Advertisement recommendation method and device, electronic equipment and medium

Abstract

The invention relates to the field of financial science and technology, and discloses an advertisement recommendation method and device, electronic equipment and a medium, and the method comprises the steps: responding to an information browsing behavior of a user in a display interface, and obtaining target multimedia information corresponding to the information browsing behavior; carrying out key feature extraction based on the target multimedia information to obtain a target key feature; performing semantic analysis on the target key features to generate target semantics corresponding to the target key features; performing advertisement matching on the basis of the target semantics, and determining a first target advertisement of which the semantics is matched with the target semantics; and displaying the first target advertisement in the display interface to recommend the first target advertisement to the user. According to the embodiment of the invention, the recommended target advertisement can meet the real demand of the user.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to an advertisement recommendation method, device, electronic device, and medium. Background Art

[0002] With the rapid development of Internet technology, online advertising has become an important means for companies to promote their products and services. Accurately pushing personalized targeted advertisements to users is crucial to improving advertising effectiveness and user experience.

[0003] However, existing advertising recommendation methods mainly rely on users' historical data. When the accumulation of user behavior trajectory data is insufficient, it is difficult to build a complete interest graph, which leads to a deviation between the recommended target advertisements and the user's actual needs. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose an advertisement recommendation method, device, electronic device and medium, aiming to solve the problem that the recommended target advertisements of existing advertisement recommendation methods deviate from the actual needs of users.

[0005] To achieve the above objectives, a first aspect of an embodiment of the present application provides an advertisement recommendation method, the method comprising:

[0006] In response to an information browsing behavior of a user in a display interface, obtaining target multimedia information corresponding to the information browsing behavior;

[0007] Extract key features based on the target multimedia information to obtain target key features;

[0008] Performing semantic analysis on the target key features to generate target semantics corresponding to the target key features;

[0009] Performing advertisement matching based on the target semantics, and determining a first target advertisement having semantics matching the target semantics;

[0010] The first target advertisement is displayed in the display interface to recommend the first target advertisement to the user.

[0011] In some embodiments, extracting key features based on the target multimedia information to obtain target key features includes:

[0012] Obtaining the user's historical information browsing behavior;

[0013] Performing user portrait analysis based on the historical information browsing behavior to generate user portrait information associated with the user;

[0014] Key features of the target multimedia information are extracted based on the user portrait information to obtain target key features.

[0015] In some embodiments, the steps of performing semantic analysis on the target key features to generate target semantics corresponding to the target key features, and the steps of performing advertisement matching based on the target semantics to determine a first target advertisement having semantics matching the target semantics are performed based on a preset advertisement recommendation model;

[0016] After performing advertisement matching based on the target semantics and determining a first target advertisement having semantics matching the target semantics, the method further includes:

[0017] Acquiring user operation information corresponding to the first target advertisement;

[0018] determining a recommendation satisfaction level of the first target advertisement based on the user operation information;

[0019] generating a training sample based on the target key feature and the first target advertisement when the recommendation satisfaction is greater than or equal to a preset satisfaction threshold;

[0020] The preset advertisement recommendation model is trained based on the training samples to obtain an updated preset advertisement recommendation model.

[0021] In some embodiments, extracting key features based on the target multimedia information to obtain target key features includes:

[0022] When the target multimedia information includes video information, obtaining voice information of the video information, and recognizing the voice information by using a voice recognition technology to obtain first text information;

[0023] Inputting the video information into a target detection model, and having the target detection model output a target video frame of the video information;

[0024] Obtaining a target image corresponding to the target video frame in the video information;

[0025] Recognize the target image using image recognition technology to obtain second text information;

[0026] Obtaining the target text information by fusing the first text information and the second text information;

[0027] Key features are extracted from the target text information to obtain target key features.

[0028] In some implementations, extracting key features from the target text information to obtain target key features includes:

[0029] Performing word segmentation processing on the target text information using natural language processing technology to obtain multiple phrases;

[0030] Inputting the multiple phrases into a preset intention recognition model, and having the preset intention recognition model output a recognition result, wherein the recognition result is used to indicate the browsing intention of the user;

[0031] Matching the multiple phrases in a preset hotspot library to determine multiple hot keywords, wherein the preset hotspot library includes relevant information about people whose influence is greater than a first preset threshold and relevant information about events whose influence is greater than a second preset threshold;

[0032] Determine target key features of the target text information based on the recognition result and the multiple hot keywords.

[0033] In some implementations, displaying the first target advertisement on the display interface includes:

[0034] Based on the information browsing behavior, determining the eye focus point of the user in the display interface;

[0035] Based on the eye focus point, determining a plurality of areas to be placed advertisements in the display interface;

[0036] Determining a target advertisement placement area among the plurality of advertisement placement areas according to the user's historical operation information on advertisement placement in each display area of the display interface;

[0037] The first target advertisement is displayed in the target advertisement to be delivered area.

