belong to the same industry, by performing sentiment analysis on the entity recognition result and the intention extraction result, and then predicting the historical click-through rate band chart according to the sentiment analysis result, the real-time dynamic information can be combined with the click-through rate, providing data support for click-through rate prediction, and thus improving the accuracy of click-through rate prediction. Further, after predicting the click-through rate, the predicted click-through rate band chart can be used to quickly adjust the advertising placement time or the recommendation of the advertising placement plan, improving the feasibility and accuracy of the advertising placement plan recommendation.

[0055] In one embodiment, in step S30, that is, performing entity recognition and intention feature extraction on the dynamic information to obtain an entity recognition result and an intention extraction result, includes:

[0056] Performing word segmentation on the dynamic information to obtain multiple dynamic information words corresponding to the dynamic information.

[0057] It can be understood that word segmentation is a method of dividing dynamic information into words, and at the same time, meaningless words such as stop words in the dynamic information (such as "of", "is", etc.) can be removed; further, the method of word segmentation can be through, such as Jieba word segmentation, natural language model word segmentation, etc. In this way, after performing word segmentation on the dynamic information, multiple dynamic information words corresponding to the dynamic information can be obtained.

[0058] Performing entity recognition on each of the dynamic information words according to a preset word vector dictionary to obtain the entity recognition result.

[0059] It can be understood that the preset word vector dictionary is a dictionary that stores words and their corresponding word vectors after pre-converting various words into word vectors, that is, at least one word and a word vector corresponding to the word are stored in the preset word vector dictionary.

[0060] In one embodiment, after performing entity recognition on each of the dynamic information words according to the preset word vector dictionary to obtain the entity recognition result, it includes:

[0061] Performing word vector conversion on each of the dynamic information words according to the preset word vector dictionary to determine an information word vector corresponding to each of the dynamic information words.

[0062] Specifically, after performing word segmentation on the dynamic information to obtain multiple dynamic information words corresponding to the dynamic information, the dictionary words corresponding to each dynamic information word are queried from the preset word vector dictionary (the dictionary words are the words stored in the preset word vector dictionary), and then the word vectors corresponding to the dictionary words are used as the word vectors of the corresponding dynamic information words, thereby improving the speed of word vector conversion.

[0063] The entity recognition results are obtained by performing entity recognition on the vectors of each information word using a preset entity recognition model.

[0064] Understandably, an entity recognition model can be a model that has been trained in advance using some training data. This entity recognition model is used to recognize entities in text information. For example, the entity recognition model can be a BERT-based recognition model or an ALBERT-based recognition model.

[0065] Specifically, after performing word segmentation on the dynamic information to obtain multiple dynamic information words corresponding to the dynamic information, the dictionary words corresponding to each dynamic information word are queried from the preset word vector dictionary (the dictionary words are the words stored in the preset word vector dictionary), and then the word vectors corresponding to the dictionary words are used as the word vectors of the corresponding dynamic information words, thereby improving the speed of word vector conversion.

[0066] Based on the preset intent definition tree, each of the dynamic information words, and the entity recognition results, intent recognition is performed on the dynamic information to obtain the intent extraction result corresponding to the entity recognition results.

[0067] Understandably, the pre-defined intent definition tree is constructed by pre-collecting several intent features and classifying the collected intents (such as according to the type of intent, the emotion of the intent, etc.).

[0068] In one embodiment, the step of performing intent recognition on the dynamic information based on a preset intent definition tree, each of the dynamic information words, and the entity recognition result to obtain the intent extraction result corresponding to the entity recognition result includes:

[0069] Entities belonging to the category of executive department entities are recorded as executive department entities.

[0070] Understandably, the executive entity category refers to the executor or executive department in dynamic information. For example, a dynamic information message is "XX promotes the vigorous development of e-commerce platforms". This dynamic information message contains two entity identification results: "XX" and "e-commerce platform". Here, "XX" is the entity identification result of the executive entity category, so "XX" is recorded as the executive entity.

[0071] In the dynamic information, the dynamic information words corresponding to the execution department entity are labeled according to the execution department entity to obtain labeled information.

[0072] Specifically, after the entity identification results belonging to the execution department entity category are recorded as execution department entities, dynamic information words corresponding to the execution department entities are queried in the dynamic information information. The queried dynamic information words are then labeled according to the execution department entities to obtain labeled information information.

