Training Sample Generation Method, Data Recommendation Method and Device for Industrial Scenarios

Abstract

Embodiments of this specification provide a method for generating training samples for industrial scenario data recommendation, a method for training a to-be-viewed data recommendation model, a method and device for recommending to-be-viewed data. Using this method and device, by specifically extracting the user's work role information and target behavior information sequence, and generating the feature part in the labeled training sample based on the feature extraction of subsequences of the work role information and target behavior information sequence, and determining the label part in the labeled training sample based on the data targeted by the next target behavior information of the subsequence, the generation of high-quality training samples for industrial scenario data recommendation is achieved. Moreover, the generated training samples can also be used to train the to-be-viewed data recommendation model for recommending to-be-viewed data.

Description

Technical Field

[0001] Embodiments of this specification generally relate to the field of industrial data mining, and in particular, to a method for generating training samples for industrial scenario data recommendation, a method for training a model, a method for recommending data to be viewed, and an apparatus therefor. Background Art

[0002] In the context of intelligent manufacturing and Industry 4.0, data mining technology is increasingly widely used in the manufacturing industry. For example, in a factory production scenario, data mining can be used in aspects such as production processes, operations, fault detection, maintenance, decision support, and product quality improvement. Currently, the scale of industrial data is increasing. For example, the industrial data that needs to be managed in a large chemical plant can include data in the factory construction stage, data in the production operation stage, data in the overhaul and maintenance stage, etc., and each stage contains a large amount of various types of data. Another example is that a digital factory (DF) is also based on a large amount of data.

[0003] Faced with a large amount of data, relevant personnel (such as factory employees) often need to accurately and quickly retrieve the data files to be viewed in scenarios such as production, operation, and overhaul, in order to help respond to various on-site situations in a timely, safe, and correct manner. However, existing data recommendation technologies often recommend information such as products, audio-visuals, and news based on interests, and are not applicable to industrial scenarios. Therefore, a data recommendation technology applicable to industrial scenarios is needed to help relevant personnel obtain the required data more quickly and accurately. Summary of the Invention

[0004] In view of the above, embodiments of this specification provide a method for generating training samples for industrial scenario data recommendation, a method for training a model for recommending data to be viewed, a method for recommending data to be viewed, and an apparatus therefor. By using this method and apparatus, by specifically extracting relevant information of a user to construct the feature part and label part of a training sample, the generation of high-quality training samples for industrial scenario data recommendation is achieved. Moreover, the generated training samples can be used to train a model for recommending data to be viewed for recommending data to be viewed. This helps relevant personnel obtain the required data from a large amount of industrial data more quickly and accurately.

[0005] According to one aspect of the embodiments of the present specification, a method for generating training samples for industrial scenario data recommendation is provided, including: obtaining the work role information of a target user and a sequence of target behavior information, wherein the target behavior information in the sequence of target behavior information includes operation information recorded for a target operation behavior and data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; extracting a subsequence composed of a plurality of sequentially arranged target behavior information from the sequence of target behavior information; performing feature extraction on the work role information and the subsequence to generate corresponding sample features; constructing a corresponding sample label based on the data targeted by the next target behavior information of the subsequence; and obtaining a labeled training sample based on the work role information and the sequence of target behavior information with the sample features and the corresponding sample labels as the feature part and the label part respectively.

[0006] According to another aspect of the embodiments of the present specification, a method for training a data-to-be-viewed recommendation model is provided, including: obtaining the labeled training sample obtained by the aforementioned training sample generation method; and using the feature part of the labeled training sample as the input of the data-to-be-viewed recommendation model, and using the corresponding label part of the labeled training sample as the expected output to train and obtain the data-to-be-viewed recommendation model.

[0007] According to yet another aspect of the embodiments of the present specification, a method for recommending data to be viewed is provided, including: obtaining a labeled training sample set, where the feature part of the labeled training sample is determined based on work role information and a sequence of target behavior information, and the label part of the labeled training sample is determined based on the data targeted by the next target behavior information of the sequence of target behavior information; obtaining the work role information of the logged-in user and the current sequence of target behavior information, wherein the target behavior information in the current sequence of target behavior information includes operation information recorded up to the current time for a target operation behavior and data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; determining the recommended view data corresponding to the work role information and the current sequence of target behavior information of the logged-in user according to the obtained labeled training sample set; and presenting the recommended view data to the logged-in user.

[0008] According to another aspect of the embodiments of the present specification, there is provided a training sample generation device for industrial scenario data recommendation, including: an information acquisition unit configured to acquire the work role information of a target user and a target behavior information sequence, wherein the target behavior information in the target behavior information sequence includes operation information recorded for a target operation behavior and data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; an information extraction unit configured to extract a subsequence composed of several sequentially arranged target behavior information from the target behavior information sequence; a feature extraction unit configured to perform feature extraction on the work role information and the subsequence to generate corresponding sample features; a label construction unit configured to construct corresponding sample labels based on the data targeted by the next target behavior information of the subsequence; and a sample generation unit configured to obtain a labeled training sample with the sample features and the corresponding sample labels as the feature part and the label part respectively.

[0009] According to yet another aspect of the embodiments of the present specification, there is provided a training device for a data to be viewed recommendation model, including: a sample acquisition unit configured to acquire the labeled training samples obtained by the training sample generation method as described above; and a model training unit configured to use the feature part in the labeled training samples as the input of the data to be viewed recommendation model and the corresponding label part in the labeled training samples as the expected output to train and obtain the data to be viewed recommendation model.

[0010] According to still another aspect of the embodiments of the present specification, there is provided a device for recommending data to be viewed, including: a sample set acquisition unit configured to acquire a labeled training sample set, where the feature part of the labeled training sample is determined based on the work role information and the target behavior information sequence, and the label part in the labeled training sample is determined based on the data targeted by the next target behavior information of the target behavior information sequence; a to-be-predicted information acquisition unit configured to acquire the work role information of a logged-in user and the current target behavior information sequence, wherein the target behavior information in the current target behavior information sequence includes the operation information recorded up to the current time for a target operation behavior and the data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; a recommended data determination unit configured to determine the recommended viewing data corresponding to the work role information and the current target behavior information sequence of the logged-in user according to the acquired labeled training sample set; and a data presentation unit configured to present the recommended viewing data to the logged-in user.

[0011] According to another aspect of the embodiments of the present specification, there is provided an electronic device, including: at least one processor, and a memory coupled to the at least one processor, where the memory stores instructions, and when the instructions are executed by the at least one processor, the at least one processor is caused to execute the training sample generation method as described above, or implement the training method as described above, or implement the method for recommending data to be viewed as described above.

[0012] According to yet another aspect of the embodiments of the present specification, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the training sample generation method as described above, or implements the training method as described above, or implements the method for recommending data to be viewed as described above.