[0038] In some implementations, after displaying the first target advertisement in the display interface to recommend the first target advertisement to the user, the method further includes:

[0039] Obtaining user behavior data regarding the first target advertisement;

[0040] performing a correlation analysis on the behavior data and the semantics of the first target advertisement to obtain a temporal correlation between the behavior data and the semantics of the first target advertisement;

[0041] Inputting the temporal correlation between the behavior data and the semantics of the first target advertisement into a preset dynamic perception model, and having the preset dynamic perception model output a perception result, wherein the perception result is used to indicate a dynamic change in the user's interest;

[0042] determining a second target advertisement based on the perception result;

[0043] The second target advertisement is displayed in the display interface to recommend the second target advertisement to the user.

[0044] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides an advertisement recommendation device, comprising:

[0045] an acquisition module, configured to acquire target multimedia information corresponding to an information browsing behavior of a user in a display interface in response to the information browsing behavior of the user;

[0046] An extraction module, configured to extract key features based on the target multimedia information to obtain target key features;

[0047] An analysis module, configured to perform semantic analysis on the target key features and generate target semantics corresponding to the target key features;

[0048] a matching module, configured to perform advertisement matching based on the target semantics, and determine a first target advertisement having semantics matching the target semantics;

[0049] The display module is configured to display the first target advertisement in the display interface to recommend the first target advertisement to the user.

[0050] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the advertising recommendation method described in the first aspect.

[0051] To achieve the above-mentioned purpose, a fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the advertising recommendation method described in the first aspect is implemented.

[0052] To achieve the above-mentioned purpose, an embodiment of the present application may provide a computer program product for implementation. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device implements the advertising recommendation method described in the first aspect above.

[0053] The advertising recommendation method, device, electronic device and medium proposed in the present application obtain target multimedia information corresponding to the information browsing behavior when the electronic device receives the information browsing behavior of the user in the display interface, so as to quickly obtain the user's current browsing purpose, perform key feature extraction based on the target multimedia information to obtain the target key features, and perform semantic analysis on the target key features to generate target semantics corresponding to the target key features, more accurately understand the theme and key information of the user's current browsing content through the target semantics, perform advertisement matching based on the target semantics, determine a first target advertisement having semantics matching the target semantics, display the first target advertisement in the display interface to recommend the first target advertisement to the user, perform advertisement matching through the target semantics to quickly and accurately obtain advertisements similar to the user's current browsing interests, so that the recommended target advertisements meet the user's current real needs and increase the user's interest in clicking on the target advertisements. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of the advertisement recommendation method provided by an embodiment of the present application;

[0055] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0056] Figure 3 yes Figure 1 Flow chart of step S102 in FIG.

[0057] Figure 4 yes Figure 3 Flow chart of step S406 in FIG.

[0058] Figure 5 is a schematic diagram of the structure of the advertisement recommendation device provided in an embodiment of the present application;

[0059] Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0061] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0063] With the rapid development of Internet technology, online advertising has become an important means for companies to promote their products and services. Accurately pushing personalized targeted advertisements to users is crucial to improving advertising effectiveness and user experience.

[0064] However, existing advertising recommendation methods mainly rely on users' historical data. When the accumulation of user behavior trajectory data is insufficient, it is difficult to build a complete interest graph, which leads to a deviation between the recommended target advertisements and the user's actual needs.

[0065] Based on this, the embodiments of the present application provide an advertisement recommendation method, device, electronic device and medium, aiming to solve the problem that the recommended target advertisements of existing advertisement recommendation methods deviate from the actual needs of users.

[0066] The advertisement recommendation method, device, electronic device, and medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the advertisement recommendation method in the embodiments of the present application is described.

[0067] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0068] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0069] The advertising recommendation method provided in the embodiment of the present application relates to the field of financial technology. The advertising recommendation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, or as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the advertising recommendation method, etc., but is not limited to the above forms.

[0070] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0071] It should be noted that in each specific embodiment of this application, when it comes to the need to perform relevant processing based on information, behavioral data, historical data, location information, and other data related to identity or characteristics, permission or consent must be obtained first, and the collection, use, and processing of such data must comply with relevant laws, regulations, and standards. In addition, when the embodiment of this application needs to obtain sensitive personal information, a separate permission or separate consent will be obtained through a pop-up window or by jumping to a confirmation page. After the separate permission or separate consent is clearly obtained, the necessary relevant data for the normal operation of the embodiment of this application will be obtained.

[0072] Figure 1 This is a flow chart of the advertisement recommendation method provided in the embodiment of the present application. Figure 1 The advertisement recommendation method provided in the embodiment of the present application may include but is not limited to steps S101 to S104.

[0073] Step S101: In response to a user's information browsing behavior in a display interface, target multimedia information corresponding to the information browsing behavior is acquired.

