Communication product processing method, user emotion processing method and device

Abstract

This specification provides a communication product processing method, a user emotion processing method and a device. The method includes: determining the avatar image sequence used historically by batch users and the ordering information sequence of historically ordered communication products; quantifying the emotion value reflected by the avatar image in the avatar image sequence to obtain the avatar emotion value sequence of each user, and the avatar emotion value sequence is used to describe the change of the user's emotion value within a preset historical time period; based on the avatar image sequence, the avatar emotion value sequence and the ordering information sequence, generating the user abstract features of each user; using the user abstract features and the communication product abstract features of each user as input features, training the communication product processing model, and obtaining the preference relationship between the user and the communication product, and the communication product abstract features are the abstract features of the product information of the full amount of communication products. In this way, the visual information of the user's social network avatar can be fully mined, so that the characterization of consumer preferences under the recommendation system is more comprehensive.

Description

Technical Field

[0001] This document relates to the field of communication technology, and in particular to a communication product processing method, a user emotion processing method and a device. Background Art

[0002] As the demands of subscribers of communication products gradually become more diversified, communication companies have also launched a variety of different communication products to meet the subscription needs of different consumers. However, too many types of communication products make it difficult for consumers to effectively select communication products that meet their expectations. Therefore, recommendation systems came into being.

[0003] The existing communication product recommendation solutions can be divided into the following two types:

[0004] Conventional recommendation scheme: Conventional recommendation is based on the order volume of communication products, and the top K communication products with the highest order volume are recommended to users in a unified manner. Under conventional recommendation, the products recommended to users will not be recommended differently due to different users.

[0005] Personalized recommendation scheme: Under the personalized recommendation scheme, the service provider collects user characteristics (age, occupation, etc.), user ordering behavior data (ordering history, search history, etc.) and communication product information (product-related attributes, etc.), and recommends to users communication products ordered by similar users or communication products similar to products they have ordered by calculating the similarity of user-item relationships.

[0006] However, consumer preferences are affected by many factors, with great differentiation and high uncertainty. Existing technologies mainly analyze and recommend based on traditional structured data. Facing complex consumer preferences, these traditional structured data cannot fully reflect consumers' consumption habits.

[0007] Therefore, there is an urgent need to introduce a more effective communication product processing solution. Summary of the invention

[0008] The embodiments of this specification provide a communication product processing method for fully mining the visual information of user social network avatars, so that the description of consumer preferences in the recommendation system is more comprehensive.

[0009] The embodiment of this specification also provides a communication product processing method, including:

[0010] Determine the avatar image sequence used historically by batch users and the ordering information sequence of the communication products ordered historically;

[0011] quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within the preset historical time period;

[0012] generating a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0013] The user abstract features and communication product abstract features of each user are used as input features to train a communication product processing model to obtain the preference relationship between the user and the communication product. The communication product abstract features are abstract features of product information of all communication products.

[0014] The embodiment of this specification also provides a method for processing user emotions, including:

[0015] Obtain the image color and image content of the target user's avatar image;

[0016] Determining a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively;

[0017] The emotion value of the target user is determined based on the color positivity weight and the content positivity weight of the avatar image.

[0018] The embodiment of this specification also provides a communication product processing device, including:

[0019] A determination module, used to determine the avatar image sequences used historically by batch users and the ordering information sequences of the communication products ordered historically;

[0020] A quantification module, used to quantify the emotion values ​​reflected by the avatar images in the avatar image sequence, and obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change of the emotion value of the user within the preset historical time period;

[0021] A processing module, configured to generate a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0022] The training module is used to take the user abstract features and communication product abstract features of each user as input features, train the communication product processing model, and obtain the preference relationship between the user and the communication product. The communication product abstract features are the abstract features of the product information of all communication products.

[0023] The embodiment of this specification also provides a user emotion processing device, including:

[0024] An acquisition module, used to acquire the image color and image content of the avatar image of the target user;

[0025] A first processing module, configured to determine a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively;

[0026] The second processing module is used to determine the emotion value of the target user based on the color positivity weight and the content positivity weight of the avatar image.

[0027] The embodiment of this specification also provides an electronic device, including: a communication interface, a processor and a memory;

[0028] The processor calls the program instructions in the memory to perform the following actions:

[0029] Determine the avatar image sequence used historically by batch users and the ordering information sequence of the communication products ordered historically;

[0030] quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within the preset historical time period;

[0031] generating a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0032] The user abstract features and communication product abstract features of each user are used as input features to train a communication product processing model to obtain the preference relationship between the user and the communication product. The communication product abstract features are abstract features of product information of all communication products.

