Advertisement pushing system and coffee machine with advertisement playing function

Abstract

The invention discloses an advertisement pushing system and a coffee machine with an advertisement playing function, and the system comprises an obtaining module which is used for obtaining the user information of a plurality of coffee purchasing users in a coffee machine putting area; the analysis module is used for analyzing the user information and determining interest tags of a plurality of coffee purchasing users; the advertisement pushing strategy module is used for determining an advertisement pushing strategy based on the interest labels of the coffee buying users; the matching module is used for inputting the advertisement pushing strategy into a pre-trained advertisement pushing model for matching based on the advertisement pushing strategy, and determining a target pushing advertisement of the coffee machine; and the playing module is used for playing the target push advertisement based on the video window of the coffee machine. Personalized demands and interests and preferences of the users are considered, and the advertisement conversion rate is increased; and the use experience and the favor degree of the coffee machine by the user are improved.

Description

Technical Field

[0001] The present invention relates to the field of advertising technology, and in particular to an advertisement push system and a coffee machine with advertisement playing function. Background Art

[0002] In the current commercial advertising landscape, coffee machines, as common commercial devices, are often equipped with ad playback capabilities to achieve advertising promotion and commercial profitability. However, existing coffee machine ad push methods often use crude delivery methods, such as looping fixed ads or simply switching ads based on preset time periods. These methods lack in-depth analysis and precise targeting of user groups within the coffee machine's delivery area.

[0003] This traditional advertising method has significant drawbacks. Firstly, because it fails to consider users' individual needs and interests, the ads delivered often don't match their needs, resulting in low conversion rates and difficulty achieving the desired promotional results. Secondly, for coffee machine operators, ineffective advertising not only wastes advertising resources but can also reduce user experience and favorability with the coffee machines by frequently displaying ads that users aren't interested in, impacting the long-term operation and profitability of the equipment.

[0004] Therefore, there is an urgent need for an advertisement push system and a coffee machine with advertisement playback. Summary of the Invention

[0005] The present invention aims to at least partially address one of the technical problems in the aforementioned technologies. To this end, a first aspect of the present invention is to provide an advertising push system that considers users' personalized needs and interests, improves advertising conversion rates, and enhances users' experience and favorable impressions of coffee machines.

[0006] A second aspect of the present invention is to provide a coffee machine with advertisement playback.

[0007] To achieve the above objectives, a first embodiment of the present invention provides an advertisement push system, comprising:

[0008] An acquisition module is used to obtain user information of several coffee-purchasing users in the coffee machine deployment area;

[0009] An analysis module, configured to analyze the user information and determine interest tags of a number of coffee-purchasing users;

[0010] An advertising push strategy module, configured to determine an advertising push strategy based on the interest tags of the plurality of coffee purchasing users;

[0011] The matching module is used to match the ad push strategy input into the pre-trained ad push model to determine the target ad push for the coffee machine;

[0012] The playback module is used to play the target push advertisement based on the video window of the coffee machine.

[0013] Preferably, the acquisition module includes:

[0014] A first acquisition submodule is used to acquire registration information of a number of coffee purchasing users in the coffee machine placement area as first information;

[0015] The second acquisition submodule is used to obtain the browsing history of several users who purchased coffee in the coffee machine deployment area as the second information;

[0016] The third acquisition submodule is used to obtain the purchase behavior of several coffee-buying users in the coffee machine deployment area as the third information;

[0017] The first determining submodule is configured to use the first information, the second information, and the third information as user information of a plurality of users who purchase coffee in the coffee machine placement area.

[0018] Preferably, the analysis module includes:

[0019] The word segmentation submodule is used to:

[0020] Randomly select a user as the target user; obtain the user information of the target user as the target user information;

[0021] Performing word segmentation processing on the target user information to obtain a first word segmentation set;

[0022] A data cleaning submodule, configured to clean the first word segmentation set to obtain a cleaned first word segmentation set;

[0023] An analysis submodule, configured to analyze the first word segmentation set and determine a number of key entity words;

[0024] Extract submodules for:

[0025] Extract attributes of several key entity words based on a preset attribute extraction model to determine the attributes corresponding to the several key entity words;

[0026] Determining entity relationships between the key entity words based on attributes corresponding to the plurality of key entity words;

[0027] The second determining submodule is configured to:

[0028] Determine the target user's interest tags based on the attributes corresponding to the key entities and key entity words and the entity relationships between the key entity words;

[0029] Traverse the user information of all users who purchased coffee and determine the interest tags of several users who purchased coffee.