[0013] According to another aspect of the embodiments of the present specification, there is provided a computer program product including a computer program, and the computer program is executed by a processor to implement the training sample generation method as described above, or implement the training method as described above, or implement the method for recommending data to be viewed as described above. Description of the Drawings

[0014] By referring to the following drawings, a further understanding of the essence and advantages of the content of the present specification can be achieved. In the drawings, similar components or features may have the same reference numerals.

[0015] Figure 1 An exemplary architecture of a training sample generation method for industrial scenario data recommendation, a training method for a data-to-be-viewed recommendation model, a method for recommending data to be viewed, and an apparatus according to the embodiments of the present specification is shown.

[0016] Figure 2 A flowchart of an example of a training sample generation method for industrial scenario data recommendation according to the embodiments of the present specification is shown.

[0017] Figure 3A A schematic diagram of an example of the construction process of sample labels according to the embodiments of the present specification is shown.

[0018] Figure 3B A schematic diagram of an example of the association relationship between industrial devices according to the embodiments of the present specification is shown.

[0019] Figure 4 A flowchart of another example of a training sample generation method for industrial scenario data recommendation according to the embodiments of the present specification is shown.

[0020] Figure 5 A flowchart of an example of a training method for a data-to-be-viewed recommendation model according to the embodiments of the present specification is shown.

[0021] Figure 6 A flowchart showing an example of a method for recommending data to be viewed according to an embodiment of this specification.

[0022] Figure 7 A flowchart showing an example of a determination process for recommending data to be viewed according to an embodiment of this specification.

[0023] Figure 8 A block diagram showing an example of a training sample generation device for industrial scenario data recommendation according to an embodiment of this specification.

[0024] Figure 9 A block diagram showing an example of a training device for a data-to-be-viewed recommendation model according to an embodiment of this specification.

[0025] Figure 10 A block diagram showing an example of a device for recommending data to be viewed according to an embodiment of this specification.

[0026] Figure 11 A schematic diagram showing an example of an electronic device according to an embodiment of this specification. Detailed implementation manners

[0027] The subject matter described herein will be discussed with reference to example embodiments below. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and is not a limitation on the scope of protection, applicability, or examples set forth in the claims. The functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the content of the embodiments of this specification. Each example may omit, substitute, or add various processes or components as needed. Additionally, the features described relative to some examples may be combined in other examples.

[0028] As used herein, the term "including" and its variants represent open terms, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other definitions may be included below, whether explicit or implicit. Unless explicitly specified in the context, the definition of a term is consistent throughout the specification.

[0029] The flowcharts used in this specification illustrate operations implemented by a system according to some embodiments in this specification. It should be clearly understood that the operations in the flowchart may not be implemented in sequence. On the contrary, the operations may be implemented in a reverse order or simultaneously. In addition, one or more other operations may be added to the flowchart. One or more operations may be removed from the flowchart.

[0030] Next, a method for generating training samples for industrial scenario data recommendation, a method for training a to-be-viewed data recommendation model, a method and apparatus for recommending to-be-viewed data according to embodiments of this specification will be described in detail with reference to the accompanying drawings.

[0031] Figure 1 An exemplary architecture 100 of a method for generating training samples for industrial scenario data recommendation, a method for training a to-be-viewed data recommendation model, a method and apparatus for recommending to-be-viewed data according to embodiments of this specification is shown.

[0032] In Figure 1 it, the network 110 is applied to interconnect the terminal device 120 and the application server 130.

[0033] The network 110 can be any type of network capable of interconnecting network entities. The network 110 can be a single network or a combination of various networks. In terms of coverage, the network 110 can be a local area network (LAN), a wide area network (WAN), etc. In terms of the bearer medium, the network 110 can be a wired network, a wireless network, etc. In terms of data exchange technology, the network 110 can be a circuit-switched network, a packet-switched network, etc.

[0034] The terminal device 120 can be any type of electronic computing device capable of connecting to the network 110, accessing servers on the network 110, processing data or signals, etc. For example, the terminal device 120 can include various desktop computers, laptop computers, smart phones, etc. In one implementation, the terminal device 120 can be used by a user. In some cases, the terminal device 120 can interact with the application server 130. It can be understood that different numbers of terminal devices 120 can be connected to the network 110. In one example, there can be multiple terminal devices 120. In one example, a user can send a data viewing request to the application server 130 through the terminal device 120. The application server 130 can recommend to the user the data files that the user needs to view. In some examples, the application server 130 can also be connected to the database server 140. The database server 140 can be used to store various data files for the application server 130 to recommend to users using terminal devices. In some examples, in the smart factory scenario, the database server 140 can be used to store various data files owned by the factory, such as 3D models of equipment, equipment manuals, process flowcharts, operation specification documents, operation demonstration videos, account books, equipment maintenance records, inventory information, etc. In some examples, the database server 140 can be an enterprise-owned database or a cloud database. It can be understood that the function of the database server 140 can also be integrated into the application server 130, which is not limited here.

[0035] In some examples, the application server 130 can also be connected to the model training server 150. In some examples, the method for generating training samples for industrial scenario data recommendation can be executed by any one of the application server 130, the database server 140, and the model training server 150, or can be executed by other servers. In some examples, the method for training the data recommendation model to be viewed can be executed by the model training server 150. In some examples, the method for recommending data to be viewed can be executed by the application server 130.

[0036] It should be understood that Figure 1 all the network entities shown are exemplary, and according to specific application requirements, any other network entities can be involved in the architecture 100.

[0037] Figure 2 A flowchart of an example 200 of a method for generating training samples for industrial scenario data recommendation according to an embodiment of the present specification is shown.

[0038] As Figure 2 shown, at 210, obtain the work role information and the target behavior information sequence of the target user.

[0039] In this embodiment, the target behavior information in the target behavior information sequence may include the recorded operation information for the target operation behavior and the data viewing information for the target data viewing behavior. The data targeted by the target data viewing behavior may include data associated with industrial equipment and / or industrial processes. In some examples, the work role information may be used to indicate operation personnel, maintenance personnel, management personnel, etc. In some examples, the work role information of the target user may be obtained through the relevant information of the logged-in user, and the target behavior information sequence may be obtained using data buried point technology. In some examples, the target user may be operation personnel, maintenance personnel, management personnel, etc. with rich experience in the corresponding industrial scenario, which helps to obtain high-quality training samples.

[0040] In some examples, the target operation behavior may be the login behavior for the data file recommendation system running on Figure 1 the application server 130 shown in the figure, and the recorded operation information may be the relevant information of the logged-in account. In some examples, the relevant information of the account may include work role information, length of service, etc. In some examples, the above data file recommendation system may also be integrated into the office automation system (OA) of enterprises, factories, etc.

[0041] In some examples, the target operation behavior may include the identification acquisition behavior and / or status switching behavior for industrial equipment. For example, each industrial equipment may have a unique identification code, and the target operation behavior may include the scanning operation for the unique identification code of the industrial equipment. For another example, the target operation behavior may include the status switching behavior of industrial equipment, such as the switching between the running and stopping states of generators, pumps, blowers, etc., and the change or switching of the opening degree or opening and closing state of valves, etc.

