CN117114772A

CN117114772A - Method, device, equipment and storage medium for mining put-in materials

Info

Publication number: CN117114772A
Application number: CN202311154693.5A
Authority: CN
Inventors: 武云珩
Original assignee: Shenzhen Dianhu Data Information Technology Co ltd
Current assignee: Shenzhen Dianhu Data Information Technology Co ltd
Priority date: 2023-09-08
Filing date: 2023-09-08
Publication date: 2023-11-24

Abstract

The invention relates to the technical field of Internet advertisements, and discloses a method, a device, equipment and a storage medium for excavating materials, which are used for improving the accuracy of a material release strategy. The method comprises the following steps: e-commerce advertisement monitoring is carried out on the plurality of target social media to obtain advertisement content data and advertisement interaction data; performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data; extracting key information from advertisement text data and advertisement image data to obtain a material key information set; extracting features and integrating the features of the advertisement interaction data to obtain target interaction features; establishing a material characteristic relation set between the target interaction characteristic and the material key information set; inputting the story feature relation set into a preset story throwing mining model to analyze the story throwing effect, and obtaining a story throwing effect analysis result; and constructing an optimal material delivery strategy corresponding to the E-commerce advertisement according to the analysis result of the material delivery effect.

Description

Method, device, equipment and storage medium for mining put-in materials

Technical Field

The present invention relates to the field of internet advertisement technologies, and in particular, to a method, an apparatus, a device, and a storage medium for mining delivered materials.

Background

With the continued development of the internet and social media, electronic commerce has grown rapidly worldwide. The electronic commerce platform and advertisers disputes turn attention to social media to promote and sell products. In this digitized era, the effectiveness of advertising material is critical to attracting potential customers, increasing conversion rates, and increasing sales.

However, social media advertising presents more challenges than traditional advertising, including vigorous competition, constantly changing user behavior and trends, algorithmic changes in the advertising platform, and the like. In this case, in order to stand out and succeed in social media, advertisers need to know exactly which advertising material can appeal to the audience and how to adjust advertising strategies to improve the impression.

Disclosure of Invention

The invention provides a method, a device, equipment and a storage medium for excavating a material, which are used for improving the accuracy of a material throwing strategy.

The first aspect of the invention provides a method for excavating put materials, which comprises the following steps:

e-commerce advertisement monitoring is carried out on the plurality of target social media to obtain advertisement content data and advertisement interaction data;

Performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data;

extracting key information from the advertisement text data and the advertisement image data to obtain a material key information set;

extracting features and integrating the features of the advertisement interaction data to obtain target interaction features;

establishing a material characteristic relation set between the target interaction characteristic and the material key information set;

inputting the material characteristic relation set into a preset material throwing mining model to analyze the material throwing effect, and obtaining a material throwing effect analysis result;

and constructing an optimal material delivery strategy corresponding to the E-commerce advertisement according to the analysis result of the material delivery effect.

With reference to the first aspect, in a first implementation manner of the first aspect of the present invention, the performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data includes:

carrying out named entity recognition on the advertisement content data based on a preset natural language processing model to obtain a plurality of named entities;

performing entity text conversion on the advertisement content data according to the plurality of named entities to generate corresponding advertisement text data;

Carrying out image feature extraction and image classification on the advertisement content data based on a preset image processing model to obtain image feature data;

and carrying out semantic segmentation on the image characteristic data to obtain advertisement image data.

With reference to the first aspect, in a second implementation manner of the first aspect of the present invention, the extracting key information from the advertisement text data and the advertisement image data to obtain a material key information set includes:

extracting keywords from the advertisement text data to obtain advertisement keywords, analyzing emotion tendencies of the advertisement text data to obtain advertisement emotion tendencies, and performing topic modeling on the advertisement text data to obtain advertisement topic information;

detecting the advertisement image data to obtain an advertisement image object, and performing color analysis on the advertisement image data to obtain advertisement image color;

and generating a material key information set according to the advertisement keywords, the advertisement emotion tendentiousness, the advertisement theme information, the advertisement image object and the advertisement image color.

With reference to the first aspect, in a third implementation manner of the first aspect of the present invention, performing feature extraction and feature integration on the advertisement interaction data to obtain a target interaction feature includes:

Performing data preprocessing on the advertisement interaction data to obtain standard interaction data;

analyzing the interaction quantity of the standard interaction data to obtain interaction quantity characteristics, wherein the interaction quantity characteristics comprise praise numbers, comment numbers, sharing numbers and click numbers;

calculating the interaction frequency data of the interaction quantity features to obtain interaction frequency features;

trend analysis is carried out on the interactive quantity features based on a time sequence analysis method, so that interactive trend features are obtained;

performing user participation analysis on the standard interaction data to obtain user participation characteristics;

and carrying out feature integration on the interaction quantity feature, the interaction frequency feature, the interaction trend feature and the user participation feature to obtain a target interaction feature.

With reference to the first aspect, in a fourth implementation manner of the first aspect of the present invention, the establishing a material feature relation set between the target interaction feature and the material key information set includes:

integrating the target interaction characteristics and the material key information set into a target data table, wherein each advertisement in the target data table corresponds to the interaction characteristics and the material key information;

Carrying out correlation analysis on the target interaction characteristics and the material key information set according to the target data table to obtain a target correlation coefficient;

and creating a material characteristic relation set between the target interaction characteristic and the material key information set according to the target correlation coefficient.

With reference to the first aspect, in a fifth implementation manner of the first aspect of the present invention, inputting the feature relation set of the material into a preset material-throwing mining model to perform material-throwing effect analysis, to obtain a material-throwing effect analysis result, includes:

performing feature coding on the material feature relation set to obtain a plurality of feature coding values;

vector conversion is carried out on the plurality of characteristic coding values to obtain a target input vector;

inputting the target input vector into a preset input material mining model, wherein the input material mining model comprises: two layers of long-short-time memory networks and two layers of fully-connected networks;

extracting features of the target input vector through the two layers of long-short-term memory networks to obtain a target feature vector;

and inputting the target feature vector into the two-layer fully-connected network to perform material throwing effect analysis, and obtaining a material throwing effect analysis result.

