Dynamic recommendation method and device for media content, equipment and medium

Abstract

The invention provides a dynamic recommendation method for media content, which is applied to the financial field and comprises the following steps: responding to a request of a target user for acquiring a to-be-recommended media text, and acquiring target voice information of the target user; extracting target voiceprint information of the target voice information, determining behavior information and attribute information of the target user, and constructing an initial user portrait according to the behavior information and the attribute information; inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain a real-time emotion, and constructing a real-time user portrait according to the real-time emotion and the initial user portrait; identifying a keyword of the target voice information, and crawling a preset number of media texts from a preset page by using the keyword to obtain a media text set; and selecting a media text from the media text set according to the real-time user portrait for recommendation.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and finance, and to a method, device, equipment and medium for dynamically recommending media content. Background Art

[0002] With the progress of society, media content is becoming more and more diverse, such as financial news, investment and financial management guides, economic analysis and commentary, and financial technology reports;

[0003] In the prior art, users can wake up mobile phones, smart speakers and other devices by inputting voice information to recommend media content. Currently, the device recommends media content by recognizing the "voice information" input by the user, and can only randomly select a text with high popularity for recommendation and playback. It is unable to recommend media content that the user is interested in based on the user's real-time portrait, resulting in inaccurate media content recommendation. For example, when a user inputs "play today's financial news", the smart speaker or mobile phone App and other devices randomly recommends one of the top ten most popular financial news for playback, but the user is not interested in the played financial news, and it is unable to recommend media content that the user is interested in based on the user's real-time portrait, resulting in inaccurate media content recommendation;

[0004] Therefore, there is an urgent need for a dynamic recommendation method for media content that can recommend accurate media content for playback based on the user's real-time portrait. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for dynamically recommending media content, the main purpose of which is to solve the technical problem in the prior art that media content of interest to a user cannot be recommended based on the user's real-time portrait, resulting in inaccurate media content recommendation.

[0006] To achieve the above object, the present invention provides a method for dynamically recommending media content, the method comprising:

[0007] In response to a request from a target user for obtaining a media text to be recommended, obtaining target voice information of the target user;

[0008] Extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determining behavior information and attribute information of the target user, and constructing an initial user portrait based on the behavior information and attribute information;

[0009] Inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile based on the real-time emotions and the initial user profile;

[0010] Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set;

[0011] Media texts are selected from the media text set for recommendation according to the real-time user portrait.

[0012] Optionally, the acquiring target voice information of the target user includes:

[0013] Acquiring initial voice information of the target user;

[0014] Using a preset noise reduction algorithm to perform noise filtering on the initial voice information to obtain noise-reduced voice information;

[0015] The noise reduction voice information is pre-emphasized to obtain the target voice information.

[0016] Optionally, the extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, and determining the behavior information and attribute information of the target user includes:

[0017] Performing voiceprint recognition on the target voice information to extract the target voiceprint features;

[0018] Performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result;

[0019] The attribute information and behavior information of the target user are obtained according to the matching result.

[0020] Optionally, performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result includes:

[0021] Calculate the similarity value between the target voiceprint information and each voiceprint information in the voiceprint database using a similarity function to obtain a similarity value set;

[0022] Sorting all similarity values ​​in the similarity value set to obtain a similarity value sorting result;

[0023] The voiceprint information with the highest similarity value is selected as the matching result.

[0024] Optionally, before inputting the target speech information into a pre-trained emotion recognition model, the method further comprises:

[0025] Constructing an initial emotion recognition model, obtaining a preset number of voice sample information, and preprocessing the voice sample information to obtain standard voice sample information;

[0026] Extracting sound features of the standard voice sample information, annotating the sound features, and obtaining an emotion training set;

[0027] The initial emotion recognition model is trained using the emotion training set to obtain the emotion recognition model.

[0028] Optionally, the initial emotion model is convolved using the emotion training set to obtain an emotion feature set;

[0029] The emotion feature set is calculated using a preset activation function to obtain an emotion prediction value, a label value of an emotion label corresponding to each emotion information in the emotion training set is obtained, and a loss value is obtained by calculating using a pre-constructed loss function according to the prediction value and the label value;

[0030] The iterative action is repeated until the loss value is less than a preset threshold, and then the training is stopped to obtain an interest recognition model.