[0030] Preferably, the analysis submodule includes:

[0031] Conversion unit for:

[0032] Perform vector conversion on the cleaned first segmentation set to obtain the vector corresponding to each segmentation;

[0033] Take any word as the first word, and take the vector corresponding to the first word as the first vector;

[0034] a calculation unit, configured to determine a discreteness corresponding to the first word segmentation based on the first vector;

[0035] a deleting unit, configured to compare the discreteness with a preset discreteness threshold, select the first participle corresponding to the discreteness being greater than or equal to the preset discreteness threshold as a discrete participle, and delete the discrete participle from the cleaned first participle set to obtain a second participle set;

[0036] Analytical unit for:

[0037] Randomly select a word from the second word set as the second word; use the vector corresponding to the second word as the second vector; calculate the similarity between the second vector and other vectors in the vectors corresponding to the second word set except the second vector, and sum up the similarities to obtain the first similarity corresponding to the second word;

[0038] Comparing the first similarity with a preset similarity threshold, and taking the second participle corresponding to the first similarity being greater than or equal to the preset similarity threshold as the key entity word;

[0039] Traverse all second participles to obtain several key entity words.

[0040] Preferably, the computing unit includes:

[0041] The first calculation subunit is configured to respectively calculate similarities between the first vector and other vectors in the vectors corresponding to the cleaned first word segmentation set except the first vector, to obtain a plurality of second similarities;

[0042] The second computing subunit is configured to:

[0043] The average of several second similarities corresponding to the first participle is used as the first average similarity;

[0044] Traverse all the segmented words in the first segmented word set after cleaning to obtain several first mean similarities;

[0045] The average of several first mean similarities is used as the second mean similarity;

[0046] The determination subunit is configured to determine the discreteness corresponding to the first word segmentation based on the first mean similarity and the second mean similarity.

[0047] Preferably, the data cleaning submodule includes:

[0048] An acquisition unit, configured to acquire a plurality of segmentations corresponding to each sentence in the first segmentation set;

[0049] The third computing unit is configured to:

[0050] Calculate the similarity between the several segmented words corresponding to each sentence and the sensitive words in the preset sensitive word library to determine the sensitive words contained in each sentence;

[0051] Calculate the sensitivity index of each sentence based on a preset algorithm according to the number of segmented words corresponding to each sentence and the sensitive words contained in each sentence;

[0052] comparing the sensitivity index with a preset sensitivity index threshold;

[0053] If the sensitivity index is greater than or equal to the preset sensitivity index threshold, the corresponding statement will be deleted;

[0054] If the sensitivity index is less than the preset sensitivity index threshold, the corresponding statement is used as the target statement to be cleaned;

[0055] Traverse all sentences in the first word segmentation set to obtain several target sentences to be cleaned;

[0056] Data cleaning unit, used for:

[0057] Get preset data cleaning rules;

[0058] The data of the first word segmentation set is cleaned based on the preset data cleaning rule to obtain a cleaned first word segmentation set.

[0059] Preferably, the preset algorithm includes:

[0060]

[0061] Among them, P c Indicates the sensitivity index corresponding to the c-th statement; N c Indicates the total number of participles in the cth sentence; N t represents the total number of sensitive words in the cth sentence; e represents a natural constant; T c represents the sum of the frequencies of all the participles in the cth sentence; T i Indicates the frequency of occurrence of the i-th participle; T kIndicates the frequency of occurrence of the kth sensitive word in the cth sentence.

[0062] Preferably, the advertisement push strategy module includes:

[0063] The fifth acquisition submodule is used to obtain the frequency of coffee purchases by each coffee-buying user within a preset time period;

[0064] A third determining submodule is configured to determine a first frequency of the interest tag corresponding to each coffee purchasing user based on the product of the interest tag of each coffee purchasing user and the frequency of coffee purchasing within a preset time period;

[0065] The fourth determining submodule is configured to:

[0066] Determining a second frequency for each interest tag based on the first frequency to obtain a plurality of second frequencies;

[0067] Comparing the plurality of second frequencies with the preset frequency thresholds respectively, and taking the interest tags corresponding to when the second frequencies are greater than or equal to the preset frequency thresholds as target interest tags, to obtain the plurality of target interest tags;

[0068] The advertisement push strategy determination submodule is used to determine the advertisement push strategy based on a plurality of target interest tags and a second frequency corresponding to the plurality of target interest tags.

[0069] Preferably, the training method of the advertisement push model includes:

[0070] Obtain the ad push training dataset;

[0071] Train the neural network model based on the advertising push training data set to obtain the initial advertising push model;

[0072] Get the ad push test dataset;

[0073] The initial advertising push model is tested based on the advertising push test data set. When the test passes, the final advertising push model is obtained.