[0042] In some examples, the data associated with industrial equipment may include at least one of the following: three-dimensional model files of industrial equipment, images related to industrial equipment, videos related to industrial equipment, document manuals related to industrial equipment. For example, the data associated with industrial equipment may include at least one of the following: three-dimensional model files of complex industrial equipment (such as in.fbx,.obj, etc. formats), various photos of industrial equipment, introduction videos, inspection videos, maintenance videos, teaching videos of industrial equipment, various operation manuals, design documents, instruction documents, maintenance records, and archived files of industrial equipment.

[0043] In some examples, the data associated with an industrial process may include at least one of the following: images related to the industrial process, videos related to the industrial process, document manuals related to the industrial process. For example, the data associated with an industrial process may include at least one of the following: various explanatory pictures related to the industrial process, demonstration pictures, explanatory videos related to the industrial process, demonstration videos, teaching videos, recorded archived videos, various operation specifications related to the industrial process, explanatory documents, operation manuals, archived documents.

[0044] In some examples, the data associated with the industrial equipment and / or the data associated with the industrial process as described above may specifically include at least one of the following: certain chapters, paragraphs in the document, certain video segments in the video file, etc.

[0045] At 220, a subsequence composed of a number of target behavior information arranged in sequence is extracted from the target behavior information sequence.

[0046] In this embodiment, the arrangement order of the target behavior information in the subsequence is generally the same as the arrangement order in the target behavior information sequence. In some examples, the target behavior information sequence may be "view the design document of barrel A, scan the device code of pump A, scan the device code of pipeline A, view the historical maintenance report of pipeline A, view the operation manual of pump B". In one example, the subsequence may be "view the design document of barrel A, scan the device code of pump A, scan the device code of pipeline A, view the historical maintenance report of pipeline A". In one example, the subsequence may be "scan the device code of pump A, scan the device code of pipeline A".

[0047] At 230, feature extraction is performed on the work role information and the subsequence to generate corresponding sample features.

[0048] In this embodiment, the work role information and the subsequence can be respectively converted into corresponding features. In some examples, various feature extraction methods can be used to achieve the conversion of the corresponding features. Then, sample features can be generated based on the features respectively corresponding to the obtained work role information and the subsequence of the target behavior information sequence. In some examples, the features respectively corresponding to the obtained work role information and the subsequence of the target behavior information sequence can be combined into sample features. In some examples, the obtained work role information and the subsequence of the target behavior information sequence or their respective corresponding features can be input into a pre-trained first feature embedding layer, and the output of the obtained first feature embedding layer is used as the sample feature.

[0049] In some examples, the obtained job role information can be "operation staff", and x0 can be used to represent the features corresponding to "operation staff". The feature sequence corresponding to the subsequence "View the design document of barrel A, scan the device code of pump A, scan the device code of pipeline A, view the historical maintenance report of pipeline A" can be [x1, x2, x3, x4]. Among them, x1 to x4 can be used to represent the features corresponding to viewing the design document of barrel A, scanning the device code of pump A, scanning the device code of pipeline A, and viewing the historical maintenance report of pipeline A respectively. In one example, the corresponding sample features can be {x0, x1, x2, x3, x4}. In one example, the corresponding sample features can also be f(x0, x1, x2, x3, x4), where f() can be used to represent the above-mentioned first feature embedding layer. Correspondingly, the feature sequence corresponding to the subsequence "Scan the device code of pump A, scan the device code of pipeline A" can be [x2, x3]. In one example, the corresponding sample features can be {x0, x2, x3}. In one example, the corresponding sample features can also be f(x0, x2, x3).

[0050] At 240, construct corresponding sample labels based on the data targeted by the next target behavior information of the subsequence.

[0051] In this embodiment, it is possible to first determine the data targeted by the next target behavior information of the subsequence. In some examples, taking the example described in step 220 of the foregoing Figure 2 embodiment as an example, the data targeted by the next target behavior information of the subsequence (such as "View the operation manual of pump B") can be "the operation manual of pump B". In some examples, feature extraction can be performed on the data targeted by the next target behavior information of the subsequence, and the obtained features can be used as sample labels. This embodiment does not limit the way of feature extraction. In some examples, the data targeted by the next target behavior information of the subsequence can be mapped to a specified category, and the corresponding specified category can be used as a sample label. For example, map "the operation manual of pump B" to "category 002003", map "the historical maintenance report of pipeline A" to "category 001012", etc.

[0052] At 250, use the sample features and the corresponding sample labels as the feature part and the label part respectively to obtain labeled training samples.

[0053] In one example, taking the foregoing Figure 2Taking the example described in step 220 of the embodiment as an example, a labeled training sample can be expressed as {x0, x1, x2, x3, x4|l5}. Among them, x0 can be used to represent the feature corresponding to "operation personnel". {x0, x1, x2, x3, x4} can be used as the feature part of this labeled training sample, and l5 can be used as the label part of this labeled training sample. In some examples, l5 can be the feature corresponding to "operation manual of pump B". In some examples, l5 can be the specified category mapped by "operation manual of pump B". Similarly, a labeled training sample can be expressed as {x0, x2, x3|l4}. The meanings of the relevant symbols can be referred to the foregoing. {x0, x2, x3} can be used as the feature part of this labeled training sample, and l4 can be used as the label part of this labeled training sample. In some examples, l4 can be the feature corresponding to "historical maintenance report of pipeline A". In some examples, l4 can be the specified category mapped by "historical maintenance report of pipeline A".

[0054] Figure 3A FIG. shows a schematic diagram of an example of the construction process 300 of sample labels according to an embodiment of the present specification. As Figure 3A shown, the association relationship can be determined between the industrial equipment and / or industrial process involved in the data targeted by the next target behavior information of the subsequence of the above target behavior information sequence and the industrial equipment and / or industrial process involved in the data targeted by the target behavior information in the above subsequence. Then, based on the data targeted by the next target behavior information and the determined association relationship, the label part in the labeled training sample is constructed.

[0055] In some examples, still taking the foregoing Figure 2Taking the example described in step 220 of the embodiment as an example, the industrial equipment (such as pump B) related to the data targeted by the next target behavior information (such as "view the operation manual of pump B") of the subsequence of the above target behavior information sequence (such as "view the design document of barrel A", "scan the device code of pump A", "scan the device code of pipeline A", "view the historical maintenance report of pipeline A") can be determined, and the correlation relationship between the industrial equipment (such as barrel A, pump A, pipeline A) related to the data targeted by the target behavior information in the above subsequence can be determined. In some examples, the correlation relationship between industrial equipment and / or industrial processes can be determined based on the analysis of all data files that can be recommended by the aforementioned data file recommendation system. For example, the correlation relationship between industrial equipment, such as direct connection relationship, indirect connection relationship, substitution relationship (such as multiple-path backup), etc., can be determined according to the equipment connection diagram, equipment layout diagram, etc. For another example, the correlation relationship between industrial processes, such as direct upstream and downstream relationship, indirect upstream and downstream relationship, substitution relationship, etc., can be determined according to the operation procedure, process specification manual, etc. In some examples, feature extraction or category mapping can be performed on the above correlation relationship to obtain relationship labels. The above operations can refer to the corresponding processing process of the data targeted by the next target behavior information. In some examples, the corresponding feature extraction result or category mapping result of the data targeted by the next target behavior information can be combined with the above relationship labels, so as to construct sample labels.