[0074] In this step, information browsing behavior refers to click, slide, or pause operations generated during the user's interaction with the display interface, such as browsing web pages, watching videos, and posting pictures and text on social software. When the electronic device detects the user's information browsing behavior on the display interface, the electronic device can use page crawling technology to extract elements in the display interface, or it can obtain streaming media data through interface calls, without limitation here, to obtain the target multimedia information corresponding to the information browsing behavior, where the target multimedia data includes interactive content such as pictures, texts, and videos currently being viewed by the user.

[0075] For example, a user is using the Ping An Bank APP and browses an article about "regular fund investment" on the homepage. The user obtains a comparative chart of regular fund investment returns, a video tutorial on regular fund investment, and related text information on the "regular fund investment" article page as target multimedia information.

[0076] Step S102: extract key features based on the target multimedia information to obtain target key features.

[0077] In this step, natural language processing technology can be used to identify text keywords to extract key features of the target multimedia information and obtain target key features; visual features can be extracted through image recognition technology to extract key features of the target multimedia information and obtain target key features; or a neural network model for key feature extraction can be pre-trained to extract key features of the target multimedia information and obtain target key features, which are not limited here.

[0078] For example, natural language processing technology is used to identify and extract keywords in the text, such as "blockchain", "decentralization", "smart contract", "cryptocurrency", "investment opportunity", etc., and to identify entity information and topic classification in the text to obtain target key features.

[0079] Step S103: Perform semantic analysis on the target key features to generate target semantics corresponding to the target key features.

[0080] In this step, the target key features are mapped to the semantic space through a word vector model or a topic model to obtain the target semantics corresponding to the target key features. For example, the BERT model (Bidirectional Encoder Representations from Transformers) can be used to generate text semantic vectors to analyze the relationship between the target key features and the advertising content.

[0081] For example, the target key feature is "the smart fixed investment curve shows a steadily rising trend". Semantic analysis of the target key feature yields the target semantics "the smart fixed investment strategy has performed stably in the past year, with low risk and a steadily rising return trend". Based on this target semantics, low-risk smart fixed investment products can be recommended to users.

[0082] Step S104: performing advertisement matching based on the target semantics, and determining a first target advertisement having semantics matching the target semantics.

[0083] In this step, an advertisement library is pre-set, which includes multiple candidate advertisements and the semantics corresponding to each candidate advertisement. The similarity between the target semantics and the semantics corresponding to the candidate advertisements in the advertisement library is calculated, and the candidate advertisement with the highest similarity is used as the first target advertisement.

[0084] In some implementations, when there are multiple advertisement placement locations, the multiple candidate advertisements may be sorted according to similarity, and a corresponding number of candidate advertisements may be selected as first target advertisements based on the number of advertisement placement locations.

[0085] Step S105: Display the first target advertisement in the display interface to recommend the first target advertisement to the user.

[0086] In this step, at least one advertisement slot to be placed is preset in the display interface, and the first target advertisement is displayed in the advertisement slot to be placed in the display interface to recommend the first target advertisement to the user.

[0087] For example, when a user browses a webpage containing information about vehicle insurance, the system first captures the user's pause on the page, obtaining voice data and product display images from the video. Speech recognition is used to extract keywords such as "premium" and "claim coverage" from the commentary, while simultaneously identifying the vehicle's appearance features in the video frame. After combining the text and visual features, a semantic model is used to generate the core semantics of "professional vehicle insurance." Once a vehicle insurance product with the same semantics is matched in the ad library, the ad is placed in the sidebar of the page the user is currently browsing.

[0088] In this implementation, when the electronic device receives the user's information browsing behavior in the display interface, it obtains the target multimedia information corresponding to the information browsing behavior to quickly obtain the user's current browsing purpose, performs key feature extraction based on the target multimedia information to obtain the target key features, and performs semantic analysis on the target key features to generate target semantics corresponding to the target key features. The target semantics are used to more accurately understand the theme and key information of the user's current browsing content, and advertisement matching is performed based on the target semantics to determine a first target advertisement having semantics that matches the target semantics. The first target advertisement is displayed in the display interface to recommend the first target advertisement to the user. Advertisement matching is performed through the target semantics to quickly and accurately obtain advertisements that are similar to the user's current browsing interests, so that the recommended target advertisements meet the user's current real needs and increase the user's interest in clicking on the target advertisements.

[0089] In some embodiments, as Figure 2 As shown, in step S102 , key features are extracted based on the target multimedia information to obtain target key features, which may include but is not limited to steps S201 to S203 .

[0090] Step S201: Obtain the user's historical information browsing behavior.

[0091] Step S202: Perform user portrait analysis based on the historical information browsing behavior to generate user portrait information associated with the user.

[0092] Step S203: extract key features of the target multimedia information based on the user portrait information to obtain target key features.