[0033] The embodiments of this specification also provide a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including multiple application programs, the one or more programs are used to perform the following actions:

[0034] Determine the avatar image sequence used historically by batch users and the ordering information sequence of the communication products ordered historically;

[0035] quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within the preset historical time period;

[0036] generating a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0037] The user abstract features and communication product abstract features of each user are used as input features to train a communication product processing model to obtain the preference relationship between the user and the communication product. The communication product abstract features are abstract features of product information of all communication products.

[0038] The embodiment of this specification also provides an electronic device, including: a communication interface, a processor and a memory;

[0039] The processor calls the program instructions in the memory to perform the following actions:

[0040] Obtain the image color and image content of the target user's avatar image;

[0041] Determining a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively;

[0042] The emotion value of the target user is determined based on the color positivity weight and the content positivity weight of the avatar image.

[0043] The embodiments of this specification also provide a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including multiple application programs, the one or more programs are used to perform the following actions:

[0044] Obtain the image color and image content of the target user's avatar image;

[0045] Determining a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively;

[0046] The emotion value of the target user is determined based on the color positivity weight and the content positivity weight of the avatar image.

[0047] Any of the above embodiments of the present specification fully mines the visual information of the user's social network avatar from the perspective of image visual features, making the characterization of consumer preferences in the recommendation system more comprehensive and maximizing the utilization of complex unstructured data. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0049] Figure 1 A flowchart of a communication product processing method provided in one embodiment of this specification;

[0050] Figure 2 A flowchart of a communication product processing method provided by another embodiment of this specification;

[0051] Figure 3 A schematic diagram of a communication product processing model provided in an embodiment of this specification;

[0052] Figure 4 A schematic diagram of a long short-term memory layer unit provided in an embodiment of the present specification;

[0053] Figure 5 A flowchart of a method for processing user emotions provided in one embodiment of this specification;

[0054] Figure 6 A schematic diagram of the structure of a communication product processing device provided in one embodiment of this specification;

[0055] Figure 7 A schematic diagram of the structure of a user emotion processing device provided in one embodiment of this specification;

[0056] Figure 8 A schematic diagram of the structure of an electronic device provided in one embodiment of this specification;

[0057] Fig. 9 A schematic diagram of the structure of an electronic device provided in another embodiment of the present specification. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this document.

[0059] In combination with the background technology, criminals disguise themselves as SPs and push a large number of marketing messages or even fraudulent text messages to terminal users through fake base stations, which seriously infringes on the interests of terminal users, and terminal users cannot accurately judge the legitimacy of the source of business messages. Based on this, this specification provides a business message processing method, which combines the signature information preset by the terminal user with the business message before sending it, so as to help users accurately identify the legitimacy of the source of the business message.

[0060] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0061] Figure 1A flowchart of a communication product processing method provided in an embodiment of this specification is shown in FIG. Figure 1 , the method may specifically include the following steps:

[0062] Step 102: Determine the avatar image sequences used historically by batch users and the ordering information sequences of the communication products ordered historically; specifically:

[0063] Acquire the head portrait images and ordering information of communication products used by batch users in a preset historical period; sort the head portrait images of each user in chronological order to obtain the head portrait image sequence of each user; sort the ordering information of each user in chronological order to obtain the ordering information sequence of each user. The preset historical period can be set flexibly.

[0064] Combination Figure 2 It should be noted that the data in step 102 can be obtained by Figure 2 Step 1 in the data collection is to obtain data by arranging them, specifically: three types of relevant information are obtained through the device, including user information, communication product information, and ordering information. User information includes the user number and the historical avatar images used by the user; the communication product information is the communication product number and the relevant information of the communication product (such as the monthly fee, the call duration, traffic, and number of short messages included in the product); the ordering information includes the user number, the communication product number, the interaction relationship between the user and the communication product, and the time when the user ordered the product. By arranging the three types of relevant information, the avatar image sequence and the ordering information sequence of the historically ordered communication products can be obtained.

[0065] Step 104: quantify the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within the preset historical time period; specifically:

[0066] S11, based on the image color and image content of the avatar image, query a pre-established positive and negative emotion scale to obtain the positivity weight corresponding to the image color and image content;

[0067] The positive and negative emotion scale contains the corresponding relationship between color attributes, RGB value ranges and positivity weights, and the corresponding relationship between the type of image content and positivity weights; accordingly, the implementation of S11 may be:

[0068] Find the value range of RGB value corresponding to the color attribute that matches the image color of the avatar image in the positive and negative emotion scale; if the RGB value of the avatar image matches the value range of the found RGB value, use the positivity weight corresponding to the value range of the found RGB value as the color positivity weight of the avatar image; find the positivity weight corresponding to the type of image content in the positive and negative emotion scale, and use it as the content positivity weight of the avatar image.