[0056] Figure 3B FIG. shows a schematic diagram of an example of the correlation relationship between industrial equipment according to an embodiment of the present specification. As Figure 3B shown, pump A is connected to barrel A through pipeline A. As a backup, pump B is connected to barrel A through pipeline B. Thus, it can be determined that the correlation relationships between pump B and barrel A, pump A, and pipeline A are direct connection relationship, substitution relationship, and indirect connection relationship, respectively. After that, the label part of the labeled training sample can be generated based on the data targeted by the next target behavior information and the determined correlation relationship. In some examples, the features corresponding to the data targeted by the next target behavior information and the determined correlation relationship can be combined into the label part of the labeled training sample. For example, the features corresponding to the direct connection relationship, indirect connection relationship, substitution relationship, direct upstream and downstream relationship, and indirect upstream and downstream relationship can be 11, 12, 13, 14, and 15, respectively. In some examples, the data targeted by the next target behavior information and the determined correlation relationship can be input into a pre-trained second feature embedding layer, and the output result of the second feature embedding layer can be used as the label part of the labeled training sample.

[0057] Figure 4The flowchart shows another example of a training sample generation method 400 for industrial scenario data recommendation according to an embodiment of the present specification.

[0058] As Figure 4 shown, at 410, obtain the job role information and the target behavior information sequence of the target user.

[0059] At 420, extract a subsequence composed of a plurality of target behavior information arranged in sequence from the target behavior information sequence.

[0060] At 430, obtain the additional status information corresponding to the target behavior information in the subsequence.

[0061] In this embodiment, the additional status information can be used to indicate the additional status at a specific moment. In some examples, the above specific moment can be the moment when the behavior indicated by each target behavior information in the target behavior information sequence occurs. The additional status information can be various status information associated with the next target behavior information. In some examples, the additional status information can be the status information of the industrial equipment targeted by the target behavior information, spare part status information, etc.

[0062] In some examples, the type of the additional status information can be determined according to the job role information and the operation information and data viewing information in the subsequence before the current target behavior information. In the example as described above, if the feature sequence corresponding to the target behavior information sequence is [x1, x2, x3, x4, x5], then the type of the additional status information corresponding to the current target behavior information (such as the information of viewing the operation manual of pump B corresponding to x4) can be determined according to the job role information (such as "operator"), the operation information corresponding to x1 (such as the information of recording and scanning the equipment code of pump A), the operation information corresponding to x2 (such as the information of recording and scanning the equipment code of pipeline A), and the data viewing information corresponding to x3 (such as the information of recording and viewing the historical maintenance report of pipeline A) (for example, the current air pressure of barrel A, the current power of pump A, etc.). Similarly, the type of the additional status information corresponding to the current target behavior information (such as the information of viewing the historical maintenance report of pipeline A corresponding to x3) can also be determined according to the job role information, the operation information corresponding to x1, and the operation information corresponding to x2 (such as the working status of pipeline A, etc.).

[0063] In some examples, the correspondence between the type of additional status information and the operation information and data viewing information in the subsequence before the work role information and the current target behavior information can be determined by a predetermined rule, can also be determined by a pre-set correspondence table, or can also be determined according to a pre-trained type determination model. In some examples, the above-mentioned type determination model can output the type of additional status information corresponding to the current target behavior information according to the input work role information, operation information and data viewing information in the subsequence before the current target behavior information.

[0064] At 440, feature extraction is performed on the obtained work role information, subsequence and corresponding additional status information to generate corresponding sample features.

[0065] In this embodiment, feature extraction of the additional status information can be further added on the basis of the feature extraction of the work role information and the subsequence as described above. In some examples, as in the foregoing examples, the sample features can be expressed as {x0, x1, x2, x3, st3, x4, st4}, where the meanings of x0 to x4 can be referred to the foregoing. st3 and st4 can be used as the features of the additional status information corresponding to the target behavior information corresponding to x3 and x4 respectively. Similarly, in one example, the sample features can be expressed as {x0, x2, x3, st3}. The meanings of the relevant symbols can be referred to the foregoing.

[0066] At 450, a corresponding sample label is constructed based on the data targeted by the next target behavior information of the subsequence.

[0067] At 460, using the sample features and the corresponding sample labels as the feature part and the label part respectively, a labeled training sample is obtained.

[0068] In some examples, the labeled training sample can be expressed as {x0, x1, x2, x3, st3, x4, st4|l5}. {x0, x1, x2, x3, st3, x4, st4} can be used as the feature part of the labeled training sample, and l5 can be used as the label part of the labeled training sample. Similarly, in one example, the labeled training sample can be expressed as {x0, x2, x3, st3|l4}. Among them, the meanings of the relevant symbols can be referred to the foregoing. {x0, x2, x3, st3} can be used as the feature part of the labeled training sample, and l4 can be used as the label part of the labeled training sample.

[0069] It should be noted that the above steps 410, 420, 450, 460 can refer to the relevant descriptions of steps 210, 220, 240, 250 in the foregoing embodiments.

[0070] In the above manner, by introducing additional state information, the dimension of the feature part of the training samples is further enriched, which helps to improve the training effect of the model.

[0071] Figure 5 FIG. 4 shows a flowchart of an example of a training method 500 for a data-to-be-viewed recommendation model according to an embodiment of the present specification.

[0072] As Figure 5 shown, at 510, obtain labeled training samples obtained according to a training sample generation method.

[0073] In this embodiment, the training sample generation method may refer to the relevant description of the foregoing Figures 2-4 embodiment. In this embodiment, a training sample set composed of multiple labeled training samples may be obtained.

[0074] At 520, use the feature part in the labeled training samples as the input of the data-to-be-viewed recommendation model, and use the corresponding label part in the labeled training samples as the expected output, and train to obtain the data-to-be-viewed recommendation model.

[0075] In this embodiment, the data-to-be-viewed recommendation model may be various machine learning models (such as a multi-layer perceptron, a neural network model, a deep learning model, etc.). In some examples, the prediction result output by the current data-to-be-viewed recommendation model according to the feature part in the input labeled training samples may be compared with the corresponding label part, the loss value may be calculated using a loss function, and then the model parameters of the current data-to-be-viewed recommendation model may be adjusted according to the calculated loss value.