[0074] To achieve the above-mentioned purpose, a second embodiment of the present invention provides a coffee machine with advertising playback, comprising: a coffee machine body, an advertising display and an advertising push system; the advertising display is embedded in the coffee machine body.

[0075] The present invention proposes an advertising push system and a coffee machine with advertising playback. The system collects information about users who purchase coffee in the area where the coffee machine is placed through an acquisition module, and combines this with interest tags determined by an analysis module and a strategy formulated by an advertising push strategy module to accurately locate user needs. Compared with traditional extensive advertising placement, the system closely matches advertising content with user interests, making it easier for advertisements to attract user attention, thereby effectively improving advertising conversion rates and bringing higher returns on investment to advertisers. The advertising push strategy module determines push strategies based on user interest tags, and uses a matching module to screen target push advertisements in a pre-trained advertising push model, thereby avoiding the playback of invalid advertisements. This precise advertising delivery method reduces the waste of advertising resources, lowers the operating costs of coffee machine operators during the advertising delivery process, and improves resource utilization efficiency. The playback module pushes ads based on precise matching targets and plays them in the coffee machine video window, avoiding pushing irrelevant or uninteresting ads to users and reducing user resistance to ads. High-quality advertising content that fits the user's interests can enhance the user's overall experience when using the coffee machine, thereby increasing the user's favorability and stickiness with the coffee machine equipment, and contributing to the long-term stable operation of the equipment. The system analyzes each user's information and assigns interest tags, which can meet the diverse needs of different users and realize personalized advertising push services. This personalized service model makes advertising push more targeted and flexible, adapts to the development trend of increasingly diversified user needs in the market, and enhances the market competitiveness of the advertising push system.

[0076] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0077] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 is a block diagram of an advertisement push system according to an embodiment of the present invention;

[0080] Figure 2 is a block diagram of an acquisition module according to one embodiment of the present invention;

[0081] Figure 3 is a block diagram of an analysis module according to one embodiment of the present invention. DETAILED DESCRIPTION

[0082] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0083] Example 1

[0084] like Figure 1 As shown, an advertisement push system includes:

[0085] An acquisition module is used to obtain user information of several coffee-purchasing users in the coffee machine deployment area;

[0086] An analysis module, configured to analyze the user information and determine interest tags of a number of coffee-purchasing users;

[0087] An advertising push strategy module, configured to determine an advertising push strategy based on the interest tags of the plurality of coffee purchasing users;

[0088] The matching module is used to match the ad push strategy input into the pre-trained ad push model to determine the target ad push for the coffee machine;

[0089] The playback module is used to play the target push advertisement based on the video window of the coffee machine.

[0090] The working principle of the above technical solution is as follows: the acquisition module first takes effect, collecting user information about coffee purchasers in the coffee machine deployment area. This user information covers multi-dimensional data such as basic user attributes (such as age and gender), consumption behavior (such as frequency of coffee purchases, preferred coffee types), etc., providing raw data support for subsequent analysis and advertising push; the user information is analyzed to determine the interest tags of several coffee purchasers; an advertising push strategy is determined based on the interest tags of several coffee purchasers; the advertising push strategy is input into a pre-trained advertising push model for matching, and the target push advertisement for the coffee machine is determined; and the target push advertisement is played based on the video window of the coffee machine.

[0091] The beneficial effects of the above technical solution are: by collecting information about users who purchase coffee in the coffee machine placement area through the acquisition module, combined with the interest tags determined by the analysis module and the strategy formulated by the advertising push strategy module, user needs can be accurately located; compared with traditional extensive advertising, the system closely matches the advertising content with user interests, making it easier for advertisements to attract user attention, thereby effectively improving the advertising conversion rate and bringing higher return on investment to advertisers; the advertising push strategy module determines the push strategy based on user interest tags, and uses the matching module to screen target push advertisements in the pre-trained advertising push model, avoiding the broadcast of invalid advertisements. This precise advertising delivery method reduces the waste of advertising resources, lowers the operating costs of coffee machine operators during the advertising delivery process, and improves resource utilization efficiency. The playback module pushes ads based on precise matching targets and plays them in the coffee machine video window, avoiding pushing irrelevant or uninteresting ads to users and reducing user resistance to ads. High-quality advertising content that fits the user's interests can enhance the user's overall experience when using the coffee machine, thereby increasing the user's favorability and stickiness with the coffee machine equipment, and contributing to the long-term stable operation of the equipment. The system analyzes each user's information and assigns interest tags, which can meet the diverse needs of different users and realize personalized advertising push services. This personalized service model makes advertising push more targeted and flexible, adapts to the development trend of increasingly diversified user needs in the market, and enhances the market competitiveness of the advertising push system.

