CN114661572A

CN114661572A - Big data analysis method and system for user habit in handling digital cloud office

Info

Publication number: CN114661572A
Application number: CN202210342488.0A
Authority: CN
Inventors: 杨喜顺
Original assignee: Individual
Current assignee: Individual
Priority date: 2022-03-31
Filing date: 2022-03-31
Publication date: 2022-06-24

Abstract

The embodiment of the disclosure discloses a method and a system for analyzing big data of user habits in handling digital cloud office; the method can be used for mining key behavior details of specified office behavior events in the online office collaboration information matched with the habit analysis conditions, so that the key behavior details of the specified office behavior events are determined, stylized annotation is carried out on the office operation habits of the specified office behavior events through the key behavior details of the specified office behavior events, so that the key behavior details and the first collaborative office behavior text set can be combined to carry out accurate and efficient analysis on the user operation habits of the specified office behavior events, high correlation between the obtained office operation habit records and the specified office behavior events is ensured, and the noise operation habits mixed with other office behavior events in the office operation habit records are avoided.

Description

Big data analysis method and system for user habit in handling digital cloud office

Technical Field

The disclosure relates to the technical field of big data, in particular to a big data analysis method and system for user habits in digital cloud office.

Background

In recent years, normalization of public health events has presented many challenges to traditional offline office work. Digital office/online cloud office has gradually pushed production work mode conversion of various industries due to the advantages of 'no contact', 'no space-time limitation', 'low construction cost', and the like. In the case of digital office/online cloud office, in order to realize sustainable development of digital office/online cloud office, in-depth analysis on office users and service optimization are required. Through research and development of the inventor, the entrance point for realizing the optimization of the digital office/online cloud office service lies in the analysis of the operation habits of the user, but the accuracy and the efficiency are difficult to guarantee in the analysis of the operation habits of the user by the related technology.

Disclosure of Invention

An object of the present disclosure is to provide a method and a system for analyzing big data of user habits in handling digital cloud office.

The technical scheme of the disclosure is realized by at least some embodiments as follows.

A big data analysis method for user habit in handling digital cloud office is applied to a digital cloud office system, and the method at least comprises the following steps: responding to the authenticated operation habit analysis request, performing cooperative operation behavior identification operation on the online office cooperation information matched with the habit analysis conditions, and acquiring a first cooperative office behavior text set; performing key behavior detail mining operation on a specified office behavior event in the online office collaboration information matching the habit analysis conditions based on the first collaborative office behavior text set, and acquiring key behavior detail mining information of the specified office behavior event, wherein the key behavior detail mining information is used for reflecting the key behavior detail of the specified office behavior event; and performing user operation habit analysis on the specified office behavior event in the online office collaboration information matched with the habit analysis conditions based on the key behavior detail mining information and the first collaborative office behavior text set, and acquiring an office operation habit record of the specified office behavior event.

The method and the device are applied to the embodiment, key behavior detail mining operation can be carried out on the specified office behavior events in the on-line office collaboration information matched with the habit analysis conditions, so that the key behavior details of the specified office behavior events are determined, stylized annotation is carried out on the office operation habits of the specified office behavior events through the key behavior details of the specified office behavior events, so that the key behavior details and the first collaborative office behavior text set can be combined to carry out accurate and efficient analysis on the user operation habits of the specified office behavior events, high correlation between the obtained office operation habit records and the specified office behavior events is ensured, and noise operation habits mixed with other office behavior events in the office operation habit records are avoided.

For an independently implementable technical solution, performing a cooperative operation behavior recognition operation on online office cooperation information matching habit analysis conditions to obtain a first cooperative office behavior text set, including: and performing cooperative operation behavior recognition operation on the online office cooperation information matched with the habit analysis conditions according to at least two cooperative operation behavior recognition units, and acquiring first cooperative office behavior text sets respectively corresponding to the at least two cooperative operation behavior recognition units.

For an independently implementable technical solution, based on the first collaborative office behavior text set, performing a key behavior detail mining operation on a specified office behavior event in the online office collaboration information matching the habit analysis conditions to obtain key behavior detail mining information of the specified office behavior event, including: performing key behavior detail mining on a first cooperative office behavior text set corresponding to the last cooperative operation behavior recognition unit to obtain key behavior detail vector distribution; and performing behavior detail translation on the key behavior detail vector distribution and the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in the at least two cooperative operation behavior recognition units to acquire key behavior detail mining information.

The method and the device are applied to the embodiment, the key behavior detail vector distribution and the first cooperative office behavior text set of the at least one cooperative operation behavior recognition unit are subjected to behavior detail translation, so that the key behavior detail content in the key behavior detail vector distribution is stylized and annotated by integrating the first cooperative office behavior text set with less information content, and further the accuracy and the reliability of key behavior detail mining can be improved.

For an independently implementable technical solution, performing behavior detail translation on the key behavior detail vector distribution and a first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in the at least two cooperative operation behavior recognition units to obtain key behavior detail mining information, including: performing detail content derivation operation on the key behavior detail vector distribution to obtain first derived behavior detail vector distribution; integrating the first derived behavior detail vector distribution and the first cooperative office behavior text set to obtain a first office behavior composite description set; and performing windowed mining operation on the first office behavior composite description set to acquire the detail mining information of the key behaviors.

The method and the device are applied to the embodiment, and the key behavior detail mining information can be completely and abundantly mined through the first office behavior composite description set.

For an independently implementable technical solution, mining information based on the key behavior details and the first collaborative office behavior text set, performing user operation habit analysis on a specified office behavior event in the online office collaboration information matching the habit analysis conditions, and acquiring an office operation habit record of the specified office behavior event, the method includes: performing personalized operation habit mining on the first collaborative office behavior text set to obtain a personalized office behavior phrase set; mining information and the personalized office behavior phrase set based on the key behavior details, and acquiring a behavior detail attention field set; and performing behavior detail translation on the behavior detail attention field set to acquire the office operation habit record.

When the method and the device are applied to the embodiment, the key behavior detail attention operation can be carried out on the personalized office behavior phrase set through key behavior detail mining information, so that the operation behavior track corresponding to the key behavior detail positioning tag is focused, the focusing algorithm (attention algorithm) is focused on the positioning tag corresponding to the key behavior detail, and the precision of the determined office operation habit record can be improved.

For an independently implementable technical solution, the first cooperative office behavior text set includes at least two first cooperative office behavior text sets obtained by performing a cooperative operation behavior recognition operation according to at least two cooperative operation behavior recognition units; the behavior detail attention field set comprises a first behavior detail attention field set corresponding to a personalized office behavior phrase set and a second behavior detail attention field set corresponding to a first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of at least two cooperative operation behavior recognition units, wherein the behavior detail attention field set is obtained based on the key behavior detail mining information and the personalized office behavior phrase set, and the method comprises the following steps: performing first key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information to obtain a first behavior detail attention field set; and respectively carrying out second key behavior detail attention operation on the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of the at least two cooperative operation behavior recognition units through the key behavior detail mining information, and acquiring second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit.

The method and the device for identifying the cooperative operation behavior are applied to the embodiment, the second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior identification unit can be accurately positioned, and confusion caused by one-to-one matching of the second behavior detail attention field sets is avoided.

For an independently implementable technical solution, the key behavior detail mining information includes a key behavior detail estimation result, wherein the first key behavior detail attention operation is performed on the personalized office behavior phrase set through the key behavior detail mining information, and the obtaining of the first behavior detail attention field set includes: calculating the key behavior detail estimation result and the visual category value of the corresponding operation behavior track in the personalized office behavior phrase set to obtain a second cooperative office behavior text set; integrating the second cooperative office behavior text set and the key behavior detail estimation result to obtain a third cooperative office behavior text set; and performing key behavior detail attention operation on the third cooperative office behavior text set to acquire the first behavior detail attention field set.

By analyzing the visual category value of the operation behavior track, the second cooperative office behavior text can be positioned from the global level, so that the integrity of the subsequently obtained first behavior detail attention field set is ensured.

For an independently implementable technical solution, performing a key behavior detail attention operation on the third cooperative office behavior text set, and acquiring the first behavior detail attention field set, includes: performing separated windowing mining operation on the third cooperative office behavior text set to obtain a separated mining list; performing cooperative operation behavior identification operation on the separated mining list to obtain a cooperative behavior identification list; and calculating the corresponding operation behavior tracks of the cooperative behavior identification list and the separated mining list to obtain the first behavior detail attention field set.