With reference to the first aspect, in a sixth implementation manner of the first aspect of the present invention, the constructing an optimal material delivery policy corresponding to the e-commerce advertisement according to the analysis result of the delivered material effect includes:

determining material interaction elements according to the analysis result of the material effect;

constructing an initial material throwing strategy corresponding to the E-commerce advertisement according to the material interaction element;

and carrying out strategy optimization on the initial material throwing strategy to obtain an optimal material throwing strategy.

The second aspect of the present invention provides a delivered material mining apparatus, including:

the monitoring module is used for carrying out E-commerce advertisement monitoring on the plurality of target social media to obtain advertisement content data and advertisement interaction data;

the analysis module is used for carrying out text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data;

the extraction module is used for extracting key information from the advertisement text data and the advertisement image data to obtain a material key information set;

the integration module is used for carrying out feature extraction and feature integration on the advertisement interaction data to obtain target interaction features;

The building module is used for building a material characteristic relation set between the target interaction characteristic and the material key information set;

the processing module is used for inputting the material characteristic relation set into a preset material throwing mining model to analyze the material throwing effect, and obtaining a material throwing effect analysis result;

and the construction module is used for constructing an optimal material delivery strategy corresponding to the E-commerce advertisement according to the analysis result of the material delivery effect.

A third aspect of the present invention provides a delivered-material mining apparatus, comprising: a memory and at least one processor, the memory having instructions stored therein; and the at least one processor calls the instruction in the memory so that the released material mining equipment executes the released material mining method.

A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored therein that, when executed on a computer, cause the computer to perform the above-described delivered-material mining method.

In the technical scheme provided by the invention, E-commerce advertisement monitoring is carried out on a plurality of target social media to obtain advertisement content data and advertisement interaction data; performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data; extracting key information from advertisement text data and advertisement image data to obtain a material key information set; extracting features and integrating the features of the advertisement interaction data to obtain target interaction features; establishing a material characteristic relation set between the target interaction characteristic and the material key information set; inputting the story feature relation set into a preset story throwing mining model to analyze the story throwing effect, and obtaining a story throwing effect analysis result; according to the analysis result of the material throwing effect, an optimal material throwing strategy corresponding to the advertisement of the electronic commerce is constructed, and the advertisement throwing material mining method can help the advertisement commerce advertiser to optimize advertisement content, images, texts and release strategies by deeply analyzing advertisement materials and interaction data, so that the click rate, conversion rate and interaction rate of the advertisement are improved. Advertisers may reduce advertising costs by locating and attracting potential customers more precisely. The optimized advertisement strategy and materials can reach the target audience more effectively, personalized advertisement materials can improve user experience, so that the audience is more likely to interact with advertisements, and the advertisement is more in line with interests and demands of the audience, so that the accuracy of the material throwing strategy is improved.

Drawings

FIG. 1 is a schematic diagram of an embodiment of a method for mining delivered materials in an embodiment of the present invention;

FIG. 2 is a flow chart of key information extraction in an embodiment of the invention;

FIG. 3 is a flow chart of feature extraction and feature integration in an embodiment of the invention;

FIG. 4 is a flowchart of establishing a material feature relation set in an embodiment of the present invention;

FIG. 5 is a schematic view of an embodiment of a device for mining delivered material according to an embodiment of the present invention;

fig. 6 is a schematic diagram of an embodiment of a material mining apparatus according to an embodiment of the present invention.

Detailed Description

The embodiment of the invention provides a method, a device, equipment and a storage medium for excavating materials, which are used for improving the accuracy of a material throwing strategy. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For easy understanding, a specific flow of an embodiment of the present invention is described below, referring to fig. 1, and an embodiment of a method for mining delivered materials in an embodiment of the present invention includes:

s101, E-commerce advertisement monitoring is carried out on a plurality of target social media to obtain advertisement content data and advertisement interaction data;

it can be understood that the execution body of the present invention may be a material delivering mining device, and may also be a terminal or a server, which is not limited herein. The embodiment of the invention is described by taking a server as an execution main body as an example.

Specifically, the server first explicitly defines the target social media platform to monitor, and the specific choice of which social media depends on the target audience and strategy of the e-commerce advertisement. Data acquisition is then performed. Typically, social media platforms provide an API (application programming interface) through which data on the platform may be accessed. These APIs may be used to obtain advertisement content data and advertisement interaction data, such as advertisement text, images, click-through rates, likes, comments, etc. It is also contemplated that web crawler technology may be used to crawl data on social media websites if APIs are not available. Data storage is then performed. This may be accomplished by a database system, such as MySQL, mongoDB or a specialized data warehouse. The purpose of data storage is to ensure the security and accessibility of data for subsequent analysis and querying. Then, data processing and analysis are performed. The data needs to be cleaned and preprocessed, including deduplication, handling missing data, and data format conversion. The advertising content data then requires further analysis and natural language processing techniques can be used to extract key information such as keywords, topics and emotions in the advertisement. Meanwhile, the advertisement interaction data also needs to be sorted and analyzed to know the influence of advertisements in the audience, such as click rate, like number, share number, comment number and the like. And finally, data visualization is performed. By using data visualization tools such as Matplotlib, tableau or Power BI, charts and reports can be created that make advertising content data and interactive data easier to understand and interpret. This helps the decision maker to better understand the performance of the advertisement, identify successful strategies, and optimize. For example, suppose an online e-commerce company places a series of advertisements on social media to promote its new product line. By monitoring these advertisements, the server obtains advertisement content data, such as advertisement text, images, and links, and advertisement interaction data, such as click-through rates, likes, and comments, using the social media platform's API. These data are stored in a special database and subjected to data cleansing and preprocessing. The server then analyzes the advertisement text using natural language processing techniques to determine which keywords and topics perform best in the advertisement. Meanwhile, the server analyzes the advertisement interaction data to identify advertisements that are most reactive to the audience. Finally, the server creates a dashboard using data visualization tools to visualize advertisement performance metrics, help the server to better understand the effectiveness of the advertisement, and optimize advertisement content and placement policies if necessary.

