Data generation method and device and electronic equipment

Abstract

The invention provides a data generation method, comprising: acquiring a feature vector of a target object, the feature vector at least representing quantitative information corresponding to at least one target feature of the target object; according to the feature vector, through a target qualitative model, generating at least one piece of qualitative information for the target object; wherein the qualitative information is information for evaluating the quantitative information of the target object.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and more specifically, to a data generating method, device and electronic device. Background Art

[0002] Some objects have various characteristics, which can be expressed by numerical values, such as specific quality data, size data, price data, speed data, etc. Although these numerical values ​​can describe the properties of the object in detail, it is difficult to retrieve the object in a qualitative way. However, for users, it is much easier to describe objects through qualitative descriptions, such as cheap, expensive, light, heavy, etc., than to accurately give specific numerical values, which is more in line with operating habits. In some scenarios, users often want to use an intuitive and qualitative way to describe objects, requiring the objects to have some qualitative description content corresponding to the quantitative numerical values. Therefore, a method is needed to convert quantitative data into qualitative information. Summary of the invention

[0003] In view of this, the present disclosure provides a data generating method, device and electronic device.

[0004] One aspect of the present disclosure provides a data generation method, including: obtaining a feature vector of a target object, the feature vector at least representing quantitative information corresponding to at least one target feature of the target object; generating at least one qualitative information for the target object through a target qualitative model based on the feature vector; wherein the qualitative information is information for evaluating the quantitative information of the target object.

[0005] According to an embodiment of the present disclosure, at least one qualitative information for a target object is generated based on a feature vector through a target qualitative model, including: determining, based on the feature vector, through a target qualitative model, a qualitative word for the target object and a degree word corresponding to the qualitative word; generating at least one qualitative information based on the degree word and the qualitative word; wherein the qualitative word represents a natural language qualitative evaluation of at least one target feature, and the degree word is a vocabulary in natural language used to represent a degree.

[0006] According to an embodiment of the present disclosure, the data generation method also includes: generating a membership degree corresponding to the qualitative information through a target qualitative model based on the feature vector, the membership degree representing the probability that the target object is evaluated as qualitative information; wherein the membership degrees corresponding to the respective qualitative information satisfy a specific relationship.

[0007] According to an embodiment of the present disclosure, the target qualitative model is obtained by the following operations: acquiring sample data, the sample data including a first sample vector and first sample qualitative information of a sample object, the first sample vector representing quantitative information corresponding to at least one target feature of the sample object, and the first sample qualitative information being information for evaluating the quantitative information of the sample object; and training a qualitative model network based on the sample data to obtain the target qualitative model.

[0008] According to an embodiment of the present disclosure, obtaining sample data includes: obtaining a first sample vector of a sample object through a large model; obtaining a plurality of qualitative words for a target object from comment information on the sample object, the qualitative words representing a natural language qualitative evaluation of the target object; clustering the plurality of qualitative words to obtain at least one qualitative word collection; and determining the first sample qualitative information corresponding to the qualitative word collection.

[0009] According to an embodiment of the present disclosure, the types of sample data include positive sample data and negative sample data, and obtaining the sample data also includes: obtaining the degree words corresponding to the qualitative words from the comment information, the degree words being words used to represent the degree in natural language; and determining the positive and negative types of the sample data according to the semantic tendency of the degree words.

[0010] According to an embodiment of the present disclosure, the sample data also includes sample time information, and the sample time information represents the time range corresponding to the first sample qualitative information. Training the qualitative model network includes: training the qualitative model network according to the first sample vector, the first sample qualitative information and the sample time information to obtain a target qualitative model, and the target qualitative model is used to generate at least one qualitative information for the target object at the current moment according to the feature vector of the product.

[0011] According to an embodiment of the present disclosure, sample data includes original sample data and similar sample data, wherein similar sample objects corresponding to the similar sample data have product correlation with original sample objects corresponding to the original sample data; the similar sample data is obtained by the following operations: determining similar sample objects based on the original sample objects, the similar sample objects are different from the original sample objects and have the same target features; obtaining a second sample vector and first sample qualitative information of the similar sample objects as the similar sample data.

[0012] According to an embodiment of the present disclosure, sample data includes original sample data and similar sample data, wherein the similar sample data and the original sample data have a time correlation; the sample data also includes sample time information, the sample time information characterizes the time range corresponding to the first sample qualitative information, and the similar sample data is obtained by the following operations: obtaining comment information of the sample object within a first time range, the first time range being different from a second time range corresponding to the original sample data of the same sample object; generating second sample qualitative information for the sample object based on the comment information within the first time range; obtaining the first sample vector, the second sample qualitative information, and the sample time information corresponding to the second time range as the similar sample data.

[0013] Another aspect of the present disclosure provides a data generating device, including: a first acquisition module, used to acquire a feature vector of a target object, the feature vector at least representing quantitative information corresponding to at least one target feature of the target object; and a first generation module, used to generate at least one qualitative information for the target object according to the feature vector through a target qualitative model; wherein the qualitative information is information for evaluating the quantitative information of the target object.