Applied to the above embodiment, the noise minimization of the resulting first action detail attention field set can be ensured by a separate windowing mining operation (deep convolution processing).

For an independently implementable technical solution, the key behavior detail mining information includes a key behavior detail estimation result, wherein second key behavior detail attention operations are respectively performed on a first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of at least two cooperative operation behavior recognition units through the key behavior detail mining information, and second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit are obtained, including: calculating the key behavior detail estimation result and a visual category value of a corresponding operation behavior track in a first cooperative office behavior text set of a pth cooperative operation behavior identification unit to obtain a fourth cooperative office behavior text set, wherein the pth cooperative operation behavior identification unit is one of the at least one cooperative operation behavior identification unit; integrating the fourth cooperative office behavior text set and the key behavior detail estimation result to obtain a fifth cooperative office behavior text set; and performing key behavior detail attention operation on the fifth cooperative office behavior text set to acquire a second behavior detail attention field set corresponding to the first cooperative office behavior text set of the p-th cooperative operation behavior recognition unit.

By considering the operation behavior trace, the integrity of the obtained second behavior detail attention field set can be ensured, and the second behavior detail attention field set is prevented from being missing as much as possible.

For an independently implementable technical solution, the method for analyzing big data of user habit in handling digital cloud office is implemented by means of an AI machine learning model, wherein the method further comprises: loading office cooperation information on a reference line to the AI machine learning model, and acquiring reference key behavior detail distribution and reference office operation habit records of specified office behavior events in the office cooperation information on the reference line; acquiring habit analysis cost deviation of the AI machine learning model based on the prior knowledge of the office cooperation information on the reference line and the reference office operation habit record; acquiring the detail deviation of the key behaviors of the AI machine learning model based on the prior knowledge of the office cooperation information on the reference line and the detail distribution of the reference key behaviors; performing global calculation on the habit resolution cost deviation and the key behavior detail deviation to obtain model quality evaluation of the AI machine learning model; improving the AI machine learning model based on the model quality assessment.

For an independently implementable solution, the specified office activity event includes a file sharing event and/or a resource encryption event.

A digital cloud office system, comprising: a memory for storing an executable computer program, a processor for implementing the above method when executing the executable computer program stored in the memory.

A computer-readable storage medium, on which a computer program is stored which, when executed, performs the above-described method.

Drawings

Fig. 1 is a schematic diagram illustrating a digital cloud office system in which embodiments of the present disclosure may be implemented.

Fig. 2 is a flowchart illustrating a method for analyzing big data of user habits in handling digital cloud office, in which an embodiment of the present disclosure may be implemented.

Fig. 3 is an architecture diagram illustrating an application environment in which a method for analyzing big data of user habits in handling digital cloud office according to an embodiment of the present disclosure may be implemented.

Detailed Description

The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. For the purpose of making the purpose, technical solutions and advantages of the present disclosure clearer, the present disclosure will be described in further detail with reference to the accompanying drawings, the described embodiments should not be construed as limiting the present disclosure, and all other embodiments obtained by a person of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present disclosure. In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing embodiments of the disclosure only and is not intended to be limiting of the disclosure.

Fig. 1 is a block diagram illustrating a communication configuration of a digital cloud office system 100 that can implement an embodiment of the present disclosure, where the digital cloud office system 100 includes a memory 101 for storing an executable computer program, and a processor 102 for implementing a big data analysis method for user habits in handling digital cloud office in the embodiment of the present disclosure when the executable computer program stored in the memory 101 is executed.

Fig. 2 is a schematic flow diagram illustrating a method for analyzing big data of user habits in coping with digital cloud office, where the method for analyzing big data of user habits in coping with digital cloud office according to an embodiment of the present disclosure may be implemented by the digital cloud office system 100 shown in fig. 1, and further may include the technical solutions described in the following relevant steps.

And Step11, responding to the authenticated operation habit analysis request, performing cooperative operation behavior identification operation on the online office cooperation information matched with the habit analysis conditions, and acquiring a first cooperative office behavior text set.

For the embodiment of the disclosure, the authenticated operation habit analysis request can be understood as a legal operation habit analysis request passing the security authentication, rather than an unknown source request, so that the security of the behavior data information of the office user can be guaranteed.

In addition, the online office collaboration information matching the habit analysis condition may be understood as the online office collaboration information to be processed, the habit analysis condition may be "number of collaborating offices exceeds X", or "office place is YYY", and the like, and the online office collaboration information may be understood as an office log containing rich information.

Further, the cooperative operation behavior recognition operation may be understood as recognition and mining processing of cooperative operation behavior characteristics, so as to obtain the first cooperative office behavior text set.

Based on the above, the first set of collaborative office behavior text may be a summary result of the collaborative operation behavior feature, and the text may be a feature vector or a feature character string.

Step12, based on the first collaborative office behavior text set, performing key behavior detail mining operation on the specified office behavior event in the online office collaboration information matching the habit analysis conditions, and acquiring key behavior detail mining information of the specified office behavior event.

For the disclosed embodiments, the key behavior details mining information is intended to reflect key behavior details of the specified office behavior event.

In addition, a given office behavioral event may be understood as a targeted behavioral event (such as some relatively important, popular office behavioral events) in the online office collaboration information.

Further, the mining operation of the key behavior details on the specified office behavior events in the online office collaboration information can be understood as mining the behavior details with significant discrimination on the specified office behavior events in the online office collaboration information, and by the design, different office operation habit records can be determined by utilizing the key behavior details in a subsequent pertinence manner.

Step13, mining information and the first cooperative office behavior text set based on the key behavior details, performing user operation habit analysis on the specified office behavior event in the online office cooperative information matched with the habit analysis conditions, and acquiring an office operation habit record of the specified office behavior event.

For the embodiment of the disclosure, the analysis of the user operation habits of the specified office behavior events in the online office collaboration information can be understood as the extraction of the user operation trends of the specified office behavior events in the online office collaboration information (for example, the user is accustomed to scanning instead of reading, the user does not pay attention to all contents on a page, and the like; for example, the user is accustomed to sliding operation of a touch screen; and for example, the user is accustomed to intelligent interaction through voice). Further, the office operation habit record may be, for example, a result of summarizing a series of operation habit preferences for a specific office behavior event, such as "frequently sliding the touch screen," tendency to communicate by writing, "tendency to make the GUI operation interface concise," and the like.

In the embodiment of the present disclosure, the online office collaboration information matching the habit analysis conditions may include at least one designated office behavior event. On the basis that the designated office behavior event is the resource encryption event, the designated office behavior event can be analyzed through ideas such as an AI machine learning model. It can be understood that the distinction degree of the key behavior details of the file sharing event or the resource encryption event in the online office collaboration information is relatively high, so that the key behavior details of the specified office behavior event can be determined, and the determination efficiency of the office operation habit record can be guaranteed.

It is understood that, for Step11, the cooperative work behavior recognition operation may be performed on the online office cooperation information matching the habit analysis condition to obtain a first cooperative work behavior text set. Further, Step11 may include the following: performing cooperative operation behavior recognition operation on the online office cooperation information matching the habit analysis conditions according to at least two cooperative operation behavior recognition units (such as a feature mining layer), and acquiring first cooperative office behavior text sets (which can also be understood as a cooperative office behavior feature map) corresponding to the at least two cooperative operation behavior recognition units respectively.

For the embodiments of the present disclosure, the first collaborative office behavior text set may be mined by an AI machine learning model, which includes at least one collaborative behavior recognition unit, which may be used to mine collaborative behavior of online office collaboration information that matches the habit analysis conditions. In some possible embodiments, the AI machine learning model may include at least two (cascaded) cooperative behavior recognition units that are associated with each other, and each cooperative behavior recognition unit may obtain a first cooperative office behavior text set of the unit, and may obtain, as a simple understanding, at least two first cooperative office behavior text sets corresponding to the at least two cooperative behavior recognition units. In some possible embodiments, the cooperative behavior identification unit may include at least one of a cooperative behavior identification node, a function transformation node (activation node), a down-sampling node (pooling node), and other model units, and the disclosed embodiments do not limit the model architecture of the cooperative behavior identification unit, and in some cases, the AI machine learning model may be a GCN model or an LSTM model.