S102, carrying out text and image analysis on advertisement content data to obtain advertisement text data and advertisement image data;

specifically, the server first, advertisement content data is information including text and image elements, and is generally used for advertising and promotion. Natural Language Processing (NLP) is a field that studies how to enable computers to understand and process natural language. The server uses a pre-set NLP model to identify named entities from the advertising content data. Named entities generally refer to words or phrases that have a particular meaning or represent a particular thing in text. For example, assume that the advertisement content data contains the following text information: "an innovation company has introduced an intelligent product. "named entity recognition based on NLP model, this method would recognize and label keywords in text, such as" Innovative company "and" Smart product ". On the basis of named entity recognition, the server performs entity text conversion. The purpose is to translate the identified named entities into more advertisement-related and abstract textual representations to better express the characteristics and features of the advertisement. For example, for the text information mentioned before: "an innovation company has introduced an intelligent product. The 'innovation company' can be converted into the 'leading technical company' through entity text conversion, and the 'intelligent product' can be converted into the 'latest technological product'. In this way, text becomes more advertised, highlighting the innovativeness and high-tech nature of advertising. And extracting image characteristics and classifying images based on a preset image processing model. The advertising content data typically contains image elements and therefore requires image analysis. A preset image processing model is used to extract features of the image and to classify the image. For example, for an image in an advertisement, the image processing model may extract image features such as color, texture, shape, etc., without describing objects in a particular image. Subsequently, the image classification model may classify the images into different categories, such as "product shows" or "nature landscapes," according to the extracted features. Finally, the advertisement image data is subjected to semantic segmentation, the image is divided into different semantic areas, and semantic information is marked for each area. This helps to understand the advertising image content in more detail. For example, for an advertisement image, semantic segmentation may segment the image into different regions, such as a "product display region" and a "background region. This helps to understand the different elements in the image.

S103, extracting key information from advertisement text data and advertisement image data to obtain a material key information set;

first, the server processes advertisement text data. Natural Language Processing (NLP) techniques are used to analyze text and extract keywords. This can be achieved by word segmentation, word frequency statistics and disabling word filtering. And meanwhile, carrying out emotion tendentiousness analysis on the advertisement text data to determine the emotion effect of the advertisement. Emotion analysis may classify text as positive, negative, or neutral emotion, typically implemented using emotion vocabulary and machine learning models. In addition, there is a need to subject the advertisement text to modeling to identify the main topics or content of the advertisement. This may be accomplished by using a topic modeling algorithm (e.g., latent Dirichlet Allocation, LDA). Next, the server processes the advertisement image data. This includes object detection, by computer vision techniques such as Convolutional Neural Network (CNN) or YOLO (You Only Look Once), to detect objects in the advertising image, such as products, characters, or other objects. At the same time, color analysis is also required to understand the dominant color of the advertising image. This can be achieved by computing pixel color histograms, color distribution statistics, or deep learning models, typically using an image processing library such as OpenCV for image data processing. Finally, key information extracted from the text data and the image data is integrated into a data structure to generate a material key information set. This set includes information such as advertisement keywords, advertisement emotional tendency, advertisement topic information, advertisement image objects, and advertisement image colors. Such information may assist the decision algorithm in selecting the material most suitable for advertisement placement and monitoring the performance of the advertisement. The decision algorithm can select materials according to advertisement keywords, emotion tendencies, theme information, image objects and color information, and optimize advertisements according to feedback data so as to improve advertisement effects.

S104, carrying out feature extraction and feature integration on the advertisement interaction data to obtain target interaction features;

specifically, firstly, the server performs data preprocessing on advertisement interaction data to obtain standard interaction data. This process may involve data cleansing, missing value handling, outlier detection, etc. Data preprocessing ensures that the data used by the server in subsequent analysis is accurate and consistent. After standard interaction data are obtained, interaction quantity analysis is carried out to obtain interaction quantity characteristics. These features include endorsements, comments, shares, clicks, etc. These features may help the server learn about how popular and interactive advertisements are on social media or other platforms. The number of interactions feature tells the server the number of interactions, but in order to more fully understand the interaction situation of the advertisement, it is necessary to calculate the interaction frequency feature. The interaction frequency data tells the server about the distribution of the interaction of the advertisements in a period of time, and can help the server to determine the liveness and interaction trend of the advertisements. Based on a time sequence analysis method, the server performs trend analysis on the interaction quantity characteristics to acquire interaction trend characteristics. These features can tell the server whether the advertisement interaction is increasing, decreasing or remaining stable, helping the server to predict future interactive performance of the advertisement. User engagement analysis involves analyzing standard interaction data to obtain characteristics about user engagement. These characteristics may include the number of users interacting, the frequency of user interactions, the time distribution of user interactions, etc. This may help the server to know which users are more interested in the advertisement and are actively engaged. And finally, integrating the interaction quantity characteristics, the interaction frequency characteristics, the interaction trend characteristics and the user participation characteristics together to obtain target interaction characteristics. This integration process may involve feature merging, normalization, etc. to ensure that different types of features can be effectively combined together, facilitating a comprehensive assessment of advertisement performance. For example, assuming a server has a social media advertisement, the analysis and feature extraction are performed by the method described above, the server gets the following target interaction features: interaction quantity characteristics: praise number=1000, comment number=200, share number=50, click number=5000. The method comprises the steps of carrying out a first treatment on the surface of the Interaction frequency characteristics: the praise numbers are distributed in one month, and continuous interaction of advertisements is displayed; interaction trend feature: according to time sequence analysis, the praise numbers show a gradually increasing trend, and the advertisement attractiveness is rising; user engagement feature: and 2 times of participation of each user on average, and information such as geographic distribution of interactive users. These targeted interactive features will help the advertising marketing team better understand the effectiveness of the advertisement and formulate optimization strategies based on such information to increase popularity and interaction of the advertisement.

