Article recommendation method and device, equipment and storage medium

Abstract

The invention provides an article recommendation method and device, equipment and a storage medium, and the article recommendation method comprises the steps: collecting a physiological signal of a target object when the target object browses a page, carrying out the emotion recognition of the physiological signal, obtaining a character emotion, recognizing the character emotion based on the physiological signal, and considering an emotion factor; hot spot information is obtained, an article emotion vector is determined according to the hot spot information, and emotion factors reflected by an article are considered; and inputting the character emotion and the article emotion vector into an analysis model to obtain a target article, and performing recommendation display on the target article, thereby improving the accuracy of recommendation display of the target article.

Description

Technical Field

[0001] This application relates to the field of computers, and in particular, to an item recommendation method, apparatus, device, and storage medium. Background Art

[0002] In recent years, with the rapid development of the Internet and the emergence of big data, recommendation systems can analyze users' behaviors and interests, provide customized recommendation content for users, improve user satisfaction and interaction experience, and have been widely used. However, current recommendation systems lack the analysis of personalized factors such as users' emotional factors. In daily life, emotion is a manifestation of a person's comprehensive state and a stress response to external stimuli. People express the same basic emotions under the same semantics. Emotional factors such as joy and satisfaction generated by consumers when browsing content play an important role in making purchase decisions. Therefore, it is necessary to combine the recognition of users' emotional factors on the basis of recommendation algorithms to improve the accuracy of recommendations. Summary of the Invention

[0003] Embodiments of this application provide an item recommendation method, apparatus, device, and storage medium to solve at least one problem existing in the related art. The technical solutions are as follows:

[0004] In a first aspect, embodiments of this application provide an item recommendation method, including:

[0005] When a target object browses a page, collect the physiological signals of the target object;

[0006] Perform emotion recognition on the physiological signals to obtain the emotion of the person;

[0007] Obtain hot information, and determine an item emotion vector according to the hot information;

[0008] Input the emotion of the person and the item emotion vector into an analysis model to obtain a target item, and recommend and display the target item.

[0009] In an implementation manner, the performing emotion recognition on the physiological signals to obtain the emotion of the person includes:

[0010] Calculate target feature values of the physiological signals respectively based on different feature calculation rules;

[0011] According to each target feature value and an identification network, determine an emotion recognition result corresponding to each target feature value;

[0012] Analyze each emotion recognition result to determine the emotion of the person.

[0013] In one embodiment, calculating the target feature values of the physiological signal based on different feature calculation rules includes at least two of the following:

[0014] Calculating the Lempel-Ziv complexity of the physiological signal based on a first feature calculation rule;

[0015] Calculating the wavelet detail coefficients of the physiological signal based on a second feature calculation rule;

[0016] Calculating the cointegration degree of the physiological signal based on a third feature calculation rule;

[0017] Decomposing and extracting the average approximate entropy of the physiological signal based on a fourth feature calculation rule;

[0018] Wherein, the target feature values include at least two of the Lempel-Ziv complexity, wavelet detail coefficients, cointegration degree, and average approximate entropy.

[0019] In one embodiment, analyzing each of the emotion recognition results to determine the emotion of the person includes:

[0020] Analyzing each of the emotion recognition results by using fuzzy integral to determine a first score corresponding to a positive emotion and a second score corresponding to a negative emotion. When the first score is greater than the second score, determining the positive emotion as the emotion of the person; otherwise, determining the negative emotion as the emotion of the person;

[0021] Or,

[0022] Determining a first quantity of positive emotions and a second quantity of negative emotions according to the emotion recognition results. When the first quantity is greater than the second quantity, determining the positive emotion as the emotion of the person; otherwise, determining the negative emotion as the emotion of the person.

[0023] In one embodiment, determining the item emotion vector according to the hot information includes:

[0024] Determining the item corresponding to the hot information;

[0025] Performing vector encoding on the item to obtain an item feature vector including a plurality of item features, and calculating the correlation degree between the hot information and each of the item features;

[0026] Inputting the target item feature with the highest correlation degree into a classification network to obtain an item emotion vector segmented based on Arousal and Valence.

[0027] In one embodiment, the method further includes:

[0028] When the target object browses the target item, collect new physiological signals and determine the new person's emotion according to the new physiological signals, or receive the evaluation score of the target item and determine the new person's emotion according to the evaluation score;

[0029] When the new person's emotion is a positive emotion, continuously recommend the target item; otherwise, input the negative emotion and the item emotion vector into the analysis model to determine a new target item and perform a recommended display.