[0031] Optionally, crawling a preset number of media texts from a preset page using the keyword to obtain a media text set includes:

[0032] Creating a URL queue, wherein the URL queue includes a plurality of URLs;

[0033] Read all URLs in the URL queue in sequence and resolve them into IP addresses;

[0034] The web page data specified by the IP address is downloaded based on the HTTP communication protocol, and the web page data is analyzed to obtain a media text set.

[0035] In order to solve the above problems, the present invention provides a dynamic recommendation device for media content, the device comprising:

[0036] A response module, used to respond to a target user's request for obtaining a media text to be recommended, and obtain target voice information of the target user;

[0037] A construction module is used to extract the target voiceprint information of the target voice information, perform similarity matching between the target voiceprint information and each voiceprint information in the voiceprint database, determine the behavior information and attribute information of the target user, and construct an initial user portrait according to the behavior information and attribute information;

[0038] An analysis module, used for inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile according to the real-time emotions and the initial user profile;

[0039] A crawling module, used for identifying keywords of the target voice information, crawling a preset number of media texts from a preset page using the keywords, and obtaining a media text set;

[0040] The recommendation module is used to select media texts from the media text set for recommendation according to the real-time user portrait.

[0041] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0042] at least one processor; and,

[0043] a memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the dynamic recommendation method for media content as described in any one of claims 1 to 7.

[0045] In order to solve the above problems, the present invention further provides a computer-readable storage medium, which, when executed by a processor, implements the method for dynamically recommending media content as described in any one of claims 1 to 7.

[0046] In the embodiment of the present invention, by performing a preprocessing operation on the initial voice information, the target human voice part can be highlighted, so that the target voice information extracted subsequently is clearer; by identifying the voiceprint information of the target user, the attribute information and behavior information of the user can be quickly obtained, and the attribute information and behavior information of the user can be accurately obtained by using the uniqueness of the voiceprint information, and the initial portrait of the user can be accurately constructed based on the attribute information and behavior information of the user, so as to improve the accuracy of the subsequent acquisition of the media text to be recommended; by identifying the real-time emotion of the user, the real-time emotion is combined with the initial user portrait to construct the real-time user portrait of the user, and the influence of the real-time emotion of the user is taken into account, so as to improve the accuracy of the media text recommendation; by identifying the keywords of the target voice, the media text set to be recommended can be accurately obtained from the preset web page, so as to reduce the workload of manual retrieval and screening, and improve the recommendation efficiency, and by using keywords to obtain media text from the web page, more comprehensive and diversified media text can be obtained, so as to avoid the one-sidedness and limitations of information, and improve the accuracy of information collection; media text is selected from the media text set for recommendation according to the real-time user portrait of the user, so as to improve the accuracy of the media text recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flow chart of a method for dynamically recommending media content provided by the present invention;

[0048] Figure 2A schematic diagram of a module provided by the present invention;

[0049] Figure 3 a schematic diagram of an electronic device provided for the invention;

[0050] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0052] The embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology and the financial field. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0053] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biological retrieval technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0054] The embodiment of the present invention provides a method for dynamic recommendation of media content, and the execution subject of the method for dynamic recommendation of media content includes but is not limited to at least one of the electronic devices such as a client and a mobile terminal that can be configured to execute the method provided by the embodiment of the present application. The APP can be a blockchain platform. The mobile terminal includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0055] In the embodiment of the present invention, the dynamic recommendation method of media content obtains the target voice information of the target user by responding to the request of the target user to obtain the media text to be recommended;

[0056] Extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determining behavior information and attribute information of the target user, and constructing an initial user portrait based on the behavior information and attribute information;

[0057] Inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile based on the real-time emotions and the initial user profile;

[0058] Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set;

[0059] Media texts are selected from the media text set for recommendation according to the real-time user portrait.

[0060] The following combination Figure 1 The implementation principle of the present invention is further explained.

[0061] Reference Figure 1 The flowchart shown describes the dynamic recommendation method of media content provided by the present invention, including:

[0062] S1. Responding to a target user's request for obtaining a media text to be recommended, obtaining target voice information of the target user.

[0063] In this embodiment, the execution subject can be an electronic device with data processing function, such as a mobile phone, tablet computer, laptop computer, desktop computer, etc., which responds to the target user's request to obtain the media text to be played, and collects the initial voice information of the target user through the voice module of the electronic device, such as a microphone, for example, the user says "Give me an article of today's financial news" on the microphone or other sound collecting device, thereby obtaining the initial voice information of the target user, pre-processing the initial voice information to obtain the target voice information, and then using the target voice information to subsequently obtain the media text to be recommended and play it.