It is understood that, for Step12, the key behavior detail mining information may be obtained by determining the key behavior details of the specified office behavior event in the online office collaboration information matching the habit analysis conditions based on the first collaborative office behavior text set. For example, the key behavior detail cooperative operation behavior recognition operation may be performed on the first cooperative office behavior text set derived by the cooperative operation behavior recognition model (for example, the first cooperative office behavior text set derived by the last cooperative operation behavior recognition unit of the cooperative operation behavior recognition model). For example, the key behavior detail recognition model of the AI machine learning model may perform a key behavior detail cooperative operation behavior recognition operation on the first cooperative office behavior text set of the second at least two cooperative operation behavior recognition units to obtain a key behavior detail vector distribution, and perform behavior detail translation (e.g., feature decoding processing) on the key behavior detail vector distribution to obtain the key behavior detail mining information.

For an independently implementable technical scheme, the accuracy of the key behavior detail mining information can be improved according to the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in the at least two cooperative operation behavior recognition units. Step12 may include: performing key behavior detail mining on a first cooperative office behavior text set corresponding to the last cooperative operation behavior recognition unit to obtain key behavior detail vector distribution; and performing behavior detail translation on the key behavior detail vector distribution and the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in the at least two cooperative operation behavior recognition units to acquire key behavior detail mining information.

For the disclosed embodiments, the key behavior detail vector distribution may be understood as a feature map of key behavior details.

For an independently implementable technical solution, the key behavior detail cooperative operation behavior recognition operation may be performed on the first cooperative office behavior text sets of the first at least two cooperative operation behavior recognition units through the key behavior detail recognition model. In some possible embodiments, the key behavior detail recognition model may include model elements such as a cooperative behavior recognition node and a down-sampling node, for example, the key behavior detail recognition model may include a cooperative behavior recognition node with a recognition window of 2 × 2, a cooperative behavior recognition node with a recognition window of 6 × 6 and an expansion index of 24, a cooperative behavior recognition node with a recognition window of 6 × 6 and an expansion index of 36, and a down-sampling node. The embodiment of the present disclosure does not limit the architecture of the key behavior detail recognition model. And identifying, down-sampling and the like on the first cooperative office behavior text set through each model unit of the key behavior detail identification model, so as to obtain the key behavior detail vector distribution.

For an independently implementable technical scheme, the accuracy of detail mining information of key behaviors can be improved according to the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in at least two cooperative operation behavior recognition units. The key behavior detail mining information may be obtained by performing behavior detail translation on the first cooperative office behavior text set of the at least one cooperative operation behavior recognition unit and the key behavior detail vector distribution (for example, performing behavior detail translation through a behavior detail translation model), and it may be understood that a positioning tag (for example, a relative position of a key behavior detail) corresponding to a key behavior detail of the specified office behavior event is obtained.

Based on the above, performing behavior detail translation on the key behavior detail vector distribution and the first cooperative office behavior text set of at least one of the at least two cooperative operation behavior recognition units to obtain key behavior detail mining information may include: performing detail content derivation operation on the key behavior detail vector distribution to obtain first derived behavior detail vector distribution; performing integration operation (splicing processing) on the first derived behavior detail vector distribution and the first cooperative office behavior text set to obtain a first office behavior composite description set (spliced office behavior description text); and performing windowing mining operation (such as convolution operation) on the first office behavior compound description set to acquire the key behavior detail mining information. For the embodiment of the present disclosure, the detail content derivation operation on the key behavior detail vector distribution may be understood as performing detail content expansion processing/detail dimension expansion processing on the key behavior detail vector distribution.

In some possible embodiments, the first cooperative office behavior text set of the at least one cooperative operation behavior recognition unit and the key behavior detail vector are distributed to perform behavior detail translation, so that fusion of detail information of different levels can be realized, the category richness of the detail information is improved, the key behavior detail content in the cooperative office behavior text set is optimized, and the positioning tag corresponding to the key behavior detail of the specified office behavior event is favorably mined.

In some possible embodiments, the key behavior detail vector distribution may be translated with behavior details (e.g., translated with a behavior detail translation model) of a first co-office behavior text set of one of the co-operating behavior recognition units (e.g., the first co-operating behavior text set of one of the first co-operating behavior recognition unit, the second co-operating behavior recognition unit, the last co-operating behavior recognition unit, etc.). In view of the fact that the distribution of the key behavior detail vectors is the collaborative office behavior text set processed by the key behavior detail recognition model, the number of the focus points of the distribution of the key behavior detail vectors is generally larger than that of the first collaborative office behavior text set, and the dimensionality of the focus points is lower than that of the first collaborative office behavior text set, so that the content coverage range is guaranteed. Detail content derivation operation can be performed on the key behavior detail vector distribution (for example, detail content expansion processing is performed by combining ideas such as sampling) to obtain first derived behavior detail vector distribution, so that the size of the first derived behavior detail vector distribution, the number of the attention points and the first cooperative office behavior text set are kept consistent. Further, the first derived behavior detail vector distribution and the first collaborative office behavior text set may be integrated (for example, the key behavior detail vector distribution after the detail content derivation operation and all detail categories of the first collaborative office behavior text set are recorded and integrated), the first office behavior composite description set is obtained, and the first office behavior composite description set is subjected to a windowing mining operation (for example, the first office behavior composite description set is subjected to a windowing mining operation through the collaborative operation behavior recognition node with the recognition window of 6 × 6), and a result of the behavior detail translation is obtained, which may be understood as key behavior detail mining information. The key behavior detail mining information can be a key behavior detail estimation result with the same dimension as that of the on-line office collaboration information matching the habit analysis condition, and can be understood as the on-line office collaboration information of the positioning label corresponding to the key behavior detail of the specified office behavior event in the on-line office collaboration information matching the habit analysis condition.

In some possible embodiments, the key behavior detail vector distribution may be translated in behavior detail with a first co-office behavior text set of several co-operating behavior recognition units. For example, the first cooperative office behavior text sets (p and w are both positive integers, and p is less than w) of the p-th cooperative operation behavior recognition unit and the w-th cooperative operation behavior recognition unit and the key behavior detail vector distribution may be subjected to behavior detail translation. The distribution size of the detail vector of the key behavior is smaller than the first cooperative office behavior text set of the w-th cooperative operation behavior recognition unit, and the size of the first cooperative office behavior text set of the w-th cooperative operation behavior recognition unit is smaller than the first cooperative office behavior text set of the p-th cooperative operation behavior recognition unit.

Furthermore, the detailed content derivation operation can be performed on the key behavior detail vector distribution to obtain a first derived behavior detail vector distribution, the first derived behavior detail vector distribution and the first cooperative office behavior text set of the w cooperative operation behavior recognition unit are subjected to integration operation (such as splicing processing, combination processing and the like), a first office behavior composite description set corresponding to the w cooperative operation behavior recognition unit is obtained, and 6 × 6 windowing mining operation is performed on the first office behavior composite description set.

Further, the cooperative office behavior text set after the windowed mining operation may be subjected to detail content derivation operation, and integrated with the first cooperative office behavior text set of the p-th cooperative operation behavior recognition unit to obtain the first office behavior composite description set corresponding to the p-th cooperative operation behavior recognition unit, and then 6-6 windowed mining operation may be performed to obtain the key behavior detail mining information. The behavior detail translation may also be performed on the key behavior detail vector distribution and the first cooperative office behavior text set of the more cooperative operation behavior recognition units, in other words, the detail content derivation operation may be performed on the key behavior detail vector distribution, the integration operation may be performed on the key behavior detail vector distribution and the first cooperative office behavior text set with a lower dimension, the detail content derivation operation may be performed again after the windowed mining operation, the integration operation may be performed on the key behavior detail vector distribution and the first cooperative office behavior text set with a higher dimension, and further, the next round of detail content derivation operation may be performed after the windowed mining operation, and the integration operation may be performed on the key behavior detail vector distribution and the first cooperative office behavior text set with a larger dimension. It is understood that the number of the first co-office activity text sets for performing the activity detail translation is not limited by the embodiments of the present disclosure.