[0093] In this implementation, user profile information associated with the user is constructed by analyzing the user's historical browsing behavior. When a user browses new content, the current multimedia content is broken down into multimodal features such as text, images, and audio. The preference tags stored in the user profile are used to calculate the matching degree. A feature dimensionality reduction algorithm is used to filter the multimodal features, retaining those features with a matching degree above a preset threshold with the user profile. This forms a set of key features that reflect the user's potential needs, namely the target key features.

[0094] Specifically, the user's historical information browsing behavior includes the user's historical information browsing behavior on the current display interface and the user's historical information browsing behavior on other display interfaces. The historical information browsing behavior is feature extracted through a clustering algorithm (such as a collaborative filtering algorithm) to form a user tag set, namely user portrait information. The user portrait information includes multi-dimensional information such as user preference tags, behavior habit tags, content attention tags, etc. The key features of the target multimedia information are extracted through the user portrait information, and features with a high degree of match with the user portrait are screened out as the target key features.

[0095] In this embodiment, by constructing user portrait information and introducing historical behavior data as a compensation mechanism in the feature extraction stage, the key features cover both long-term interest preferences and real-time browsing intentions to improve the accuracy of feature extraction, thereby making the target advertisements more in line with the real needs of users.

[0096] In some embodiments, the steps of performing semantic analysis on the target key features to generate target semantics corresponding to the target key features, and the steps of performing advertisement matching based on the target semantics to determine the first target advertisement having semantics that matches the target semantics are performed based on a preset advertisement recommendation model. After performing advertisement matching based on the target semantics to determine the first target advertisement having semantics that matches the target semantics in step S104, the advertisement recommendation method provided in the embodiment of the present application also includes but is not limited to steps S301 to S304.

[0097] Step S301: Acquire user operation information corresponding to the first target advertisement.

[0098] Step S302: Determine the recommendation satisfaction of the first target advertisement based on the user operation information.

[0099] Step S303: When the recommendation satisfaction is greater than or equal to a preset satisfaction threshold, generate a training sample based on the target key feature and the first target advertisement.

[0100] Step S304: training the preset advertisement recommendation model based on the training samples to obtain an updated preset advertisement recommendation model.

[0101] In this implementation, after the first target advertisement is displayed, user operation information on the target first advertisement is collected in real time, such as operation data such as clicks and stays on the target first advertisement. The user's recommendation satisfaction with the first target advertisement is calculated based on the user operation information. When the user's recommendation satisfaction with the first target advertisement reaches a preset satisfaction threshold, the target key features and the first target advertisement are constructed as training samples, and the preset advertisement recommendation model is trained, so that the preset advertisement recommendation model can continuously learn the correlation between user behavior and advertising effect. In the case of insufficient data accumulation, the recommendation strategy is gradually optimized through real-time feedback data, thereby effectively reducing the recommendation deviation caused by the lack of user behavior data.

[0102] Among them, a weighted algorithm can be used to comprehensively calculate the click-through rate and the length of stay to obtain the user's recommendation satisfaction with the first target advertisement. The preset advertisement recommendation model is a machine learning model with semantic analysis and advertisement matching functions, such as a recommendation algorithm based on a deep neural network. The specific preset advertisement recommendation model can be selected according to actual conditions and is not limited here.

[0103] It should be noted that the preset satisfaction threshold can be set according to actual conditions and is not limited here.

[0104] In this implementation, the recommendation results are directly linked to user feedback to form a closed-loop advertising recommendation optimization process. Through incremental training, the preset advertising recommendation model can quickly adapt to changes in user interests and improve the accuracy and timeliness of advertising delivery.

[0105] In some embodiments, as Figure 3 As shown, in step S102 , key features are extracted based on the target multimedia information to obtain target key features, which may include but is not limited to steps S401 to S406 .

[0106] Step S401: When the target multimedia information includes video information, obtain voice information of the video information, and recognize the voice information by using a voice recognition technology to obtain first text information.

[0107] Step S402: input the video information into a target detection model, and the target detection model outputs a target video frame of the video information.

[0108] Step S403: Obtain a target image corresponding to the target video frame in the video information.

[0109] Step S404: Recognize the target image using image recognition technology to obtain second text information.

[0110] Step S405: Obtain the target text information by fusing the first text information and the second text information.

[0111] Step S406: extract key features from the target text information to obtain target key features.

[0112] In this implementation, the target multimedia information may include text information, video information and picture information. When the target multimedia information includes video information, the voice information in the video information is converted into text information, i.e., first text information, through voice recognition technology. The shot switching point is detected by the target detection model, the target key frame in each shot is determined, and the target video image information corresponding to each target key frame is obtained. The target video image information is converted into text information, i.e., second text information, through image recognition technology. The first text information and the second text information can be merged using a text alignment method based on an attention mechanism to obtain the target text information. The text keywords in the target text information are identified through natural language processing technology to extract key features of the target multimedia information and obtain target key features.