S105, establishing a material characteristic relation set between the target interaction characteristic and the material key information set;

specifically, first, the server integrates the target interaction characteristics and the material key information into a target data table. This target data table will contain the interactive features and material key information corresponding to each advertisement. Each row represents an advertisement and each column represents a feature. Next, the server uses statistical methods to analyze the correlation between the target interaction characteristics and the material key information. One of the most common methods is to calculate the correlation coefficient, typically using pearson correlation coefficients or spearman correlation coefficients. The correlation coefficient measures the strength of a linear relationship between two variables, its value being between-1 and 1, where 1 represents a complete positive correlation, -1 represents a complete negative correlation, and 0 represents no correlation. For example, the server calculates a correlation coefficient between the endorsement number and the click number to determine whether there is a correlation between the endorsement number and the advertisement click number. If the correlation coefficient is 0.8, it can be derived that there is a strong positive correlation between the click number and the endorsement number, i.e., as the endorsement number increases, the click number increases. And creating a material characteristic relation set by the server according to the result of the correlation analysis. This set will include correlation indicators between the target interaction characteristics and the material key information, as well as a description of the relationship between them. For example, a correlation coefficient between the number of clicks and the number of clicks of 0.8 indicates that there is a strong positive correlation between the two, i.e., an increase in the number of clicks is typically accompanied by an increase in the number of clicks. The correlation coefficient between the advertisement theme and the comment count is-0.2, which indicates that a weak negative correlation relationship exists between the advertisement theme and the comment count, namely the variation trend between the advertisement theme and the comment count is inconsistent. By creating a collection of material characteristics relationships, the server more clearly knows which material characteristics have relevance to the target interaction characteristics, which helps the server better optimize the advertising strategy. For example, suppose that the server performs the above analysis for a batch of travel advertisements. Through correlation analysis, the server obtains a part of the following material characteristic relation set: the correlation coefficient between the endorsement number and the click number is 0.85, indicating that the likelihood of clicking on the advertisement increases when the user endorses the advertisement; the correlation coefficient between the advertisement theme and the sharing number is 0.45, which indicates that a certain positive correlation exists between the advertisement theme and the sharing number; the correlation coefficient between the image color and the comment count is-0.3, which indicates that there is a certain degree of negative correlation between the advertisement image color and the comment count, i.e., some colors may reduce the comment count.

S106, inputting the feature relation set of the materials into a preset material throwing mining model to analyze the material throwing effect, and obtaining a material throwing effect analysis result;

specifically, first, the server performs feature encoding on the feature relation set of the material. The purpose of feature encoding is to convert various types of features into numerical form so that the deep learning model can understand and process. This can be done using various encoding methods, such as one-hot encoding, tag encoding, word embedding, etc., with the particular choice depending on the type of feature. After encoding, the server obtains a plurality of feature code values. These feature code values are then converted into a target input vector. This vector is an array of values that integrates the information of all features together. The dimension of the vector is typically equal to the number of features. This process can be seen as a stacking of features, combining the encoded features into a vector in a certain order. The server then inputs the target input vector into a pre-set drop material mining model, which typically includes two layers of short-to-long memory networks (LSTM) and two layers of fully connected networks. Two-layer long and short term memory networks (LSTM) are a type of recurrent neural network, particularly suited for modeling of sequential data, such as time-sequential or text data. Here, the task of LSTM is to perform feature extraction on the target input vector. It can capture time correlation and sequence information in the input vector, helping to better understand the relationship between material features. The two-layer fully-connected network is used for further processing the target feature vector output by the LSTM so as to analyze the effect of the delivered materials. The fully-connected network models a nonlinear relation through learning weights and deviations so as to finally output an effect analysis result of the delivered materials. And finally, the server obtains the analysis result of the effect of the delivered material through the combination of the two layers of LSTM and the two layers of fully connected networks. The result may be a numerical value, such as a value representing click through rate, conversion rate, or other effectiveness metric of the advertisement. The relevance between the material characteristics and the advertisement effect is reflected, and the advertisement marketing team can be helped to evaluate and optimize the selection of advertisement materials. For example, assume that the server has an advertising material mining model that has been pre-trained and has good performance. The server wants to analyze the effects of a collection of social media advertisements that contain text, images, and interactive data. First, the server encodes the characteristics of the text keywords, emotion tendencies, subject information, image objects, image colors, and the like of the advertisement as numerical values. These encoded values are then combined into a target input vector. Next, the server inputs the target input vector into a preset advertisement material mining model. The LSTM layer of the model will learn to extract key information from these features, such as which keywords are related to the number of interactions, or which image colors are related to the advertisement click-through rate. The fully connected layer will then further process these features to generate a final advertising effectiveness analysis. For example, the model may output a predicted click-through rate for an advertisement, if the value is high, indicating that the advertisement material has good interactive effects, and vice versa. This result can be used to select advertising material to optimize the advertising strategy to improve the effectiveness and ROI of the advertisement.

And S107, constructing an optimal material delivery strategy corresponding to the E-commerce advertisement according to the analysis result of the material delivery effect.