For an independently implementable solution, reference may be made to the following for the description of the behavior detail translation model: the cooperative behavior recognition model may include 4 cooperative behavior recognition units, and the key behavior detail vector distribution may perform behavior detail translation with the first cooperative office behavior text set of the first cooperative behavior recognition unit and the second cooperative behavior recognition unit. In other words, the behavior detail translation model may perform the behavior detail translation on the key behavior detail vector distribution, the first set of co-office behavior texts of the first co-operating behavior recognition unit, and the first set of co-office behavior texts of the second co-operating behavior recognition unit. In some embodiments of the possibilities, the size of the distribution of the detail vectors of the key behaviors is smaller than the first text set of co-office behaviors of the second recognition unit, and the size of the first text set of co-office behaviors of the second recognition unit is smaller than the first text set of co-office behaviors of the first recognition unit. The detail content derivation operation can be carried out on the key behavior detail vector distribution to obtain a first derived behavior detail vector distribution, the 2 x 2 windowing mining operation is carried out on the first cooperative office behavior text set of the second cooperative operation behavior recognition unit, the integration operation is carried out on the first derived behavior detail vector distribution and the first cooperative office behavior text set of the second cooperative operation behavior recognition unit after the 2 x 2 windowing mining operation, a first office behavior composite description set is obtained, and the 6 x 6 windowing mining operation is carried out on the first office behavior composite description set. Further, the detail content derivation operation may be performed on the cooperative office behavior text set after the 6 × 6 windowing mining operation, and the integration operation may be performed on the cooperative office behavior text set after the first cooperative office behavior recognition unit, so as to obtain the key behavior detail mining information.

By means of the design, behavior detail translation can be performed on the key behavior detail vector distribution and the first cooperative office behavior text set of the at least one cooperative operation behavior recognition unit, stylized annotation can be performed on key behavior detail content in the key behavior detail vector distribution by integrating the first cooperative office behavior text set with less information, and accuracy and reliability of key behavior detail mining are improved.

In some possible embodiments, when the user operation habit resolution is performed on the designated office behavior event, the user operation habit resolution may be performed through an information resolution unit in the AI machine learning model. For example, the information analysis unit may perform user operation habit analysis on a first cooperative office behavior text set (for example, a first cooperative office behavior text set corresponding to a last cooperative office behavior recognition unit), and perform behavior detail translation on a personalized office behavior phrase set derived by the information analysis unit to obtain an office operation habit record of a specified office behavior event. Further, the office operation habit record of the designated office behavior event may be determined by using a dynamic visual graph, for example, the office operation habit record may be output through a knowledge graph or a feature relationship network.

For an independently implementable solution, Step13 may include the following: performing personalized operation habit mining on the first collaborative office behavior text set to obtain a personalized office behavior phrase set; mining information and the personalized office behavior phrase set based on the key behavior details, and acquiring a behavior detail attention field set; and performing behavior detail translation on the behavior detail attention field set to acquire the office operation habit record.

For the embodiment of the disclosure, personalized operation habit mining may be performed on the first collaborative office behavior text set through the information parsing unit, so as to obtain a personalized office behavior phrase set. In some possible embodiments, the first cooperative office behavior text set includes at least two first cooperative office behavior text sets obtained by performing a cooperative operation behavior recognition operation according to at least two cooperative operation behavior recognition units, and the information parsing unit may perform personalized operation habit mining on the first cooperative office behavior text sets of the at least two cooperative operation behavior recognition units to obtain a personalized office behavior phrase set. The information analysis unit may include model units such as a cooperative behavior identification node and a down-sampling node, for example, the information analysis unit may include a cooperative behavior identification node with an identification window of 2 × 2, a cooperative behavior identification node with an identification window of 6 × 6 and an expansion index of 12, a cooperative behavior identification node with an identification window of 6 × 6 and an expansion index of 24, a cooperative behavior identification node with an identification window of 6 × 6 and an expansion index of 36, and a down-sampling node. The embodiment of the present disclosure does not limit the architecture of the information parsing unit. And identifying, down-sampling and the like on the first cooperative office behavior text set through each model unit of the information analysis unit, so as to obtain the personalized office behavior phrase set.

It is understood that the parsing precision can be improved by mining information (e.g., key behavior detail estimation results) through key behavior details. In some possible embodiments, the behavior detail attention field set may be obtained through the key behavior detail mining information and the personalized office behavior phrase set, for example, the key behavior detail attention field set may be obtained by performing a key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information. For example, the key behavior detail mining information and the personalized office behavior phrase set can be imported into a style annotation submodel of the AI machine learning model, and a behavior detail attention field set can be obtained. Further, office operation habit records (for example, office operation habit records are obtained through behavior detail translation) after key behavior detail style annotations (for example, enhancement processing on different key behavior details) can be obtained according to the behavior detail attention field set, so that office operation habit records with relatively high accuracy and reliability can be obtained.

In some possible embodiments, the style annotation processing may be performed on the key behavior detail content according to the first co-office behavior text set of at least one of the at least two co-operating behavior recognition units. For example, the number of detail categories can be expanded through the first collaborative office behavior text set, style annotation processing is performed on key behavior detail content, and further optimized key behavior detail vector distribution is obtained. The behavior detail attention field set includes a first behavior detail attention field set corresponding to the personalized office behavior phrase set and a second behavior detail attention field set corresponding to a first co-operating office behavior text set of at least one of the at least two co-operating behavior recognition units.

For an independently implementable technical solution, on the basis of the above contents, the step of obtaining a behavior detail attention field set based on the key behavior detail mining information and the personalized office behavior phrase set may include the following contents: performing first key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information to obtain a first behavior detail attention field set; and respectively carrying out second key behavior detail attention operation on the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of the at least two cooperative operation behavior recognition units through the key behavior detail mining information, and acquiring second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit.

For example, the first key behavior detail attention operation and the second key behavior detail attention operation may be performed one by one through the style annotation submodel. The behavior detail translation can be carried out on the first behavior detail attention field set and the second behavior detail attention field set, and the office operation habit records are obtained, so that the office operation habit records with relatively high analysis accuracy and reliability can be obtained.

For an independently implementable technical solution, performing a first key behavior detail attention operation on the personalized office behavior phrase set (office behavior description phrases with significant distinction) through the key behavior detail mining information, and acquiring a first behavior detail attention field set, may include the following contents: calculating the key behavior detail estimation result and the visual category value of the corresponding operation behavior track in the personalized office behavior phrase set to obtain a second cooperative office behavior text set; integrating the second cooperative office behavior text set and the key behavior detail estimation result to obtain a third cooperative office behavior text set; and performing key behavior detail attention operation on the third cooperative office behavior text set to acquire the first behavior detail attention field set.

For the disclosed embodiments, the first key behavior detail attention operation may be performed by a style annotation submodel. The style annotation submodel can calculate the key behavior detail estimation result and the visual category value of the corresponding operation behavior track in the personalized office behavior phrase set to obtain a second collaborative office behavior text set. The key behavior detail estimation result is the same as the dimension of the on-line office collaboration information matching the habit analysis condition, the key behavior detail estimation result is different from the dimension of the personalized office behavior phrase set, but the personalized office behavior phrase set can be obtained through the on-line office collaboration information matching the habit analysis condition through mining, personalized operation habit mining and other processes, the operation behavior track in the personalized office behavior phrase set is in an association relationship with the operation behavior track in the on-line office collaboration information matching the habit analysis condition, and therefore the operation behavior track in the personalized office behavior phrase set is in an association relationship with the operation behavior track in the key behavior detail estimation result.

For example, the operation behavior trajectory in the personalized office behavior phrase set is obtained by performing cooperative operation behavior recognition operations such as a cooperative office behavior text set aggregation windowing mining operation in the on-line office cooperation information matching the habit analysis conditions, for example, in the windowing mining operation with a recognition window of 6 × 6, 9 operation behavior trajectories in the on-line office cooperation information matching the habit analysis conditions correspond to one operation behavior trajectory in the personalized office behavior phrase set. Therefore, based on the relation between the operation behavior tracks in the personalized office behavior phrase set and the operation behavior tracks in the key behavior detail estimation result, the visual category values of the operation behavior tracks in the personalized office behavior phrase set and the corresponding operation behavior tracks in the key behavior detail estimation result are calculated (for example, weighted summation, multiplication and the like), and a second collaborative office behavior text set is obtained.