[0076] In some examples, there may be more than one sub-label in the label part of the labeled training samples. For example, a first sub-label for indicating the data to be recommended and a second sub-label for indicating the association relationship between the industrial equipment and / or industrial process involved in the data to be recommended and the data involved in the target behavior information corresponding to the corresponding feature part. In these examples, the data-to-be-viewed recommendation model may use each sub-label in the label part as the expected output, calculate a comprehensive loss value according to the prediction results corresponding to each sub-label output, and then adjust the model parameters of the current data-to-be-viewed recommendation model according to the calculated comprehensive loss value. Thus, the data-to-be-viewed recommendation model can learn the association relationship between industrial equipment and / or industrial processes, so as to more accurately achieve high-quality data recommendation for industrial scenarios, and reduce the time waste of relevant personnel in searching for required information in a large number of data files.

[0077] Utilize Figures 1-5The training sample generation method for industrial scenario data recommendation and the training method of the data to be viewed recommendation model disclosed in [reference], by specifically extracting relevant information of users to construct the feature part and label part of the training samples, realizes the generation of high-quality training samples for industrial scenario data recommendation. Moreover, the generated training samples can be used to train the data to be viewed recommendation model for recommending the data to be viewed.

[0078] Figure 6 FIG. 4 shows a flowchart of an example of a method 600 for recommending data to be viewed according to an embodiment of the present specification.

[0079] As Figure 6 shown, at 610, a labeled training sample set is obtained.

[0080] In this embodiment, the feature part of the labeled training samples in the labeled training sample set can be determined based on the work role information and the target behavior information sequence, and the label part can be determined based on the data targeted by the next target behavior information of the target behavior information sequence.

[0081] In some examples, the labeled training sample set can be obtained according to the training sample generation method described in Figures 2-4 the embodiment.

[0082] At 620, the work role information and the current target behavior information sequence of the logged-in user are obtained.

[0083] In this embodiment, the target behavior information in the current target behavior information sequence may include the operation information for the target operation behavior and the data viewing information for the target data viewing behavior recorded up to the current time. The data targeted by the target data viewing behavior may include data associated with industrial equipment and / or industrial processes.

[0084] In this embodiment, the logged-in user may be the user to whom the data recommendation is to be made. The specific meanings and explanations of the above work role information and current target behavior information sequence may refer to the relevant descriptions in Figures 2-4 the embodiment.

[0085] At 630, according to the obtained labeled training sample set, the recommended viewing data corresponding to the work role information and the current target behavior information sequence of the logged-in user is determined.

[0086] In some examples, the work role information and the current target behavior information sequence of the obtained logged-in user can be input into the data to be viewed recommendation model to obtain the recommended viewing data.

[0087] In these examples, the data to-be-viewed recommendation model can be trained using a labeled training sample set, and it can select recommended view data from candidate data to-be-viewed according to the input work role information of the logged-in user and the current target behavior information sequence. In some examples, the above candidate data to-be-viewed can be all data files that can be recommended by the data file recommendation system as described above. In some examples, the data to-be-viewed recommendation model can be obtained according to the training method described in Figure 5 the embodiment.

[0088] In some examples, the above candidate data to-be-viewed can also be data that matches the work role information selected from all data files that can be recommended by the data file recommendation system as described above. In some examples, for the same industrial device (such as device p), for "operators", the candidate data to-be-viewed can include the design document, 3D model file, operation manual, etc. of the industrial device; while for "maintenance personnel", the candidate data to-be-viewed can include the design document, maintenance record, spare part inventory, maintenance manual, etc. of the industrial device. In these examples, all data files that can be recommended by the data file recommendation system as described above can be classified in advance to distinguish the work roles that each data file matches. It can be understood that in these examples, a data file can belong to multiple categories, for example, it can be recommended to both "operators" and "managers". Thus, for each work role information, candidate data to-be-viewed that matches the work role information can be selected from all data files that can be recommended by the data file recommendation system as described above.

[0089] Figure 7 A flowchart showing an example of the determination process 700 of recommended view data according to an embodiment of the present specification is shown.

[0090] As Figure 7 shown, at 710, according to the determination method of the feature part of the labeled training sample, the work role information of the logged-in user and the current target behavior information sequence obtained are converted into corresponding current user features.

[0091] In this embodiment, the method of converting the obtained work role information and the current target behavior information sequence of the logged-in user into the current features of the user may be the same as the method of determining the feature part of the labeled training samples based on the work role information and the target behavior information sequence. In some examples, the work role information may be converted into corresponding features by using the correspondence table between work roles and feature encodings, and the method of feature extraction may be used to implement the conversion of the features corresponding to the current target behavior information sequence, so that the current features of the user can be obtained based on the features corresponding to the work role information and the features corresponding to the current target behavior information sequence. For example, the current features of the user can be directly combined. For another example, the pre-trained first feature embedding layer may be used to output the current features of the user according to the features corresponding to the input work role information and the features corresponding to the current target behavior information sequence.

[0092] At 720, based on the k-Nearest Neighbor (KNN) algorithm, the recommended view data is determined from the data indicated by the label part in the labeled training sample set.

[0093] In this embodiment, the labeled training samples to which the K feature parts closest to the current features of the user belong may be determined from the labeled training sample set. Then, the recommended view data is determined from the data indicated by the label parts of the determined K labeled training samples.

[0094] In some examples, based on the voting mechanism, the data with the highest number of votes or the top few (e.g., 3) data with the highest number of votes among the data indicated by the label parts of the K labeled training samples may be determined as the recommended view data.

[0095] In some examples, the recommended view data can be further determined by combining the correspondence between the data indicated by the tag part and the job role information. In these examples, a correspondence table between the data and the job role information can be obtained in advance. The correspondence table can be used to indicate the degree of association between the candidate data and each job role information. For example, the degrees of association between the maintenance manual of a certain device and the operation personnel, maintenance personnel, and management personnel can be 0.3, 1, and 0.3 respectively; while the degrees of association between the operation manual of the device and the operation personnel, maintenance personnel, and management personnel can be 1, 0.1, and 0.4 respectively. In some examples, the recommended view data can be determined by combining the voting situation and the above correspondence table. In one example, if 3 / 5 in the voting result (K = 5) indicates a recommendation for the maintenance manual of device M, 2 / 5 indicates a recommendation for the operation manual of device M, and the job role of the logged-in user is an operation personnel, then the operation manual of device M (2 / 5 × 1 = 0.4) rather than the maintenance manual of device M (3 / 5 × 0.3 = 0.18) can be determined as the recommended view data. It can be understood that the voting situation and the above correspondence table can also be combined to determine more than one recommended view data sorted from high to low according to the comprehensive score.

[0096] Back to Figure 6 , at 640, present the recommended view data to the logged-in user.

[0097] In this embodiment, the at least one recommended view data determined can be presented to the logged-in user in various ways (such as in list form, pop-up form, etc.). In some examples, if the recommended view data is a document, video, 3D model, etc., the storage address of the recommended view data, such as the Uniform Resource Identifier (URI), access link, etc., can be provided to the logged-in user. In some examples, if the recommended view data is a specific part of a document, video, etc., an identifier for indicating the above specific part can also be provided to the logged-in user.