[0014] Another aspect of the present disclosure provides an electronic device, including: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described above.

[0015] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above method when executed.

[0016] Another aspect of the present disclosure provides a computer program product, which includes computer executable instructions, and when the instructions are executed, are used to implement the method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0018] Figure 1 An exemplary system architecture diagram to which the data generation method according to an embodiment of the present disclosure can be applied is schematically shown;

[0019] Figure 2 The flowchart of the data generation method according to the embodiment of the present disclosure is schematically shown;

[0020] Figure 3A flowchart for determining qualitative information according to an embodiment of the present disclosure is schematically shown;

[0021] Figure 4 Another flow chart of the data generation method according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 5 A flowchart of a method for training a target qualitative model according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 6 A flowchart of obtaining sample data in a training method of a target qualitative model according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 7 Another flowchart of obtaining sample data in the training method of the target qualitative model according to an embodiment of the present disclosure is schematically shown;

[0025] Figure 8 Another flowchart of training a target qualitative network in a training method for training a target model according to an embodiment of the present disclosure is schematically shown;

[0026] Fig. 9 A flowchart for obtaining similar sample data in a method for training a target qualitative model according to an embodiment of the present disclosure is schematically shown;

[0027] Fig.10 Another flowchart for obtaining similar sample data in the training method of the target qualitative model according to an embodiment of the present disclosure is schematically shown;

[0028] Fig.11 A block diagram schematically shows a data generating device according to an embodiment of the present disclosure;

[0029] Fig.12 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0030] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0031] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0032] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0033] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0034] In the embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security, and national security.

[0035] An embodiment of the present disclosure provides a data generation method, including: obtaining a feature vector of a target object, the feature vector at least representing quantitative information corresponding to at least one target feature of the target object; generating at least one qualitative information for the target object through a target qualitative model based on the feature vector; wherein the qualitative information is information for evaluating the quantitative information of the target object.

[0036] Figure 1 The exemplary system architecture 100 to which the data generation method according to the embodiment of the present disclosure can be applied is schematically shown. It should be noted that: Figure 1 What is shown is merely an example of a system architecture to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0037] like Figure 1As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0038] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software.

[0039] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices with communication and interaction functions, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0040] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0041] It should be noted that the data generation method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the data generation system provided in the embodiment of the present disclosure can generally be set in the server 105. The data generation method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Correspondingly, the data generation system provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Alternatively, the data generation method provided in the embodiment of the present disclosure can also be performed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or can also be performed by other terminal devices that are different from the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the data generation system provided by the embodiment of the present disclosure may also be arranged in the first terminal device 101, the second terminal device 102 or the third terminal device 103, or in other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0042] For example, the data to be processed may be originally stored in any one of the first terminal device 101, the second terminal device 102 or the third terminal device 103 (for example, the first terminal device 101, but not limited thereto), or stored in an external storage device and may be imported into the first terminal device 101. Then, the first terminal device 101 may locally execute the data generation method provided by the embodiment of the present disclosure, or send the data to be processed to other terminal devices, servers, or server clusters, and the other terminal devices, servers, or server clusters that receive the data to be processed may execute the data generation method provided by the embodiment of the present disclosure.

[0043] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0044] Figure 2 The flowchart of the data generating method according to the embodiment of the present disclosure is schematically shown.

[0045] like Figure 2 As shown, the method includes operations S210 to S220.

[0046] In operation S210, a feature vector of a target object is obtained, where the feature vector at least represents quantitative information corresponding to at least one target feature of the target object.

[0047] The feature vector of the target object may be a vector including various attributes or feature information of the commodity, wherein each dimension of the vector represents an attribute or feature of the target object, and the value of each dimension of the vector represents the specific content of the corresponding attribute or feature.

[0048] Attributes and features may include, but are not limited to, product name, brand, category, specific model, specific price, material, size, weight, number of CPU cores, memory capacity, etc.

[0049] The feature vector is the content corresponding to these attributes or features, which is processed and converted into a specific vector form to facilitate processing by models, etc. For example, for text features (product name, brand, category, operating system, etc.), the corresponding encoding technology can be used to encode the text content into the corresponding numerical value, which is used as the value of the corresponding dimension in the feature vector. For the quantitative information of numerical features (price, weight, memory capacity, etc.), it can be used directly or after row normalization as the value of the corresponding dimension in the feature vector.

[0050] For example, an object has the following attributes and features: object name: "abc", brand: "brand B", category: "mobile phone", price: 2999, weight: "1800 grams". Assume that the code value corresponding to "abc" is 56; suppose there are three brands "brand A", "brand B", "brand C", and the corresponding codes are (0, 1, 2); suppose there are three series "mobile phone", "laptop", "desktop", and the corresponding codes are (0, 1, 2). In the feature vector, the order of dimensions is product name, brand, category, price, and weight, so the corresponding feature vector can be: [56, 1, 0, 2999, 1800]. In the subsequent description of feature vectors, for the convenience of reading and understanding, the values ​​of the dimensions of qualitative attributes will be directly replaced by corresponding text instead of the corresponding encoding values. For example, the above feature vector will be described as [abc, brand B, mobile phone, 2999, 1800], but in actual operation, its feature vector should be [56, 1, 0, 2999, 1800].

