Commodity video content scoring method and system based on multiple features

Abstract

The invention provides a commodity video content scoring method and system based on multiple features, and the method comprises the steps: extracting voice information in a commodity promotion video, and converting the voice information into text information; performing named entity recognition on the text information to obtain key features appearing in the commodity promotion video; performing image recognition on the obtained commodity promotion video, analyzing a commodity display mode, lens application diversity and angle selection, and constructing corresponding features; performing normalization processing on each feature by adopting a maximum and minimum value normalization method to obtain a normalized feature; and a linear regression model is constructed based on the normalized features, each feature corresponds to a weight, the linear regression model determines the weight of each feature through learning, and finally the commodity promotion video content is scored according to the output of the linear regression model to obtain a scoring result. According to the invention, merchants are helped to screen videos with higher attraction and conversion potential, so that the user experience and the commodity conversion rate are improved.

Description

Technical Field

[0001] The present application relates to the technical field of video content scoring, and in particular to a commodity video content scoring method and system based on multiple features. Background Art

[0002] In the current e-commerce and video marketing fields, product promotion videos have become an important means for merchants to attract consumers' attention, showcase product features, and promote sales. However, the existing methods for evaluating the content quality and effectiveness of product promotion videos have the following shortcomings:

[0003] (1) Reliance on subjective evaluation: Traditional methods for rating product promotion videos often rely on users’ subjective feedback, such as ratings and comments. Although this method can reflect the intuitive feelings of some users, it is limited by the differences in users’ personal preferences, emotional states, and evaluation criteria, making it difficult to form an objective and unified rating standard.

[0004] (2) Lack of intelligent analysis: With the development of big data and artificial intelligence technology, although some platforms have begun to try to use data analysis to evaluate video effects, most of these methods remain at the simple data statistics level, such as the number of views and likes, and fail to deeply explore the quality characteristics of the video content itself, such as the accuracy of voice information, the richness of text information, and the accuracy of image recognition.

[0005] (3) Single feature: Existing scoring methods often only focus on one or a few features of the video, such as video length, viewing completion rate, etc., while ignoring the rich multi-feature information in the video content, or lack in-depth analysis of the video content. Multi-feature information is crucial for evaluating the quality and attractiveness of video content. Otherwise, it is difficult to effectively improve the user experience.

[0006] In view of this, the present application proposes a multi-feature-based product video content scoring method, system, device and storage medium, which can deeply mine the multi-feature information in the video content, build an objective, accurate and comprehensive video scoring system, and provide valuable reference for merchants. Summary of the invention

[0007] In order to solve the problems of strong subjectivity, low intelligence and single features in the existing product promotion video scoring methods, the present application provides a product video content scoring method, system, device and storage medium based on multiple features to solve the above technical defects.

[0008] According to the first aspect of the present application, a commodity video content scoring method based on multiple features is proposed, and the method comprises the following steps:

[0009] S1. Obtain a product promotion video, extract voice information from the product promotion video, and convert the voice information into text information through voice recognition technology;

[0010] S2. Use natural language processing technology to perform named entity recognition on text information to obtain key features that appear in product promotion videos;

[0011] S3, performing image recognition on the product promotion video obtained in step S1, analyzing the display method of the product, the diversity of lens application and the angle selection, and constructing product image recognition features, lens application and angle selection features;

[0012] S4. Using the maximum and minimum normalization method to normalize the key features, product image recognition features, lens application and angle selection features to obtain normalized features;

[0013] S5. Construct a linear regression model based on the normalized features, where each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain a scoring result.

[0014] Preferably, the product video content scoring method based on multiple features provided in the present application further includes the following steps:

[0015] S6. Combining the user feedback data, the video monetization capability, and the rating result obtained in step S5, a final rating of the product promotion video is comprehensively evaluated.