Specifically, the server first determines, according to the analysis result of the impression material effect, the material interaction element affecting the advertisement effect. These elements may be keywords, emotional tendency, topic information in the advertisement text, or may be objects, colors, etc. in the advertisement image. The analysis results may reveal which elements have a significant impact on the advertising effectiveness, e.g., a particular keyword may attract more clicks, or a color may increase user interaction. And constructing an initial e-commerce advertisement material putting strategy based on the determined material interaction elements. This strategy involves selecting appropriate advertising text, images and other material elements to maximize the use of the identified interactive factors. For example, if a particular keyword is associated with a high click through rate, a policy may include emphasizing the keyword in the advertisement text. The initial material delivery strategy may be a starting point, but it may be further improved by strategy optimization. Policy optimization is an iterative process with the goal of fine-tuning and improving the advertisement placement policy based on actual effectiveness data to achieve optimal effectiveness. And (3) adopting an A/B test method to deliver advertisement materials of different versions to different audience groups, and then comparing the effects of the advertisement materials. By analyzing the different versions of advertising material, it is possible to determine which version performs best and learn from it which elements have a positive effect. In the process of advertisement putting, advertisement effect data including click rate, conversion rate, interaction quantity and the like are monitored in real time. If some materials or strategies are found to be not as expected, the advertisement delivery strategy is adjusted and optimized in time. Machine learning algorithms or deep learning models are used to automate optimization of advertisement placement strategies. The models can analyze a large amount of data, and rapidly adjust advertisement materials and delivery strategies so as to achieve the best effect. For example, suppose an e-commerce company operates a series of social media advertisements to promote its new products. Based on previous material mining and effect analysis, the server finds that a certain keyword (e.g., "unique design") in the advertisement text is significantly related to click-through rate, while a certain color (e.g., blue) in the image is related to the number of interactions. Thus, the server builds an initial material delivery strategy: highlighting "unique design" keywords in the advertisement text. The advertising image is designed using a blue palette. The server then begins running the advertisement and adopts the A/B test method. The server creates two versions of the advertisement, one version designed according to the initial strategy and the other version of the advertisement is compared and tested, and keyword emphasis and blue color palette are removed. After a period of operation and data collection, the server finds that the original version performs better and has higher click rate and conversion rate. Over time, the e-commerce company continuously monitors advertising effectiveness and further improves advertising material delivery strategies through real-time data feedback and algorithm optimization. The server automatically adjusts keyword weights in the advertisement text, selects different image colors, and optimizes advertisement materials according to different platforms and audience groups. Through the continuously improved process, the server finally realizes the optimal material delivery strategy, and the advertising effect and the ROI are improved.

In the embodiment of the invention, E-commerce advertisement monitoring is carried out on a plurality of target social media to obtain advertisement content data and advertisement interaction data; performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data; extracting key information from advertisement text data and advertisement image data to obtain a material key information set; extracting features and integrating the features of the advertisement interaction data to obtain target interaction features; establishing a material characteristic relation set between the target interaction characteristic and the material key information set; inputting the story feature relation set into a preset story throwing mining model to analyze the story throwing effect, and obtaining a story throwing effect analysis result; according to the analysis result of the material throwing effect, an optimal material throwing strategy corresponding to the advertisement of the electronic commerce is constructed, and the advertisement throwing material mining method can help the advertisement commerce advertiser to optimize advertisement content, images, texts and release strategies by deeply analyzing advertisement materials and interaction data, so that the click rate, conversion rate and interaction rate of the advertisement are improved. Advertisers may reduce advertising costs by locating and attracting potential customers more precisely. The optimized advertisement strategy and materials can reach the target audience more effectively, personalized advertisement materials can improve user experience, so that the audience is more likely to interact with advertisements, and the advertisement is more in line with interests and demands of the audience, so that the accuracy of the material throwing strategy is improved.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Carrying out named entity recognition on advertisement content data based on a preset natural language processing model to obtain a plurality of named entities;

(2) According to the named entities, performing entity text conversion on the advertisement content data to generate corresponding advertisement text data;

(3) Carrying out image feature extraction and image classification on advertisement content data based on a preset image processing model to obtain image feature data;

(4) And carrying out semantic segmentation on the image characteristic data to obtain advertisement image data.

In particular, server named entity recognition is a natural language processing technique for recognizing specific types of named entities in text, such as person names, place names, organization names, dates, and the like. First, the server selects a pre-set natural language processing model, such as BERT, GPT-3, or a model specific to NER, such as BERT-CRF, for named entity recognition of the advertisement text data. Advertisement content data including advertisement text is prepared. The text may be the title, description, banner, etc. of the advertisement. Processing the advertisement text data by using the selected NER model, and identifying and extracting named entities in the text. The model will annotate the entities in the text and categorize them, such as brand names, product models, geographic locations, etc. After obtaining a plurality of named entities, the next task is to convert these entities into corresponding advertisement text data. The identified named entities are classified, for example, brand names, product models, geographic locations, etc. Each entity is then matched to a corresponding advertisement text location, such as a title, description, etc. Named entities are combined into different portions of advertisement text data, e.g., titles, descriptions, etc., using text generation techniques such as template filling, natural language generation models (e.g., GPT strings), etc. And finally, generating complete advertisement text data according to the matching and the generated result. These text data will be used in the advertisement. For image portions in advertising content, a server a pre-set image processing model, typically using Convolutional Neural Networks (CNNs) or other deep learning models, extracts image features and classifies images. And extracting the characteristics of the advertisement image by using a preset image processing model. These features may include edges, textures, colors, and the like. The extracted features will facilitate subsequent image classification and semantic segmentation. The advertisement image is classified using another preset image classification model to determine the object or scene category contained in the image. This helps to understand the content of the advertisement image, such as the product, scene, or theme in the advertisement. Finally, semantic segmentation techniques may be applied to further understand the content of the advertising image, particularly the semantic information of different objects or regions in the image. The advertisement image is segmented into a plurality of semantic regions using a preset semantic segmentation model, such as U-Net or FCN. These regions correspond to different semantic categories such as products, backgrounds, text, etc. In combination with the result of the semantic segmentation, structured information containing advertising image data may be generated. These data can be used to more precisely understand the meaning and association of the different parts of the advertisement.

In a specific embodiment, as shown in fig. 2, the process of performing step S103 may specifically include the following steps:

s201, extracting keywords from advertisement text data to obtain advertisement keywords, analyzing emotion tendencies of the advertisement text data to obtain advertisement emotion tendencies, and performing topic modeling on the advertisement text data to obtain advertisement topic information;

s202, object detection is carried out on advertisement image data to obtain advertisement image objects, and color analysis is carried out on the advertisement image data to obtain advertisement image colors;

s203, generating a material key information set according to the advertisement keywords, advertisement emotion tendentiousness, advertisement theme information, advertisement image objects and advertisement image colors.