[0113] In some other implementations, the video information may further include subtitle information, and the subtitle information may be used as supplementary information for speech recognition, that is, the subtitle information may be used to supplement the first text information.

[0114] The target detection model may be a pre-trained neural network model, for example, a target detection model may be obtained by training an object detection framework based on a convolutional neural network.

[0115] In this implementation, when video information is identified as a processing object, its voice track is converted into text-based dialogue content through a speech recognition model, solving the problem that audio information is difficult to directly participate in text analysis. The target detection model scans the video stream frame by frame to screen out key frames containing high-value information, avoiding redundant processing of non-key frames. The images corresponding to the key frames are identified through image recognition technology and converted into text information describing the content of the picture, thus realizing the textual expression of visual information. The speech recognition text is fused with the image recognition text to form a complete content description covering both audio and visual modes. The fused text is subjected to natural language processing technology for keyword extraction and semantic association analysis to generate a set of key features that can reflect the core theme of the video, avoiding the problem of feature loss caused by ignoring key audio or image content, thereby improving the relevance of advertising recommendations and user satisfaction.

[0116] In some embodiments, as Figure 4 As shown, in step S406 , key features are extracted from the target text information to obtain target key features, which may include but is not limited to steps S501 to S504 .

[0117] Step S501: perform word segmentation processing on the target text information using natural language processing technology to obtain multiple phrases.

[0118] Step S502: input the multiple phrases into a preset intention recognition model, and the preset intention recognition model outputs a recognition result, where the recognition result is used to indicate the user's browsing intention.

[0119] Step S503: Match the multiple phrases in a preset hotspot library to determine multiple hot keywords. The preset hotspot library includes relevant information of people whose influence is greater than a first preset threshold and relevant information of events whose influence is greater than a second preset threshold.

[0120] Step S504: Determine target key features of the target text information based on the recognition result and the multiple hot keywords.

[0121] In this implementation, word segmentation processing using natural language processing technology refers to the process of dividing continuous text into independent semantic units, such as breaking down "New Family Health Insurance Press Conference" into "New", "Family", "Health Insurance", and "Press Conference"; inputting multiple phrases into a preset intent recognition model and matching them with a preset hotspot library, and analyzing the association model between phrases through the preset intent recognition model. For example, the intent recognition result of "Follow Health Insurance Trends" is output, and hot keywords related to "Health Insurance" are screened out through the preset hotspot library, such as "Classic Health Insurance Claims Cases", and finally the intent recognition result is combined with the hot keywords to form a multi-dimensional feature vector containing the user's subjective focus and objective hotspots as the target key features.

[0122] Among them, the preset intent recognition model is a pre-trained semantic classifier based on machine learning training, which can be implemented specifically using a bidirectional long short-term memory network model. The preset hotspot library refers to a database that stores social hotspot information, which can be constructed through high-frequency topic data on social media. For example, the first preset threshold can be set to public figures with more than 1 million fans, and the second preset threshold can be set to hot events with more than 100,000 discussions per day.

[0123] It should be noted that the first preset threshold and the second preset threshold can be set according to actual conditions and are not limited here.

[0124] In this implementation, by integrating real-time intent recognition with hot information matching, key features containing individual preferences and group trends are generated under limited data conditions to accurately identify the user's current focus. At the same time, the recommendation dimension is expanded by combining social hot information, effectively improving the matching accuracy of advertising content and user needs.

[0125] In some implementations, displaying the first target advertisement in the display interface in step S105 may include, but is not limited to, steps S601 to S604.

[0126] Step S601: Determine the eye focus point of the user in the display interface based on the information browsing behavior.

[0127] Step S602: Based on the eye focus point, determine a plurality of areas to be placed for advertisements in the display interface.

[0128] Step S603: Determine a target advertisement delivery area among the plurality of advertisement delivery areas according to the user's historical operation information on advertisement delivery in each display area in the display interface.

[0129] Step S604: Display the first target advertisement in the target advertisement delivery area.

[0130] In this implementation, the eye focus point can refer to the interface position coordinates where the user's current visual attention is focused, the position coordinates where the user's finger stays on the interface for the longest time, or the position coordinates where the user uses an external device such as a mouse to stay on the interface for the longest time. The area for advertising delivery refers to a set of candidate positions in the interface that meet the requirements for advertising display. The interface segmentation algorithm combined with heat map analysis can be used to determine multiple areas for advertising delivery on the current interface to avoid advertisements blocking the core content that the user is browsing. The historical operation information is the user's click-through rate, stay time and interaction frequency data on advertisements in different interface areas.

[0131] In this implementation, when a user browses information, the coordinates of the current gaze point are identified by collecting the user's eye focus in real time, and a heat map of visual attention distribution is generated. According to the heat map, the core content area that the user is currently paying attention to is avoided, and several candidate advertising positions are delineated. The user's historical click-through rate and average stay time data for advertisements in different positions are retrieved to determine the area where the first target advertisement is to be delivered, and the first target advertisement is displayed in the area where the first target advertisement is to be delivered.