[0016] Preferably, in step S5, a linear regression model is constructed based on the normalized features, and a regression method is used to learn the weight corresponding to each feature, and the expression is:

[0017]

[0018] In the formula, represents the predicted value, b represents the bias term, x1x2......x n Represents each feature; w1w2......w n Indicates the weight corresponding to each feature;

[0019] The mean square error including the L2 regularization term is defined as the loss function to evaluate the difference between the predicted value of the linear regression model and the true label value and to control the complexity of the linear regression model. The mean square error expression is:

[0020]

[0021] In the formula, N represents the number of samples, y i represents the true label value of the i-th sample, Represents the predicted value of the i-th sample;

[0022] The expression of the loss function is:

[0023]

[0024] In the formula, Loss represents the loss function, MSE represents the mean square error, λ represents the regularization parameter, and w i represents the weight of the i-th sample.

[0025] Further preferably, in step S6, the user feedback data includes: the number of readings, the number of likes, the number of favorites, the number of comments and the number of reposts; the video monetization capability includes: the number of purchasers and the number of clicks.

[0026] Preferably, in step S2, the key features include: product name, brand, price, color, size specification, logistics introduction, applicable population, applicable scenarios, brand endorsement, celebrity endorsement, new model launch and product efficacy information.

[0027] Further preferably, in step S2, the key feature is constructed in the following manner:

[0028] Calculate the number of occurrences of the key feature category, the first key feature value x1 is the number of occurrences of the category;

[0029] Determine whether a specific key feature among the key features appears, if it appears, the second key feature value x2 is 1, if it does not appear, the second key feature value x2 is 0, wherein the specific key feature includes: product name and brand;

[0030] The number of occurrences of the key feature information is analyzed, and the third key feature value x3 is the number of occurrences of the information.

[0031] Preferably, in step S2, image recognition is performed on the product promotion video obtained in step S1, and analyzing the display method of the products includes: counting the number of different types of products appearing in the video and analyzing the display method of the products in the video.

[0032] In a second aspect, the present application provides a commodity video content rating system based on multiple features, the system comprising:

[0033] A video acquisition module is configured to acquire a product promotion video, extract voice information in the product promotion video, and convert the voice information into text information through voice recognition technology;

[0034] A key feature acquisition module is configured to use natural language processing technology to perform named entity recognition on text information to obtain key features appearing in the product promotion video;

[0035] The remaining feature acquisition modules are configured to perform image recognition on the product promotion video obtained by the video acquisition module, analyze the display method of the product, the diversity of lens application and the angle selection, and construct product image recognition features, lens application and angle selection features;

[0036] A normalization module is configured to use a maximum and minimum value normalization method to normalize key features, product image recognition features, lens application and angle selection features to obtain normalized features;

[0037] The scoring module builds a linear regression model based on normalized features, in which each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain the scoring result.

[0038] In a third aspect, the present application provides a terminal device, including a processor, a memory, and a computer program stored in the memory, wherein the computer program is executed by the processor to implement a multi-feature-based commodity video content scoring method as described above.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, any of the above-mentioned methods for scoring product video content based on multiple features is implemented.

[0040] Compared with the prior art, the beneficial results of the present invention are:

[0041] (1) Improving the objectivity and accuracy of ratings: By comprehensively using multiple technologies such as speech recognition, natural language processing, and image recognition, the present invention can comprehensively extract and analyze multi-feature information in product promotion videos, thereby overcoming the limitations of traditional reliance on subjective evaluation or simple statistics and improving the objectivity and accuracy of ratings.

[0042] (2) Realize intelligent scoring: This method uses machine learning algorithms (such as linear regression models) to automatically learn and determine the weights of each feature, realize intelligent and automated scoring, reduce the workload of manual scoring, and improve the efficiency and consistency of scoring.

[0043] (3) Comprehensive evaluation of video content quality: By considering multiple dimensions such as product display methods, lens diversity, and angle selection, the present invention can more comprehensively evaluate the quality of video content and provide merchants with more detailed improvement suggestions.

[0044] (4) Support dynamic adjustment and optimization: By combining user feedback data and video monetization capabilities, the present invention can dynamically adjust the scoring model to make the scoring results closer to actual market feedback, while supporting merchants to optimize video content based on the scoring results.