Specifically, the server first extracts keywords from the advertisement text using natural language processing techniques. These keywords represent the core concept of the advertisement and help the server to better understand the advertisement content. Keyword extraction may employ TF-IDF (Word frequency-inverse document frequency) analysis or a Word vector model, such as Word2Vec or BERT. Emotion analysis is a technique for determining emotion tendencies in advertisement text. Through emotion analysis, the server knows whether the emotion color of the advertisement is positive, negative or neutral. Emotion analysis may be implemented using an emotion dictionary or a machine learning model (e.g., a deep learning model). Topic modeling helps a server understand the main topics or topics in the advertisement text. This can be achieved by techniques such as Latent Dirichlet Allocation (LDA) or non-Negative Matrix Factorization (NMF). Topic modeling helps a server identify key topics in advertisement text, thereby better grasping core information of advertisements. Next, the server processes the advertisement image data. Object detection is a technique for identifying an object or object in an advertisement image. The server uses Convolutional Neural Networks (CNNs) or other deep learning models for object detection. This helps the server to learn about key objects in the advertising image, such as products, characters, etc. Color analysis is used to determine the dominant color in the advertising image. This may be achieved by cluster analysis or color histogram analysis. Knowing the dominant color of the advertisement image helps to maintain a consistent visual style in the advertisement design. And finally, the server combines information such as advertisement keywords, advertisement emotion tendentiousness, advertisement theme information, advertisement image objects, advertisement image colors and the like into a material key information set. This set provides advertising marketers with a tool to gain insight into advertising material, helping to better formulate advertising strategies, creative designs, and targeting.

In a specific embodiment, as shown in fig. 3, the process of executing step S104 may specifically include the following steps:

s301, carrying out data preprocessing on advertisement interaction data to obtain standard interaction data;

s302, carrying out interaction quantity analysis on standard interaction data to obtain interaction quantity characteristics, wherein the interaction quantity characteristics comprise praise numbers, comment numbers, sharing numbers and click numbers;

s303, calculating interaction frequency data of the interaction quantity characteristics to obtain the interaction frequency characteristics;

s304, carrying out trend analysis on the interaction quantity characteristics based on a time sequence analysis method to obtain interaction trend characteristics;

s305, carrying out user participation analysis on the standard interaction data to obtain user participation characteristics;

s306, carrying out feature integration on the interaction quantity features, the interaction frequency features, the interaction trend features and the user participation features to obtain target interaction features.

Specifically, the server first collects data from the advertising interaction data source. Such data may include information on the endorsement number, comment number, share number, click number, etc. of the advertisement, as well as related time stamps and user information. The data needs to be pre-processed, including removing duplicate data, processing missing values, processing outliers, etc., before further analysis can take place. Next, the standard interaction data is analyzed to obtain the interaction quantity characteristics. The interactive quantity features typically include endorsements, comments, shares, and clicks. The server uses statistical methods such as summation, mean, median, maximum and minimum, etc. to calculate these features. The interaction frequency feature may provide more insight that reflects how often interactions occur over a period of time. Calculating the interaction frequency requires combining the interaction quantity feature with a time stamp. The average number of interactions per day, week or month, and the trend of the interactions can be calculated. The interactive trend feature can be obtained by a time series analysis method. The server uses methods such as sliding window averaging, exponential smoothing, ARIMA model, etc. to analyze the trend of the interactive quantitative features. This will help the server know whether the number of interactions is rising, falling or remaining stable. User engagement analysis aims to learn how an audience segment of an advertisement interacts with the advertisement. The server analyzes the server's behavior on advertisements based on the user's characteristics, such as gender, age, geographic location, etc. This may be achieved by means of group analysis, user classification, association analysis, etc. And finally, integrating the interaction quantity feature, the interaction frequency feature, the interaction trend feature and the user participation feature together to obtain the target interaction feature. This can be done through data consolidation and feature engineering. Ensuring that appropriate feature choices and feature engineering techniques are selected to ensure that the final feature set has sufficient information to predict or analyze advertisement interactions.

In a specific embodiment, as shown in fig. 4, the process of performing step S105 may specifically include the following steps:

s401, integrating the target interaction characteristics and the material key information set into a target data table, wherein each advertisement in the target data table corresponds to the interaction characteristics and the material key information;

s402, carrying out correlation analysis on the target interaction characteristics and the material key information set according to the target data table to obtain a target correlation coefficient;

s403, creating a material characteristic relation set between the target interaction characteristic and the material key information set according to the target correlation coefficient.

Specifically, the server first creates a target data table that will contain the interactive characteristics and material key information for each advertisement. Each row represents an advertisement and the columns include different interactive features (e.g., endorsements, comments, shares, clicks, etc.) and material key information (e.g., advertisement type, creative description, release time, etc.). Ensuring that the structure and format of the data table is suitable for subsequent analysis. And integrating the interaction characteristic data and the material key information data into a target data table to ensure that each advertisement is associated with the corresponding interaction characteristic and material information. This may be accomplished by a unique identifier of the advertisement or advertisement ID to ensure consistency and accuracy of the data. Correlation analysis can then be performed to understand the degree of correlation between the interaction characteristics and the material key information. The relevance analysis may help the server determine which interaction features have significant relevance to the material key information and which may not. Correlation coefficients (such as pearson correlation coefficients or spearman rank correlation coefficients) are typically used to measure the degree of correlation between two variables. The correlation coefficient has a value between-1 and 1, a negative value indicating a negative correlation, a positive value indicating a positive correlation, and 0 indicating no correlation. The closer the correlation coefficient is to-1 or 1, the stronger the correlation, and the closer to 0 the weaker the correlation. And creating a material characteristic relation set by the server according to the result of the target correlation coefficient. This set will include material critical information that is highly relevant to the interactive features. And selecting material key information highly related to the interaction characteristic based on the threshold value of the correlation coefficient. The server determines the appropriate correlation threshold based on the analyzed objectives and domain knowledge. For example, feature selection: once the server determines the high-relevance material key information, the high-relevance material key information can be selected into a material characteristic relation set. These features will be used in subsequent analysis for modeling, prediction, or other decision-making tasks; visual analysis: correlation between interactive features and material key information is visualized using data visualization tools such as scatter plots, thermodynamic diagrams, and the like. This may help the server more intuitively understand the data relationships; and (3) establishing a model: based on the selected set of story feature relationships, a machine learning model or statistical model may be built for further analysis or prediction. For example, the server may attempt to predict the amount of interaction (e.g., number of clicks) of the advertisement based on the selected material characteristics; verification and optimization: after the model is built, verification and optimization are required to ensure the accuracy and generalization capability of the model. This may include the use of cross-validation, adjustment of model parameters, and the like.