[0069] S12, determining the emotion value corresponding to the avatar image based on the positivity weights corresponding to the image color and the image content; specifically:

[0070] Determine the color positivity weight and content positivity weight of the same user in each time unit of a preset historical period; average the color positivity weight and content positivity weight in each time unit to obtain the emotion value corresponding to the avatar used by the user in the time unit.

[0071] S13. Generate an avatar emotion value sequence based on the emotion value corresponding to each avatar image, wherein the avatar emotion value sequence is used to describe the change of the emotion value of each time unit of the same user within the preset historical time period.

[0072] See below Figure 2 Step 2: Data preprocessing, S11-S13 are described in detail in combination with Table 1 and Table 2 below:

[0073] First, the image features of user avatars are different from traditional data types. Therefore, this embodiment innovatively constructs a positive and negative emotional scale for comparing avatar emotional standards and pre-processes avatar visual information (content category, color). Specifically:

[0074] (1) Based on the user avatar information, construct the avatar emotion value sequence of user i at time t Where T is the sequence length of the user avatar emotion value sequence.

[0075] ①. From the perspective of image visual features, based on relevant psychological research, a set of positive and negative emotion labeling scales was compiled. The positive and negative emotion labeling scale consists of two sets of image visual feature subscales, namely the image color scale and the image content scale. Specifically:

[0076] Table 1 shows the image color positive and negative emotion scale compiled in this embodiment. The image color positive and negative emotion scale constructs an image color classification rule based on the RGB value of the user's avatar image. The rule divides the image color into ten categories and assigns a corresponding color positivity weight w to each category of image color. cThe image color categories are sorted from high to low in terms of positivity weight as follows: green, yellow, orange, cyan, red, white, blue, purple, black, and gray.

[0077] Table 1 Image color positive and negative emotion scale

[0078] color Positive Weight Positive / Negative Represents mood R G B Other Rules green 0.95 positive Relaxed, happy, comfortable, hopeful >200 R+B<128 yellow 0.93 positive Energetic, happy, excited >200 >200 <20 orange color 0.85 positive Active, lively >230 128<,<200 <20 blue 0.79 positive calm, relax, sky <20 >200 >200 red 0.63 positive Love, romance, evil, blood >200 G+B<128 White 0.57 positive Peace, Innocence, Loneliness, Boredom >220 >220 >220 blue 0.44 negative Melancholy, indifference >200 R+G<128 Purple 0.38 negative Uneasy, solemn >128 >20 >200 black 0.19 negative Disappointment, fear, death, misfortune <80 <80 <80 grey 0.07 negative Melancholy, boredom R=G=B

[0079] Table 2 shows the image content positive and negative emotion scale compiled by this patent. The image content positive and negative emotion scale presets the image content into six categories: animals, cartoons, people, scenery, objects, and others (arranged from high to low in terms of positive weight), and assigns a corresponding content positive weight w to each category of image content. s .

[0080] Table 2 Image content positive and negative emotion scale

[0081] content Positive Weight Positive / Negative animal 0.88 positive Cartoon 0.71 positive figure 0.6 positive landscape 0.55 positive object 0.41 negative other 0.32 negative

[0082] ②. Quantify the emotions of social user avatars based on the emotion scale and preset rules, and label the user's emotion value according to the user's historical avatar emotions.

[0083] ③. According to the image color and image content scale rules, identify the color and content category of the user's current avatar and historical avatars, and assign color positivity weights to the avatar used by user i at time t accordingly and content positivity weight Here, time t is the t-th time unit in the previous text, and the length of the time unit can be flexibly configured, such as 1 week, 1 month, etc., which is not limited here.

[0084] ④. Average the color positivity weight and content positivity weight of each user's avatar to obtain the emotion value of the avatar used by user i at time t

[0085] ⑤. Convert all user avatar emotion value sequences into equal-length digital sequences Where T is the sequence length of all user avatar emotion value sequences.

[0086] Similarly, the head portrait image sequence and the order information sequence in step 102 can also be processed in a similar manner, which is briefly described as follows:

[0087] (2) Generate a user's historical communication product subscription information sequence by ordering time. Then, convert the obtained user's historical communication product subscription information sequence into a sequence of equal length, expressed as Among them, O itis the serial number of the communication product ordered by user i at time t, and T′ is the length of the user's historical communication product ordering information sequence.