[0051] It should be noted that the feature vector contains quantitative information of the object, which is usually described by specific numerical values, such as the weight and price of the object. The type of information contained in the feature vector may also include other information, such as the brand of the object, the type of the object, etc. According to the embodiment of the present disclosure, the feature vector at least represents the quantitative information corresponding to at least one target feature of the target object. That is, in the feature vector used in the embodiment of the present disclosure, the feature or attribute represented by at least one dimension is described in a quantitative manner, and the specific numerical value of this dimension can undoubtedly quantitatively reflect a certain feature or attribute of the target object. For example, in the feature vector [abc, brand B, mobile phone, 2999, 1800], the fourth dimension represents the price, and its value is 2999, which directly and quantitatively reflects the price value of the target object.

[0052] In operation S220, at least one qualitative information for the target object is generated according to the feature vector through a target qualitative model, wherein the qualitative information is information for evaluating the quantitative information of the target object.

[0053] Specifically, the feature vector is input into the target qualitative model to generate at least one qualitative information for the target object. Qualitative information is information that evaluates the quantitative information of the target object, and for the sake of convenience of description, it is also referred to as evaluation information. Generally, qualitative information is an adjective, name or adverb in a natural language that evaluates the quantitative information of the target object.

[0054] The quantitative information targeted by the qualitative information may be in each attribute or feature represented by the feature vector. For example, after the feature vector [abc, brand B, mobile phone, 2999, 1800] is input into the compensatory qualitative model, the evaluation information of the quantitative information "2999" for the attribute "price" is obtained, for example, the evaluation word obtained is "moderate". The quantitative information targeted by the qualitative information may not be a specific one of the attributes or features represented by the feature vector. For example, the dimensions of a feature vector represent the brand, the number of CPU cores, and the memory capacity in turn. Then, the target qualitative model can also reasonably infer the evaluation information for "overall performance" based on the quantitative information (the number of CPU cores, memory capacity, etc.) in the feature vector.

[0055] In some embodiments, the generated evaluation information can be used as a new attribute tag of the target object and added to the data of the target object, so that qualitative retrieval of the target object can be quickly realized. For example, if the evaluation information obtained for a mobile phone is "cheap", then the label "cheap" will be added to the data of the mobile phone. When the user performs a qualitative search, such as "recommend me a cheap mobile phone", the mobile phone can be matched through the label to realize further retrieval, recommendation and other processing.

[0056] Figure 3 The flowchart of determining qualitative information according to an embodiment of the present disclosure is schematically shown.

[0057] like Figure 3 As shown, in Figure 2 Based on the embodiment, operation S220 may include operations S310 to S320.

[0058] In operation S310, a qualitative word for a target object and a degree word corresponding to the qualitative word are determined according to the feature vector through a target qualitative model, wherein the qualitative word represents a natural language qualitative evaluation of at least one target feature, and the degree word is a word used to represent a degree in a natural language.

[0059] In operation S320, at least one qualitative information is generated according to the degree word and the qualitative word.

[0060] In some cases, the scope of evaluation information is relatively large and cannot accurately evaluate the target object. Therefore, in the embodiment of the present disclosure, the target qualitative model generates content qualitative words and degree words corresponding to the qualitative words based on the feature vector. Then, based on the qualitative words and degree words, quantitative information for evaluating the target object is generated.

[0061] For example, with respect to price, in the user's cognition, mobile phones in the price range of 0 to 2000 may all be cheap mobile phones, but the classification is still not accurate enough. In the disclosed embodiment, degree words corresponding to "cheap" can be generated, such as degree words such as "a little" (lower degree), "medium" (medium degree), and "very" (higher degree). Then, based on the evaluation words and degree words, a qualitative evaluation of "price" is generated. For example, after the feature vector of a 1999 yuan mobile phone is input into the target qualitative model, the evaluation word obtained is: "cheap", and the corresponding degree word obtained is: "a little", then its evaluation information is "a little cheap"; after the feature vector of a 300 yuan mobile phone is input into the target qualitative model, the evaluation word obtained is also: "cheap", and the corresponding degree word obtained is "very", and the generated evaluation information is "very cheap". In this way, the evaluation words are further subdivided to improve the granularity of the generated evaluation information.

[0062] Figure 4 Another flow chart of the data generating method according to an embodiment of the present disclosure is schematically shown.

[0063] like Figure 4 As shown, based on any of the foregoing embodiments, the data generating method may further include operation S410.

[0064] In operation S410, the membership degree corresponding to the qualitative information is generated through the target qualitative model according to the feature vector, and the membership degree represents the probability that the target object is evaluated as qualitative information. Among them, the membership degrees corresponding to the various qualitative information satisfy a specific relationship.