[0092] Example 2

[0093] like Figure 2 As shown, the acquisition module includes:

[0094] A first acquisition submodule is used to acquire registration information of a number of coffee purchasing users in the coffee machine placement area as first information;

[0095] The second acquisition submodule is used to obtain the browsing history of several users who purchased coffee in the coffee machine deployment area as the second information;

[0096] The third acquisition submodule is used to obtain the purchase behavior of several coffee-buying users in the coffee machine deployment area as the third information;

[0097] The first determining submodule is configured to use the first information, the second information, and the third information as user information of a plurality of users who purchase coffee in the coffee machine placement area.

[0098] In this embodiment, the specific implementation method of browsing history and purchase behavior information is that during the user registration process, various permission options will pop up to remind the user to make a choice. Only when the user selects the permission to allow the acquisition of browsing history information and the permission to obtain purchase behavior information can the user's relevant information be obtained; when the user chooses to refuse to obtain permission, the acquisition module will not be able to obtain the user's relevant information.

[0099] The beneficial effects of this technical solution are as follows: the first acquisition submodule captures user registration information, covering basic attributes such as age, gender, and occupation, providing a framework for user profiling; the second acquisition submodule collects browsing history, which can reflect the user's potential interests and preferences; and the third acquisition submodule records purchase behavior, intuitively demonstrating the user's consumption tendencies. These three submodules work together to collect data from different dimensions, comprehensively covering user characteristics. Compared with single data collection methods, the user information obtained is more complete and multi-dimensional, providing rich data support for subsequent precise analysis.

[0100] Example 3

[0101] like Figure 3 As shown, the analysis module includes:

[0102] The word segmentation submodule is used to:

[0103] Randomly select a user as the target user; obtain the user information of the target user as the target user information;

[0104] Performing word segmentation processing on the target user information to obtain a first word segmentation set;

[0105] A data cleaning submodule, configured to clean the first word segmentation set to obtain a cleaned first word segmentation set;

[0106] An analysis submodule, configured to analyze the first word segmentation set and determine a number of key entity words;

[0107] Extract submodules for:

[0108] Extract attributes of several key entity words based on a preset attribute extraction model to determine the attributes corresponding to the several key entity words;

[0109] Determining entity relationships between the key entity words based on attributes corresponding to the plurality of key entity words;

[0110] The second determining submodule is configured to:

[0111] Determine the target user's interest tags based on the attributes corresponding to the key entities and key entity words and the entity relationships between the key entity words;

[0112] Traverse the user information of all users who purchased coffee and determine the interest tags of several users who purchased coffee.

[0113] In this embodiment, the preset attribute extraction model is a pre-trained model used to extract attributes, and the specific attribute extraction model is not limited.

[0114] The working principle of the above technical solution is: segment the text information in the user information to determine the first segmentation set; perform data cleaning on the first segmentation set to improve the accuracy of the data in the first segmentation set; analyze and screen the segmentations in the first segmentation set to determine a number of key entity words; extract attributes of the key entity words to determine the attributes corresponding to the key entity words; determine the entity relationship between the key entity words based on the attributes corresponding to the key entity words; determine the interest tags of the target user by combining the key entities, the attributes corresponding to the key entity words and the entity relationships between the key entity words.

[0115] The beneficial effects of the above technical solution are as follows: the word segmentation submodule performs detailed word segmentation on user information, the data cleaning submodule removes interference information, the analysis submodule identifies key entity words, the extraction submodule further defines the attributes and entity relationships of key entity words, and finally, the second determination submodule generates interest tags. This series of steps, which progresses step by step, accurately captures user interests and ensures that the generated interest tags are more closely aligned with the user's true interests and needs. Pushing ads based on these precise interest tags significantly improves the relevance of ads to users, boosting their appeal and conversion rates. This approach not only focuses on the surface vocabulary in user information but also deeply explores the underlying meaning and connections behind user information by identifying key entity words, their attributes, and their entity relationships. This in-depth analysis adds more dimensions to user profiles, giving the ad push system a more comprehensive and in-depth understanding of users. For example, by analyzing the entity relationships within a user's coffee brands, flavor preferences, and related browsing history, it can uncover potential consumer upgrade needs or interest in related products, providing a richer content selection for ad push. The use of a pre-set attribute extraction model enables the analysis module to automatically process and analyze user information, improving analysis efficiency and accuracy. At the same time, this model-based analysis approach is versatile and scalable, adapting to user information of varying types and formats. As user data accumulates and changes, the system can continuously improve its analysis of user interests through model optimization and adjustment, maintaining a high level of adaptability and intelligence. Accurate interest tags mean users will see more ads that align with their interests, reducing the distraction of irrelevant ads. This not only enhances the user experience while using the coffee machine, but also allows users to feel the system's attention and respect for their personalized needs, thereby increasing user satisfaction and favorability with the ad push system and the coffee machine itself, and promoting long-term user usage and engagement.