[0045] (5) Improving user experience and conversion rate: By accurately evaluating the quality of video content, the present invention helps merchants screen out videos that are more attractive and have greater conversion potential, thereby improving user experience and product conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0047] Figure 1 is a flowchart of a method for scoring product video content based on multiple features according to the present application;

[0048] Figure 2 is a structural diagram of a product video content rating system based on multiple features according to the present application;

[0049] Figure 3 It is a structural diagram of a computer system suitable for implementing an electronic device of an embodiment of the present application. DETAILED DESCRIPTION

[0050] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0051] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0052] Figure 1 The flowchart of the commodity video content scoring method based on multiple features of the present application is shown as follows: Figure 1 As shown, the method comprises the following steps:

[0053] S1. Obtain a product promotion video, extract voice information from the product promotion video, and convert the voice information into text information through voice recognition technology. In this embodiment, firstly, the address of the product promotion video segment needs to be obtained in the database, and then the product promotion video is obtained.

[0054] The voice information is converted into text information through voice recognition technology, which specifically includes the following steps:

[0055] (1) Voice signal acquisition: Voice signals in product promotion videos are collected through microphones and other audio devices and converted into electrical signals, i.e., audio data.

[0056] (2) Preprocessing: Preprocess the collected audio signals, including removing noise, eliminating echoes, adjusting volume, etc., to improve the accuracy of speech recognition.

[0057] (3) Feature extraction: Extract feature vectors that can represent speech characteristics from the preprocessed audio signal. These feature vectors usually include important information of speech signals such as frequency, energy, and harmonics. Common feature extraction methods include Mel-frequency cepstral coefficients (MFCC) and linear predictive coding (LPC).

[0058] (4) Acoustic model training: Use a large amount of annotated speech data to train the acoustic model. The acoustic model is mainly used to perform pattern matching on speech signals and compare the feature extraction results with pre-trained templates to determine the most likely recognition result. Among them, the acoustic models used in this application include hidden Markov models (HMMs) and deep neural networks (DNNs).

[0059] (5) Language model training: The language model is used to correct and modify the language context of the recognition results. The language model is built on a large amount of text data, and predicts and supplements the semantic information in the recognition results by analyzing the vocabulary, syntax and semantic rules in the text. Among them, the language models used in this application include n-gram models and recurrent neural networks (RNNs).

[0060] (6) Decoding process: In the decoding process, the feature-extracted speech signal is decoded using the acoustic model and the language model and converted into a recognition result in text form. In the decoding process, the present application uses a dynamic time warping (DTW) algorithm or a probability-based search algorithm (such as the Viterbi algorithm) to find the best matching recognition result.

[0061] (7) Post-processing of results: Further post-processing of the recognition results, including grammatical analysis, semantic analysis, lexical analysis, etc., to improve the accuracy of the recognition results and the ability of semantic understanding.

[0062] Continue to refer Figure 1 The product video content scoring method based on multiple features provided in this application also includes the following steps:

[0063] S2. Use natural language processing technology to perform named entity recognition on text information and obtain key features that appear in product promotion videos.

[0064] Among them, natural language processing technology is used to perform named entity recognition on text information, which specifically includes the following steps:

[0065] (1) Word segmentation: The text information is segmented according to the language characteristics. In Chinese, continuous sequences of Chinese characters need to be divided into words; in English, the text is usually segmented into words. Words can also be restored to their basic form or their roots can be extracted to unify different variants of the same concept.

[0066] (2) Select named entity recognition tools or models: Select appropriate NER tools or libraries, such as spaCy, NLTK, HanLP, etc. These tools usually provide pre-trained named entity recognition models. If the existing model does not meet the requirements, you can use the annotated dataset to train a custom NER model. The deep learning models that can be used in this application include LSTM+CRF, BERT, etc.

[0067] (3) Named Entity Recognition: The processed text is input into the NER model, and the model will identify the named entities in the text, such as product names, etc. The annotation modes that can be used in this application include BIO (Begin, Inside, Outside), etc., which are used to mark the beginning, inside, and outside of the entity.

[0068] (4) Result extraction and post-processing: Extract all recognized named entities and their types from the output of the NER model. Perform further processing on the results, such as removing duplicate entities and merging adjacent entities of the same type.