[0132] In this implementation, by integrating the user's eye focus in the display interface with historical behavior data, the advertising display position is dynamically adjusted according to the user's real-time attention. At the same time, the area selection strategy is optimized in combination with historical operation data, so that the advertising content can be intelligently delivered in the area around the user's current visual focus, which not only improves the advertising click-through rate, but also avoids interference with the main browsing content, achieving a balance between accurate recommendation and user experience.

[0133] In some implementations, after the first target advertisement is displayed in the display interface in step S105, the advertisement recommendation method provided in the embodiment of the present application further includes but is not limited to steps S701 to S705.

[0134] Step S701: Obtain user behavior data on the first target advertisement.

[0135] Step S702: performing a correlation analysis on the behavior data and the semantics of the first target advertisement to obtain a temporal correlation between the behavior data and the semantics of the first target advertisement.

[0136] Step S703: Input the temporal correlation between the behavior data and the semantics of the first target advertisement into a preset dynamic perception model, and the preset dynamic perception model outputs a perception result, which is used to indicate the dynamic change of the user's interest.

[0137] Step S704: Determine a second target advertisement based on the perception result.

[0138] Step S705: Display the second target advertisement in the display interface to recommend the second target advertisement to the user.

[0139] In this implementation, behavioral data is a record of operations generated during the user's interaction with the advertisement, and may include data such as click-through rate, dwell time, and sliding trajectory, which are used to characterize the user's immediate feedback on the first target advertisement content.

[0140] Temporal correlation refers to the strength of the association between behavioral data and advertising semantics in the time dimension. Specifically, a time series analysis algorithm can be used to calculate the dynamic matching degree between the changing trend of behavioral data and the semantic features of the first target advertisement to capture the trajectory of user interest migration.

[0141] The preset dynamic perception model refers to a machine learning model used to identify dynamic interest patterns. Specifically, it can be implemented using a time series prediction model based on a long short-term memory network. By analyzing time series correlation data, the characteristics of user interest changes are extracted. The perception result refers to the dynamic change trend of user interests.

[0142] In this implementation, when the first target ad is displayed in the user's browsing interface, the user's interactive behavior in the ad area is continuously collected. For example, when a user repeatedly clicks on the product details page in the ad but does not complete the purchase, this behavioral data is temporally associated with the "promotional activity" feature in the ad semantics. By cross-analyzing the timestamp-marked behavioral sequence with the ad keywords, a time-varying interest intensity curve is generated. After this curve is input into the dynamic perception model, the model identifies the trend of increasing user price sensitivity, which in turn triggers an adjustment to the ad recommendation strategy. Based on this trend, an ad for a similar product with a higher discount rate is matched as the second target ad, and the ad content in the interface is immediately updated when the user ends the current browsing operation.

[0143] In this implementation, a dynamic perception mechanism based on time-series correlation analysis is established to quickly identify the direction of user interest migration during the current browsing process, and adjust the advertising delivery strategy accordingly to shorten the response delay of interest change detection and advertising content update, thereby improving advertising click-through rate and user engagement.

[0144] Figure 5 This is a schematic diagram of the structure of the advertisement recommendation device provided in the embodiment of the present application. Figure 5 The embodiment of the present application further provides an advertisement recommendation device 800, which can implement the above-mentioned advertisement recommendation method. The advertisement recommendation device 800 includes:

[0145] An acquisition module 801 is configured to acquire target multimedia information corresponding to an information browsing behavior of a user in a display interface in response to the information browsing behavior of the user;

[0146] An extraction module 802 is configured to extract key features based on the target multimedia information to obtain target key features;

[0147] An analysis module 803 is configured to perform semantic analysis on the target key features to generate target semantics corresponding to the target key features;

[0148] A matching module 804 is configured to perform advertisement matching based on the target semantics, and determine a first target advertisement having semantics that matches the target semantics;

[0149] The display module 805 is configured to display the first target advertisement in the display interface to recommend the first target advertisement to the user.

[0150] In some implementations, the extraction module 802 includes:

[0151] A first acquisition submodule is used to acquire the user's historical information browsing behavior;

[0152] A portrait generation submodule, configured to perform user portrait analysis based on the historical information browsing behavior and generate user portrait information associated with the user;

[0153] The feature extraction submodule is used to extract key features of the target multimedia information based on the user portrait information to obtain target key features.