[0064] In this embodiment, in response to a target user's request for obtaining a media text to be recommended, the request includes initial voice information output by the target user for the media text to be recommended, the initial voice information is a recording file of the target user or the sound of the user speaking in real time, and further, due to the influence of the recording device (such as a microphone) and the recording environment, the initial voice information contains some noise, such as the voice of a non-target user, environmental noise, etc. In order not to affect the subsequent extraction of human voice information in the initial voice information, the embodiment of the present invention pre-processes the initial voice information to obtain the target voice information;

[0065] For example, in the financial field, if a user wants to listen to today's financial news, he or she can say "Give me today's financial news" to the collection device (mobile phone microphone). The background will pre-process the user's input "Give me today's financial news" to remove environmental noise, etc., and obtain a clean "Give me today's financial news".

[0066] In one embodiment, the acquiring the target voice information of the target user includes:

[0067] Acquiring initial voice information of the target user;

[0068] Using a preset noise reduction algorithm to perform noise filtering on the initial voice information to obtain noise-reduced voice information;

[0069] The noise reduction voice information is pre-emphasized to obtain the target voice information.

[0070] Specifically, in this embodiment, the preset noise reduction algorithm is an LMS algorithm, and the initial voice information is denoised by the LMS algorithm to obtain the noise-reduced voice information. In order to ensure the accuracy of the subsequent acquisition of the target voice information, the target voice information in the noise-reduced voice information is highlighted, and the noise-reduced voice information is pre-emphasized to enhance the human voice part in the noise-reduced voice information to obtain the target voice information; the pre-emphasis operation can be performed by a function, and the function formula includes:

[0071] y(t)=x(t)-μx(t-1)

[0072] Wherein, the y(t) is the target speech information, x(t) is the noise reduction speech information, and μ is the adjustment value of the pre-emphasis operation.

[0073] In this embodiment, the value range of μ is [0.9, 1.0].

[0074] In this embodiment, by performing a preprocessing operation on the initial voice information, the target vocal part can be highlighted, so that the target voice information extracted subsequently is clearer.

[0075] S2. Extract the target voiceprint information of the target voice information, perform similarity matching between the target voiceprint information and each voiceprint information in the voiceprint database, determine the behavior information and attribute information of the target user, and construct an initial user portrait based on the behavior information and attribute information.

[0076] In this embodiment, before extracting the voiceprint information of the target voice information and performing similarity matching on the voiceprint information, a voiceprint database is constructed, and the voiceprint database includes the voiceprint information left by the user when registering, and each voiceprint information corresponds to the user's attribute information and behavior information. The attribute information includes the user's gender, age, education level, etc., and the behavior information includes the user's historical browsing media text records, etc. The target voiceprint information is matched with each voiceprint information in the voiceprint database for similarity, the behavior information and attribute information of the target user are confirmed, and the initial user portrait of the user is constructed based on the behavior information and the attribute information.

[0077] For example, in the financial field, the user's input "Give me an article of today's financial news" is collected and pre-processed, and then voiceprint recognition is performed on "Give me an article of today's financial news", the target voiceprint information is extracted, and similarity matching is performed on each voiceprint information in the voiceprint database to confirm the user's attribute information and behavior information at the time of registration, and build an initial user portrait based on the information left by the user when registering.

[0078] In one embodiment, extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, and determining the behavior information and attribute information of the target user include:

[0079] Performing voiceprint recognition on the target voice information to extract the target voiceprint features;

[0080] Performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result;

[0081] The attribute information and behavior information of the target user are obtained according to the matching result.

[0082] Specifically, the target voice information is identified by using a preset voiceprint recognition technology, and the voiceprint features of the target voice information are extracted to obtain the target voiceprint information, which includes a voiceprint feature vector. The target voiceprint information is matched with all the voiceprint information in a voiceprint database one by one for similarity, and the voiceprint information with the highest similarity value is selected. The attribute information and behavior information of the target user are determined through the information pre-stored in the voiceprint database.

[0083] For example, in the financial field, users will leave voiceprint information when registering on an APP or a smart speaker, including voiceprint features and attribute information corresponding to the voiceprint features (such as gender: male, age: 35, education: undergraduate, etc.), and the voiceprint information is stored in a voiceprint database. When the user wakes up the smart speaker by inputting the target voice information "Give me an article of today's financial news", the voiceprint features of "Give me an article of today's financial news" are extracted, and the voiceprint features are compared with the voiceprint features stored in the database for similarity, and the voiceprint features with the highest matching value in the database are found, and the user's attribute information and behavior information are determined based on the voiceprint features with the highest matching value.