[0073] The labeled information is input into the preset intent definition tree, so that the intent of the labeled information is identified through the preset intent definition tree, and the intent extraction result is obtained.

[0074] Specifically, after annotating the dynamic information words corresponding to the executing department entity according to the executing department entity to obtain annotated information information, the annotated information information is input into a preset intent definition tree. The preset intent definition tree performs intent recognition on the annotated information information, that is, determines the intent features in the dynamic information words in the context of the executing department entity, and obtains the intent extraction result. For example, in the above description, when the dynamic information information is "XX promotes the vigorous development of e-commerce platforms", the corresponding executing department entity is "XX". After annotating "XX" with the executing department entity, the preset intent definition tree can identify the intent features in the annotated information information as "promote" and "vigorously develop", thus obtaining the corresponding intent extraction result.

[0075] Furthermore, the annotation proposed in the above steps using the executive department entity is only an example. When there is no executive department entity, for example, when the dynamic information is "At this stage, the e-sports industry should maintain good development", there is only the entity identification result of "e-sports industry", and the entity identification result is not the executive department entity. In this case, the "e-sports industry" entity can also be annotated, and the subsequent steps can be performed to extract the intent extraction result as "maintain".

[0076] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0077] In one embodiment, a click-through rate (CTR) prediction device is provided, which corresponds one-to-one with the CTR prediction methods described in the above embodiments. For example... Figure 3As shown, the click-through rate prediction device includes a historical data acquisition module 10, an information crawling module 20, an entity intent extraction module 30, an industry tag comparison module 40, an industry sentiment analysis module 50, and a click-through rate prediction module 60. Detailed descriptions of each functional module are as follows:

[0078] Historical data acquisition module 10 is used to acquire historical click-through rate (CTR) band charts; each historical CTR band chart is associated with a target industry tag and a historical statistical time period.

[0079] The information crawling module 20 is used to crawl dynamic information within a preset time range using crawler technology; the preset time range is determined based on the historical statistical time.

[0080] The entity intent extraction module 30 is used to perform entity recognition and intent feature extraction on the dynamic information to obtain entity recognition results and intent extraction results.

[0081] The industry label comparison module 40 is used to determine the information industry label based on the entity recognition result, and to determine whether the information industry label is the same as the target industry label;

[0082] The industry sentiment analysis module 50 is used to perform industry sentiment analysis on the entity recognition result and the intent extraction result when the information industry label is the same as the target industry label, so as to obtain the sentiment analysis result.

[0083] Click-through rate prediction module 60 is used to predict the click-through rate of the historical click-through rate band map based on the sentiment analysis results, and obtain a predicted click-through rate band map.

[0084] Preferably, the entity intent extraction module 30 includes:

[0085] The word segmentation processing unit is used to segment the dynamic information to obtain multiple dynamic information words corresponding to the dynamic information.

[0086] An entity recognition unit is used to perform entity recognition on each of the dynamic information words according to a preset word vector dictionary, and obtain the entity recognition result;

[0087] The intent recognition unit is used to perform intent recognition on the dynamic information based on a preset intent definition tree, each of the dynamic information words and the entity recognition result, and to obtain the intent extraction result corresponding to the entity recognition result.

[0088] For specific limitations regarding the click-through rate (CTR) prediction device, please refer to the limitations of the CTR prediction method above, which will not be repeated here. Each module in the aforementioned CTR prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0089] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data used by the click-through rate prediction method in the above embodiment. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a click-through rate prediction method.

[0090] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the click-through rate prediction method in the above embodiment.

[0091] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the click-through rate prediction method described in the above embodiment.

[0092] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0094] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A click-through rate prediction method, characterized in that, include: Obtain a historical click-through rate (CTR) band chart; each historical CTR band chart is associated with a target industry tag and a historical statistical time period. Web crawling technology is used to crawl dynamic information within a preset time range; the preset time range is determined based on the historical statistical time. Entity recognition and intent feature extraction are performed on the dynamic information to obtain entity recognition results and intent extraction results; Based on the entity recognition results, an information industry label is determined, and it is determined whether the information industry label is the same as the target industry label; When the information industry label is the same as the target industry label, industry sentiment analysis is performed on the entity recognition result and the intent extraction result to obtain the sentiment analysis result. Based on the sentiment analysis results, click-through rate (CTR) prediction is performed on the historical CTR band map to obtain a predicted CTR band map. After determining whether the information industry label is the same as the target industry label, the method further includes: When the information industry label is different from the target industry label, the historical click-through rate band chart is recorded as the predicted click-through rate band chart; The predicted click-through rate (CTR) band chart includes an upward predicted CTR band chart and a downward predicted CTR band chart; the step of predicting the CTR based on the historical CTR band chart according to the sentiment analysis results to obtain the predicted CTR band chart includes: When the sentiment analysis results represent positive sentiment, an upward prediction is made on the historical click-through rate band map to obtain an upward prediction band map. When the sentiment analysis results represent reverse sentiment, a downward prediction is made on the historical click-through rate band map to obtain a downward prediction band map.