In some possible embodiments, the key behavior detail estimation result may be a binary chart (for example, the visual category value (category description value) of the positioning tag corresponding to the key behavior detail is "V1" or a value close to "V1", and the visual category value of the remaining positioning tags is "V0" or a value close to "V0"), the visual category value of the corresponding operation behavior trajectory is calculated, the visual category value of the operation behavior trajectory of the key behavior detail positioning tag in the personalized office behavior phrase set can be optimized (the operation behavior trajectory of the key behavior detail positioning tag in the personalized office phrase set is multiplied by "V1" or a value close to "V1"), and the visual category value of the operation behavior trajectory of the remaining positioning tag is attenuated (the operation behavior of the key behavior detail positioning tag in the personalized office phrase set is multiplied by "V0" or a value close to "V0"), and acquiring a second cooperative office behavior text set.

For an independently implementable technical solution, the style annotation sub-model may perform an integration operation on the second cooperative office behavior text set and the key behavior detail estimation result (for example, all detail categories may be cached), and acquire a third cooperative office behavior text set. Based on this, the third cooperative office behavior text set is subjected to a key behavior detail attention operation, and a first behavior detail attention field set is obtained, which may exemplarily include the following: performing separated windowing mining operation on the third cooperative office behavior text set to obtain a separated mining list; performing cooperative operation behavior identification operation on the separated mining list to obtain a cooperative behavior identification list; and calculating the corresponding operation behavior tracks of the cooperative behavior identification list and the separated mining list to obtain the first behavior detail attention field set.

For the embodiment of the present disclosure, a 6 × 6 separated windowing mining operation may be performed on the third collaborative office behavior text set to obtain a separated mining list, and a collaborative behavior recognition operation may be performed on the separated mining list, for example, the collaborative behavior recognition list may be obtained by processing a down-sampling node (e.g., an integrated down-sampling node), a classification node, a function transformation node (e.g., a RELU function transformation node), a classification node, and a function transformation node (e.g., a normalized function transformation node) one by one, and further, the first behavior detail attention field set may be obtained by calculating visual category values of corresponding operation behavior tracks of the collaborative behavior recognition list and the separated mining list.

In addition, the separated windowing mining operation on the third cooperative office behavior text set can be understood as that the third cooperative office behavior text set is subjected to deep mining processing, so that a separated mining list is obtained. Further, the separate mining inventory may be understood as a behavioral text convolution result.

For an independently implementable solution, the second key behavior detail attention operation can be performed by a style annotation submodel. Performing, by the key behavior detail mining information, a second key behavior detail attention operation on a first cooperative office behavior text set of at least one of the at least two cooperative operation behavior recognition units, respectively, to obtain second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit, where the second behavior detail attention field sets may include the following contents: calculating the key behavior detail estimation result and a visual category value of a corresponding operation behavior track in a first cooperative office behavior text set of a pth cooperative operation behavior identification unit to obtain a fourth cooperative office behavior text set, wherein the pth cooperative operation behavior identification unit is one of the at least one cooperative operation behavior identification unit; integrating the fourth cooperative office behavior text set and the key behavior detail estimation result to obtain a fifth cooperative office behavior text set; and performing key behavior detail attention operation on the fifth cooperative office behavior text set to acquire a second behavior detail attention field set corresponding to the first cooperative office behavior text set of the p-th cooperative operation behavior recognition unit.

For the embodiment of the present disclosure, the operation behavior track may be a streamlined visual behavior record obtained by integrating the behavior segments, such as a scriptured and nodulated user office behavior track.

For the embodiment of the present disclosure, the style annotation submodel may calculate the result of estimating the details of the key behavior and the visual category value of the corresponding operation behavior trajectory in the first cooperative office behavior text set of the p-th cooperative operation behavior recognition unit, and obtain a fourth cooperative office behavior text set. The dimension of the key behavior detail estimation result is the same as that of the on-line office collaboration information matching the habit analysis condition, and the dimension of the key behavior detail estimation result is different from that of the first cooperative office behavior text set, but the first cooperative office behavior text set is obtained by mining and other processing through the on-line office collaboration information matching the habit analysis condition, and the operation behavior trajectory in the first cooperative office behavior text set is associated with the operation behavior trajectory in the on-line office collaboration information matching the habit analysis condition. The visual category values of the operation behavior tracks in the first cooperative office behavior text set and the corresponding operation behavior tracks in the key behavior detail estimation result can be calculated to obtain a fourth cooperative office behavior text set.

It is understood that the style annotation submodel may integrate the fourth set of collaborative office behavior texts with the estimation result of the key behavior details (for example, all detail categories may be cached), and obtain a fifth set of collaborative office behavior texts. And performing key behavior detail attention operation on the fifth cooperative office behavior text set to obtain a second behavior detail attention field set. In some possible embodiments, the fifth collaborative office activity text set may be subjected to a 6 × 6 separate windowing mining operation, and the result of the separate windowing mining operation is processed one by a down-sampling node (e.g., an integrated down-sampling node), a classification node, a function transformation node (e.g., a RELU function transformation node), a classification node, and a function transformation node (e.g., a normalized function transformation node), and further, the processing result may be calculated with the visual category value of the corresponding operation behavior trajectory of the content of the 6 × 6 separate windowing mining operation, so as to obtain the second behavior detail attention field set.

In some possible embodiments, the second behavior detail attention field set corresponding to the plurality of cooperative behavior recognition units may be acquired through the above process. The key behavior detail attention operation can be carried out on the first cooperative office behavior text sets of the plurality of cooperative operation behavior identification units, and the key behavior detail contents in the plurality of first cooperative office behavior text sets can be integrated, so that the key behavior detail contents are optimized, a focusing algorithm is further facilitated to focus on the positioning labels corresponding to the key behavior details, and the quality of analyzing the operation habits of the user can be improved by the colleagues. For example, a second behavior detail attention field set corresponding to a first co-operation behavior recognition unit and a second behavior detail attention field set corresponding to a second co-operation behavior recognition unit may be obtained. The number of the second behavior detail attention field sets is not limited by the embodiments of the present disclosure.

In some embodiments of the possibilities, a behavior detail translation may be performed on the first set of behavior detail attention fields and the second set of behavior detail attention fields. In some embodiments of the possibilities, the behavioral detail translation may be performed by a behavioral detail translation model. For example, a detail content derivation operation may be performed on a first behavior detail attention field set, and the first behavior detail attention field set after the detail content derivation operation is completed and a second behavior detail attention field set with a lower dimension (a second behavior detail attention field set corresponding to a cooperative operation behavior recognition unit of a higher-level cooperative operation behavior recognition unit) are integrated, and a 6 × 6 windowing mining operation may be performed on a cooperative office behavior text set after the integration operation is completed. And performing detail content derivation operation again after the windowing mining operation, performing integration operation with a second behavior detail attention field set with a higher dimension (the second behavior detail attention field set corresponding to the cooperative operation behavior identification unit of the cooperative operation behavior identification unit with a lower hierarchy), and performing 6 × 6 windowing mining operation. It can be understood that the office operation habit record can be obtained after a plurality of rounds of detail content derivation operation, integration operation and windowing mining operation.

In some embodiments of the possibilities, the second set of behavior detail attention fields includes a second set of behavior detail attention fields corresponding to a first one of the co-operating behavior recognition units and a second set of behavior detail attention fields corresponding to a second one of the co-operating behavior recognition units. The behavior detail translation model can conduct detail content derivation operation on the first behavior detail attention field set, conduct 2 x 2 windowing mining operation on the second behavior detail attention field set corresponding to the second cooperative operation behavior recognition unit, conduct integration operation on the first behavior detail attention field set after the detail content derivation operation and the second behavior detail attention field set after the 2 x 2 windowing mining operation, and conduct 6 x 6 windowing mining operation on the cooperative office behavior text set after the integration operation. Further, the cooperative office behavior text set after the 6 × 6 windowed mining operation may be subjected to detail content derivation operation, and integrated with the second behavior detail attention field set corresponding to the first cooperative operation behavior recognition unit, so as to perform the 6 × 6 windowed mining operation, so as to obtain the office operation habit record.