[0084] In one embodiment, performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result includes:

[0085] Calculate the similarity value between the target voiceprint information and each voiceprint information in the voiceprint database using a similarity function to obtain a similarity value set;

[0086] Sorting all similarity values ​​in the similarity value set to obtain a similarity value sorting result;

[0087] The voiceprint information with the highest similarity value is selected as the matching result.

[0088] In another embodiment, when all similarity values ​​do not exceed a preset threshold, the user's voiceprint information is re-registered in the voiceprint database.

[0089] The similarity calculation function is:

[0090]

[0091] Wherein, x represents the target voiceprint information, y i represents the i-th voiceprint information in the voiceprint database, sim(x,y i ) represents the similarity value between the target voiceprint information and the i-th voiceprint information in the voiceprint database, and n represents that there are n voiceprint information in the voiceprint database.

[0092] In this embodiment, after the attribute information and the behavior information of the target user are extracted, an initial user portrait is constructed according to the attribute information and the behavior information.

[0093] Specifically, the attribute information and behavior information are preprocessed, including processing missing values ​​and outliers, to obtain standard information, extract features of the standard information (such as age, gender, keywords of media text, topics, etc.), integrate the extracted features, and construct an initial user portrait.

[0094] In another embodiment, the attribute information and the behavior information are updated periodically.

[0095] In this embodiment, by identifying the voiceprint information of the target user, the user's attribute information and behavior information can be quickly obtained, and the user's attribute information and behavior information can be accurately obtained by utilizing the uniqueness of the voiceprint information. The user's initial portrait can be accurately constructed based on the user's attribute information and behavior information, thereby improving the accuracy of subsequent acquisition of the media text to be recommended.

[0096] S3. Input the target voice information into a trained emotion recognition model, perform emotion spectrum analysis on the target voice information to obtain real-time emotions, and construct a real-time user portrait based on the real-time emotions and the initial user portrait.

[0097] In this embodiment, when the user needs to obtain the media text to be recommended, the user's emotion is an important feature, and the spectral characteristics of the speech have obvious differences in different emotions. Therefore, by performing emotional spectrum analysis on the target speech information, the user's real-time emotion can be accurately obtained.

[0098] In this embodiment, the target voice information is input into a trained emotion recognition model, the spectral features of the target voice information are extracted, the spectral features are analyzed to obtain real-time emotions, and a real-time user profile is constructed based on the real-time emotions and the initial user profile.

[0099] In one embodiment, before inputting the target speech information into a pre-trained emotion recognition model, the method further comprises:

[0100] Constructing an initial emotion recognition model, obtaining a preset number of voice sample information, and preprocessing the voice sample information to obtain standard voice sample information;

[0101] In this embodiment, the emotion recognition model is a convolutional neural network, and a preset amount of voice sample information is obtained. The voice sample information can be obtained from a publicly available voice database or recorded on-site by a voice acquisition device. The voice sample information is preprocessed, and the preprocessing includes noise reduction and enhancement, pre-emphasis, short-time analysis, and windowing and framing of the voice sample information. The voice sample information is preprocessed by one or more of the above preprocessing methods to obtain standard voice sample information.

[0102] Extracting sound features of the standard voice sample information, annotating the sound features, and obtaining an emotion training set;

[0103] In this embodiment, sound feature extraction is performed for each standard speech sample, and the sound feature includes an audio spectrum. The sound feature is labeled to include emotional states such as happy, frustrated, angry, and calm, to obtain an emotional training set.

[0104] The initial emotion recognition model is trained using the emotion training set to obtain the emotion recognition model.

[0105] In another embodiment, the using the emotion training set to train the initial emotion recognition model to obtain the emotion recognition model includes:

[0106] Using the emotion training set to perform a convolution operation on the initial emotion model to obtain an emotion feature set;

[0107] The emotion feature set is calculated using a preset activation function to obtain an emotion prediction value, a label value of an emotion label corresponding to each emotion information in the emotion training set is obtained, and a loss value is obtained by calculating using a pre-constructed loss function according to the prediction value and the label value;

[0108] The iterative action is repeated until the loss value is less than a preset threshold, and then the training is stopped to obtain an interest recognition model.