[0065] Different users' evaluations of the same product are subjective, and different people have different understandings. For example, for the quantitative information of "price", some users think that 3,000 yuan is cheap, while some users think that 3,000 yuan is expensive. In this case, the target qualitative model can also generate at least one qualitative information and the corresponding membership degree of the qualitative information. The membership degree represents the probability that the target object is evaluated as the corresponding qualitative information.

[0066] For example, if a mobile phone with a price of 3,500 yuan is fed into the target qualitative model, the content obtained for “price” can be:

[0067]

[0068] From the above table, we can see that for a mobile phone priced at 3,500 yuan, the probability of being evaluated as “cheap” is 20%, the probability of being evaluated as “moderate” is 70%, and the probability of being evaluated as “expensive” is 10%.

[0069] Combination Figure 3 In the embodiment shown, after a mobile phone with a price of 3,500 yuan is input into the target qualitative model, the content obtained for "price" may also be:

[0070]

[0071] Each qualitative word has a corresponding degree word and a corresponding membership degree under the degree word. In this way, the coverage and accuracy of the generated qualitative information are further improved, and the qualitative information of the target object can be obtained more intuitively. The qualitative information can also be further analyzed and processed through the distribution of membership degree, such as finding the most appropriate qualitative word through data distribution.

[0072] According to the embodiment of the present disclosure, the membership degrees corresponding to the various qualitative information satisfy a specific relationship. For example, the sum of the membership degrees corresponding to the various qualitative information is not greater than 100%.

[0073] In some embodiments, the qualitative information may be screened according to the membership degree. When the membership degree corresponding to the qualitative information is less than or equal to a preset membership degree threshold, the qualitative information is regarded as invalid qualitative information.

[0074] Figure 5 The flowchart of the training method of the target qualitative model according to the embodiment of the present disclosure is schematically shown.

[0075] like Figure 5 As shown, the training process of the target qualitative model used in any of the aforementioned embodiments may include operations S510 to S520.

[0076] In operation S510, sample data is acquired, the sample data including a first sample vector and first sample qualitative information of the sample object, the first sample vector represents quantitative information corresponding to at least one target feature of the sample object, and the first sample qualitative information is information evaluating the quantitative information of the sample object.

[0077] The first sample vector of the sample object is similar to the feature vector of the target object. Each dimension of the first sample vector represents an attribute or feature of the sample object. The value of each dimension of the vector represents the specific content of the corresponding attribute or feature. Please refer to the above for a detailed description, which will not be repeated here. The first sample qualitative information is pre-acquired evaluation information for the sample object, for example, it can be obtained and analyzed from the user's historical evaluation of the sample object.

[0078] In operation S520, a qualitative model network is trained according to the sample data to obtain a target qualitative model. Specifically, the qualitative model network is trained through a pair of (first sample vector, first sample qualitative information). The input of the qualitative model network is the first sample vector, and the output is the predicted content of the qualitative information. The parameters of the qualitative model network are continuously adjusted based on the calculation of the loss function, etc., until the preset conditions are met (for example, the value of the loss function is less than or equal to the preset threshold, the number of iterations is greater than or equal to the preset number of iterations, etc.), and the target qualitative model is obtained.

[0079] The qualitative model network can be a combination of one or more of a fully connected neural network, a convolutional neural network, a recurrent neural network, a generative adversarial network, an autoencoder, a graph neural network, a Transformer network, a reinforcement learning network, a U-Net, a residual network, a temporal convolutional network, and a capsule network. Those skilled in the art can select a suitable model network for training according to actual conditions, and the present disclosure does not limit this.

[0080] It should be noted that the target qualitative model is obtained to predict the target feature corresponding to the first sample qualitative information in the sample data. For example, if the first sample qualitative information is all evaluations of price, then the target qualitative model obtained is used to evaluate and predict the target feature "price"; when the first sample qualitative information is evaluation information for the two items "price" and "weight", the target qualitative model obtained is used to evaluate and predict the two target features "price" and "weight" respectively.

[0081] Figure 6The flowchart of obtaining sample data in the training method of the target qualitative model according to the embodiment of the present disclosure is schematically shown.

[0082] like Figure 6 As shown, in Figure 5 Based on the illustrated embodiment, operation S510 may include operations S610 to S640.

[0083] In operation S610, a first sample vector of a sample object is acquired through a large model.

[0084] The training data of the big model is huge and comes from rich sources. By obtaining products with corresponding qualitative information through the big model, target samples can be obtained from a wider range, and the samples are more representative. For example, by asking the big model "please recommend several moderately priced desktop computers" or "please recommend several affordable notebooks", etc., we can obtain sample objects with the qualitative information of "moderately priced" and then obtain the first sample vector of the sample object.

[0085] In operation S620, multiple qualitative words for the target object are obtained from the comment information for the sample object, and the qualitative words represent the natural language qualitative evaluation of the target object. The multiple qualitative words for the target object can be evaluation words for the same target feature quantitative information of the same sample object. There may be multiple comment information for the same sample object. Through semantic analysis and other methods, evaluation words for the same target feature quantitative information of the target object are extracted from the comment information. For example, for the target feature of price, evaluation words such as expensive, affordable, and great price are obtained.