[0088] (3) These three data information tables are linked based on the user number and the communication product number to generate a new data table. The fields in this data table include user number, user avatar emotion value sequence, user history avatar sequence, communication product number, communication product related information (such as monthly fee, call duration, traffic, and short message quantity characteristics included in the product), the interaction relationship between the user and the communication product, and the user's history communication product subscription information sequence.

[0089] Based on this, this embodiment achieves feature enhancement and reduces the cost of company data collection by serializing and modeling relevant historical information while keeping the number of original features within a reasonable range.

[0090] Before executing step 106, see Figure 2 , also includes:

[0091] Step 3: Divide the input features X and output Y of the communication product recommendation model. The input feature X includes the user avatar emotion value sequence, the user's historical avatar sequence, the communication product number, the communication product, and the user's historical communication product ordering information sequence; the output Y is the interaction relationship between the user and the communication product.

[0092] Step 4: Divide into training set and test set.

[0093] Step 106: Generate user abstract features for each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0094] Combination Figure 3 After obtaining the avatar image sequence, the avatar emotion value sequence and the order information sequence, the three can be processed separately to obtain corresponding sequence features and then fused to obtain user abstract features. Specifically:

[0095] For the avatar image sequence: a convolutional neural network is used to extract features of the avatar images in the avatar image sequence of the target user, and a long short-term memory neural network is used to identify the temporal features of the avatar images to obtain the historical avatar abstract feature matrix of the target user.

[0096] For the avatar emotion value sequence: through the embedding layer, each emotion value in the target user's avatar emotion value sequence is mapped into a real-valued vector, and is input into the long short-term memory network layer in time sequence to obtain the user avatar emotion value sequence characteristics of the target user.

[0097] For the ordering information sequence: through the embedding layer, each ordering information in the ordering information sequence of the target user is mapped into a real-valued vector, and is input into the long short-term memory network layer in time sequence to obtain the user history communication product ordering sequence features of the target user; the historical avatar abstract feature matrix, the user avatar emotion value sequence features and the user history communication product ordering sequence features are fused to obtain the user abstract features of the target user.

[0098] Combination Figure 2 Step 5 shown: Building a communication product recommendation model, and Figure 3 The following is an explanation of the generation process of user abstract features:

[0099] ① For user avatar image sequence data, input all historical avatar images X of user i it (t=1,2,…T), use the convolution kernel K of size k1×k2 to convolve the image to obtain the feature map A it , the following convolution formula is used to calculate the convolution layer output, that is, the feature map A it The value of the mth row and nth column in the table is:

[0100]

[0101] For feature map A it Take the maximum pooling method, that is, take the maximum value in the sliding window for pooling operation to obtain the feature map B it Then, the feature map B it Perform a flattening operation to obtain a one-dimensional vector a it At this point, we get the historical avatar abstract feature matrix F of user i i ={a i1 ,a i2 ,…,a iT}.

[0102] ② For the user avatar emotion value sequence data, the emotion value E it Mapped to a real-valued vector Get the user i's avatar emotion value sequence data This sequence data is input into the long short-term memory network layer according to the time sequence relationship to describe the time sequence characteristics of the user's avatar's emotions changing over time. The structure of the long short-term memory unit is as follows: Figure 3 shown.

[0103] ③. For the user's historical communication product ordering sequence, the same processing scheme as step ② is adopted to obtain the deep abstract features of the user's historical communication product ordering sequence.

[0104] ④. Fusion the features generated in step ①, step ②, and step ③ to generate user abstract features.

[0105] in, Figure 3 The schematic diagram of the long short-term memory layer unit in Figure 4 shown.

[0106] Based on this, this embodiment realizes the re-mining of deep abstract features of social platform user emotions based on long short-term memory neural network and graph convolutional neural network: graph convolutional neural network extracts abstract features of social platform user avatars; based on long short-term memory neural network, the temporal features of user avatars are identified to realize the re-mining of emotional abstract features of social users, making it easier for the model to capture effective information in the features.

[0107] Similarly, for the abstract features of communication products, the generation process is as follows:

[0108] ① For the communication product number and communication product related information, these feature values ​​are mapped to corresponding real-valued vectors through the embedding layer to generate corresponding embedded features.

[0109] ②. Fusion the outputs of the above steps to generate abstract features of communication products.

[0110] Step 108: Use the user abstract features and communication product abstract features of each user as input features to train a communication product processing model to obtain a preference relationship between the user and the communication product. The communication product abstract features are abstract features of product information of all communication products. Figure 3 As shown, specifically:

[0111] (3) Fusion of user abstract features and communication product abstract features, input of the fusion features into the fully connected layer, and output

[0112] (4) Use x′ as the input of the factorization layer, and finally output Y=[y uc ]∈{0,1} indicates whether user u has an interactive preference relationship with communication product c.