[0116] Example 4

[0117] Conversion unit for:

[0118] Perform vector conversion on the cleaned first segmentation set to obtain the vector corresponding to each segmentation;

[0119] Take any word as the first word, and take the vector corresponding to the first word as the first vector;

[0120] a calculation unit, configured to determine a discreteness corresponding to the first word segmentation based on the first vector;

[0121] a deleting unit, configured to compare the discreteness with a preset discreteness threshold, select the first participle corresponding to the discreteness being greater than or equal to the preset discreteness threshold as a discrete participle, and delete the discrete participle from the cleaned first participle set to obtain a second participle set;

[0122] Analytical unit for:

[0123] Randomly select a word from the second word set as the second word; use the vector corresponding to the second word as the second vector; calculate the similarity between the second vector and other vectors in the vectors corresponding to the second word set except the second vector, and sum up the similarities to obtain the first similarity corresponding to the second word;

[0124] Comparing the first similarity with a preset similarity threshold, and taking the second participle corresponding to the first similarity being greater than or equal to the preset similarity threshold as the key entity word;

[0125] Traverse all second participles to obtain several key entity words.

[0126] In this embodiment, the similarity may be calculated using cosine similarity or Euclidean distance.

[0127] In this embodiment, the preset similarity threshold is specifically set based on industry experience.

[0128] The beneficial effects of the above technical solution are: the conversion unit converts the cleaned first segmentation set into a vector, and the calculation unit determines the discreteness of each segmentation; the deletion unit deletes the discrete segmentations according to the preset discreteness threshold to obtain the second segmentation set; this can remove information that is irrelevant to the user's core interests and has high discreteness, so that the data is more focused on the user's key points of interest and improves data quality; the analysis unit calculates the similarity between the second segmentation vector and other vectors to screen out key entity words with high similarity to other segmentations. This helps the system focus on more representative and relevant information in the user's interests and further improves the relevance of the data; by removing discrete segmentations and screening key entity words, the user's interest characteristics can be more accurately reflected; the interest tags determined based on these key entity words can more accurately depict the user's interest preferences, making advertising push more in line with user needs; removing discrete information and screening key entity words can effectively reduce the impact of interference factors on interest tags, avoid interest tag deviations caused by irrelevant information, and improve the accuracy and reliability of interest tags.

[0129] Example 5

[0130] Computing unit, including:

[0131] The first calculation subunit is configured to respectively calculate similarities between the first vector and other vectors in the vectors corresponding to the cleaned first word segmentation set except the first vector, to obtain a plurality of second similarities;

[0132] The second computing subunit is configured to:

[0133] The average of several second similarities corresponding to the first participle is used as the first average similarity;

[0134] Traverse all the segmented words in the first segmented word set after cleaning to obtain several first mean similarities;

[0135] The average of several first mean similarities is used as the second mean similarity;

[0136] The determination subunit is configured to determine the discreteness corresponding to the first word segmentation based on the first mean similarity and the second mean similarity.

[0137] In this embodiment,

[0138]

[0139] Among them, F represents the discreteness corresponding to the first participle; represents the first mean similarity; Represents the second mean similarity.

[0140] The beneficial effect of the above technical solution is that the first calculation subunit calculates the similarity between the first vector and other vectors to obtain several second similarities, which provides basic data for evaluating the degree of association between the first participle and other participles. The second calculation subunit further takes the average of the several second similarities corresponding to the first participle as the first mean similarity, traverses all participles to obtain multiple first mean similarities, and then calculates the average to obtain the second mean similarity. This multi-level calculation method can comprehensively consider the similarity relationship between each participle and other participles, as well as the overall average of the similarities between all participles, so as to measure the discreteness of each participle more comprehensively and accurately. Compared with a single calculation method, the result is more reliable; the discreteness obtained by comprehensive calculation can more effectively identify those relatively discrete participles with low correlation with other participles. In subsequent processing, the deletion unit can more accurately remove these discrete participles based on the comparison of the discreteness with the preset threshold, retaining the participles with higher relevance to user interests, thereby improving the accuracy of key entity word screening. This helps build more accurate user interest tags, making advertising push more in line with users' real interests and needs, and improving the relevance and conversion rate of advertising.