In some possible embodiments, instead of performing style annotation processing on the key behavior detail content according to the first collaborative office behavior text set of the at least two collaborative operation behavior recognition units, the personalized office behavior phrase set and the key behavior detail mining information may be loaded to the style annotation submodel to obtain the behavior detail attention field set, and perform behavior detail translation on the behavior detail attention field set, for example, 6 × 6 windowing mining operations may be performed on the behavior detail attention field set to obtain the office operation habit record.

By means of the design, key behavior detail attention operation can be conducted on the personalized office behavior phrase set through key behavior detail mining information, operation behavior tracks corresponding to key behavior detail positioning labels are focused on, a focusing algorithm is focused on the positioning labels corresponding to the key behavior details, and therefore accuracy and reliability of office operation habit recording can be improved.

For an independently implementable solution, the AI machine learning model can be refined before it is used. It will be appreciated that the method may also include the following: loading office cooperation information on a reference line to the AI machine learning model, and acquiring reference key behavior detail distribution and reference office operation habit records of specified office behavior events in the office cooperation information on the reference line; acquiring habit analysis cost deviation of the AI machine learning model based on the prior knowledge of the office cooperation information on the reference line and the reference office operation habit record; acquiring the detail deviation of the key behaviors of the AI machine learning model based on the priori knowledge of the office cooperation information on the reference line and the reference key behavior detail distribution; performing global calculation (weighted summation processing) on the habit resolution cost deviation and the key behavior detail deviation to obtain model quality evaluation of the AI machine learning model; improving the AI machine learning model based on the model quality assessment.

For the embodiments of the present disclosure, at least one designated office activity event may be included in the above reference online office collaboration information, and the designated office activity event may include a file sharing event and/or a resource encryption event. The office collaboration information on the reference line may have a priori knowledge (tag information or label information), wherein the a priori knowledge may include a key behavior detail experience basis, which may be a two-dimensional distribution list having a visual category value of "V1" for key behavior detail location tags of a specified office behavior event and a visual category value of "V2" for the remaining location tags. The priori knowledge can also comprise a user operation habit experience basis, and the user operation habit experience basis can analyze a log content set corresponding to the appointed office behavior event through the differentiative information of the appointed office behavior event.

In some possible embodiments, the AI machine learning model may process the on-line reference office collaboration information to obtain a reference key behavior detail distribution and a reference office operation habit record of a specified office behavior event. The reference key behavior detail distribution is a content set which is obtained through the AI machine learning model and used for determining key behavior details of the appointed office behavior event, and the reference office operation habit record can be a content set which is obtained through the AI machine learning model and used for analyzing the office place corresponding to the appointed office behavior event. There may be some deviation between the reference critical behavior detail distribution and the reference office operation habit record.

It is understood that the habit resolution cost bias (which can be understood as a loss difference) of the AI machine learning model can be determined by combining the reference office operation habit record and the user operation habit experience. In some possible embodiments, the hinge cost may be determined by referring to the visual category value of the corresponding operation behavior trajectory in the office operation habit record and the user operation habit experience basis, and the hinge cost is determined as a habit resolution cost deviation (habit resolution function) of the AI machine learning model.

In some embodiments of the possibilities, the key behavior detail bias of the AI machine learning model may be determined based on the reference key behavior detail distribution and the key behavior detail empirical basis. Further, the visual category value of the key behavior detail positioning tag of the designated office behavior event based on which the key behavior detail experience is based is "V1" or close to "V1", the visual category value of the remaining positioning tags is "V0" or close to "V0", the reference key behavior detail distribution is a distribution list, and the key behavior detail deviation can be determined according to the visual category value of the corresponding operation behavior track.

For an independently implementable technical scheme, the habit resolution cost deviation and the key behavior detail deviation can be globally calculated, and the model quality evaluation of the AI machine learning model can be obtained.

In some possible embodiments, the AI machine learning model may be improved by the model quality assessment. The model quality evaluation can be subjected to feedback debugging, and the values of model variables in the AI machine learning model are adjusted through an optimization algorithm, so that the model quality evaluation is minimized. And when the improvement indexes are met, acquiring an improved AI machine learning model. It is to be understood that the improvement index may include an improvement round number, and it is to be understood that, when the number of rounds performed in the above improvement processing loop matches the set round number, the improvement may be performed to obtain an improved AI machine learning model. The improvement index may include that the model quality evaluation is less than or equal to the set evaluation value or falls within a specified numerical range, for example, through multiple rounds of improvement, so that the model quality evaluation does not exceed the set evaluation value or fall within the specified numerical range, the improvement may be completed, and the improved AI machine learning model may be obtained. The improvement index may further include that the analysis precision exceeds a set evaluation value, and it can be understood that when the model is debugged in combination with the debugging reference set, the analysis precision of the obtained specified office behavior event exceeds the set evaluation value, the improvement may be completed, and the improved AI machine learning model is obtained.

Applied to the implementation of the disclosure, the method can perform key behavior detail mining operation on the specified office behavior event in the on-line office collaboration information matching the habit analysis conditions, perform behavior detail translation on the key behavior detail vector distribution and the first cooperative office behavior text set of the at least one cooperative operation behavior recognition unit, perform stylized annotation on the key behavior detail content in the key behavior detail vector distribution through the first cooperative office behavior text set, thereby determining the key behavior detail of the specified office behavior event, obtain key behavior detail mining information, further perform key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information, focus on the operation behavior track corresponding to the key behavior detail positioning tag, and focus on the positioning tag corresponding to the key behavior detail by the focusing algorithm, thereby improving the precision of the recording of the office operation habit.

It is understood that the online office collaboration information matching the habit analysis conditions may include at least one designated office activity event, and the designated office activity event may include a file sharing event and/or a resource encryption event.

In addition, the cooperative operation behavior recognition model can be used for performing cooperative operation behavior recognition operation on the online office cooperation information matching the habit analysis conditions, and a first cooperative office behavior text set (behavor _ text _ 1-behavor _ text _ 4) of the four cooperative operation behavior recognition units is obtained. The size of the first cooperative office behavior text set behavor _ text _1 is 25% of the on-line office collaboration information matching the habit analysis condition, the size of the first cooperative office behavior text set behavor _ text _2 is 12.5% of the on-line office collaboration information matching the habit analysis condition, the size of the first cooperative office behavior text set behavor _ text _3 is 6.25% of the on-line office collaboration information matching the habit analysis condition, and the size of the first cooperative office behavior text set behavor _ text _4 is 6.25% of the on-line office collaboration information matching the habit analysis condition.

In some possible embodiments, the first collaborative office behavior text set behavior _ text _4 may be input into a key behavior detail recognition model, and the key behavior detail recognition model may include a collaborative operation behavior recognition node with a recognition window of 2 × 2, a collaborative operation behavior recognition node with a recognition window of 6 × 6 and an expansion index of 24, a collaborative operation behavior recognition node with a recognition window of 6 × 6 and an expansion index of 36, and a down-sampling node, and by processing of the above model unit, a key behavior detail vector distribution may be obtained. Further, the key behavior detail vector distribution and the first cooperative office behavior text set behavior _ text _1 and the first cooperative office behavior text set behavior _ text _2 may be input into the behavior detail translation model to perform behavior detail translation, and key behavior detail mining information, for example, a key behavior detail estimation result, may be understood as on-line office cooperation information of a positioning tag corresponding to a key behavior detail of a specified office behavior event in the on-line office cooperation information that matches the habit analysis condition is marked.

It can be understood that the first cooperative office behavior text set behavior _ text _4 may be input to the information parsing unit for processing, where the information parsing unit may include a cooperative operation behavior identification node identifying a window 2 × 2, a cooperative operation behavior identification node identifying a window 6 × 6 and having an expansion index of 12, a cooperative operation behavior identification node identifying a window 6 × 6 and having an expansion index of 24, a cooperative operation behavior identification node identifying a window 6 × 6 and having an expansion index of 36, and a down-sampling node, and a personalized office behavior phrase set may be obtained through processing by the cooperative operation behavior identification unit. Further, the key behavior detail content of the personalized office behavior phrase set can be optimized through the key behavior detail estimation result, and the analysis precision is improved.