[0154] In some embodiments, the steps of performing semantic analysis on the target key features to generate target semantics corresponding to the target key features, and the steps of performing advertisement matching based on the target semantics to determine a first target advertisement having semantics matching the target semantics are performed based on a preset advertisement recommendation model. The advertisement recommendation apparatus 800 further includes:

[0155] an operation acquisition module, configured to acquire user operation information corresponding to the first target advertisement;

[0156] a determination module, configured to determine a recommendation satisfaction level of the first target advertisement based on the user operation information;

[0157] A sample generating module, configured to generate a training sample based on the target key feature and the first target advertisement when the recommendation satisfaction is greater than or equal to a preset satisfaction threshold;

[0158] An updating module is used to train the preset advertisement recommendation model based on the training samples to obtain an updated preset advertisement recommendation model.

[0159] In some implementations, the extraction module 802 further includes:

[0160] a first recognition submodule, configured to, when the target multimedia information includes video information, obtain voice information of the video information, and recognize the voice information using a voice recognition technology to obtain first text information;

[0161] a video frame determination submodule, configured to input the video information into a target detection model, and have the target detection model output a target video frame of the video information;

[0162] An image acquisition submodule, configured to acquire a target image corresponding to the target video frame in the video information;

[0163] A second recognition submodule is configured to recognize the target image using image recognition technology to obtain second text information;

[0164] An information fusion submodule, configured to obtain the target text information by fusing the first text information and the second text information;

[0165] The feature extraction submodule is used to extract key features of the target text information to obtain target key features.

[0166] In some embodiments, the feature extraction submodule includes:

[0167] A word segmentation unit is used to perform word segmentation processing on the target text information using natural language processing technology to obtain multiple phrases;

[0168] an intention recognition unit, configured to input the plurality of phrases into a preset intention recognition model, and output a recognition result from the preset intention recognition model, wherein the recognition result is used to indicate the browsing intention of the user;

[0169] a hotspot matching unit, configured to match the plurality of phrases in a preset hotspot library to determine a plurality of hotspot keywords, wherein the preset hotspot library includes information related to persons whose influence is greater than a first preset threshold and information related to events whose influence is greater than a second preset threshold;

[0170] A feature determination unit is used to determine target key features of the target text information based on the recognition result and the multiple hot keywords.

[0171] In some embodiments, the display module 805 includes:

[0172] A focus determination submodule, configured to determine an eye focus point of the user in the display interface based on the information browsing behavior;

[0173] An area determination submodule, configured to determine a plurality of areas to be placed advertisements in the display interface based on the eye focus point;

[0174] A delivery determination submodule, configured to determine a first target advertisement delivery area among the plurality of advertisement delivery areas based on the user's historical operation information on delivering advertisements in each display area of the display interface;

[0175] The advertisement display submodule is configured to display the first target advertisement in the area to be delivered of the first target advertisement.

[0176] In some implementations, the advertisement recommendation device 800 further includes:

[0177] A behavior acquisition module, configured to acquire user behavior data regarding the first target advertisement;

[0178] a semantic analysis module, configured to perform a correlation analysis on the behavior data and the semantics of the first target advertisement to obtain a temporal correlation between the behavior data and the semantics of the first target advertisement;

[0179] a dynamic perception module, configured to input the temporal correlation between the behavior data and the semantics of the first target advertisement into a preset dynamic perception model, and have the preset dynamic perception model output a perception result, wherein the perception result is used to indicate the dynamic change of the user's interest;

[0180] an advertisement determining module, configured to determine a second target advertisement based on the perception result;

[0181] An advertisement display module is configured to display the second target advertisement in the display interface to recommend the second target advertisement to the user.

[0182] The specific implementation of the advertisement recommendation device 800 is substantially the same as the specific embodiment of the advertisement recommendation method described above, and will not be described in detail herein.

[0183] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-described advertising recommendation method when executing the computer program. The electronic device can be any intelligent terminal, including a desktop computer, a tablet computer, a mobile phone, and an in-vehicle computer.

[0184] See also Figure 6 , Figure 6 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. The electronic device includes:

[0185] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0186] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the advertising recommendation method of the embodiments of this application.

[0187] Input / output interface 903, used to implement information input and output;

[0188] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0189] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0190] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0191] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned advertisement recommendation method is implemented.

[0192] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0193] In addition, the embodiments of the present application may be implemented by providing a computer program product. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device implements the advertisement recommendation method in the above embodiment.

[0194] The advertising recommendation method, device, electronic device and medium provided in the embodiments of the present application obtain target multimedia information corresponding to the information browsing behavior when the electronic device receives the information browsing behavior of the user in the display interface, so as to quickly obtain the current browsing purpose of the user, perform key feature extraction based on the target multimedia information to obtain the target key features, perform semantic analysis on the target key features, generate target semantics corresponding to the target key features, more accurately understand the theme and key information of the content currently browsed by the user through the target semantics, perform advertisement matching based on the target semantics, determine a first target advertisement having semantics matching the target semantics, display the first target advertisement in the display interface to recommend the first target advertisement to the user, perform advertisement matching through the target semantics to quickly and accurately obtain advertisements similar to the current browsing interests of the user, so that the recommended target advertisements meet the current real needs of the user and increase the user's interest in clicking on the target advertisements.