[0030] In one implementation, the analysis model is obtained through the following steps:

[0031] Obtain the person emotion training data and the item emotion vector training data;

[0032] Use the person emotion training data and the item emotion vector training data to train the initialization model in the blockchain to obtain the analysis model through federated learning.

[0033] In a second aspect, an embodiment of the present application provides an item recommendation device, including:

[0034] A collection module, configured to collect the physiological signals of the target object when the target object browses a page;

[0035] An identification module, configured to perform emotion recognition on the physiological signals to obtain the person's emotion;

[0036] A determination module, configured to obtain hot information and determine the item emotion vector according to the hot information;

[0037] A recommendation module, configured to input the person's emotion and the item emotion vector into the analysis model to obtain a target item, and perform a recommended display of the target item.

[0038] In one implementation, the recommendation module is further configured to:

[0039] When the target object browses the target item, collect new physiological signals and determine the new person's emotion according to the new physiological signals, or receive the evaluation score of the target item and determine the new person's emotion according to the evaluation score;

[0040] When the new person's emotion is a positive emotion, continuously recommend the target item; otherwise, input the negative emotion and the item emotion vector into the analysis model to determine a new target item and perform a recommended display.

[0041] In a third aspect, an embodiment of the present application provides an electronic device, including: a processor and a memory, where instructions are stored in the memory, and the instructions are loaded and executed by the processor to implement the method in any one of the above aspects.

[0042] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, which when executed implements the method in any one of the above aspects.

[0043] The beneficial effects in the above technical solutions at least include:

[0044] By collecting the physiological signals of the target object when browsing the page, performing emotion recognition on the physiological signals to obtain the human emotion, identifying the human emotion based on the physiological signals and considering the emotional factors; obtaining the hot information, determining the item emotion vector according to the hot information and considering the emotional factors reflected by the item; inputting the human emotion and the item emotion vector into the analysis model to obtain the target item, and recommending and displaying the target item, the accuracy of the recommended display of the target item is improved.

[0045] The above summary is only for the purpose of the specification and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present application will be readily apparent by reference to the drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In the drawings, unless otherwise specified, the same reference numerals throughout the several views denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present application and should not be regarded as limiting the scope of the present application.

[0047] Figure 1 It is a schematic flow chart of the steps of an item recommendation method according to an embodiment of the present application;

[0048] Figure 2 It is a schematic emotion diagram based on the segmentation of Arousal and Valence;

[0049] Figure 3 It is a structural block diagram of an item recommendation device according to an embodiment of the present application;

[0050] Figure 4 It is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] In the following, only some exemplary embodiments are briefly described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and the description are considered to be exemplary in nature and not restrictive.

[0052] In related technologies, traditional federated learning relies on a trusted central server. Each institution, enterprise, and organization collaborates to train a global model while ensuring that the data does not leave the local area. The entire training process is vulnerable to the performance and failures of the central server. In addition, some studies have shown that unencrypted intermediate parameters can be used to infer important information in the training data, and the private data of participants faces the risk of exposure. Therefore, in the process of model training, it is particularly important to adopt a suitable encryption scheme for local model updates and maintain the global model on distributed nodes.

[0053] Referring to Figure 1 , a flowchart of an item recommendation method according to an embodiment of the present application is shown. The item recommendation method may at least include steps S100-S400:

[0054] S100. When the target object browses a page, collect the physiological signals of the target object.

[0055] S200. Perform emotion recognition on the physiological signals to obtain the emotion of the person.

[0056] S300. Obtain hot information and determine the item emotion vector according to the hot information.

[0057] S400. Input the emotion of the person and the item emotion vector into an analysis model to obtain a target item, and recommend and display the target item.

[0058] The item recommendation method according to the embodiment of the present application can be executed by an electronic control unit, a controller, a processor, etc. of terminals such as a computer, a mobile phone, a tablet, and a vehicle-mounted terminal, or can also be executed by a cloud server.

[0059] The technical solution of the embodiment of the present application collects the physiological signals of the target object when the target object browses a page, performs emotion recognition on the physiological signals to obtain the emotion of the person, recognizes the emotion of the person based on the physiological signals, and considers the emotional factor; obtains hot information and determines the item emotion vector according to the hot information, and considers the emotional factor reflected by the item; inputs the emotion of the person and the item emotion vector into an analysis model to obtain a target item, and recommends and displays the target item, thereby improving the accuracy of the recommendation and display of the target item.