[0141] Example 6

[0142] Data cleaning submodule, including:

[0143] An acquisition unit, configured to acquire a plurality of segmentations corresponding to each sentence in the first segmentation set;

[0144] The third computing unit is configured to:

[0145] Calculate the similarity between the several segmented words corresponding to each sentence and the sensitive words in the preset sensitive word library to determine the sensitive words contained in each sentence;

[0146] Calculate the sensitivity index of each sentence based on a preset algorithm according to the number of segmented words corresponding to each sentence and the sensitive words contained in each sentence;

[0147] comparing the sensitivity index with a preset sensitivity index threshold;

[0148] If the sensitivity index is greater than or equal to the preset sensitivity index threshold, the corresponding statement will be deleted;

[0149] If the sensitivity index is less than the preset sensitivity index threshold, the corresponding statement is used as the target statement to be cleaned;

[0150] Traverse all sentences in the first word segmentation set to obtain several target sentences to be cleaned;

[0151] Data cleaning unit, used for:

[0152] Get preset data cleaning rules;

[0153] The data of the first word segmentation set is cleaned based on the preset data cleaning rule to obtain a cleaned first word segmentation set.

[0154] The beneficial effect of the above technical solution is that the third calculation unit calculates the similarity between the word segmentation of each sentence and sensitive words in the preset sensitive word library, effectively identifying sentences containing sensitive information. When the sensitivity index is greater than or equal to the preset sensitivity index threshold, the corresponding sentence is deleted, preventing the user's sensitive information (such as personal identity information, private data, etc.) from being leaked or improperly used, thereby protecting the user's privacy and information security and enhancing the user's trust in the system.

[0155] Example 7

[0156] Preset algorithms, including:

[0157]

[0158] Among them, P c Indicates the sensitivity index corresponding to the c-th statement; N c Indicates the total number of participles in the cth sentence; N t represents the total number of sensitive words in the cth sentence; e represents a natural constant; T c represents the sum of the frequencies of all the participles in the cth sentence; T i Indicates the frequency of occurrence of the i-th participle; T k Indicates the frequency of occurrence of the kth sensitive word in the cth sentence.

[0159] Example 8

[0160] Advertisement push strategy module, including:

[0161] The fifth acquisition submodule is used to obtain the frequency of coffee purchases by each coffee-buying user within a preset time period;

[0162] A third determining submodule is configured to determine a first frequency of the interest tag corresponding to each coffee purchasing user based on the product of the interest tag of each coffee purchasing user and the frequency of coffee purchasing within a preset time period;

[0163] The fourth determining submodule is configured to:

[0164] Determining a second frequency for each interest tag based on the first frequency to obtain a plurality of second frequencies;

[0165] Comparing the plurality of second frequencies with the preset frequency thresholds respectively, and taking the interest tags corresponding to when the second frequencies are greater than or equal to the preset frequency thresholds as target interest tags, to obtain the plurality of target interest tags;

[0166] The advertisement push strategy determination submodule is used to determine the advertisement push strategy based on a plurality of target interest tags and a second frequency corresponding to the plurality of target interest tags.

[0167] In this embodiment, a plurality of target interest tags and a plurality of second frequencies corresponding to the target interest tags are used as an advertisement push strategy.

[0168] The above technical solution operates as follows: The fifth acquisition submodule obtains the frequency of coffee purchases by each coffee-purchasing user within a preset time period. This preset time period can be set based on actual circumstances, such as a week or a month. By collecting purchase frequency data, it is possible to understand the user's coffee consumption activity and provide basic data for subsequent analysis. The third determination submodule determines the first frequency of each coffee-purchasing user's interest tag by multiplying the frequency of coffee purchases within the preset time period. The logic here is that the more frequent a user's purchases, the greater the impact of the interests represented by their interest tags on their behavior. By multiplying the two, the correlation between each user's interest tag and their purchasing behavior is quantified, resulting in the first frequency of each user's interest tag. The fourth determination submodule first statistically determines the second frequency of each interest tag based on the first frequency of each user's interest tag. Specifically, the first frequency of the same interest tag across all users is accumulated to obtain a number of second frequencies. This provides a comprehensive understanding of the popularity of each interest tag within the user population. Then, several second frequencies are compared with the preset frequency threshold. When the second frequency is greater than or equal to the preset frequency threshold, it means that the interest tag has a high popularity and attention in the user group, and it is used as the target interest tag, thereby obtaining several target interest tags. The preset frequency threshold is a standard set according to actual conditions and business needs, and is used to filter out important interest tags. The advertising push strategy determination submodule determines the advertising push strategy based on several target interest tags and the second frequencies corresponding to several target interest tags. Specifically, the target interest tags reflect the areas of general interest of the user group, and the second frequency reflects the popularity of these interest tags. Based on this information, the system can formulate corresponding advertising push strategies. For example, for target interest tags with high popularity, the push frequency and exposure of related advertisements are increased; for different target interest tags, appropriate advertising content is selected for push to increase the attractiveness and conversion rate of the advertisements and achieve accurate advertising delivery.