Further, the key behavior detail estimation result and the personalized office behavior phrase set can be input into a style annotation submodel, and a first behavior detail attention field set is obtained. Further, style annotation processing can be performed on the key behavior detail content through the first cooperative office behavior text set behavior _ text _1 and the first cooperative office behavior text set behavior _ text _2, for example, the key behavior detail estimation result and the first cooperative office behavior text set behavior _ text _2 can be input into a style annotation sub-model, a second behavior detail attention field set corresponding to the first cooperative office behavior text set behavior _ text _2 is obtained, the key behavior detail estimation result and the first cooperative office behavior text set behavior _ text _1 are input into the style annotation sub-model, and a second behavior detail attention field set corresponding to the first cooperative office behavior text set behavior _ text _1 is obtained.

Further, the first behavior detail attention field set, the second behavior detail attention field set corresponding to the first cooperative office behavior text set behavior _ text _2, and the second behavior detail attention field set corresponding to the first cooperative office behavior text set behavior _ text _1 may be loaded to the behavior detail translation model for processing, so as to obtain an office operation habit record as accurate and complete as possible.

For an independently implementable technical solution, after acquiring the office operation habit record of the specified office activity event, the method further comprises: carrying out potential interaction demand identification on the office operation habit record to obtain a potential interaction demand field aiming at the specified office action event; and updating the designated cloud office service through the potential interaction requirement field.

In some possible embodiments, the potential interaction requirement identification may be used to mine an implicit operation requirement of the office user from the office operation habit record, for example, the office operation habit record includes "there are multiple click/withdrawal operations for the GUI area B by the user a", and then the determined potential interaction requirement field may be "there is an interaction function graphics module optimization for the GUI area B for the user a". In this way, the program code corresponding to the GUI area B can be debugged, optimized and updated through the potential interaction requirement field, thereby ensuring that the subsequent office operation requirements of the user are matched as much as possible.

For an independently implementable technical solution, performing potential interaction demand identification on the office operation habit record to obtain a potential interaction demand field for the specified office behavior event, and may be implemented by the following technical solution.

Step31, acquiring an operation habit emphasis content queue from the office operation habit record, and determining an improvised preference emphasis content queue according to the operation habit emphasis content queue, wherein the operation habit emphasis content queue comprises L groups of operation habit emphasis contents which are associated, and L is an integer not less than 1; the impromptu preference profile queue includes L sets of impromptu preference profile content for which there is a connection.

Step32, determining an operation habit phrase set queue through a first phrase mining sub-thread covered by an operation habit analysis thread by combining the operation habit emphasis content queue, wherein the operation habit phrase set queue comprises L operation habit phrase sets; and calling a second phrase mining sub-thread covered by the operation habit analysis thread to determine an improvised preference phrase set queue in combination with the improvised preference emphasis content queue, wherein the improvised preference phrase set queue comprises L improvised preference phrase sets.

Step33, combining the operation habit phrase set queue and the improvised preference phrase set queue, calling an implicit demand mining module covered by the operation habit analysis thread to determine an implicit demand label corresponding to the operation habit emphasis content; and determining a potential interaction requirement field of the operation habit emphasis content queue according to the implicit requirement label.

In steps 31-33, the operation habit analysis thread can be a convolutional neural network, related functional units can be understood as a network layer or a sub-model, the impromptu preference can be a temporary favorite habit preference in the operation habits, and the impromptu preference in the operation habits is considered, so that noise in determining the implicit demand label can be reduced as much as possible, and the precision and the reliability of the implicit demand label are guaranteed. In this way, the potential interaction requirement field of the operation habit emphasis content queue can be determined from the preset relational database through the implicit requirement tag.

For an independently implementable technical solution, the determining, by combining the operation habit phrase set queue and the improvised preference phrase set queue, an implicit requirement mining module covered by the operation habit analysis thread to determine an implicit requirement tag corresponding to the operation habit emphasis content queue includes: calling a first interaction interface attention sub-thread covered by the operation habit analysis thread to determine L first multi-dimensional arrays by combining the operation habit phrase set queue, wherein each first multi-dimensional array corresponds to one operation habit phrase set; calling a second interaction interface attention sub-thread covered by the operation habit analysis thread to determine L second multi-dimensional arrays by combining the improvised preference phrase set queue, wherein each second multi-dimensional array corresponds to one improvised preference phrase set; combining the L first multi-dimensional arrays and the L second multi-dimensional arrays to obtain L target multi-dimensional arrays, wherein each target multi-dimensional array comprises a first multi-dimensional array and a second multi-dimensional array; and calling the implicit demand mining module covered by the operation habit analysis thread to determine the implicit demand label corresponding to the operation habit emphasis content queue by combining the L target multi-dimensional arrays. By the design, the operation habit phrase set is recorded through the multidimensional data, the data processing and mining efficiency can be improved, and the implicit requirement label can be determined efficiently.

For an independently implementable technical solution, the invoking, in combination with the operation habit phrase set queue, a first interactive interface attention sub-thread covered by the operation habit analysis thread to determine L first multidimensional arrays includes: for each group of operation habit phrase sets in the operation habit phrase set queue, calling an overall optimization unit covered by the first interaction interface attention sub-thread to determine a first overall optimization phrase set, wherein the first interaction interface attention sub-thread belongs to the operation habit analysis thread; for each group of operation habit phrase sets in the operation habit phrase set queue, calling a targeted optimization unit covered by the attention sub-thread of the first interactive interface to determine a first targeted optimization phrase set; for each group of operation habit phrase sets in the operation habit phrase set queue, combining the first overall optimization phrase set and the first pertinence optimization phrase set, and calling a phrase mining unit covered by the attention sub-thread of the first interactive interface to determine a first linkage phrase set; and for each group of operation habit phrase sets in the operation habit phrase set queue, combining the first linkage phrase set and the operation habit phrase set, and calling a first pertinence optimization unit covered by the attention sub-thread of the first interactive interface to determine a first multi-dimensional array.

For an independently implementable solution, said determining, in conjunction with said improvised set of preference phrases queue, L second multidimensional arrays by invoking a second interactive interface attention sub-thread encompassed by said operational habit analysis thread, comprises: for each group of improvised preference phrase sets in the improvised preference phrase set queue, calling an overall optimization unit covered by a second interactive interface attention sub-thread to determine a second overall optimization phrase set, wherein the second interactive interface attention sub-thread belongs to the operation habit analysis thread; for each group of improvised preference phrase sets in the improvised preference phrase set queue, calling a pertinence optimization unit covered by the attention sub-thread of the second interactive interface to determine a second pertinence optimization phrase set; for each group of improvised preference phrase sets in the improvised preference phrase set queue, determining a second set of linkage phrases by invoking a phrase mining unit covered by the second interactive interface attention sub-thread in combination with the second global optimized phrase set and the second targeted optimized phrase set; for each group of improvised preference phrase sets in the improvised preference phrase set queue, in conjunction with the second set of linked phrases and the improvised preference phrase sets, invoking a second targeted optimization unit encompassed by the second interactive interface interest sub-thread to determine a second multi-dimensional array.

It can be understood that, through the above, the detail expression quality of the first multidimensional array and the second multidimensional array can be ensured, and the scale of the first multidimensional array and the second multidimensional array can be reduced, thereby avoiding the additional calculation resource overhead.

For an independently implementable embodiment, L is an integer greater than 1. Based on this, the step of calling the implicit demand mining module covered by the operation habit analysis thread to determine the implicit demand label corresponding to the operation habit emphasis content queue by combining the L target multidimensional arrays includes: calling response attention sub-threads covered by the operation habit analysis thread to determine a linkage multi-dimensional array by combining the L target multi-dimensional arrays, wherein the linkage multi-dimensional array is determined according to the L target multi-dimensional arrays and L delay coefficients, and each target multi-dimensional array corresponds to one delay coefficient; and calling the implicit demand mining module covered by the operation habit analysis thread to determine the implicit demand label corresponding to the operation habit emphasis content queue by combining the linkage multidimensional array.

For an independently implementable technical solution, the invoking of the responding attention sub-thread covered by the operation habit analysis thread by combining the L target multi-dimensional arrays to determine the linkage multi-dimensional array includes: calling a first thread branch covered by the response attention sub-thread to determine L first local character strings by combining the L target multi-dimensional arrays, wherein the response attention sub-thread belongs to the operation habit analysis thread; calling a second thread branch covered by the response attention sub-thread to determine L second local character strings by combining the L first local character strings; determining L delay coefficients according to the L second local character strings, wherein each delay coefficient corresponds to a target multi-dimensional array; and determining the linkage multi-dimensional array according to the L target multi-dimensional arrays and the L delay coefficients. By the design, the precision and the reliability of the linkage multidimensional array can be ensured.