[0060] In one embodiment, in step S100, when the target object views the page content on the browsing page, for example, when shopping for items on the browsing page, physiological signals of the target object are collected, including but not limited to electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, etc. Optionally, EEG signals can be collected through an EEG signal detector (Emotiv Epoc X electroencephalogram detection and analysis instrument), and ECG signals can be collected through an ECG signal collector. The former, the Emotiv Epoc X electroencephalogram detection and analysis instrument, has a total of 14 channels for the user to wear, that is, the obtained EEG signals have signals of 14 channels; the ECG signal collector is two ECG electrode patches, which are respectively pasted on the wrist pulses of the left and right hands. At this time, ECG signals can be collected. Meanwhile, when the target object browses the page, corresponding behavior data can be collected to generate log data, such as behaviors like browsing, clicking on products, staying, commenting, liking, and collecting. The data is uniformly cleaned, escaped, reorganized, merged, split, and globally processed with unified features, and finally stored in a data warehouse for use by recommendation systems, machine learning, etc., as the data source for determining the consumption category to which the target object belongs, user preferences, etc.

[0061] In one embodiment, the processing process of the present application embodiment taking the physiological signal as the EEG signal is described, and the processing method of the ECG signal is similar. Optionally, after the EEG signals are collected, preprocessing can be performed. The preprocessing includes but is not limited to denoising the EEG signals through independent component analysis (ICA), using a Butterworth filter to extract effective EEG signals, and then performing the next processing on the effective EEG signals.

[0062] In one embodiment, step S200 includes steps S210 - S230:

[0063] S210. Calculate the target feature values of the physiological signals respectively based on different feature calculation rules.

[0064] Optionally, the target feature values include Lempel - Ziv complexity, wavelet detail coefficients, cointegration relationship degree, and average approximate entropy. Corresponding step S210 includes steps S2101 - S2014. In other embodiments, the target feature values include at least two of Lempel - Ziv complexity, wavelet detail coefficients, cointegration relationship degree, and average approximate entropy. At this time, the corresponding step S210 includes at least two steps among steps S2101 - S2014. Specifically:

[0065] S2101. Calculate the Lempel - Ziv complexity of the physiological signals based on the first feature calculation rule.

[0066] S2102. Calculate the wavelet detail coefficients of the physiological signals based on the second feature calculation rule.

[0067] S2103. Calculate the cointegration degree of the physiological signal based on the third feature calculation rule.

[0068] S2104. Decompose and extract the average approximate entropy of the physiological signal based on the fourth feature calculation rule.

[0069] In the embodiments of the present application, corresponding algorithms or tools such as Lempel-Ziv complexity, wavelet detail coefficients, cointegration degree, and EMD decomposition to calculate the average approximate entropy can be pre-configured as the first feature calculation rule, the second feature calculation rule, the third feature calculation rule, and the fourth feature calculation rule respectively, and then these rules are used to calculate the Lempel-Ziv complexity, wavelet detail coefficients, cointegration degree, and average approximate entropy of the physiological signal respectively.

[0070] S220. Determine the emotion recognition result corresponding to each target eigenvalue according to each target eigenvalue and the recognition network.

[0071] Optionally, the Lempel-Ziv complexity, wavelet detail coefficients, cointegration degree, and average approximate entropy of the physiological signal are respectively input into a pre-trained recognition network, and based on the output result of the recognition network, the emotion recognition result corresponding to each target eigenvalue is determined, that is, the emotion recognition results corresponding to the Lempel-Ziv complexity, wavelet detail coefficients, cointegration degree, and average approximate entropy of the physiological signal respectively.

[0072] S230. Analyze each emotion recognition result to determine the emotion of the person.

[0073] Optionally, step S230 includes step S2301 or S2302:

[0074] S2301. Analyze each emotion recognition result using fuzzy integral to determine the first score corresponding to the positive emotion and the second score corresponding to the negative emotion. When the first score is greater than the second score, determine the positive emotion as the emotion of the person, otherwise determine the negative emotion as the emotion of the person.