[0169] Example 9

[0170] The training method for the ad push model includes:

[0171] Obtain the ad push training dataset;

[0172] Train the neural network model based on the advertising push training data set to obtain the initial advertising push model;

[0173] Get the ad push test dataset;

[0174] The initial advertising push model is tested based on the advertising push test data set. When the test passes, the final advertising push model is obtained.

[0175] To achieve the above-mentioned purpose, a second embodiment of the present invention provides a coffee machine with advertising playback, comprising: a coffee machine body, an advertising display and an advertising push system; the advertising display is embedded in the coffee machine body.

[0176] The beneficial effects of the above technical solution are: by collecting information about users who purchase coffee in the coffee machine placement area through the acquisition module, combined with the interest tags determined by the analysis module and the strategy formulated by the advertising push strategy module, user needs can be accurately located; compared with traditional extensive advertising, the system closely matches the advertising content with user interests, making it easier for advertisements to attract user attention, thereby effectively improving the advertising conversion rate and bringing higher return on investment to advertisers; the advertising push strategy module determines the push strategy based on user interest tags, and uses the matching module to screen target push advertisements in the pre-trained advertising push model, avoiding the broadcast of invalid advertisements. This precise advertising delivery method reduces the waste of advertising resources, lowers the operating costs of coffee machine operators during the advertising delivery process, and improves resource utilization efficiency. The playback module pushes ads based on precise matching targets and plays them in the coffee machine video window, avoiding pushing irrelevant or uninteresting ads to users and reducing user resistance to ads. High-quality advertising content that fits the user's interests can enhance the user's overall experience when using the coffee machine, thereby increasing the user's favorability and stickiness with the coffee machine equipment, and contributing to the long-term stable operation of the equipment. The system analyzes each user's information and assigns interest tags, which can meet the diverse needs of different users and realize personalized advertising push services. This personalized service model makes advertising push more targeted and flexible, adapts to the development trend of increasingly diversified user needs in the market, and enhances the market competitiveness of the advertising push system.

[0177] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An advertisement push system, characterized in that: include: An acquisition module is used to obtain user information of several coffee-purchasing users in the coffee machine deployment area; An analysis module, configured to analyze the user information and determine interest tags of a number of coffee-purchasing users; An advertising push strategy module, configured to determine an advertising push strategy based on the interest tags of the plurality of coffee purchasing users; The matching module is used to match the ad push strategy input into the pre-trained ad push model to determine the target ad push for the coffee machine; The playback module is used to play the target push advertisement based on the video window of the coffee machine.

2. The advertisement push system according to claim 1, wherein: Get modules, including: A first acquisition submodule is used to acquire registration information of a number of coffee purchasing users in the coffee machine placement area as first information; The second acquisition submodule is used to obtain the browsing history of several users who purchased coffee in the coffee machine deployment area as the second information; The third acquisition submodule is used to obtain the purchase behavior of several coffee-buying users in the coffee machine deployment area as the third information; The first determining submodule is configured to use the first information, the second information, and the third information as user information of a plurality of users who purchase coffee in the coffee machine placement area.

3. The advertisement push system according to claim 1, wherein: Analysis modules, including: The word segmentation submodule is used to: Randomly select a user as the target user; obtain the user information of the target user as the target user information; Performing word segmentation processing on the target user information to obtain a first word segmentation set; A data cleaning submodule, configured to clean the first word segmentation set to obtain a cleaned first word segmentation set; An analysis submodule, configured to analyze the first word segmentation set and determine a number of key entity words; Extract submodules for: Extract attributes of several key entity words based on a preset attribute extraction model to determine the attributes corresponding to the several key entity words; Determining entity relationships between the key entity words based on attributes corresponding to the plurality of key entity words; The second determining submodule is configured to: Determine the target user's interest tags based on the attributes corresponding to the key entities and key entity words and the entity relationships between the key entity words; Traverse the user information of all users who purchased coffee and determine the interest tags of several users who purchased coffee.