[0069] (5) Evaluation and optimization: Use the test set to evaluate the performance of the NER model, including indicators such as accuracy, recall, and F1 score. Optimize the model based on the evaluation results, which may include adjusting model parameters, changing the model architecture, or adding training data.

[0070] (6) Application to product promotion videos: Extract key features (such as product name, brand, etc.) identified by NER as key information of product promotion videos. Use this key information to further analyze product promotion videos.

[0071] In this embodiment, according to the translated text information, natural language processing technology is used to perform NER (named entity recognition) to identify and extract important information such as product names, categories, brands, and marketing selling points that appear in the video. This information is crucial for evaluating the key features of video content because they reveal the uniqueness and appeal of the product. For example, if the video emphasizes the ultra-long standby time of a certain mobile phone or the natural ingredients of a certain cosmetic, these key selling points can be accurately captured, thereby affecting the final scoring results. Specific key features include: product name, brand, price, color, size specifications, logistics introduction, applicable population, applicable scenarios, brand endorsement, celebrity endorsement, new models on the market, and product efficacy information. The following table is a detailed explanation of these key features:

[0072]

[0073]

[0074] Further preferably, in step S2, the key feature is constructed in the following manner:

[0075] (1) Calculate the number of occurrences of the key feature category, and the first key feature value x1 is the number of occurrences of the category. Specifically, the number of occurrences of the key feature category extracted from the product selling point information is counted as x1. For example, if a product promotion video contains the product name, product category, color, and price of the product, then the first key feature value x1 = 4.

[0076] (2) Determine whether a specific key feature among the key features appears. If it appears, the second key feature value x2 is 1; if it does not appear, the second key feature value x2 is 0. The specific key features include: product name and brand. That is, whether the product name and brand information appear is counted as x2. If it appears, x2 = 1; if it does not appear, x2 = 0.

[0077] (3) Analyze the number of occurrences of key feature information. The third key feature value x3 is the number of occurrences of information. That is, the number of occurrences of product selling point information extraction features is counted as x3. For example, if the product has four product functions and two applicable groups, then x3 = 6.

[0078] Continue to refer Figure 1 The product video content scoring method based on multiple features provided in this application also includes the following steps:

[0079] S3. Perform image recognition on the product promotion video obtained in step S1, analyze the product display method, lens application diversity and angle selection, and construct product image recognition features, lens application and angle selection features.

[0080] This embodiment involves an in-depth understanding and analysis of the content of a product promotion video. The product images that appear in the video will be identified, and different types of product display methods will be analyzed. In addition, the use of lenses and angle selection in the video will be evaluated to ensure that the video content can attract the audience's attention and effectively convey product information. By comprehensively analyzing multiple aspects of the video content, the quality and attractiveness of the video can be more comprehensively evaluated, thereby affecting the final scoring results. The following table is a detailed explanation of the product image recognition features, lens application and angle selection features, as well as a construction description.

[0081]

[0082]

[0083] Continue to refer Figure 1 The product video content scoring method based on multiple features provided in this application also includes the following steps:

[0084] S4. Using the maximum and minimum normalization method to normalize the key features, product image recognition features, lens application and angle selection features to obtain normalized features;

[0085] S5. Construct a linear regression model based on the normalized features, where each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain a scoring result.

[0086] In a specific embodiment, the currently constructed multiple features cannot be used directly, and the features need to be normalized, which is mainly used to eliminate the dimensional influence between data features so that different features are comparable. Specifically, normalization usually linearly converts the original data to a specific range (such as 0 to 1 or -1 to 1) so that the data conforms to a certain specific distribution, which can make different features or data comparable in order of magnitude and avoid certain features from having too much influence on the final result due to their large absolute values.

[0087] This application uses the Min-Max Normalization method for normalization. This method maps the original data to the [0,1] interval. Specifically, the original value of each feature is subtracted from the minimum value of the feature, and then divided by the difference between the maximum and minimum values ​​of the feature. The expression is:

[0088]

[0089] Where, X norm represents the normalized data value, whose value range is [0,1]; X represents a value in the original data; X min Indicates the minimum value in the sample data; X max Indicates the maximum value in the sample data.