[0075] Optionally, the positive emotion can be positive emotions such as happy, excited, pleasant, satisfied, etc., and the negative emotion is negative emotions such as unhappy, unpleasant, angry, fearful, sad, depressed, calm, etc. In the embodiments of the present application, the emotion recognition result may include the first probability of the positive emotion and the second probability of the negative emotion. Analyze each emotion recognition result using the method of fuzzy integral. For example, corresponding weights can be set for each emotion recognition result and weighted calculation can be performed in combination with the first probability and the second probability to determine the first score corresponding to the positive emotion and the second score corresponding to the negative emotion. Then, when the first score is greater than the second score, determine the positive emotion as the emotion of the person, otherwise determine the negative emotion as the emotion of the person.

[0076] S2302. Determine the first quantity of positive emotions and the second quantity of negative emotions according to the emotion recognition result. When the first quantity is greater than the second quantity, determine that the positive emotion is the human emotion; otherwise, determine that the negative emotion is the human emotion.

[0077] Optionally, the first quantity of positive emotions and the second quantity of negative emotions can be determined according to the emotion recognition result. For example, if there are three emotion recognition results all being positive emotions, at this time the first quantity is 3 and the second quantity is 1, and at this time determine that the positive emotion is the human emotion.

[0078] It should be noted that when using the Emotiv Epoc X electroencephalogram detection analyzer to collect electroencephalogram signals with a total of 14 channels, that is, the electroencephalogram signals have signals of 14 channels. Taking the Lempel-Ziv complexity as an example, in the emotion recognition result corresponding to the Lempel-Ziv complexity, there are sub-emotion recognition results of 14 channels, and each sub-emotion recognition result forms a separate LIBSVM classifier. At this time, when the third probability of the positive emotion of the sub-emotion recognition result is greater than the probability threshold, it is considered that the sub-emotion recognition result represents a positive emotion; otherwise, it represents a negative emotion. The emotion represented by the emotion recognition result can be determined by a voting method. For example, among the 14 sub-emotion recognition results, 8 are positive emotions and 6 are negative emotions, and 8 is greater than 6, so this emotion recognition result represents a positive emotion.

[0079] In one implementation manner, determining the item emotion vector according to the hot information in step S300 includes steps S310 - S330:

[0080] S310. Determine the item corresponding to the hot information.

[0081] Optionally, the hot information of each platform on the network can be obtained through big data technology, the hot information can be extracted, the current popular keyword sentences can be determined, and they can be corresponded to the commodities one by one, so as to determine the item corresponding to the hot information.

[0082] S320. Perform vector encoding on the item to obtain an item feature vector containing several item features, and calculate the correlation degree between the hot information and each item feature.

[0083] Optionally, classify the items and use a series of text tags to describe the item features, such as text tags like health, reading, internet celebrity, household, food, etc. Vectorize the text tags using word2vector to obtain an item feature vector containing several item features, for example, [health, reading, travel, internet celebrity...]. Then, denoise and normalize the item feature vector pair representing the commodity vector, extract linear features using machine learning methods, and extract high-dimensional features using a deep dnn network, so as to extract each item feature in the item feature vector. Finally, based on the hot features extracted from the hot information, calculate the correlation between the item features and the hot information (specifically, the hot features corresponding to the hot information) through the cos(|a,b|) formula (a is the item feature, b is the hot feature), thereby obtaining the correlation degree between the hot information and each item feature.

[0084] S330. Input the target item feature with the highest correlation degree into the classification network to obtain an item emotion vector segmented based on Arousal and Valence.

[0085] Then, input the target item feature with the highest correlation degree into the trained classification network, such as the CNN-LSTM network, to obtain an item emotion vector segmented based on Arousal and Valence. As Figure 2 shown, segment Arousal and Valence into multi-dimensional emotions with 0.5 to obtain the item emotion vector.

[0086] In one implementation, in step S400, input the person emotion and the item emotion vector into the trained analysis model. The analysis model outputs the target item that meets the person emotion based on the person emotion factor and the item emotion factor, and then recommends and displays the target item on the page for the target object (user) to browse, improving the accuracy of item recommendation and being beneficial to improving the browsing experience of the target object. It should be noted that the target item can be one or more items of the same or similar type.

[0087] In one implementation, step S400 can also be to input the person emotion, the item emotion vector, and supplementary data into the trained analysis model to obtain a more accurate target item. Among them, the supplementary data can be behavior data and / or hot information.

[0088] Optionally, the analysis model can be obtained through steps S410 - S420:

[0089] S410. Obtain the person emotion training data and the item emotion vector training data.