2. The click-through rate prediction method as described in claim 1, characterized in that, The process of performing entity recognition and intent feature extraction on the dynamic information to obtain entity recognition results and intent extraction results includes: The dynamic information is segmented to obtain multiple dynamic information words corresponding to the dynamic information. Entity recognition is performed on each of the dynamic information words according to a preset word vector dictionary to obtain the entity recognition result; Based on the preset intent definition tree, each of the dynamic information words, and the entity recognition results, intent recognition is performed on the dynamic information to obtain the intent extraction result corresponding to the entity recognition results.

3. The click-through rate prediction method as described in claim 2, characterized in that, The step of performing entity recognition on each of the dynamic information words according to a preset word vector dictionary to obtain the entity recognition result includes: Based on a preset word vector dictionary, each dynamic information word is converted into a word vector to determine the information word vector corresponding to each dynamic information word. The entity recognition results are obtained by performing entity recognition on the vectors of each information word using a preset entity recognition model.

4. The click-through rate prediction method as described in claim 2, characterized in that, The step of performing intent recognition on the dynamic information based on a preset intent definition tree, each of the dynamic information words, and the entity recognition result to obtain the intent extraction result corresponding to the entity recognition result includes: Record the entity identification results belonging to the category of executive department entity as executive department entity; In the dynamic information, the dynamic information words corresponding to the execution department entity are labeled according to the execution department entity to obtain labeled information; The labeled information is input into the preset intent definition tree, so that the intent of the labeled information is identified through the preset intent definition tree, and the intent extraction result is obtained.

5. A click-through rate prediction device, characterized in that, include: The historical data acquisition module is used to acquire historical click-through rate (CTR) band charts; each historical CTR band chart is associated with a target industry tag and a historical statistical time period. The information crawling module is used to crawl dynamic information within a preset time range using web crawling technology; the preset time range is determined based on the historical statistical time. The entity intent extraction module is used to perform entity recognition and intent feature extraction on the dynamic information to obtain entity recognition results and intent extraction results. The industry tag comparison module is used to determine the information industry tag based on the entity recognition result, and to determine whether the information industry tag is the same as the target industry tag; The industry sentiment analysis module is used to perform industry sentiment analysis on the entity recognition result and the intent extraction result when the information industry tag is the same as the target industry tag, so as to obtain the sentiment analysis result; The click-through rate prediction module is used to predict the click-through rate of the historical click-through rate band map based on the sentiment analysis results, and obtain the predicted click-through rate band map. After determining whether the information industry label is the same as the target industry label, the method further includes: When the information industry label is different from the target industry label, the historical click-through rate band chart is recorded as the predicted click-through rate band chart; The predicted click-through rate (CTR) band chart includes an upward predicted CTR band chart and a downward predicted CTR band chart; the step of predicting the CTR based on the historical CTR band chart according to the sentiment analysis results to obtain the predicted CTR band chart includes: When the sentiment analysis results represent positive sentiment, an upward prediction is made on the historical click-through rate band map to obtain an upward prediction band map. When the sentiment analysis results represent reverse sentiment, a downward prediction is made on the historical click-through rate band map to obtain a downward prediction band map.

6. The click-through rate prediction device as described in claim 5, characterized in that, The entity intent extraction module includes: The word segmentation processing unit is used to segment the dynamic information to obtain multiple dynamic information words corresponding to the dynamic information. An entity recognition unit is used to perform entity recognition on each of the dynamic information words according to a preset word vector dictionary, and obtain the entity recognition result; The intent recognition unit is used to perform intent recognition on the dynamic information based on a preset intent definition tree, each of the dynamic information words and the entity recognition result, and to obtain the intent extraction result corresponding to the entity recognition result.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the click-through rate prediction method as described in any one of claims 1 to 4.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the click-through rate prediction method as described in any one of claims 1 to 4.

Images