[0109] In this embodiment, the convolution operation is:

[0110]

[0111] Wherein, ω' represents the number of channels of the emotion feature set, ω represents the number of channels of the emotion training set, k is the size of the convolution kernel, f is the stride of the convolution operation, and p is the preset data zero-padding matrix.

[0112] In this embodiment, the loss function includes:

[0113]

[0114] Among them, L ce represents the loss value, N is the number of data in the emotion training set, i is a positive integer, y i is the label value, p i is the predicted value.

[0115] In this embodiment, the real-time user portrait is constructed by combining the initial user portrait and the real-time emotion.

[0116] For example, in the financial field, emotion recognition is performed on the user input "Give me an article of today's financial news", the target speech is input into the trained emotion recognition model, the spectrum of the target speech is extracted, and the user's real-time emotion is obtained by analyzing the spectrum. Based on the user's real-time emotion and the initial user portrait, a real-time user portrait of the user is constructed.

[0117] In this embodiment, the real-time emotion of the user is identified and combined with the initial user portrait to construct a real-time user portrait of the user. The real-time emotional impact of the user's emotions is taken into consideration, thereby improving the accuracy of media text recommendations.

[0118] S4. Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set.

[0119] In this embodiment, in order to obtain the text information of the target voice information and facilitate the subsequent extraction of keywords, the keywords are used to crawl a preset web page to obtain a preset number of media texts. In this embodiment of the present invention, the target voice information needs to be converted into text. Therefore, text recognition processing is performed on the target voice information to obtain the target text data, extract the keywords of the target text data, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set.

[0120] In detail, the target voice information is converted into target text information through a preset text conversion technology, and the keywords of the target text information are extracted. The preset text conversion technology includes but is not limited to any one of a hidden Markov model (HMM), a deep neural network (DNN), and an end-to-end model. The keywords extracted from the target text information can use any one of TF-IDF (Term Frequency-Inverse Document Frequency) and LDA (Latent Dirichlet Allocation).

[0121] For example, in the financial field, the keyword "give me today's financial news" input by the user is identified, such as the keyword: financial news; all media texts about "financial news" are crawled on the specified web page using the URL.

[0122] In one embodiment, crawling a preset number of media texts from a preset page using the keyword to obtain a media text set includes:

[0123] Creating a URL queue, wherein the URL queue includes a plurality of URLs;

[0124] Read all URLs in the URL queue in sequence and resolve them into IP addresses;

[0125] The web page data specified by the IP address is downloaded based on the HTTP communication protocol, and the web page data is analyzed to obtain a media text set.

[0126] In this embodiment, the URL is called a uniform resource locator, which is a concise representation of the location and access methods of various resources in the news forum website, and is also called the address of various resources in the news forum website. The URL consists of a protocol, a host name, a port, a path, a query string, a hash element, etc. The protocol represents the protocol for accessing resources and services, such as http, ftp, mailto, and file; the host name represents the fully qualified domain name of the host where the resource is located, such as www.01cj.com (Financial News Network). The port (port) represents the TCP port number used by the protocol, wherein the commonly used port of the HTTP communication protocol is 80, which is generally omitted by default; the path (path) represents the directory / file path name of the resource; the query string represents the query string passed in the URL; the hash symbol element represents the file offset specified by the URL, including a hash symbol (#) plus a position related to the file offset, and the parsing into an IP address is to extract the protocol (protocol), host name (hostname), port (port), path (path), etc. to obtain the IP address.

[0127] After the IP address is obtained through parsing, the web page data specified by the IP address is downloaded based on the HTTP communication protocol, and the web page data is analyzed to obtain a media text set.

[0128] In this embodiment, by identifying the keywords of the target voice, the set of media texts to be recommended can be accurately obtained from the preset web page, which reduces the workload of manual retrieval and screening and improves the recommendation efficiency. By using keywords to obtain media texts from web pages, more comprehensive and diverse media texts can be obtained, avoiding the one-sidedness and limitations of information and improving the accuracy of information collection.

[0129] S5. Select media texts from the media text set for recommendation according to the real-time user portrait.

[0130] In this embodiment, the media text set is classified, each emotional feature corresponds to a type of media text, and multiple category media text sets are obtained. According to the real-time user portrait, any one media text in the corresponding category media text set is selected from the multiple category media text sets for recommendation.