[0113] Based on this, this embodiment constructs an inter-system interaction relationship identification model, identifies the interaction dependency between the user-communication product subsystem, refers to the time series of emotion changes and communication product orders, and uses an effective factor decomposition machine to optimize the ranking effect, and outputs the purchased product preference ranking of users in different categories.

[0114] See also Figure 2 , after completing the training of the communication product processing model, it also includes:

[0115] Step 6: Complete the training of the communication product recommendation model proposed in step 5 based on the training set. Optimize the model parameter set with the goal of minimizing the training error.

[0116] Step 7: Based on the test set and the communication product recommendation model trained in step 4, obtain the recommendation results corresponding to the training set, and use the root mean square error evaluation method to evaluate the model performance. The root mean square error calculation can be used as follows:

[0117]

[0118] Where m is the size of the training set, is the prediction result, y i For actual results.

[0119] In another feasible embodiment, it also includes:

[0120] Step 8: recommending a communication product, including: recommending a solution for a communication product, and recommending a solution for a communication product to a certain user (group), specifically:

[0121] (1) Combine the trained communication product abstract features and save them into the database.

[0122] (2) The trained user features are abstracted and combined into a user feature matrix and saved in the database.

[0123] (3) For a specified user (such as a new user), calculate the cosine similarity between the current user feature and the entire user feature matrix, select the TopK users with the largest similarity, and select the communication products with the highest subscription rate among these TopK users as recommendations. Specifically:

[0124] Based on the user abstract features of the designated user, similar users whose similarity with the designated user meets preset conditions are searched from the database; based on the preference relationship between the user and the communication product, the communication product preferred by the similar user is determined and recommended to the designated user.

[0125] (4) For a specified product (such as a new product), the cosine similarity calculation method is also used to match the new product features with the communication product feature matrix, and the TopK products with the greatest similarity are recommended to the corresponding users, thereby solving the product cold start problem. Specifically:

[0126] Based on the communication product abstract features of the specified communication product, a communication product whose similarity with the specified communication product meets preset conditions is searched; based on the preference relationship between the user and the communication product, a user group that has a preference relationship with the found communication product is determined, and the specified communication product is recommended to the user group.

[0127] Based on this, this embodiment uses cosine similarity to calculate the similarity between new products and existing businesses, and recommends new products to different user groups based on the similarity to solve the product cold start problem; combines local outlier recognition and popular product and marginal product feature co-occurrence methods to mine products that are less purchased by a certain category of social users but have greater purchasing appeal; at the same time, association rules are used to identify the connection between different products, and recommend products that have not yet been ordered but have potential preferences and needs to users. The present invention has a good recommendation effect, can fully consider the social user's avatar and emotional information, and enable social users to obtain product recommendation results with high satisfaction.

[0128] It can be seen that this embodiment, starting from the perspective of image visual features and based on relevant psychological research, has compiled a set of positive and negative emotion labeling scales, fully mined the visual information of users' social network avatars, and made the portrayal of consumer preferences under the recommendation system more comprehensive, thus maximizing the utilization of complex unstructured data.

[0129] Figure 5 A flowchart of a method for processing user emotions provided in an embodiment of this specification is shown in FIG. Figure 5 , the method may specifically include the following steps:

[0130] Step 502: Obtain the image color and image content of the target user's avatar image;

[0131] Step 504: Based on the image color and image content of the avatar image, respectively determine the color positivity weight and content positivity weight of the avatar image; specifically:

[0132] Starting from the perspective of image visual features and based on relevant psychological research, a set of positive and negative emotion labeling scales are compiled in advance, such as Tables 1 and 2 above; then, the range of RGB values ​​corresponding to the color attributes that match the image color of the avatar image in the positive and negative emotion scale is searched; if the RGB value of the avatar image matches the range of RGB values ​​searched for, the positivity weight corresponding to the range of RGB values ​​searched for is used as the color positivity weight of the avatar image; the positivity weight corresponding to the type of image content in the positive and negative emotion scale is searched, and used as the content positivity weight of the avatar image.

[0133] Step 506: Determine the emotion value of the target user based on the color positivity weight and the content positivity weight of the avatar image.

[0134] Since the implementation of steps 502 to 506 has been Figure 1 It is described in the embodiment, so step 502 to step 506 will not be described in detail here.

[0135] It can be seen that this embodiment studies the relationship between social software user avatars and emotions based on the positive and negative emotion scale method, scores the positive and negative emotions of avatar pictures, and realizes an avatar emotion standard comparison library.