[0195] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0196] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0198] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0199] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0200] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0201] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units 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 units, which can be electrical, mechanical or other forms.

[0202] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0204] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0205] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. An advertisement recommendation method, characterized in that: The method comprises: In response to an information browsing behavior of a user in a display interface, obtaining target multimedia information corresponding to the information browsing behavior; Extract key features based on the target multimedia information to obtain target key features; Performing semantic analysis on the target key features to generate target semantics corresponding to the target key features; Performing advertisement matching based on the target semantics, and determining a first target advertisement having semantics matching the target semantics; The first target advertisement is displayed in the display interface to recommend the first target advertisement to the user.

2. The method according to claim 1, characterized in that The extracting key features based on the target multimedia information to obtain the target key features includes: Obtaining the user's historical information browsing behavior; Performing user portrait analysis based on the historical information browsing behavior to generate user portrait information associated with the user; Key features of the target multimedia information are extracted based on the user portrait information to obtain target key features.

3. The method according to claim 1, characterized in that The steps of performing semantic analysis on the target key feature to generate target semantics corresponding to the target key feature, and the steps of performing advertisement matching based on the target semantics to determine a first target advertisement having semantics matching the target semantics are performed based on a preset advertisement recommendation model; After performing advertisement matching based on the target semantics and determining a first target advertisement having semantics matching the target semantics, the method further includes: Acquiring user operation information corresponding to the first target advertisement; determining a recommendation satisfaction level of the first target advertisement based on the user operation information; generating a training sample based on the target key feature and the first target advertisement when the recommendation satisfaction is greater than or equal to a preset satisfaction threshold; The preset advertisement recommendation model is trained based on the training samples to obtain an updated preset advertisement recommendation model.

4. The method according to claim 1, wherein The extracting key features based on the target multimedia information to obtain the target key features includes: When the target multimedia information includes video information, obtaining voice information of the video information, and recognizing the voice information by using a voice recognition technology to obtain first text information; Inputting the video information into a target detection model, and having the target detection model output a target video frame of the video information; Obtaining a target image corresponding to the target video frame in the video information; Recognize the target image using image recognition technology to obtain second text information; Obtaining the target text information by fusing the first text information and the second text information; Key features are extracted from the target text information to obtain target key features.

5. The method according to claim 4, characterized in that The step of extracting key features from the target text information to obtain target key features includes: Performing word segmentation processing on the target text information using natural language processing technology to obtain multiple phrases; Inputting the multiple phrases into a preset intention recognition model, and having the preset intention recognition model output a recognition result, wherein the recognition result is used to indicate the browsing intention of the user; Matching the multiple phrases in a preset hotspot library to determine multiple hot keywords, wherein the preset hotspot library includes relevant information about people whose influence is greater than a first preset threshold and relevant information about events whose influence is greater than a second preset threshold; Determine target key features of the target text information based on the recognition result and the multiple hot keywords.

6. The method according to claim 1, characterized in that The displaying the first target advertisement on the display interface includes: Based on the information browsing behavior, determining the eye focus point of the user in the display interface; Based on the eye focus point, determining a plurality of areas to be placed advertisements in the display interface; Determining a target advertisement placement area among the plurality of advertisement placement areas according to the user's historical operation information on advertisement placement in each display area of the display interface; The first target advertisement is displayed in the target advertisement to be delivered area.

7. The method according to claim 1, characterized in that After displaying the first target advertisement on the display interface to recommend the first target advertisement to the user, the method further includes: Obtaining user behavior data regarding the first target advertisement; performing a correlation analysis on the behavior data and the semantics of the first target advertisement to obtain a temporal correlation between the behavior data and the semantics of the first target advertisement; Inputting the temporal correlation between the behavior data and the semantics of the first target advertisement into a preset dynamic perception model, and having the preset dynamic perception model output a perception result, wherein the perception result is used to indicate a dynamic change in the user's interest; determining a second target advertisement based on the perception result; The second target advertisement is displayed in the display interface to recommend the second target advertisement to the user.

8. An advertisement recommendation device, characterized in that: The device comprises: an acquisition module, configured to acquire target multimedia information corresponding to an information browsing behavior of a user in a display interface in response to the information browsing behavior of the user; An extraction module, configured to extract key features based on the target multimedia information to obtain target key features; An analysis module, configured to perform semantic analysis on the target key features and generate target semantics corresponding to the target key features; a matching module, configured to perform advertisement matching based on the target semantics, and determine a first target advertisement having semantics matching the target semantics; The display module is configured to display the first target advertisement in the display interface to recommend the first target advertisement to the user.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the advertisement recommendation method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the advertisement recommendation method according to any one of claims 1 to 7 is implemented.

Images