[0131] For example, in the financial field, all media texts about "financial news" are classified into categories, such as a media text set corresponding to happiness, a media text set corresponding to anger, a media text set corresponding to calmness, and a media text set corresponding to frustration. According to the user's real-time user portrait, a media text from several media text sets is selected for recommendation.

[0132] In the embodiment of the present invention, by performing a preprocessing operation on the initial voice information, the target human voice part can be highlighted, so that the target voice information extracted subsequently is clearer; by identifying the voiceprint information of the target user, the attribute information and behavior information of the user can be quickly obtained, and the attribute information and behavior information of the user can be accurately obtained by using the uniqueness of the voiceprint information, and the initial portrait of the user can be accurately constructed based on the attribute information and behavior information of the user, so as to improve the accuracy of the subsequent acquisition of the media text to be recommended; by identifying the real-time emotion of the user, the real-time emotion is combined with the initial user portrait to construct the real-time user portrait of the user, and the influence of the real-time emotion of the user is taken into account, so as to improve the accuracy of the media text recommendation; by identifying the keywords of the target voice, the media text set to be recommended can be accurately obtained from the preset web page, so as to reduce the workload of manual retrieval and screening, and improve the recommendation efficiency, and by using keywords to obtain media text from the web page, more comprehensive and diversified media text can be obtained, so as to avoid the one-sidedness and limitations of information, and improve the accuracy of information collection; media text is selected from the media text set for recommendation according to the real-time user portrait of the user, so as to improve the accuracy of the media text recommendation.

[0133] Reference Figure 2 FIG. 1 is a schematic diagram of functional modules of a dynamic media content recommendation device 100 according to the present invention.

[0134] A response module 110 is used to respond to a target user's request for obtaining a media text to be recommended, and obtain target voice information of the target user;

[0135] In one embodiment, the acquiring the target voice information of the target user includes:

[0136] Acquiring initial voice information of the target user;

[0137] Using a preset noise reduction algorithm to perform noise filtering on the initial voice information to obtain noise-reduced voice information;

[0138] The noise reduction voice information is pre-emphasized to obtain the target voice information.

[0139] A construction module 120 is used to extract target voiceprint information of the target voice information, perform similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determine the behavior information and attribute information of the target user, and construct an initial user portrait according to the behavior information and attribute information;

[0140] In one embodiment, extracting the target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, and determining the behavior information and attribute information of the target user include:

[0141] Performing voiceprint recognition on the target voice information to extract the target voiceprint features;

[0142] Performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result;

[0143] The attribute information and behavior information of the target user are obtained according to the matching result.

[0144] In one embodiment, performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result includes:

[0145] Calculate the similarity value between the target voiceprint information and each voiceprint information in the voiceprint database using a similarity function to obtain a similarity value set;

[0146] Sorting all similarity values ​​in the similarity value set to obtain a similarity value sorting result;

[0147] The voiceprint information with the highest similarity value is selected as the matching result.

[0148] The analysis module 130 is used to input the target voice information into the trained emotion recognition model, perform emotion spectrum analysis on the target voice information to obtain real-time emotions, and construct a real-time user profile according to the real-time emotions and the initial user profile;

[0149] In one embodiment, before inputting the target speech information into a pre-trained emotion recognition model, the method further comprises:

[0150] Constructing an initial emotion recognition model, obtaining a preset number of voice sample information, and preprocessing the voice sample information to obtain standard voice sample information;

[0151] Extracting sound features of the standard voice sample information, annotating the sound features, and obtaining an emotion training set;

[0152] The initial emotion recognition model is trained using the emotion training set to obtain the emotion recognition model.

[0153] In one embodiment, the emotion training set is used to perform a convolution operation on the initial emotion model to obtain an emotion feature set;

[0154] The emotion feature set is calculated using a preset activation function to obtain an emotion prediction value, a label value of an emotion label corresponding to each emotion information in the emotion training set is obtained, and a loss value is obtained by calculating using a pre-constructed loss function according to the prediction value and the label value;

[0155] The iterative action is repeated until the loss value is less than a preset threshold, and then the training is stopped to obtain an interest recognition model.

[0156] A crawling module 140, configured to identify keywords of the target voice information, and crawl a preset number of media texts from a preset page using the keywords to obtain a media text set;

[0157] In one embodiment, crawling a preset number of media texts from a preset page using the keyword to obtain a media text set includes:

[0158] Creating a URL queue, wherein the URL queue includes a plurality of URLs;

[0159] Read all URLs in the URL queue in sequence and resolve them into IP addresses;

[0160] The web page data specified by the IP address is downloaded based on the HTTP communication protocol, and the web page data is analyzed to obtain a media text set.