[0136] Figure 6 This is a schematic diagram of a communication product processing device provided in an embodiment of this specification, see Figure 6 , the above-mentioned device may specifically include:

[0137] A determination module 601 is used to determine the avatar image sequences used historically by batch users and the ordering information sequences of the communication products ordered historically;

[0138] A quantification module 602 is used to quantify the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change of the emotion value of the user within the preset historical time period;

[0139] A processing module 603, configured to generate a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0140] The training module 604 is used to train the communication product processing model using the user abstract features and communication product abstract features of each user as input features to obtain the preference relationship between the user and the communication product. The communication product abstract features are abstract features of the product information of all communication products.

[0141] Optionally, the determination module 601 is specifically configured to:

[0142] Acquire avatar images and ordering information of communication products used by batch users within a preset historical period; sort the avatar images of each user in chronological order to obtain a avatar image sequence of each user; sort the ordering information of each user in chronological order to obtain a ordering information sequence of each user.

[0143] Optionally, the quantization module 602 is specifically configured to:

[0144] Based on the image color and image content of the avatar image, a pre-established positive and negative emotion scale is queried to obtain the positivity weights corresponding to the image color and image content; based on the positivity weights corresponding to the image color and image content, the emotion value corresponding to the avatar image is determined; based on the emotion values ​​corresponding to each avatar image, an avatar emotion value sequence is generated, and the avatar emotion value sequence is used to describe the changes in the emotion values ​​of each time unit of the same user within the preset historical time period.

[0145] Optionally, the positive and negative emotion scale stores a correspondence between color attributes, a range of RGB values ​​and a positivity weight, and a correspondence between a type of image content and a positivity weight;

[0146] The quantization module 602 is specifically used for:

[0147] Find the value range of RGB value corresponding to the color attribute that matches the image color of the avatar image in the positive and negative emotion scale; if the RGB value of the avatar image matches the value range of the found RGB value, use the positivity weight corresponding to the value range of the found RGB value as the color positivity weight of the avatar image; find the positivity weight corresponding to the type of image content in the positive and negative emotion scale, and use it as the content positivity weight of the avatar image.

[0148] Optionally, the quantization module 602 is specifically configured to:

[0149] Determine the color positivity weight and content positivity weight of the same user in each time unit of a preset historical period; average the color positivity weight and content positivity weight in each time unit to obtain the emotion value corresponding to the avatar used by the user in the time unit.

[0150] Optionally, the processing module 603 is specifically configured to:

[0151] A convolutional neural network is used to extract features of the avatar images in the avatar image sequence of the target user, and the time series features of the avatar images are identified through a long short-term memory neural network to obtain the historical avatar abstract feature matrix of the target user; through an embedding layer, each emotion value in the avatar emotion value sequence of the target user is mapped to a real-valued vector, and is input into the long short-term memory network layer in time series to obtain the user avatar emotion value sequence feature of the target user; through an embedding layer, each order information in the order information sequence of the target user is mapped to a real-valued vector, and is input into the long short-term memory network layer in time series to obtain the user historical communication product order sequence feature of the target user; the historical avatar abstract feature matrix, the user avatar emotion value sequence features and the user historical communication product order sequence features are fused to obtain the user abstract feature of the target user.

[0152] Optionally, the device further comprises:

[0153] The first recommendation module is used to search for a communication product whose similarity with the specified communication product meets preset conditions based on the communication product abstract features of the specified communication product; determine a user group that has a preference relationship with the found communication product based on the preference relationship between the user and the communication product, and recommend the specified communication product to the user group.

[0154] Optionally, the device further comprises:

[0155] The second recommendation module is used to find similar users whose similarity with the designated user meets preset conditions based on the user abstract characteristics of the designated user; based on the preference relationship between the user and the communication product, determine the communication product preferred by the similar user and recommend it to the designated user.

[0156] It can be seen that this embodiment, starting from the perspective of image visual features and based on relevant psychological research, has compiled a set of positive and negative emotion labeling scales, fully mined the visual information of users' social network avatars, and made the portrayal of consumer preferences under the recommendation system more comprehensive, thus maximizing the utilization of complex unstructured data.