[0086] In operation S630, a plurality of qualitative words are clustered to obtain at least one qualitative word collection. The qualitative words are clustered according to a preset clustering rule to obtain at least one qualitative word collection, and the semantic meanings of the qualitative words in the same qualitative word collection are the same or similar. For example, when the target feature is "price", qualitative words such as "like", "people-friendly", "cost-effective", "cheap", and "cheap" are classified into the same qualitative word collection, and their semantic meanings are similar or the same as "cheap" in the corresponding evaluation scenario; qualitative words such as "moderate", "fair", "appropriate", and "balanced" for price evaluation are classified into a qualitative word collection, and their semantic meanings are similar or the same as "moderate" in the corresponding evaluation scenario; qualitative words such as "not cheap", "expensive", and "luxury" for price evaluation are classified into a qualitative word collection, and their semantic meanings are similar or the same as "expensive" in the corresponding evaluation scenario.

[0087] When there are multiple qualitative word collections obtained for the same target feature, the qualitative word collection with the largest total number of occurrences of evaluation words in the qualitative word collections can be selected as the qualitative word collection corresponding to the target feature.

[0088] In operation S640, the first sample qualitative information corresponding to the qualitative word collection is determined. The first sample qualitative information may include qualitative words corresponding to the core meaning represented by the qualitative word collection, for example, a qualitative word collection including qualitative words such as "not cheap", "expensive", and "luxury" may correspond to the first sample qualitative information "expensive".

[0089] Figure 7 Another flowchart for obtaining sample data in the training method of the target qualitative model according to an embodiment of the present disclosure is schematically shown.

[0090] like Figure 7 As shown, in Figure 6 Based on the illustrated embodiment, obtaining sample data may further include operations S710 to S720.

[0091] In operation S710, a degree word corresponding to the qualitative word is obtained from the comment information, where the degree word is a word used to represent a degree in a natural language.

[0092] In operation S720, the positive and negative types of the sample data are determined according to the semantic tendency of the degree word.

[0093] In the comments of users on objects, adjectives or adverbs that affect the degree of evaluation usually appear before the evaluation words, such as quite, relatively, slightly, and slightly. These words will affect the specific meaning of the evaluation words. For example, in an evaluation of "this mobile phone is very cheap", the evaluation word is "cheap". Before "cheap", there is a degree word "very", which strengthens the meaning of the word "cheap"; in an evaluation of "this mobile phone is not very cheap", the evaluation word obtained by text processing is "cheap", but before "cheap", there is a degree word "not very", and its semantic tendency is negative, which reduces the meaning of the word "cheap".

[0094] According to the semantic tendency of the degree word, determine the positive and negative type of the sample data. At this time, the training sample can be regarded as a triple consisting of (vector, qualitative information, degree word), in which the semantic tendency of the degree word determines the positive and negative type of the sample. When the semantic tendency is positive, the sample data is a positive sample. After the qualitative model network trained with the positive sample, the output result should be close to the corresponding first sample qualitative information after the first sample vector is input. When the semantic tendency is negative, the sample data is a negative sample. After the qualitative model network trained with the negative sample, the output result should be far away from the corresponding first sample qualitative information after the first sample vector is input.

[0095] For example, a computer costs 3,000 yuan and is evaluated as "this computer is very cheap". The qualitative word obtained is "cheap", and the tone factor is "very", which is a positive semantic tendency. Then it is a positive sample. After the trained model inputs a product vector of a computer costing 3,000 yuan, the result obtained should be close to "cheap". A computer costs 12,000 yuan and is evaluated as "this computer is not very cheap". Here, the qualitative word obtained is still "cheap", but the tone factor is "not very", which is a negative semantic tendency. Then this sample is a negative sample. After the trained model inputs a product vector of a computer costing 12,000 yuan, the result obtained should be far away from the result of "cheap".

[0096] Figure 8 Another flowchart of training a target qualitative network in a training method for target model training according to an embodiment of the present disclosure is schematically shown.

[0097] like Figure 8 As shown, based on any of the foregoing embodiments, the sample data further includes sample time information, the sample time information represents the time range corresponding to the first sample qualitative information, and the training method of the target model training may further include operation S810.

[0098] Users' evaluations of some objects are time-sensitive. The same quantitative information may have different qualitative descriptions at different times. A price that was relatively expensive two years ago may be considered moderate this year, and a display with low resolution today may have been considered high-resolution two years ago. To improve the accuracy and representativeness of the sample, the sample data may also include sample time information. The sample time information represents the time range corresponding to the first sample qualitative information. For example, it may be the release time of the comment information corresponding to the first sample qualitative information, or it may be the time information contained in the comment information corresponding to the first sample qualitative information.

[0099] Combination Figure 6 In the illustrated embodiment, multiple time ranges may be pre-divided. In operation S620, multiple qualitative words for the target object are obtained from the comment information for the sample object within the first time range. In operation S630, the multiple qualitative words are clustered to obtain at least one qualitative word collection. In operation S640, the first sample qualitative information corresponding to the qualitative word collection is determined. The first sample vector, the first sample qualitative information, and the first time range are then used as sample data.