[0090] In step S5, a linear regression model is constructed based on the normalized features, and the weight corresponding to each feature is learned using a regression method, and the expression is:

[0091]

[0092] In the formula, represents the predicted value, b represents the bias term, x1, x2...x n Represents each feature; w1, w2...w n Indicates the weight corresponding to each feature;

[0093] The mean square error including the L2 regularization term is defined as the loss function to evaluate the difference between the predicted value of the linear regression model and the true label value and to control the complexity of the linear regression model. The mean square error expression is:

[0094]

[0095] In the formula, N represents the number of samples, y i represents the true label value of the i-th sample, Represents the predicted value of the i-th sample;

[0096] The expression of the loss function is:

[0097]

[0098] Where Loss represents the loss function, MSE represents the mean square error, and λ represents the regularization parameter, which is used to control the influence of the regularization term. i Represents the weight of the i-th sample. By minimizing the above loss function with L2 regularization term, the loss function with L2 regularization term can be minimized through optimization algorithms such as gradient descent, so as to learn the weight of each feature and predict the label data.

[0099] Through the above steps, the corresponding linear regression model is constructed, so that the new product promotion video can be given a corresponding score by extracting the corresponding features and inputting them into the model.

[0100] In a specific embodiment, the product video content scoring method based on multiple features provided by the present application further includes the following steps:

[0101] S6. Combining the user feedback data, the video monetization capability, and the rating result obtained in step S5, a final rating of the product promotion video is comprehensively evaluated.

[0102] The quality of a product promotion video can be evaluated through multiple indicators, including user feedback data such as the number of views, likes, favorites, comments, and reposts. These indicators have a relatively small weight in the score. At the same time, the number of purchases and clicks of the product promotion link are directly related to the monetization ability of the video and should have a greater weight. In addition, the audio content in the video is analyzed using a large language model, and the clarity of the video explanation is evaluated by selecting different products for the model. Finally, YES can combine manual subjective scoring to consider factors such as the comfort, fluency, and degree of purchase desire stimulation of the video to score the product promotion short video.

[0103] Further references Figure 2As an implementation of the above method, in the second aspect, the present application provides an embodiment of a structure diagram of a commodity video content rating system 200 based on multiple features, which can be specifically applied to various electronic devices. The system 200 includes the following modules:

[0104] The video acquisition module 210 is configured to acquire the product promotion video, extract the voice information in the product promotion video, and convert the voice information into text information through voice recognition technology;

[0105] The key feature acquisition module 220 is configured to perform named entity recognition on text information using natural language processing technology to obtain key features appearing in the product promotion video;

[0106] The remaining feature acquisition module 230 is configured to perform image recognition on the product promotion video obtained by the video acquisition module, analyze the display method of the product, the diversity of lens application and the angle selection, and construct product image recognition features, lens application and angle selection features;

[0107] The normalization module 240 is configured to use a maximum and minimum value normalization method to normalize the key features, product image recognition features, lens application and angle selection features to obtain normalized features;

[0108] The scoring module 250 constructs a linear regression model based on the normalized features, wherein each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain a scoring result.

[0109] In a third aspect, the present application provides a terminal device, including a processor, a memory, and a computer program stored in the memory, wherein the computer program is executed by the processor to implement a multi-feature-based commodity video content scoring method as described above.

[0110] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, any of the above-mentioned methods for scoring product video content based on multiple features is implemented.

[0111] Reference below Figure 3 , which shows a schematic diagram of the structure of a computer system 300 of a terminal device or server suitable for implementing an embodiment of the present application. Figure 3 The terminal device or server shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0112] like Figure 3As shown, the computer system 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage part 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the computer system 300 are also stored. The CPU 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0113] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage section 308 as needed.