[0157] Figure 7 This is a schematic diagram of a user emotion processing device provided in an embodiment of this specification, see Figure 7 , the above-mentioned device may specifically include:

[0158] An acquisition module 701 is used to acquire the image color and image content of the avatar image of the target user;

[0159] The first processing module 702 is used to determine the color positivity weight and content positivity weight of the avatar image based on the image color and image content of the avatar image, respectively; specifically:

[0160] Starting from the perspective of image visual features and based on relevant psychological research, a set of positive and negative emotion labeling scales are compiled in advance, such as Tables 1 and 2 above; then, the range of RGB values ​​corresponding to the color attributes that match the image color of the avatar image in the positive and negative emotion scale is searched; if the RGB value of the avatar image matches the range of RGB values ​​searched for, the positivity weight corresponding to the range of RGB values ​​searched for is used as the color positivity weight of the avatar image; the positivity weight corresponding to the type of image content in the positive and negative emotion scale is searched, and used as the content positivity weight of the avatar image.

[0161] The second processing module 703 is used to determine the emotion value of the target user based on the color positivity weight and the content positivity weight of the avatar image.

[0162] It can be seen that this embodiment studies the relationship between social software user avatars and emotions based on the positive and negative emotion scale method, scores the positive and negative emotions of avatar pictures, and realizes an avatar emotion standard comparison library.

[0163] In addition, for the above-mentioned device implementation, since it is basically similar to the method implementation, the description is relatively simple, and the relevant parts can refer to the partial description of the method implementation. It should be noted that among the various components of the device in this specification, the components are logically divided according to the functions to be implemented, but this specification is not limited to this, and the various components can be re-divided or combined as needed.

[0164] Figure 8 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this specification, see Figure 8 , the electronic device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and of course may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a communication product processing device at the logical level. Of course, in addition to the software implementation method, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0165] The network interface, processor and memory can be connected to each other through a bus system. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0166] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include read-only memory and random access memory, and provide instructions and data to the processor. The memory may include high-speed random access memory (RAM), and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0167] The processor is used to execute the program stored in the memory, and specifically perform:

[0168] Determine the avatar image sequence used historically by batch users and the ordering information sequence of the communication products ordered historically;

[0169] quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within the preset historical time period;

[0170] generating a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence;

[0171] The user abstract features and communication product abstract features of each user are used as input features to train a communication product processing model to obtain the preference relationship between the user and the communication product. The communication product abstract features are abstract features of product information of all communication products.

[0172] The above is as in this manual Figure 6 The method performed by the communication product processing device or the manager (Master) node disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of this specification can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of this specification can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in a decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0173] The communication product processing device can also perform Figure 1-4 and implement the method executed by the manager node.

[0174] Based on the same invention, the embodiment of this specification also provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes Figure 1-4 The corresponding embodiment provides a communication product processing method.

[0175] Fig. 9 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this specification, see Fig. 9 , the electronic device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and of course may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a user emotion processing device at the logical level. Of course, in addition to the software implementation method, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0176] The network interface, processor and memory can be connected to each other through a bus system. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig. 9 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0177] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include read-only memory and random access memory, and provide instructions and data to the processor. The memory may include high-speed random access memory (RAM), and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0178] The processor is used to execute the program stored in the memory, and specifically perform:

[0179] Obtain the image color and image content of the target user's avatar image;

[0180] Determining a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively;

[0181] The emotion value of the target user is determined based on the color positivity weight and the content positivity weight of the avatar image.

[0182] The above is as in this manual Figure 7 The method performed by the user emotion processing device or the manager (Master) node disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of this specification can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of this specification can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0183] The user emotion processing device can also perform Figure 5 and implement the method executed by the manager node.

[0184] Based on the same invention, the embodiment of this specification also provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes Figure 5 The corresponding embodiment provides a method for processing user emotions.

[0185] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0186] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0187] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0188] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0189] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0191] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0192] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0193] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0194] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device 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, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0195] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0196] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A communication product processing method, characterized in that: include: Determine the avatar image sequence used historically by batch users and the ordering information sequence of the communication products ordered historically; quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user, wherein the avatar emotion value sequence is used to describe the change in the emotion value of the user within a preset historical time period; generating a user abstract feature of each user based on the avatar image sequence, the avatar emotion value sequence and the order information sequence; The user abstract features and communication product abstract features of each user are used as input features to train a communication product processing model to obtain a preference relationship between the user and the communication product, wherein the communication product abstract features are abstract features of product information of all communication products; Among them, the generation of user abstract features of each user based on the avatar image sequence, the avatar emotion value sequence and the ordering information sequence includes: using a convolutional neural network to extract features of the avatar images in the avatar image sequence of the target user, and identifying the temporal features of the avatar images through a long short-term memory neural network to obtain the historical avatar abstract feature matrix of the target user; mapping each emotion value in the avatar emotion value sequence of the target user into a real-valued vector through an embedding layer, and inputting it into the long short-term memory network layer in time sequence to obtain the user avatar emotion value sequence features of the target user; mapping each order information in the order information sequence of the target user into a real-valued vector through an embedding layer, and inputting it into the long short-term memory network layer in time sequence to obtain the user historical communication product ordering sequence features of the target user; fusing the historical avatar abstract feature matrix, the user avatar emotion value sequence features and the user historical communication product ordering sequence features to obtain the user abstract features of the target user.