[0090] Optionally, before model training, physiological signals can be collected during user page browsing for emotion recognition, such as in step S200, to obtain person emotion training data. At the same time, based on current hot information, such as in step S300, item emotion vector training data is determined. The person emotion training data and the item emotion vector training data together constitute the training data for model training. In some embodiments, the person emotion training data and the item emotion vector training data can use open-source data sets, such as the DEAP data set.

[0091] S420. Use the person emotion training data and the item emotion vector training data to train the initialization model in the blockchain, so as to obtain an analysis model through federated learning training.

[0092] In the embodiments of the present application, the blockchain is used to replace the parameter server in traditional federated learning, and the collaborative training data is stored in a decentralized manner to solve security problems such as the single-point failure of the central server. At the same time, homomorphic encryption and secret sharing are combined to achieve traceable privacy protection for federated learning. For example, the system may include a task publisher, participants, a committee, and a blockchain. The task publisher publishes a federated learning model training task, constructs an initialization model W and publishes it to the blockchain. Nodes interested in this task can apply to participate in the model training, download the initialization model W from the blockchain and train it using their respective training data. After training, the model uploads the encrypted local update to the blockchain. The committee downloads all local updates, aggregates all local updates on the blockchain to determine the aggregated model, and uploads the aggregated model as the global update to the blockchain. The task publisher downloads the aggregated model to obtain the final model, that is, the analysis model obtained through federated learning training.

[0093] 1) Publisher. According to the task requirements, construct an initialization model W, such as an LSTM model, and at the same time explain the requirements of the federated learning model training task (requirements for storage, computing power, etc.). The task publisher uploads the initialization model W to the blockchain. Nodes interested and meeting the requirements can apply to join and participate in this federated learning task. As more and more nodes join and conduct model training, the task publisher can finally obtain a high-value machine learning model, that is, the analysis model.

[0094] 2) Participant. Nodes that apply to join the model training task and are approved by the task publisher have a need for this initialization model, but are unable to complete the entire training task alone due to insufficient local storage, computing power, or limited data resources. The participant trains the initialization model W locally, and uploads the trained initialization model W as a local update to the blockchain. When uploading, it is necessary to declare the size of the local data volume and attach the corresponding training time consumption, so as to indicate the size of its own data contribution.

[0095] 3) Committee. The system sets the number of members in the committee to half of the number of all participants. The initial committee is randomly composed of participants designated by the task publisher and serves as the leader of the committee in turn, responsible for aggregating each local update, that is, aggregating the models after each local training and distributing data contribution rewards to the corresponding participants. When a committee member who has taken turns as the leader for a round or is not online when not serving as the leader, the system will elect half of the nodes to form a new committee according to the reputation value of the participants from high to low.

[0096] 4) Blockchain. Since the consortium chain, as a special type of blockchain, has access control and meets the setting of pre-auditing users in advance and having a relatively stable set of participants, the consortium chain is used to replace the parameter server in traditional federated learning to store local updates and global updates. All participants jointly maintain the blockchain ledger, and the failure or withdrawal of a single node does not affect other participants' access to information, which improves the data disaster tolerance ability. In addition, due to the characteristics of transparency, traceability, and non-repudiation of the blockchain, the system's historical reputation evaluation of each participant is also recorded on the blockchain, and the reputation-based federated learning committee election scheme becomes more open, transparent, and reliable.

[0097] In one implementation manner, the item recommendation method of the embodiment of the present application may further include steps S510 - S520:

[0098] S510. When the target object browses the target item, collect new physiological signals and determine the new person's emotion according to the new physiological signals, or receive the evaluation score of the target item and determine the new person's emotion according to the evaluation score.

[0099] Optionally, when the target object browses the target item shown in the recommendation, continuously and real-time collect new physiological signals and determine the new person's emotion, as the processing method in step S200, which will not be elaborated here. Or, when the target object browses the target item shown in the recommendation, provide an interaction page for the target object to give a score feedback on the target item. For example, a score threshold can be set. When the score is greater than the score threshold, the new person's emotion can be determined as a positive emotion, otherwise as a negative emotion.

[0100] S520. When the new person's emotion is a positive emotion, continuously recommend the target item; otherwise, input the negative emotion and the item emotion vector into the analysis model to determine a new target item and perform a recommendation display.