[0114] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and / or installed from the removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, the above functions defined in the method of the present application are executed. It should be noted that the computer-readable medium described in the present application can be a computer-readable signal medium or a computer-readable medium or any combination of the above two. The computer-readable medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, an apparatus or a device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, an apparatus or a device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0115] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0116] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0117] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A method for scoring product video content based on multiple features, characterized in that: The method comprises the following steps: S1. Obtain a product promotion video, extract voice information from the product promotion video, and convert the voice information into text information using voice recognition technology; S2. Performing named entity recognition on the text information using natural language processing technology to obtain key features appearing in the product promotion video; S3, performing image recognition on the product promotion video obtained in step S1, analyzing the display method of the product, the diversity of lens application and the angle selection, and constructing product image recognition features, lens application and angle selection features; S4, using a maximum and minimum normalization method to normalize the key features, product image recognition features, lens application and angle selection features to obtain normalized features; S5. Construct a linear regression model based on the normalized features, wherein each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain a scoring result.

2. The method for scoring product video content based on multiple features according to claim 1, characterized in that: The following steps are also included: S6. Combining the user feedback data, the video monetization capability, and the rating result obtained in step S5, a final rating of the product promotion video is comprehensively evaluated.

3. The method for scoring product video content based on multiple features according to claim 1, characterized in that: In step S5, a linear regression model is constructed based on the normalized features, and a regression method is used to learn the weight corresponding to each feature, and the expression is: In the formula, represents the predicted value, b represents the bias term, x1, x2...x n Indicate each feature; w1、w2......w n Indicates the weight corresponding to each feature; The mean square error including the L2 regularization term is defined as the loss function to evaluate the difference between the predicted value and the true label value of the linear regression model and to control the complexity of the linear regression model. The mean square error expression is: In the formula, N represents the number of samples, y i represents the true label value of the i-th sample, Represents the predicted value of the i-th sample; The expression of the loss function is: In the formula, Loss represents the loss function, MSE represents the mean square error, λ represents the regularization parameter, and w i represents the weight of the i-th sample.

4. The method for scoring commodity video content based on multiple features according to claim 2, characterized in that: In step S6, the user feedback data includes: the number of readings, the number of likes, the number of favorites, the number of comments and the number of reposts; the video monetization capability includes: the number of purchasers and the number of clicks.

5. The method for scoring product video content based on multiple features according to claim 1, characterized in that: In step S2, the key features include: product name, brand, price, color, size specification, logistics introduction, applicable population, applicable scenarios, brand endorsement, celebrity endorsement, new model launch and product efficacy information.

6. The method for scoring product video content based on multiple features according to claim 5, characterized in that: In step S2, the key features are constructed in the following manner: Calculate the number of occurrences of the category of the key feature, the first key feature value x1 is the number of occurrences of the category; Determine whether a specific key feature among the key features appears, if it appears, the second key feature value x2 is 1, if it does not appear, the second key feature value x2 is 0, wherein the specific key feature includes: product name and brand; The number of occurrences of the key feature information is analyzed, and the third key feature value x3 is the number of occurrences of the key feature information.

7. The method for scoring product video content based on multiple features according to claim 1, characterized in that: In step S2, image recognition is performed on the product promotion video obtained in step S1, and the display method of the products is analyzed, including: counting the number of different types of products appearing in the video and analyzing the display method of the products in the video.

8. A product video content rating system based on multiple features, characterized in that: The system comprises: A video acquisition module, configured to acquire a product promotion video, extract voice information from the product promotion video, and convert the voice information into text information through voice recognition technology; A key feature acquisition module is configured to perform named entity recognition on the text information using natural language processing technology to obtain key features appearing in the product promotion video; The remaining feature acquisition modules are configured to perform image recognition on the product promotion video obtained by the video acquisition module, analyze the display method of the product, the diversity of lens application and the angle selection, and construct product image recognition features, lens application and angle selection features; A normalization module is configured to normalize the key features, product image recognition features, lens application and angle selection features using a maximum and minimum value normalization method to obtain normalized features; The scoring module constructs a linear regression model based on the normalized features, wherein each feature corresponds to a weight. The linear regression model determines the weight of each feature through learning, and finally scores the product promotion video content according to the output of the linear regression model to obtain a scoring result.

9. A terminal device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory, wherein the computer program is executed by the processor to implement the commodity video content scoring method based on multiple features as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, wherein a computer program is stored in the medium, and when the computer program is executed by a processor, the method for scoring product video content based on multiple features as described in any one of claims 1 to 7 is implemented.

Images