[0161] The recommendation module 150 is used to select media texts from the media text set for recommendation according to the real-time user portrait.

[0162] Reference Figure 3 FIG. 1 is a schematic diagram of a preferred embodiment of an electronic device 1 of the present invention.

[0163] The electronic device 1 includes but is not limited to: a memory 11, a processor 12, a display 13 and a network interface 14. The electronic device 1 is connected to a network through the network interface 14 to obtain raw data. The network may be a wireless or wired network such as an intranet, the Internet, a global system of mobile communication (GSM), a wideband code division multiple access (WCDMA), a 4G network, a 5G network, Bluetooth, Wi-Fi, a call network, etc.

[0164] Among them, the memory 11 includes at least one type of readable medium, and the readable medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a hard disk or memory of the electronic device 1. In other embodiments, the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped with the electronic device 1. Of course, the memory 11 can also include both the internal storage unit of the electronic device 1 and its external storage device. In this embodiment, the memory 11 is generally used to store the operating system and various application software installed on the electronic device 1, such as the program code of the dynamic recommendation program 10, etc. In addition, the memory 11 can also be used to temporarily store various types of data that have been output or are to be output.

[0165] In some embodiments, the processor 12 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 12 is generally used to control the overall operation of the electronic device 1, such as performing control and processing related to data interaction or communication. In this embodiment, the processor 12 is used to run the program code stored in the memory 11 or process data, such as running the program code of the dynamic recommendation program 10.

[0166] The display 13 may be referred to as a display screen or a display unit. In some embodiments, the display 13 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an organic light-emitting diode (OLED) touch device. The display 13 is used to display information processed in the electronic device 1 and to display a visual working interface, such as displaying the results of data statistics.

[0167] The network interface 14 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), and the network interface 14 is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0168] Figure 3 Only the electronic device 1 having the components 11 - 14 and the dynamic recommendation program 10 is shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0169] Optionally, the electronic device 1 may further include a user interface, which may include a display (Display), an input unit such as a keyboard (Keyboard), and the optional user interface may also include a standard wired interface and a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an organic light-emitting diode (Organic Light-Emitting Diode, OLED) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0170] The electronic device 1 may further include a radio frequency (RF) circuit, a sensor, an audio circuit, etc., which will not be described in detail here.

[0171] In the above embodiment, the processor 12 may implement the following steps when executing the dynamic recommendation program 10 stored in the memory 11:

[0172] In response to a request from a target user for obtaining a media text to be recommended, obtaining target voice information of the target user;

[0173] Extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determining behavior information and attribute information of the target user, and constructing an initial user portrait based on the behavior information and attribute information;

[0174] Inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile based on the real-time emotions and the initial user profile;

[0175] Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set;

[0176] Media texts are selected from the media text set for recommendation according to the real-time user portrait.

[0177] For a detailed description of the above steps, please refer to Figure 1 An illustration of a flowchart of an embodiment of a method for dynamically recommending storybooks.

[0178] In addition, an embodiment of the present invention further proposes a computer-readable medium, which may be non-volatile or volatile. The computer-readable medium may be any one or any combination of a hard disk, a multimedia card, an SD card, a flash memory card, an SMC, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, etc. The computer-readable medium includes a data storage area and a program storage area, the data storage area stores data created according to the use of the blockchain node, and the program storage area stores a dynamic recommendation program 10, and the dynamic recommendation program 10 implements the following operations when executed by the processor:

[0179] In response to a request from a target user for obtaining a media text to be recommended, obtaining target voice information of the target user;

[0180] Extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determining behavior information and attribute information of the target user, and constructing an initial user portrait based on the behavior information and attribute information;

[0181] Inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile based on the real-time emotions and the initial user profile;

[0182] Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set;

[0183] Media texts are selected from the media text set for recommendation according to the real-time user portrait.

[0184] The specific implementation of the computer-readable medium of the present invention is substantially the same as the specific implementation of the above-mentioned method for dynamically recommending media content, and will not be described in detail herein.

[0185] In another embodiment, the method for dynamically adjusting the task execution time interval provided by the present invention can further ensure the privacy and security of all the above data, and all the above data can also be stored in a blockchain node. For example, feature dimensions and feature embeddings can all be stored in blockchain nodes.