In a specific embodiment, the process of executing step S106 may specifically include the following steps:

(1) Performing feature coding on the material feature relation set to obtain a plurality of feature coding values;

(2) Vector conversion is carried out on the plurality of characteristic code values to obtain a target input vector;

(3) Inputting a target input vector into a preset material throwing mining model, wherein the material throwing mining model comprises: two layers of long-short-time memory networks and two layers of fully-connected networks;

(4) Extracting features of the target input vector through two layers of long-short-term memory networks to obtain a target feature vector;

(5) And inputting the target feature vector into two layers of fully-connected networks to perform material throwing effect analysis, and obtaining a material throwing effect analysis result.

Specifically, first, feature coding is performed on a feature relation set of materials, and each feature is converted into a digital form. This can be accomplished by various Encoding techniques, such as One-Hot Encoding (Encoding): for converting discrete features into binary vectors; label Encoding (Label Encoding): for encoding the ordered classification features into integers; embedded coding (Embedding Encoding): for mapping classification features to points in a continuous vector space; numerical normalization (Numeric Normalization): the numerical features are normalized to ensure that they have similar dimensions. The encoded features are combined into a target input vector. This vector should be a numerical vector containing all the features and can be used as an input to the model. For example, if the server has 10 features, the target input vector will be a vector containing 10 elements. The cast material mining model is typically a deep learning model for analyzing the effects of cast material. This model may include multiple layers, such as two layers of long short-term memory networks (LSTM) and two layers of fully connected networks. Two-layer long and short term memory networks (LSTM) are a deep learning model suitable for sequential data, typically used to extract features of time sequential data. In this model, the LSTM layer may be used to analyze the sequence features of the target input vector, extracting the relation of relevant material features to time. Two-layer fully-connected networks are typically used at the output layer of the deep learning model for classification or regression tasks. In this model, the fully connected layer may be used to analyze features extracted from the LSTM layer, ultimately outputting results of the effect analysis of the delivered material. And analyzing the effect of delivering the materials by the server through delivering the material mining model. Specific analysis tasks may include: and (3) effect prediction: the effect of delivering the material is predicted by using the model, such as click rate, conversion rate and the like of the advertisement. Effect classification: advertising effectiveness is divided into different categories such as success, failure, general, etc. Effect regression: continuous effectiveness metrics such as sales of advertisements, ROIs, etc. are predicted. For example, assume that there is a home electronics sub-business company and the server analyzes the impact of different advertising stories on product sales. The material characteristic relation set of the server comprises advertisement types, word descriptions, picture characteristics, release time and the like. First, the server encodes these features, for example, converting advertisement types into binary vectors using one-hot encoding, mapping textual descriptions into continuous vectors using embedded encoding, normalizing picture features, and converting publication times into numerical representations. These encoded features are then combined into a target input vector. This vector contains a numerical representation of all features. Next, the server uses a pre-set drop material mining model, which includes two layers of LSTM and two layers of fully connected networks. The model receives a target input vector, and firstly analyzes the relation between the text description and time through an LSTM layer to extract sequence characteristics. These features are then passed to the fully connected network for analysis of the advertising effectiveness. Finally, the model may predict the effectiveness of different advertising materials, such as advertisement click-through rates or sales.

In a specific embodiment, the process of executing step S107 may specifically include the following steps:

(1) Determining material interaction elements according to the analysis result of the material effect;

(2) Constructing an initial material throwing strategy corresponding to the E-commerce advertisement according to the material interaction element;

(3) And carrying out strategy optimization on the initial material throwing strategy to obtain an optimal material throwing strategy.

Specifically, first, a material effect analysis is put in. This step involves collecting and analyzing data of past advertising campaigns, including click-through rates, conversion rates, number of interactions, sales, etc. By carefully studying this data, it is possible to determine which material elements succeed in advertising. These material elements may be text content, images, video, button patterns, colors, presentation modes, etc. Next, an initial material delivery strategy is constructed. The server explicitly defines the target audience for the advertisement. This includes characteristics of the target audience, such as age, gender, geographic location, interests, etc. The determination of the target audience is helpful to the accurate positioning of the advertisement, ensuring that the advertisement is only shown to the most interested people. Second, it is important to select an advertising platform that is appropriate for the target audience. Different advertising platforms are suitable for different types of audience. For example, social media advertisements may be more suitable for young users, while search engine advertisements may be suitable for users with explicit needs. An advertising budget is set. The advertisement budget determines the number and frequency of advertisements placed by the server. The server ensures that the advertising campaign is running in financial range and can achieve the desired advertising effect. And making advertisement content. Based on the previously determined material interaction elements, the server creates advertisement text, images, videos, and other materials that attract the target audience. These materials should be consistent with the brand image and advertising objectives of the server. Advertisement scheduling and targeting determines the time and location of advertisement presentation. The server selects a time period for placement of the advertisement to ensure that the advertisement is presented during the time that the target audience is active. At the same time, it is also important to select the display position of the advertisement, and different positions have different effects on the attention of the user. The strategy is then optimized. The aim is to continuously improve advertising effectiveness and adapt to changes in market and user behavior. In the policy optimization stage, the A/B test is a powerful tool. It allows the server to compare two or more different advertising policies or material versions simultaneously. By comparing the test results, the server determines which strategy or material performs best and adjusts accordingly. In addition, data analysis is also critical to the optimization strategy. The server continuously monitors advertisement effect data including click-through rate, conversion rate, interaction data, etc. Through data analysis, the server identifies potential problems and makes adjustments and improvements in time. User feedback is also an important source of information for optimizing policies. And collecting feedback and comments of the user, and knowing the comments and suggestions of the server on the advertisement. User feedback can help the server find ad deficiencies and improve. Meanwhile, market trends were analyzed. Market conditions often change and competitors' actions can also affect advertising effectiveness. Accordingly, servers continue to pay attention to market trends, dynamically adjusting advertising policies based on the market to maintain competitive advantages. Finally, through continuous optimization work, the server gradually determines the optimal material delivery strategy. This optimal strategy should be able to maximize advertising effectiveness, such as click through rate, conversion rate, ROI, etc., while remaining within the budget. Only through continuous analysis, experiments and feedback, the performance of the advertisement can be continuously improved, and the success of the advertisement activity is ensured. This process is dynamic and requires constant adjustment and improvement based on market changes and user feedback.