4. The advertisement push system according to claim 3, wherein: Analysis submodules include: Conversion unit for: Perform vector conversion on the cleaned first segmentation set to obtain the vector corresponding to each segmentation; Take any word as the first word, and take the vector corresponding to the first word as the first vector; a calculation unit, configured to determine a discreteness corresponding to the first word segmentation based on the first vector; a deleting unit, configured to compare the discreteness with a preset discreteness threshold, select the first participle corresponding to the discreteness being greater than or equal to the preset discreteness threshold as a discrete participle, and delete the discrete participle from the cleaned first participle set to obtain a second participle set; Analytical unit for: Randomly select a word from the second word set as the second word; use the vector corresponding to the second word as the second vector; calculate the similarity between the second vector and other vectors in the vectors corresponding to the second word set except the second vector, and sum up the similarities to obtain the first similarity corresponding to the second word; Comparing the first similarity with a preset similarity threshold, and taking the second participle corresponding to the first similarity being greater than or equal to the preset similarity threshold as the key entity word; Traverse all second participles to obtain several key entity words.

5. The advertisement push system according to claim 4, wherein: Computing unit, including: The first calculation subunit is configured to respectively calculate similarities between the first vector and other vectors in the vectors corresponding to the cleaned first word segmentation set except the first vector, to obtain a plurality of second similarities; The second computing subunit is configured to: The average of several second similarities corresponding to the first participle is used as the first average similarity; Traverse all the segmented words in the first segmented word set after cleaning to obtain several first mean similarities; The average of several first mean similarities is used as the second mean similarity; The determination subunit is configured to determine the discreteness corresponding to the first word segmentation based on the first mean similarity and the second mean similarity.

6. The advertisement push system according to claim 5, wherein: Data cleaning submodule, including: An acquisition unit, configured to acquire a plurality of segmentations corresponding to each sentence in the first segmentation set; The third computing unit is configured to: Calculate the similarity between the several segmented words corresponding to each sentence and the sensitive words in the preset sensitive word library to determine the sensitive words contained in each sentence; Calculate the sensitivity index of each sentence based on a preset algorithm according to the number of segmented words corresponding to each sentence and the sensitive words contained in each sentence; comparing the sensitivity index with a preset sensitivity index threshold; If the sensitivity index is greater than or equal to the preset sensitivity index threshold, the corresponding statement will be deleted; If the sensitivity index is less than the preset sensitivity index threshold, the corresponding statement is used as the target statement to be cleaned; Traverse all sentences in the first word segmentation set to obtain several target sentences to be cleaned; Data cleaning unit, used for: Get preset data cleaning rules; The data of the first word segmentation set is cleaned based on the preset data cleaning rule to obtain a cleaned first word segmentation set.

7. The advertisement push system according to claim 6, wherein: Preset algorithms, including: Among them, P c Indicates the sensitivity index corresponding to the c-th statement; N c Indicates the total number of participles in the cth sentence; N t represents the total number of sensitive words in the cth sentence; e represents a natural constant; T c represents the sum of the frequencies of all the participles in the cth sentence; T i Indicates the frequency of occurrence of the i-th participle; T k Indicates the frequency of occurrence of the kth sensitive word in the cth sentence.

8. The advertisement push system according to claim 1, wherein: Advertisement push strategy module, including: The fifth acquisition submodule is used to obtain the frequency of coffee purchases by each coffee-buying user within a preset time period; A third determining submodule is configured to determine a first frequency of the interest tag corresponding to each coffee purchasing user based on the product of the interest tag of each coffee purchasing user and the frequency of coffee purchasing within a preset time period; The fourth determining submodule is configured to: Determining a second frequency for each interest tag based on the first frequency to obtain a plurality of second frequencies; Comparing the plurality of second frequencies with the preset frequency thresholds respectively, and taking the interest tags corresponding to when the second frequencies are greater than or equal to the preset frequency thresholds as target interest tags, to obtain the plurality of target interest tags; The advertisement push strategy determination submodule is used to determine the advertisement push strategy based on a plurality of target interest tags and a second frequency corresponding to the plurality of target interest tags.

9. The advertisement push system according to claim 1, wherein: The training method for the ad push model includes: Obtain the ad push training dataset; The neural network model is trained based on the advertising push training data set to obtain the initial advertising push model; Get the ad push test dataset; The initial advertising push model is tested based on the advertising push test data set. When the test passes, the final advertising push model is obtained.

10. A coffee machine with advertising playback, characterized in that: include: A coffee machine body, an advertising display, and an advertising push system according to any one of claims 1 to 9; The advertising display is embedded in the coffee machine body.

Images