[0100] In operation S810, a qualitative model network is trained based on the first sample vector, the first sample qualitative information, and the sample time information to obtain a target qualitative model, and the target qualitative model is used to generate at least one qualitative information for the target object at the current moment based on the feature vector of the product. Since the time information corresponding to the sample data is added during the training process, the qualitative information generated by the target qualitative model also has corresponding time information, and the time information is usually the current moment when the qualitative information is generated or the time range of the current moment.

[0101] Fig. 9 The flowchart of obtaining similar sample data in the training method of the target qualitative model according to the embodiment of the present disclosure is schematically shown.

[0102] According to an embodiment of the present disclosure, based on any of the aforementioned embodiments, the sample data may include original sample data and similar sample data, wherein similar sample objects corresponding to the similar sample data have product correlation with original sample objects corresponding to the original sample data.

[0103] Acquiring sample data may include operations S910 to S920.

[0104] In operation S910, similar sample objects are determined based on the original sample objects. The similar sample objects are different from the original sample objects and have the same target features. Figure 6 In the illustrated embodiment, the original sample object may be a sample object obtained by obtaining a first sample vector through a large model.

[0105] In operation S920, the second sample vector and the first sample qualitative information of the similar sample object are obtained as similar sample data. The second sample vector of the similar sample object is a vector containing the same target feature and at least one different target feature. For example, the original sample object is notebook A, and the corresponding first sample vector contains three target features, the first target feature is type: "gaming notebook", the second target feature is price: "4500 yuan", and the third target feature is number of CPU cores: "6 cores". The second sample vector corresponding to the similar sample object needs to contain at least the same target feature as the first sample vector, and at least one target feature with different content. For example, the similar sample object is notebook B, and the second sample vector contains: type: "gaming notebook", price: "5500 yuan", number of CPU cores: "8 cores". The first sample qualitative information of the similar sample object, the first sample qualitative information here refers to the quantitative feature targeted by the sample qualitative information is the same as the type of the quantitative feature targeted by the sample qualitative information in the original sample data. For example, the sample qualitative information corresponding to the aforementioned notebook A and notebook B are both qualitative information about prices, rather than targeting different target features, such as one is qualitative information about prices and the other is qualitative information about the number of CPU cores.

[0106] Objects with product relevance often have the same or similar evaluation words. For example, the performance evaluation of a series of gaming laptops is often relatively high, and the weight evaluation of a series of business laptops is often relatively light. In the disclosed embodiment, similar sample objects are determined based on product relevance, that is, similar sample objects are different from the original sample objects and have at least one common target feature. For example, the original sample object is notebook A, and the type is "gaming notebook", and the type of notebook B is also "gaming notebook", then notebook B can be a similar sample object of notebook A. Based on this, the coverage of the sample is expanded to make the sample more comprehensive.

[0107] Fig.10 Another flowchart for obtaining similar sample data in the training method of the target qualitative model according to an embodiment of the present disclosure is schematically shown.

[0108] According to an embodiment of the present disclosure, based on any of the aforementioned embodiments, the sample data includes original sample data and similar sample data, wherein the similar sample data and the original sample data have a time correlation; the sample data also includes sample time information, and the sample time information represents the time range corresponding to the first sample qualitative information.

[0109] The sample data also includes sample time information, and the sample time information represents the time range corresponding to the first sample qualitative information. Acquiring similar sample data may include operations S1010 to S1030.

[0110] In operation S1010, comment information of a sample object within a first time range is obtained, where the first time range is different from a second time range corresponding to original sample data of the same sample object.

[0111] In operation S1020, second sample qualitative information for the sample object is generated according to the comment information within the first time range. The specific process of generating the second sample qualitative information is similar to the specific process of generating the first sample qualitative information. Figure 6 The second sample qualitative information is generated by the generation process of the first sample qualitative information in the illustrated embodiment.

[0112] In operation S1030, the first sample vector, the second sample qualitative information, and the sample time information corresponding to the second time range are obtained as similar sample data.

[0113] The first sample qualitative information and the second sample qualitative information are qualitative information of the original sample object in different time dimensions. For example, for the same mobile phone A, its feature vector is the first sample vector, and the evaluation in 2020 is "high resolution", and the evaluation in 2023 is "average resolution". Then, the original sample data corresponding to mobile phone A can be (first sample vector, "high resolution", 2020), and the corresponding similar sample data can be (first sample vector, "average resolution", 2023).

[0114] Fig. 9 and Fig.10 The embodiments shown can also be used in combination. For example, after the original sample data is generated based on the comment information of mobile phone A in time period 1, mobile phone B, a similar sample object of mobile phone A, is obtained, and then the sample data based on the comment information of mobile phone B in time period 1 is obtained as similar sample data. At the same time, the data obtained based on the comment information of mobile phone A in time period 2 is also used as another similar sample data. This further expands the coverage of the sample and makes the sample more comprehensive.

[0115] Fig.11 The block diagram schematically shows a data generating device according to an embodiment of the present disclosure.