[0186] It should be noted that the blockchain referred to in the present invention is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, platform product service layer, and application service layer.

[0187] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0188] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, an electronic device, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0189] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection of the present invention.

Claims

1. A method for dynamic recommendation of media content, characterized in that: The method comprises: In response to a request from a target user for obtaining a media text to be recommended, obtaining target voice information of the target user; Extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, determining behavior information and attribute information of the target user, and constructing an initial user portrait based on the behavior information and attribute information; Inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile based on the real-time emotions and the initial user profile; Identify keywords of the target voice information, and use the keywords to crawl a preset number of media texts from a preset page to obtain a media text set; Media texts are selected from the media text set for recommendation according to the real-time user portrait.

2. The method for dynamically recommending media content according to claim 1, characterized in that: The acquiring the target voice information of the target user includes: Acquiring initial voice information of the target user; Using a preset noise reduction algorithm to perform noise filtering on the initial voice information to obtain noise-reduced voice information; The noise reduction voice information is pre-emphasized to obtain the target voice information.

3. The method for dynamically recommending media content according to claim 1, characterized in that: The step of extracting target voiceprint information of the target voice information, performing similarity matching between the target voiceprint information and each voiceprint information in a voiceprint database, and determining behavior information and attribute information of the target user includes: Performing voiceprint recognition on the target voice information to extract the target voiceprint features; Performing similarity matching between the target voiceprint feature and each voiceprint feature in the voiceprint database to obtain a matching result; The attribute information and behavior information of the target user are obtained according to the matching result.

4. The method for dynamically recommending media content according to claim 3, characterized in that: The similarity matching of the target voiceprint feature with each voiceprint feature in the voiceprint database to obtain a matching result includes: Calculate the similarity value between the target voiceprint information and each voiceprint information in the voiceprint database using a similarity function to obtain a similarity value set; Sorting all similarity values ​​in the similarity value set to obtain a similarity value sorting result; The voiceprint information with the highest similarity value is selected as the matching result.

5. The method for dynamically recommending media content according to claim 1, characterized in that: Before inputting the target speech information into the pre-trained emotion recognition model, the method further includes: Constructing an initial emotion recognition model, obtaining a preset number of voice sample information, and preprocessing the voice sample information to obtain standard voice sample information; Extracting sound features of the standard voice sample information, annotating the sound features, and obtaining an emotion training set; The initial emotion recognition model is trained using the emotion training set to obtain the emotion recognition model.

6. The method for dynamic recommendation of media content according to claim 5, characterized in that: The method comprises: performing a convolution operation on the initial emotion model using the emotion training set to obtain an emotion feature set; The emotion feature set is calculated using a preset activation function to obtain an emotion prediction value, a label value of an emotion label corresponding to each emotion information in the emotion training set is obtained, and a loss value is obtained by calculating using a pre-constructed loss function according to the prediction value and the label value; The iterative action is repeated until the loss value is less than a preset threshold, and then the training is stopped to obtain an interest recognition model.

7. The method for dynamically recommending media content according to claim 1, characterized in that: The method of crawling a preset number of media texts from a preset page using the keyword to obtain a media text set includes: Creating a URL queue, wherein the URL queue includes a plurality of URLs; Read all URLs in the URL queue in sequence and resolve them into IP addresses; The web page data specified by the IP address is downloaded based on the HTTP communication protocol, and the web page data is analyzed to obtain a media text set.

8. A dynamic recommendation device for media content, characterized in that: The device comprises: A response module, used to respond to a target user's request for obtaining a media text to be recommended, and obtain target voice information of the target user; A construction module is used to extract the target voiceprint information of the target voice information, perform similarity matching between the target voiceprint information and each voiceprint information in the voiceprint database, determine the behavior information and attribute information of the target user, and construct an initial user portrait according to the behavior information and attribute information; An analysis module, used for inputting the target voice information into a trained emotion recognition model, performing emotion spectrum analysis on the target voice information to obtain real-time emotions, and constructing a real-time user profile according to the real-time emotions and the initial user profile; A crawling module, used for identifying keywords of the target voice information, crawling a preset number of media texts from a preset page using the keywords, and obtaining a media text set; The recommendation module is used to select media texts from the media text set for recommendation according to the real-time user portrait.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the dynamic recommendation method for media text as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for dynamically recommending media texts as described in any one of claims 1 to 7 is implemented.

Images