[0101] In the embodiments of the present application, when the new person's emotion is a positive emotion, the target item is continuously recommended to maintain the positive emotion of the current target object, which is beneficial to improving the browsing and shopping experience of the target object; otherwise, it indicates that the new person's emotion is a negative emotion at this time. The negative emotion and the item emotion vector are input into the analysis model to determine the new target item and re-recommend and display it, and the display content of the current page is adjusted to improve the recognition accuracy of the model and positively improve the model.

[0102] It should be noted that, for example, when the target object selects a suitable item and adds it to the shopping cart or makes a purchase, it can be determined according to the physiological signal that the new person's emotion is a negative emotion at this time. The content of the recommendation display can be adjusted in a timely manner to continuously maintain a good browsing and shopping experience for the target object. For example, the item recommended at this time is of other types that have not been added to the shopping cart or purchased, and the target object shows a positive emotion again at this time.

[0103] Through the method of the embodiments of the present application, at least the following effects can be achieved:

[0104] 1) The average emotion recognition rate for binary classification of Valance can reach 82.63%, and the average emotion recognition rate for binary classification of Arousal can reach 74.88%, with a high average emotion recognition rate.

[0105] 2) A blockchain is designed to replace the parameter server in traditional federated learning, and the collaborative training data is stored in a decentralized manner to solve security problems such as the single-point failure of the central server. At the same time, a federated learning method for traceable privacy protection is realized by combining homomorphic encryption and secret sharing. The present application adopts a suitable encryption scheme for local model updates in the blockchain. Secret sharing divides the secret into several shares through specific operations and distributes them to multiple participants. The secret recovery is jointly carried out by multiple participants in cooperation, and a single secret cannot be decrypted, ensuring the privacy of user data.

[0106] 3) Combine the person emotion factor and the item emotion factor to improve the accuracy of recommendation and improve the browsing experience of the target object.

[0107] Refer to Figure 3 , which shows the structural block diagram of the item recommendation device according to an embodiment of the present application. The device may include:

[0108] An acquisition module, configured to acquire the physiological signal of the target object when the target object browses the page;

[0109] An identification module, configured to perform emotion recognition on the physiological signal to obtain the person emotion;

[0110] A determination module, configured to obtain hot information and determine the item emotion vector according to the hot information;

[0111] A recommendation module for inputting the human emotion and the item emotion vector into an analysis model to obtain a target item and performing recommended display of the target item.

[0112] In one implementation, the recommendation module is further configured to:

[0113] When the target object browses the target item, collect new physiological signals and determine a new human emotion according to the new physiological signals, or receive an evaluation score of the target item and determine a new human emotion according to the evaluation score;

[0114] When the new human emotion is a positive emotion, continuously recommend the target item; otherwise, input the negative emotion and the item emotion vector into the analysis model to determine a new target item and perform recommended display.

[0115] For the functions of the modules in the devices of the embodiments of the present application, reference may be made to the corresponding descriptions in the above methods, which will not be elaborated herein.

[0116] Refer to Figure 4 , which shows a structural block diagram of an electronic device according to an embodiment of the present application. The electronic device includes: a memory 310 and a processor 320. Instructions that can run on the processor 320 are stored in the memory 310. The processor 320 loads and executes the instructions to implement the item recommendation method in the above embodiment. Among them, the number of the memory 310 and the processor 320 may be one or more.

[0117] In one implementation, the electronic device further includes a communication interface 330 for communicating with external devices and performing data interaction and transmission. If the memory 310, the processor 320, and the communication interface 330 are implemented independently, the memory 310, the processor 320, and the communication interface 330 may be connected to each other through a bus and complete communication with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 4 only a thick line is shown in

[0118] but it does not mean that there is only one bus or one type of bus.

[0119] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the item recommendation method provided in the above embodiment.

[0120] An embodiment of the present application further provides a chip, which includes a processor for calling and running instructions stored in a memory, so that a communication device equipped with the chip executes the method provided in the embodiment of the present application.

[0121] An embodiment of the present application further provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, the output interface, the processor, and the memory are connected through an internal connection path. The processor is configured to execute code in the memory, and when the code is executed, the processor is configured to execute the method provided in the embodiment of the application.

[0122] It should be understood that the above processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0123] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may further include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus random access memory (DR RAM).

[0124] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0125] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0126] In addition, the terms "first" and "second" are used only for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of these features. In the description of this application, "a plurality of" means two or more, unless otherwise specifically defined.