The method for excavating the put-in material in the embodiment of the present invention is described above, and the device for excavating the put-in material in the embodiment of the present invention is described below, referring to fig. 5, where an embodiment of the device for excavating the put-in material in the embodiment of the present invention includes:

the monitoring module 501 is configured to monitor a plurality of target social media for e-commerce advertisements, so as to obtain advertisement content data and advertisement interaction data;

the analysis module 502 is configured to perform text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data;

an extracting module 503, configured to extract key information from the advertisement text data and the advertisement image data, to obtain a material key information set;

the integrating module 504 is configured to perform feature extraction and feature integration on the advertisement interaction data to obtain a target interaction feature;

the establishing module 505 is configured to establish a material feature relation set between the target interaction feature and the material key information set;

the processing module 506 is configured to input the feature relation set of the material into a preset material throwing mining model to analyze the effect of the material throwing, so as to obtain an analysis result of the effect of the material throwing;

And the construction module 507 is configured to construct an optimal material delivery strategy corresponding to the e-commerce advertisement according to the analysis result of the material delivery effect.

E-commerce advertisement monitoring is carried out on a plurality of target social media through the cooperative cooperation of the components, so that advertisement content data and advertisement interaction data are obtained; performing text and image analysis on the advertisement content data to obtain advertisement text data and advertisement image data; extracting key information from advertisement text data and advertisement image data to obtain a material key information set; extracting features and integrating the features of the advertisement interaction data to obtain target interaction features; establishing a material characteristic relation set between the target interaction characteristic and the material key information set; inputting the story feature relation set into a preset story throwing mining model to analyze the story throwing effect, and obtaining a story throwing effect analysis result; according to the analysis result of the material throwing effect, an optimal material throwing strategy corresponding to the advertisement of the electronic commerce is constructed, and the advertisement throwing material mining method can help the advertisement commerce advertiser to optimize advertisement content, images, texts and release strategies by deeply analyzing advertisement materials and interaction data, so that the click rate, conversion rate and interaction rate of the advertisement are improved. Advertisers may reduce advertising costs by locating and attracting potential customers more precisely. The optimized advertisement strategy and materials can reach the target audience more effectively, personalized advertisement materials can improve user experience, so that the audience is more likely to interact with advertisements, and the advertisement is more in line with interests and demands of the audience, so that the accuracy of the material throwing strategy is improved.

Fig. 5 above describes the material-throwing mining apparatus in the embodiment of the present invention in detail from the point of view of the modularized functional entity, and the material-throwing mining device in the embodiment of the present invention is described in detail from the point of view of hardware processing.

Fig. 6 is a schematic structural diagram of a delivered material mining device according to an embodiment of the present invention, where the delivered material mining device 600 may have a relatively large difference due to different configurations or performances, and may include one or more processors (central processing units, CPU) 610 (e.g., one or more processors) and a memory 620, and one or more storage media 630 (e.g., one or more mass storage devices) storing application programs 633 or data 632. Wherein the memory 620 and the storage medium 630 may be transitory or persistent storage. The program stored on the storage medium 630 may include one or more modules (not shown), each of which may include a series of instruction operations for delivering material in the material mining apparatus 600. Still further, the processor 610 may be configured to communicate with the storage medium 630 to execute a series of instruction operations in the storage medium 630 on the delivered-material mining device 600.

The drop material mining apparatus 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input/output interfaces 660, and/or one or more operating systems 631, such as Windows Server, mac OS X, unix, linux, freeBSD, and the like. It will be appreciated by those skilled in the art that the structure of the launched material mining device shown in fig. 6 does not constitute a limitation of the launched material mining device, and may include more or fewer components than shown, or may combine certain components, or may have a different arrangement of components.

The invention also provides a put material mining device, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and when the computer readable instructions are executed by the processor, the processor executes the steps of the put material mining method in the above embodiments.

The present invention also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, or may be a volatile computer readable storage medium, where instructions are stored in the computer readable storage medium, where the instructions, when executed on a computer, cause the computer to perform the steps of the method for delivering material.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The method for mining the put material is characterized by comprising the following steps:

2. The method for mining delivered material according to claim 1, wherein the text and image analysis of the advertisement content data to obtain advertisement text data and advertisement image data comprises:

3. The method for mining delivered material according to claim 1, wherein the extracting key information from the advertisement text data and the advertisement image data to obtain a material key information set includes:

4. The method for mining delivered material according to claim 1, wherein the performing feature extraction and feature integration on the advertisement interaction data to obtain target interaction features includes:

5. The method for mining delivered material according to claim 1, wherein the establishing a set of material feature relationships between the target interaction feature and the set of material key information includes:

6. The method for mining delivered stories according to claim 1, wherein the step of inputting the stories feature relation set into a preset delivered stories mining model to analyze delivered stories effects, and obtaining a delivered stories effect analysis result, comprises the steps of:

7. The method for mining delivered material according to claim 1, wherein the constructing the optimal material delivery strategy corresponding to the e-commerce advertisement according to the delivered material effect analysis result comprises:

8. The utility model provides a put in material excavating gear which characterized in that, put in material excavating gear includes:

9. A delivered material mining apparatus, characterized in that the delivered material mining apparatus includes: a memory and at least one processor, the memory having instructions stored therein;

The at least one processor invoking the instructions in the memory to cause the delivered material mining apparatus to perform the delivered material mining method of any of claims 1-7.

10. A computer readable storage medium having instructions stored thereon, which when executed by a processor, implement the delivered-material mining method of any of claims 1-7.