[0098] In some examples, after presenting the recommended view data to the logged-in user, it can also be detected whether there is negative feedback from the logged-in user regarding the recommended view data. In response to detecting negative feedback from the logged-in user regarding the recommended view data, difficult samples can also be generated based on the obtained job role information and the current target behavior information sequence of the logged-in user to further train the recommended model for view data.

[0099] In these examples, negative feedback may include, for example, not opening the file where the recommended view data is located, or the viewing time of the recommended view data being less than a predetermined time threshold (e.g., 5 seconds). In some examples, the feature part of the above-mentioned difficult samples may be determined based on the work role information of the logged-in user and the current target behavior information sequence, and the label part of the above-mentioned difficult samples may be determined based on the next view data of the logged-in user. In some examples, the next view data of the logged-in user may be the data actually viewed by the logged-in user after being recommended the above-mentioned recommended view data. In some examples, the next view data of the logged-in user may be the data that should be viewed by the logged-in user determined after manual evaluation.

[0100] Using Figures 6-7 the method for recommending data to be viewed disclosed in, by obtaining the work role information and the current target behavior information sequence of the logged-in user, and using the corresponding labeled training sample set, provides a more convenient and faster KNN method or a method for more accurately and efficiently determining the recommended view data by using the trained data-to-be-viewed recommendation model.

[0101] Figure 8 FIG. shows a block diagram of an example 800 of a training sample generation device for industrial scenario data recommendation according to an embodiment of the present specification. This device embodiment may correspond to Figures 2-4 the method embodiment shown, and this device may be specifically applied to various electronic devices.

[0102] As Figure 8 shown, the training sample generation device 800 for industrial scenario data recommendation may include an information acquisition unit 810, an information extraction unit 820, a feature extraction unit 830, a label construction unit 840, and a sample generation unit 850.

[0103] The information acquisition unit 810 is configured to acquire the work role information and the target behavior information sequence of the target user, where the target behavior information in the target behavior information sequence includes the operation information recorded for the target operation behavior and the data viewing information for the target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes.

[0104] In some examples, the target operation behaviors include identification acquisition behaviors and / or status switching behaviors for industrial devices, and the data associated with the industrial devices includes at least one of the following: three-dimensional models of the industrial devices, images related to the industrial devices, videos related to the industrial devices, and documentation manuals related to the industrial devices. The data associated with the industrial processes includes at least one of the following: images related to the industrial processes, videos related to the industrial processes, and documentation manuals related to the industrial processes.

[0105] In some examples, the information acquisition unit 810 is further configured to: acquire additional status information corresponding to the target behavior information in the subsequence.

[0106] The information extraction unit 820 is configured to extract a subsequence composed of a plurality of target behavior information arranged in sequence from the target behavior information sequence.

[0107] The feature extraction unit 830 is configured to perform feature extraction on the work role information and the subsequence to generate corresponding sample features.

[0108] In some examples, the feature extraction unit 830 is further configured to: perform feature extraction on the work role information, the subsequence, and the corresponding additional status information to generate corresponding sample features.

[0109] The label construction unit 840 is configured to construct a corresponding sample label based on the data targeted by the next target behavior information of the subsequence.

[0110] In some examples, the label construction unit 840 is further configured to: determine the association relationship between the industrial devices and / or industrial processes involved in the data targeted by the next target behavior information of the subsequence and the industrial devices and / or industrial processes involved in the data targeted by the target behavior information in the subsequence; and construct the sample label based on the data targeted by the next target behavior information and the determined association relationship.

[0111] The sample generation unit 850 is configured to obtain labeled training samples with the sample features and the corresponding sample labels as the feature part and the label part respectively.

[0112] Figure 9 A block diagram showing an example of a training device 900 for a data recommendation model to be viewed according to an embodiment of this specification is shown. This device embodiment can correspond to Figure 5 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0113] As Figure 9As shown, the training device 900 of the data-to-be-viewed recommendation model may include a sample acquisition unit 910 and a model training unit 920.

[0114] The sample acquisition unit 910 is configured to acquire labeled training samples.

[0115] The model training unit 920 is configured to use the feature part in the labeled training samples as the input of the data-to-be-viewed recommendation model, and use the corresponding label part in the labeled training samples as the expected output, and train to obtain the data-to-be-viewed recommendation model.

[0116] Figure 10 FIG. shows a block diagram of an example of a device 1000 for recommending data to be viewed according to an embodiment of the present specification. This device embodiment may correspond to Figures 6-7 the method embodiment shown, and this device may be specifically applied to various electronic devices.

[0117] As Figure 10 shown, the device 1000 for recommending data to be viewed may include a sample set acquisition unit 1010, a to-be-predicted information acquisition unit 1020, a recommended data determination unit 1030, and a data presentation unit 1040.

[0118] The sample set acquisition unit 1010 is configured to acquire a labeled training sample set. The feature part of the labeled training sample is determined based on the work role information and the target behavior information sequence, and the label part in the labeled training sample is determined based on the data targeted by the next target behavior information of the target behavior information sequence.

[0119] The to-be-predicted information acquisition unit 1020 is configured to acquire the work role information and the current target behavior information sequence of the logged-in user. Among them, the target behavior information in the current target behavior information sequence includes the operation information for the target operation behavior and the data viewing information for the target data viewing behavior recorded up to the current time. The data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes.

[0120] The recommended data determination unit 1030 is configured to determine the recommended viewing data corresponding to the work role information and the current target behavior information sequence of the logged-in user according to the acquired labeled training sample set.

[0121] In some examples, the recommended data determination unit 1030 is further configured to: convert the obtained work role information of the logged-in user and the current target behavior information sequence into corresponding current user features in the same way as the determination of the feature part of the labeled training samples; and determine the recommended view data from the data indicated by the label part in the labeled training sample set by means of the K-nearest neighbor algorithm.

[0122] In some examples, the recommended data determination unit 1030 is further configured to: input the obtained work role information of the logged-in user and the current target behavior information sequence into the to-be-viewed data recommendation model to obtain the recommended view data, and the to-be-viewed data recommendation model is trained by using the labeled training sample set.

[0123] The data presentation unit 1040 is configured to present the recommended view data to the logged-in user.

[0124] In some examples, the apparatus 1000 for recommending to-be-viewed data may further include: a difficult sample generation unit 1050, configured to generate difficult samples based on the obtained work role information of the logged-in user and the current target behavior information sequence in response to detecting negative feedback from the logged-in user for the recommended view data, so as to further train the to-be-viewed data recommendation model.

[0125] It should be noted that Figures 8-10 The operations of the respective units in the described training sample generation apparatus for industrial scenario data recommendation, the training apparatus for the to-be-viewed data recommendation model, and the apparatus for recommending to-be-viewed data may refer to the descriptions of the corresponding steps above Figures 2-7 in the corresponding steps.

[0126] The above refers to Figures 1 to 10 to describe the embodiments of the training sample generation method for industrial scenario data recommendation, the training method for the to-be-viewed data recommendation model, the method and apparatus for recommending to-be-viewed data according to the embodiments of this specification.