[0116] like Fig.11 As shown, the data generating device 1100 includes a first acquisition module 1110 and a first generation module 1120 .

[0117] The first acquisition module 1110 is used to acquire a feature vector of the target object, where the feature vector at least represents quantitative information corresponding to at least one target feature of the target object. In some embodiments, the first acquisition module 1110 can be used to perform the above operation S210, which will not be described in detail herein.

[0118] The first generating module 1120 is used to generate at least one qualitative information for the target object through the target qualitative model according to the feature vector, wherein the qualitative information is information for evaluating the quantitative information of the target object. In some embodiments, the acquiring module 1120 can be used to perform the above operation S220, which is not described in detail here.

[0119] According to an embodiment of the present disclosure, the first generating module may further include a determining module and a second generating module.

[0120] The determination module is used to determine the qualitative words for the target object and the degree words corresponding to the qualitative words according to the feature vector through the target qualitative model. In some embodiments, the determination module can be used to perform the above operation S310, which will not be described in detail here.

[0121] The second generating module is used to generate at least one qualitative information according to the degree word and the qualitative word. In some embodiments, the second generating module can be used to perform the above operation S320, which will not be described in detail here.

[0122] According to an embodiment of the present disclosure, the data generating device may further include a third generating module.

[0123] The third generation module is used to generate the membership degree corresponding to the qualitative information through the target qualitative model according to the feature vector, and the membership degree represents the probability that the target object is evaluated as qualitative information; wherein the membership degrees corresponding to the various qualitative information satisfy a specific relationship. In some embodiments, the third generation module can be used to perform the above operation S410, which is not described in detail here.

[0124] According to an embodiment of the present disclosure, the data generating device may further include a training module for obtaining a target qualitative model, which will not be elaborated here.

[0125] The training module may include a second acquisition module and a model training module.

[0126] The second acquisition module is used to acquire sample data, the sample data including a first sample vector and first sample qualitative information of the sample object, the first sample vector represents quantitative information corresponding to at least one target feature of the sample object, and the first sample qualitative information is information for evaluating the quantitative information of the sample object. In some embodiments, the second acquisition module can be used to perform the above operation S510, which is not described in detail here.

[0127] The model training module is used to train the qualitative model network according to the sample data to obtain the target qualitative model. In some embodiments, the model training module can be used to perform the above operation S520, which will not be described in detail here.

[0128] According to an embodiment of the present disclosure, the second acquisition module may include a third acquisition module, a fourth acquisition module, a second obtaining module, and a first determining module.

[0129] The third acquisition module is used to acquire the first sample vector of the sample object through the large model. In some embodiments, the third acquisition module can be used to perform the above operation S610, which will not be described in detail here.

[0130] The fourth acquisition module is used to acquire multiple qualitative words for the target object from the comment information for the sample object, wherein the qualitative words represent the natural language qualitative evaluation of the target object. In some embodiments, the fourth acquisition module can be used to perform the above operation S620, which will not be described in detail here.

[0131] The second obtaining module is used to cluster the multiple qualitative words to obtain at least one qualitative word collection. In some embodiments, the second obtaining module can be used to perform the above operation S630, which will not be described in detail here.

[0132] The first determination module is used to cluster the multiple qualitative words to obtain at least one qualitative word collection. In some embodiments, the first determination module can be used to perform the above operation S640, which will not be described in detail here.

[0133] According to an embodiment of the present disclosure, the types of sample data include positive sample data and negative sample data, and the second acquisition module may also be a fifth acquisition module or a second determination module.

[0134] The fifth acquisition module is used to acquire the degree word corresponding to the qualitative word from the comment information, and the degree word is a word used to represent the degree in natural language. In some embodiments, the first determination module can be used to perform the above operation S710, which will not be described in detail here.

[0135] The second determination module is used to determine the positive and negative types of the sample data according to the semantic tendency of the degree word. In some embodiments, the second determination module can be used to perform the above operation S720, which will not be described in detail here.

[0136] According to an embodiment of the present disclosure, the sample data also includes sample time information, and the sample time information represents the time range corresponding to the first sample qualitative information. The model training module is also used to train the qualitative model network according to the first sample vector, the first sample qualitative information and the sample time information to obtain a target qualitative model, and the target qualitative model is used to generate at least one qualitative information for the target object at the current moment according to the feature vector of the product. In some embodiments, the model training module can be used to perform the above operation S810, which will not be described in detail here.

[0137] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.

[0138] For example, 1100 includes any multiple of the acquisition module 1110 and the generation module 1120 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present disclosure, 1100 includes at least one of the acquisition module 1110 and the generation module 1120 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in a suitable combination of any of them. Alternatively, at least one of the acquisition module 1110 and the generation module 1120 included in 1100 may be at least partially implemented as a computer program module, and when the computer program module is executed, the corresponding function may be executed.