2. The method according to claim 1, characterized in that The step of determining the avatar image sequences used historically by batch users and the ordering information sequences of the communication products ordered historically comprises: Obtaining avatar images used by batch users within a preset historical period and ordering information of communication products; Sort the avatar images of each user in chronological order to obtain a sequence of avatar images of each user; The ordering information of each user is sorted in chronological order to obtain the ordering information sequence of each user.

3. The method according to claim 1, characterized in that The step of quantifying the emotion values ​​reflected by the avatar images in the avatar image sequence to obtain an avatar emotion value sequence for each user includes: Based on the image color and image content of the avatar image, a pre-established positive and negative emotion scale is queried to obtain the positivity weight corresponding to the image color and image content; Determine the emotion value corresponding to the avatar image based on the positivity weights corresponding to the image color and the image content; Based on the emotion values ​​corresponding to the avatar images, an avatar emotion value sequence is generated, where the avatar emotion value sequence is used to describe the change of the emotion value of each time unit of the same user within the preset historical time period.

4. The method according to claim 3, characterized in that The positive and negative emotion scale contains the corresponding relationship between color attributes, RGB value ranges and positivity weights, and the corresponding relationship between image content types and positivity weights; The step of querying a pre-established positive and negative emotion scale based on the image color and image content of the avatar image to obtain the positivity weight corresponding to the image color and image content includes: Find the value range of the RGB value corresponding to the color attribute that matches the image color of the avatar image with the positive and negative emotion scale; If the RGB value of the avatar image matches the value range of the RGB value being searched, the positivity weight corresponding to the value range of the RGB value being searched is used as the color positivity weight of the avatar image; The positivity weight corresponding to the type of the image content in the positive and negative emotion scale is found and used as the content positivity weight of the avatar image.

5. The method according to claim 4, characterized in that The step of determining the emotion value corresponding to the avatar image based on the positivity weights corresponding to the image color and the image content includes: Determine the color positivity weight and content positivity weight of the same user in each time unit of a preset historical period; The color positivity weight and content positivity weight in each time unit are averaged to obtain the emotion value corresponding to the avatar used by the user in the time unit.

6. The method according to any one of claims 1 to 5, characterized in that Also includes: Based on the communication product abstract features of the specified communication product, searching for a communication product whose similarity with the specified communication product meets a preset condition; Based on the preference relationship between the user and the communication product, a user group having a preference relationship with the found communication product is determined, and the designated communication product is recommended to the user group.

7. The method according to any one of claims 1 to 5, characterized in that Also includes: Based on the user abstract features of the specified user, searching for similar users whose similarity with the specified user meets preset conditions; Based on the preference relationship between the user and the communication product, the communication product preferred by the similar user is determined and recommended to the designated user.

8. A method for processing user emotions, characterized in that: include: Obtain the image color and image content of the target user's avatar image; Determining a color positivity weight and a content positivity weight of the avatar image based on an image color and an image content of the avatar image, respectively; Determining the emotion value of the target user based on the color positivity weight and the content positivity weight of the avatar image; Wherein, the color positivity weight and content positivity weight of the avatar image are determined respectively based on the image color and image content of the avatar image, including: searching for the value range of RGB values ​​corresponding to the color attribute that matches the image color of the avatar image in the pre-established positive and negative emotion scale; if the RGB value of the avatar image matches the value range of the searched RGB value, then the positivity weight corresponding to the value range of the searched RGB value is used as the color positivity weight of the avatar image; searching for the positivity weight corresponding to the type of image content in the positive and negative emotion scale, and using it as the content positivity weight of the avatar image; wherein the positive and negative emotion scale contains the correspondence between the color attribute, the value range of RGB value and the positivity weight, and the correspondence between the type of image content and the positivity weight; The method of determining the target user's emotion value based on the color positivity weight and content positivity weight of the avatar image includes: determining the color positivity weight and content positivity weight of the same user in each time unit of a preset historical period; averaging the color positivity weight and content positivity weight in each time unit to obtain the emotion value corresponding to the avatar used by the user in the time unit.

9. An electronic device, characterized in that: include: Communications interfaces, processors, and memory; The processor calls the program instructions in the memory to execute the steps of the method according to any one of claims 1 to 8.

Images