[0127] The training sample generation apparatus for industrial scenario data recommendation, the training apparatus for the to-be-viewed data recommendation model, and the apparatus for recommending to-be-viewed data according to the embodiments of this specification may be implemented in hardware, or may be implemented by software or a combination of hardware and software. Taking software implementation as an example, as a logically meaningful apparatus, it is formed by the processor of the device where it is located reading the corresponding computer program instructions in the memory into the memory for operation. In the embodiments of this specification, the training sample generation apparatus for industrial scenario data recommendation, the training apparatus for the to-be-viewed data recommendation model, and the apparatus for recommending to-be-viewed data may be implemented by using an electronic device, for example.

[0128] Figure 11 A schematic diagram showing an example of an electronic device 1100 according to an embodiment of this specification is shown.

[0129] As Figure 11 shown, the electronic device 1100 may include at least one processor 1110, a memory (e.g., non-volatile memory) 1120, a memory 1130, and a communication interface 1140, and the at least one processor 1110, the memory 1120, the memory 1130, and the communication interface 1140 are connected together via a bus 1150. The at least one processor 1110 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above elements implemented in software form).

[0130] In one embodiment, computer-executable instructions are stored in a memory, which when executed cause at least one processor 1110 to: obtain the job role information and the target behavior information sequence of a target user, wherein the target behavior information in the target behavior information sequence includes the operation information recorded for a target operation behavior and the data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; extract a subsequence composed of several target behavior information arranged in sequence from the target behavior information sequence; perform feature extraction on the job role information and the subsequence to generate corresponding sample features; construct a corresponding sample label based on the data targeted by the next target behavior information of the subsequence; and obtain a labeled training sample based on the job role information and the target behavior information sequence with the sample features and the corresponding sample labels as the feature part and the label part respectively; or cause at least one processor 1110 to: obtain a labeled training sample obtained by the aforementioned training sample generation method; and use the feature part in the labeled training sample as the input of the to-be-viewed data recommendation model, and use the corresponding label part in the labeled training sample as the expected output to train the to-be-viewed data recommendation model; or cause at least one processor 1110 to: obtain a set of labeled training samples, where the feature part of the labeled training sample is determined based on the job role information and the target behavior information sequence, and the label part in the labeled training sample is determined based on the data targeted by the next target behavior information of the target behavior information sequence; obtain the job role information and the current target behavior information sequence of the logged-in user, wherein the target behavior information in the current target behavior information sequence includes the operation information recorded so far for a target operation behavior and the data viewing information for a target data viewing behavior, and the data targeted by the target data viewing behavior includes data associated with industrial equipment and / or industrial processes; determine the recommended view data corresponding to the job role information and the current target behavior information sequence of the logged-in user according to the obtained set of labeled training samples; and present the recommended view data to the logged-in user.

[0131] It should be understood that the computer-executable instructions stored in the memory, when executed, cause at least one processor 1110 to perform the various operations and functions described above in the respective embodiments of this specification in combination with Figures 1-7 the description.

[0132] According to one embodiment, a program product such as a computer-readable medium is provided. The computer-readable medium may have instructions (i.e., the elements implemented in software as described above), which when executed by a computer, cause the computer to perform the various operations and functions described above in the respective embodiments of this specification in combination with Figures 1-7 the description.

[0133] Specifically, a system or device equipped with a readable storage medium can be provided, on which software program code for implementing the functions of any one of the above embodiments is stored, and the computer or processor of the system or device is caused to read and execute the instructions stored in the readable storage medium.

[0134] In this case, the program code read from the readable medium itself can implement the functions of any one of the above embodiments, so the machine-readable code and the readable storage medium storing the machine-readable code constitute a part of the present invention.

[0135] The computer program code required for the operations of each part of this specification can be written in any one or more programming languages, including object-oriented programming languages such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB, NET, and Python, conventional procedural programming languages such as C language, Visual Basic 2003, Perl, COBOL 2002, PHP, and ABAP, dynamic programming languages such as Python, Ruby, and Groovy, or other programming languages. This program code can run on the user's computer, or run on the user's computer as an independent software package, or part of it runs on the user's computer and another part runs on a remote computer, or all of it runs on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer in any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service, such as software as a service (SaaS).

[0136] Examples of the readable storage medium include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROM. Optionally, the program code can be downloaded from a server computer or the cloud via a communication network.

[0137] The above has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require the specific order or consecutive order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0138] Not all steps and units in the above-mentioned processes and system structure diagrams are necessary, and some steps or units can be ignored according to actual needs. The execution order of each step is not fixed and can be determined according to needs. The device structures described in the above embodiments can be physical structures or logical structures, that is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities respectively, or some components in multiple independent devices may be jointly implemented.

[0139] The term "exemplary" used throughout this specification means "serving as an example, instance, or illustration" and does not mean "preferred" or "advantageous" over other embodiments. For the purpose of providing an understanding of the described technology, the detailed description includes specific details. However, the technology can be implemented without these specific details. In some instances, well-known structures and devices are shown in block diagram form to avoid obscuring the concepts of the described embodiments.

[0140] The optional implementation manners of the embodiments of this specification have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of this specification are not limited to the specific details in the above implementation manners. Within the scope of the technical concept of the embodiments of this specification, various simple modifications can be made to the technical solutions of the embodiments of this specification, and these simple modifications all fall within the protection scope of the embodiments of this specification.

[0141] The above description of the content of this specification is provided to enable any ordinary person skilled in the art to implement or use the content of this specification. For ordinary persons skilled in the art, various modifications to the content of this specification are obvious, and the general principles defined herein can also be applied to other variations without departing from the protection scope of the content of this specification. Therefore, the content of this specification is not limited to the examples and designs described herein, but is consistent with the broadest scope that conforms to the principles and novel features disclosed herein.

Claims

1. A method for generating training samples for industrial scenario data recommendation, comprising: Obtaining the work role information and the target behavior information sequence of a target user, wherein the target behavior information in the target behavior information sequence includes the operation information recorded for the target operation behavior and the data viewing information for the target data viewing behavior, the target operation behavior includes an identification scanning operation and / or a status switching behavior for an industrial device, and the data targeted by the target data viewing behavior includes data associated with the industrial device and / or the industrial process; Extracting from the target behavior information sequence a subsequence composed of a plurality of sequentially arranged target behavior information; Performing feature extraction on the work role information and the subsequence to generate corresponding sample features; Constructing a corresponding sample label based on the data targeted by the next target behavior information of the subsequence; and Using the sample features and the corresponding sample labels as the feature part and the label part respectively to obtain a labeled training sample; Before performing the feature extraction on the work role information and the subsequence to generate corresponding sample features, the training sample generation method further includes: Obtaining additional status information corresponding to the target behavior information in the subsequence, wherein the type of the additional status information is determined according to the work role information and the operation information and data viewing information before the target behavior information in the subsequence; The performing feature extraction on the work role information and the subsequence to generate corresponding sample features includes: Performing feature extraction on the work role information, the subsequence, and the corresponding additional status information to generate corresponding sample features.