[0139] Fig.12 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. Fig.12 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0140] like Fig.12As shown, the electronic device 1200 according to an embodiment of the present disclosure includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage part 1208 to a random access memory (RAM) 1203. The processor 1201 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1201 may also include an onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0141] In RAM 1203, various programs and data required for the operation of electronic device 1200 are stored. Processor 1201, ROM 1202 and RAM 1203 are connected to each other via bus 1204. Processor 1201 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1202 and / or RAM 1203. It should be noted that the program can also be stored in one or more memories other than ROM 1202 and RAM 1203. Processor 1201 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0142] According to an embodiment of the present disclosure, the electronic device 1200 may further include an input / output (I / O) interface 1205, which is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the input / output (I / O) interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the input / output (I / O) interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed, so that a computer program read therefrom is installed into the storage portion 1208 as needed.

[0143] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0144] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0145] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. For example, it may include, but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0146] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 1202 and / or the RAM 1203 described above and / or one or more memories other than the ROM 1202 and the RAM 1203 .

[0147] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains a program code for executing the method provided by the embodiment of the present disclosure. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the data generation method provided by the embodiment of the present disclosure.

[0148] When the computer program is executed by the processor 1201, the above functions defined in the system / device of the embodiment of the present disclosure are executed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0149] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0150] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0151] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments of the present disclosure can be combined and / or combined in a variety of ways, even if such a combination or combination is not explicitly recorded in the present disclosure. In particular, without departing from the spirit and teaching of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0152] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A data generation method, comprising: Acquire a feature vector of a target object, wherein the feature vector at least represents quantitative information corresponding to at least one target feature of the target object; Generate at least one qualitative information for the target object through a target qualitative model according to the feature vector; The qualitative information is information for evaluating the quantitative information of the target object.

2. The method according to claim 1, generating at least one qualitative information for the target object according to the feature vector through a target qualitative model, comprising: According to the feature vector, determining the qualitative words for the target object and the degree words corresponding to the qualitative words through the target qualitative model; Generate at least one qualitative information according to the degree word and the qualitative word; The qualitative word represents a natural language qualitative evaluation of at least one of the target features, and the degree word is a word used to represent the degree in natural language.

3. The method according to claim 1, further comprising: According to the feature vector, generating a membership degree corresponding to the qualitative information through the target qualitative model, wherein the membership degree represents the probability that the target object is evaluated as the qualitative information; The membership degrees corresponding to the various qualitative information satisfy a specific relationship.

4. According to the method of claim 1, the target qualitative model is obtained by the following operations: Acquire sample data, the sample data including a first sample vector and first sample qualitative information of a sample object, the first sample vector representing quantitative information corresponding to at least one target feature of the sample object, and the first sample qualitative information being information evaluating the quantitative information of the sample object; A qualitative model network is trained according to the sample data to obtain the target qualitative model.

5. The method according to claim 4, wherein obtaining sample data comprises: Acquire a first sample vector of the sample object through a large model; Acquire a plurality of qualitative words for the target object from the comment information for the sample object, wherein the qualitative words represent a natural language qualitative evaluation of the target object; Clustering the multiple qualitative words to obtain at least one qualitative word collection; Determine the first sample qualitative information corresponding to the qualitative word collection.

6. The method according to claim 5, wherein the types of the sample data include positive sample data and negative sample data, and the acquiring of the sample data includes: Acquire, from the review information, a degree word corresponding to the qualitative word, wherein the degree word is a word used to represent a degree in a natural language; The positive and negative types of the sample data are determined according to the semantic inclination of the degree word.

7. According to the method of claim 4, the sample data further includes sample time information, the sample time information represents the time range corresponding to the first sample qualitative information, and the training qualitative model network comprises: A qualitative model network is trained according to the first sample vector, the first sample qualitative information and the sample time information to obtain a target qualitative model, wherein the target qualitative model is used to generate at least one qualitative information for the target object at the current moment according to a feature vector of a product.

8. The method according to claim 5, wherein the sample data comprises original sample data and similar sample data, wherein: There is product correlation between the similar sample objects corresponding to the similar sample data and the original sample objects corresponding to the original sample data; The similar sample data is obtained by the following operations: Determine a similar sample object according to the original sample object, wherein the similar sample object is different from the original sample object and has the same target feature; A second sample vector of the similar sample object and the first sample qualitative information are obtained as similar sample data.

9. The method according to claim 5, wherein the sample data comprises original sample data and similar sample data, wherein: There is a time correlation between the similar sample data and the original sample data; The sample data also includes sample time information, and the sample time information represents the time range corresponding to the first sample qualitative information. The similar sample data is obtained by the following operations: Acquire comment information of the sample object within a first time range, where the first time range is different from a second time range corresponding to original sample data of the same sample object; generating second sample qualitative information for the sample object according to the comment information within the first time range; The first sample vector, the second sample qualitative information, and sample time information corresponding to the second time range are obtained as the similar sample data.

10. A data generating device, comprising: A first acquisition module is used to acquire a feature vector of a target object, wherein the feature vector at least represents quantitative information corresponding to at least one target feature of the target object; A first generating module, configured to generate at least one qualitative information for the target object according to the feature vector and through a target qualitative model; The qualitative information is information for evaluating the quantitative information of the target object.

Images