Fig. 3 is an architecture diagram illustrating an application environment of the method for analyzing the big data of the user habit for handling the digital cloud office according to the embodiment of the present disclosure, where the application environment of the method for analyzing the big data of the user habit for handling the digital cloud office may include a digital cloud office system 100 and an online office operation terminal 200 that communicate with each other. Based on this, the digital cloud office system 100 and the online office operation end 200 implement or partially implement the method for analyzing the big data of the user habit in handling digital cloud office according to the embodiment of the present disclosure during operation.

The embodiments of the present disclosure have been described above with reference to the accompanying drawings, and have at least the following advantages: the method can be used for mining key behavior details of specified office behavior events in the online office collaboration information matched with the habit analysis conditions, so that the key behavior details of the specified office behavior events are determined, stylized annotation is carried out on the office operation habits of the specified office behavior events through the key behavior details of the specified office behavior events, so that the key behavior details and the first collaborative office behavior text set can be combined to carry out accurate and efficient analysis on the user operation habits of the specified office behavior events, high correlation between the obtained office operation habit records and the specified office behavior events is ensured, and the noise operation habits mixed with other office behavior events in the office operation habit records are avoided.

The above description is only for the preferred embodiment of the present disclosure, and is not intended to limit the scope of the present disclosure.

Claims

1. A big data analysis method for user habit in handling digital cloud office is characterized by being applied to a digital cloud office system, and at least comprising the following steps:

responding to the authenticated operation habit analysis request, performing cooperative operation behavior identification operation on the online office cooperation information matched with the habit analysis conditions, and acquiring a first cooperative office behavior text set;

based on the first collaborative office behavior text set, performing key behavior detail mining operation on a specified office behavior event in the online office collaboration information matching the habit analysis conditions, and acquiring key behavior detail mining information of the specified office behavior event, wherein the key behavior detail mining information is used for reflecting key behavior details of the specified office behavior event;

and performing user operation habit analysis on the specified office behavior event in the online office collaboration information matched with the habit analysis conditions based on the key behavior detail mining information and the first collaborative office behavior text set, and acquiring an office operation habit record of the specified office behavior event.

2. The method of claim 1, wherein performing a collaborative behavior recognition operation on online office collaboration information matching the habit analysis conditions to obtain a first set of collaborative office behavior texts comprises:

and performing cooperative operation behavior recognition operation on the online office cooperation information matched with the habit analysis conditions according to at least two cooperative operation behavior recognition units, and acquiring first cooperative office behavior text sets respectively corresponding to the at least two cooperative operation behavior recognition units.

3. The method as claimed in claim 2, wherein performing a key behavior detail mining operation on a specific office behavior event in the online office collaboration information matching the habit analysis conditions based on the first collaborative office behavior text set to obtain key behavior detail mining information of the specific office behavior event comprises:

performing key behavior detail mining on a first cooperative office behavior text set corresponding to the last cooperative operation behavior recognition unit to obtain key behavior detail vector distribution;

and performing behavior detail translation on the key behavior detail vector distribution and a first cooperative office behavior text set of at least one cooperative operation behavior recognition unit in the at least two cooperative operation behavior recognition units to acquire key behavior detail mining information.

4. The method as claimed in claim 3, wherein performing behavior detail translation on the key behavior detail vector distribution and the first co-office behavior text set of at least one of the at least two co-operating behavior recognition units to obtain key behavior detail mining information comprises:

performing detail content derivation operation on the key behavior detail vector distribution to obtain first derived behavior detail vector distribution;

integrating the first derived behavior detail vector distribution and the first cooperative office behavior text set to obtain a first office behavior composite description set;

and performing windowed mining operation on the first office behavior composite description set to acquire the detail mining information of the key behaviors.

5. The method as claimed in claim 1, wherein the step of performing user operation habit analysis on a specified office behavior event in the online office collaboration information matching the habit analysis conditions based on the key behavior detail mining information and the first collaborative office behavior text set to obtain an office operation habit record of the specified office behavior event comprises:

performing personalized operation habit mining on the first collaborative office behavior text set to obtain a personalized office behavior phrase set;

mining information and the personalized office behavior phrase set based on the key behavior details, and acquiring a behavior detail attention field set;

and performing behavior detail translation on the behavior detail attention field set to acquire the office operation habit record.

6. The method according to claim 5, wherein the first text sets of cooperative office activities include at least two first text sets of cooperative office activities resulting from a cooperative operation behavior recognition operation performed by at least two cooperative operation behavior recognition units; the behavior detail attention field set comprises a first behavior detail attention field set corresponding to the personalized office behavior phrase set and a second behavior detail attention field set corresponding to a first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of the at least two cooperative operation behavior recognition units;

acquiring a behavior detail attention field set based on the key behavior detail mining information and the personalized office behavior phrase set, wherein the behavior detail attention field set comprises the following steps:

performing first key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information to obtain a first behavior detail attention field set;

and respectively carrying out second key behavior detail attention operation on the first cooperative office behavior text set of at least one cooperative operation behavior recognition unit of the at least two cooperative operation behavior recognition units through the key behavior detail mining information, and acquiring second behavior detail attention field sets corresponding to the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit.

7. The method of claim 6, wherein the key behavior detail mining information comprises a key behavior detail estimation result, wherein performing a first key behavior detail attention operation on the personalized office behavior phrase set through the key behavior detail mining information to obtain a first behavior detail attention field set comprises:

calculating the key behavior detail estimation result and the visual category value of the corresponding operation behavior track in the personalized office behavior phrase set to obtain a second cooperative office behavior text set;

integrating the second cooperative office behavior text set and the key behavior detail estimation result to obtain a third cooperative office behavior text set;

performing key behavior detail attention operation on the third cooperative office behavior text set to obtain the first behavior detail attention field set;

performing a key behavior detail attention operation on the third cooperative office behavior text set, and acquiring the first behavior detail attention field set, including:

performing separated windowing mining operation on the third cooperative office behavior text set to obtain a separated mining list;

performing cooperative operation behavior identification operation on the separated mining list to obtain a cooperative behavior identification list;

and calculating the corresponding operation behavior tracks of the cooperative behavior identification list and the separated mining list to obtain the first behavior detail attention field set.

8. The method as claimed in claim 6, wherein the key behavior detail mining information includes a key behavior detail estimation result, wherein performing a second key behavior detail attention operation on a first cooperative office behavior text set of at least one of the at least two cooperative operation behavior recognition units respectively through the key behavior detail mining information to obtain a second behavior detail attention field set corresponding to each of the first cooperative office behavior text sets of the at least one cooperative operation behavior recognition unit comprises:

calculating the key behavior detail estimation result and a visual category value of a corresponding operation behavior track in a first cooperative office behavior text set of a p-th cooperative operation behavior identification unit to obtain a fourth cooperative office behavior text set, wherein the p-th cooperative operation behavior identification unit is one of the at least one cooperative operation behavior identification unit;

integrating the fourth cooperative office behavior text set and the key behavior detail estimation result to obtain a fifth cooperative office behavior text set;

and performing key behavior detail attention operation on the fifth cooperative office behavior text set to obtain a second behavior detail attention field set corresponding to the first cooperative office behavior text set of the pth cooperative operation behavior identification unit.

9. The method of claim 1, wherein the method for analyzing big data of user habits in handling digital cloud office is implemented by means of an AI machine learning model, the method further comprising:

loading office cooperation information on a reference line to the AI machine learning model, and acquiring reference key behavior detail distribution and reference office operation habit records of specified office behavior events in the office cooperation information on the reference line;

acquiring habit analysis cost deviation of the AI machine learning model based on the prior knowledge of the office cooperation information on the reference line and the reference office operation habit record;

acquiring the detail deviation of the key behaviors of the AI machine learning model based on the priori knowledge of the office cooperation information on the reference line and the reference key behavior detail distribution;

performing global calculation on the habit resolution cost deviation and the key behavior detail deviation to obtain model quality evaluation of the AI machine learning model;

improving the AI machine learning model based on the model quality assessment.

10. A digital cloud office system, comprising:

a memory for storing an executable computer program, a processor for implementing the method of any one of claims 1-9 when executing the executable computer program stored in the memory.