[0127] Any process or method description represented in the flowchart or described in other ways herein can be understood to represent a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process. And the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in the reverse order according to the functions involved, rather than in the order shown or discussed.

[0128] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a sequenced list of executable instructions for implementing a logical function, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in combination with the instruction execution system, apparatus, or device for these instructions.

[0129] It should be understood that each part of this application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the method in the above embodiments can be completed by a program instructing relevant hardware, and this program can be stored in a computer-readable storage medium. When this program is executed, it includes one or a combination of the steps of the method embodiment.

[0130] In addition, each functional unit in various embodiments of the present application may be integrated into a processing module, or each unit may exist physically alone, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a magnetic disk, an optical disc, or the like.

[0131] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of various changes or substitutions thereof, and these should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An item recommendation method, characterized in that, Including: When on the target object browsing page, collect the physiological signals of the target object; Perform emotion recognition on the physiological signals to obtain the human emotion; Obtain hot information, and determine the item emotion vector according to the hot information; Input the human emotion and the item emotion vector into an analysis model to obtain a target item, and recommend and display the target item.

2. The article recommendation method according to claim 1, wherein: The performing emotion recognition on the physiological signals to obtain the human emotion includes: Based on different feature calculation rules, calculate the target feature values of the physiological signals respectively; According to each target feature value and the recognition network, determine the emotion recognition result corresponding to each target feature value; Analyze each of the emotion recognition results to determine the human emotion.

3. The article recommendation method according to claim 2, wherein: The calculating the target feature values of the physiological signals respectively based on different feature calculation rules includes at least two of the following: Based on the first feature calculation rule, calculate the Lempel-Ziv complexity of the physiological signals; Based on the second feature calculation rule, calculate the wavelet detail coefficients of the physiological signals; Based on the third feature calculation rule, calculate the cointegration relationship degree of the physiological signals; Based on the fourth feature calculation rule, decompose and extract the average approximate entropy of the physiological signals; Wherein, the target feature values include at least two of Lempel-Ziv complexity, wavelet detail coefficients, cointegration relationship degree, and average approximate entropy.

4. The article recommendation method according to claim 2, wherein: The analyzing each of the emotion recognition results to determine the human emotion includes: Use fuzzy integral to analyze each of the emotion recognition results, determine the first score corresponding to the positive emotion and the second score corresponding to the negative emotion. When the first score is greater than the second score, determine the positive emotion as the human emotion, otherwise determine the negative emotion as the human emotion; Or, According to the emotion recognition results, determine the first quantity of the positive emotion and the second quantity of the negative emotion. When the first quantity is greater than the second quantity, determine the positive emotion as the human emotion, otherwise determine the negative emotion as the human emotion.

5. The article recommendation method according to claim 1, wherein: The determining the item emotion vector according to the hot information includes: Determine the item corresponding to the hot information; Perform vector encoding on the item to obtain an item feature vector containing several item features, and calculate the correlation between the hot information and each item feature; Input the target item feature with the highest correlation into a classification network to obtain an item emotion vector segmented based on Arousal and Valence.

6. The article recommendation method according to any one of claims 1-5, characterized in that: The method further includes: When the target object browses the target item, collect new physiological signals and determine the new human emotion according to the new physiological signals, or receive the evaluation score of the target item and determine the new human emotion according to the evaluation score; When the new human emotion is a positive emotion, continuously recommend the target item, otherwise, input the negative emotion and the item emotion vector into the analysis model to determine a new target item and recommend and display it.

7. The article recommendation method according to any one of claims 1-5, characterized in that: The analysis model is obtained by the following method: Obtain human emotion training data and item emotion vector training data; Train the initialization model in the blockchain using the character emotion training data and the item emotion vector training data to obtain an analysis model through federated learning.

8. An item recommendation device, characterized in that, It includes: A collection module for collecting the physiological signals of the target object when the target object browses a page; An identification module for performing emotion recognition on the physiological signals to obtain character emotions; A determination module for obtaining hot information and determining an item emotion vector according to the hot information; A recommendation module for inputting the character emotion and the item emotion vector into the analysis model to obtain a target item and recommending and displaying the target item.

9. An electronic device, characterized in that, It includes: A processor and a memory, wherein instructions are stored in the memory, and the instructions are loaded and executed by the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, in which a computer program is stored, and when the computer program is executed, the method according to any one of claims 1-7 is implemented.

Images