2. The training sample generation method according to claim 1, wherein The data associated with the industrial device includes at least one of the following: a three-dimensional model of the industrial device, an image related to the industrial device, a video related to the industrial device, a document manual related to the industrial device, The data associated with the industrial process includes at least one of the following: an image related to the industrial process, a video related to the industrial process, a document manual related to the industrial process.

3. The training sample generation method according to claim 1, wherein, The constructing a corresponding sample label based on the data targeted by the next target behavior information of the subsequence includes: Determining the association relationship between the industrial device and / or industrial process involved in the data targeted by the next target behavior information of the subsequence and the industrial device and / or industrial process involved in the data targeted by the target behavior information in the subsequence; and Constructing the sample label based on the data targeted by the next target behavior information and the determined association relationship.

4. A method for training a data-to-be-viewed recommendation model, comprising: Obtaining the labeled training sample obtained by the training sample generation method according to any one of claims 1 to 3; And Using the feature part in the labeled training sample as the input of the data-to-be-viewed recommendation model, and using the corresponding label part in the labeled training sample as the expected output to train and obtain the data-to-be-viewed recommendation model.

5. A method for recommending data to be viewed, comprising: Obtain a labeled training sample set, where the feature part of the labeled training sample is determined based on the work role information, the target behavior information sequence, and the corresponding additional status information, and the label part in the labeled training sample is determined based on the data targeted by the next target behavior information in the target behavior information sequence; Obtain the work role information, the current target behavior information sequence, and the corresponding additional status information of the logged-in user. Among them, the target behavior information in the current target behavior information sequence includes the operation information for the target operation behavior recorded so far and the data viewing information for the target data viewing behavior. The type of the obtained additional status information is determined based on the work role information of the logged-in user and the operation information and data viewing information before the corresponding target behavior information in the current target behavior information sequence. The target operation behavior includes an identification scanning operation and / or a status switching behavior for an industrial device, and the data targeted by the target data viewing behavior includes data associated with the industrial device and / or the industrial process; Determine the recommended viewing data corresponding to the work role information, the current target behavior information sequence, and the corresponding additional status information of the logged-in user according to the obtained labeled training sample set; and Present the recommended viewing data to the logged-in user.

6. The method according to claim 5, wherein, The determining the recommended viewing data corresponding to the work role information, the current target behavior information sequence, and the corresponding additional status information of the logged-in user according to the obtained labeled training sample set includes: Convert the obtained work role information, current target behavior information sequence, and corresponding additional status information of the logged-in user into corresponding current user features according to the determination method of the feature part of the labeled training sample; and Determine the recommended viewing data from the data indicated by the label part in the labeled training sample set based on the K-nearest neighbor algorithm.

7. The method according to claim 5, wherein, The determining the recommended viewing data corresponding to the work role information, the current target behavior information sequence, and the corresponding additional status information of the logged-in user according to the obtained labeled training sample set includes: Input the obtained work role information, current target behavior information sequence, and corresponding additional status information of the logged-in user into a recommended data to be viewed model, and obtain the recommended viewing data. The recommended data to be viewed model is trained using the labeled training sample set.

8. The method according to claim 7, wherein, The method further includes: In response to detecting negative feedback from the logged-in user for the recommended viewing data, generate difficult samples based on the obtained work role information and current target behavior information sequence of the logged-in user to further train the recommended data to be viewed model.

9. A training sample generation device for industrial scenario data recommendation, including: An information acquisition unit, configured to acquire the work role information and the target behavior information sequence of a target user, wherein the target behavior information in the target behavior information sequence includes the recorded operation information for a target operation behavior and the data viewing information for a target data viewing behavior, the target operation behavior includes an identification scanning operation and / or a status switching behavior for an industrial device, and the data targeted by the target data viewing behavior includes data associated with the industrial device and / or the industrial process; and An information extraction unit, configured to extract a subsequence composed of a plurality of target behavior information arranged in sequence from the target behavior information sequence; A feature extraction unit, configured to perform feature extraction on the work role information and the subsequence to generate corresponding sample features; A label construction unit, configured to construct a corresponding sample label based on the data targeted by the next target behavior information of the subsequence; A sample generation unit, configured to obtain a labeled training sample by using the sample features and the corresponding sample labels as the feature part and the label part respectively; The information acquisition unit is further configured to acquire additional status information corresponding to the target behavior information in the subsequence, wherein the type of the additional status information is determined according to the work role information and the operation information and data viewing information before the target behavior information in the subsequence; and The feature extraction unit is further configured to perform feature extraction on the work role information, the subsequence, and the corresponding additional status information to generate corresponding sample features.

10. A training device for a data-to-be-viewed recommendation model, comprising: A sample acquisition unit, configured to acquire the labeled training samples obtained by the training sample generation method according to any one of claims 1 to 3; And A model training unit, configured to use the feature part of the labeled training sample as the input of the data-to-be-viewed recommendation model, and use the corresponding label part of the labeled training sample as the expected output to train the data-to-be-viewed recommendation model.

11. A device for recommending data to be viewed, comprising: A sample set acquisition unit, configured to acquire a labeled training sample set, wherein the feature part of the labeled training sample is determined based on the work role information, the target behavior information sequence, and the corresponding additional status information, and the label part of the labeled training sample is determined based on the data targeted by the next target behavior information of the target behavior information sequence; A unit for obtaining information to be predicted, configured to obtain the work role information of the logged-in user, the current target behavior information sequence, and the corresponding additional status information. Among them, the target behavior information in the current target behavior information sequence includes the operation information for the target operation behavior recorded up to the current time and the data viewing information for the target data viewing behavior. The type of the obtained additional status information is determined according to the work role information of the logged-in user and the operation information and data viewing information before the corresponding target behavior information in the current target behavior information sequence. The target operation behavior includes an identification scanning operation and / or a status switching behavior for an industrial device. The target operation behavior includes an identification scanning operation and / or a status switching behavior for an industrial device. The data targeted by the target data viewing behavior includes data associated with an industrial device and / or an industrial process; A unit for determining recommended data, configured to determine recommended viewing data corresponding to the work role information of the logged-in user, the current target behavior information sequence, and the corresponding additional status information according to the obtained labeled training sample set; and A data presentation unit, configured to present the recommended viewing data to the logged-in user.

12. An electronic device, comprising: At least one processor, a memory coupled to the at least one processor, and a computer program stored on the memory. The at least one processor executes the computer program to implement the training sample generation method according to any one of claims 1 to 3, or implement the training method according to claim 4, or implement the method for recommending data to be viewed according to any one of claims 5 to 8.

13. A computer-readable storage medium storing a computer program, which when executed by a processor, implements the training sample generation method according to any one of claims 1 to 3, or implements the training method according to claim 4, or implements the method for recommending data to be viewed according to any one of claims 5 to 8.

Images