CN116503879A

CN116503879A - Threat behavior identification method and device applied to e-commerce platform

Info

Publication number: CN116503879A
Application number: CN202310575525.7A
Authority: CN
Inventors: 于敏
Original assignee: Guangdong Junsi Information Technology Co ltd
Current assignee: GUANGZHOU FTEMALL TECHNOLOGY Co.,Ltd.
Priority date: 2023-05-22
Filing date: 2023-05-22
Publication date: 2023-07-28
Anticipated expiration: 2043-05-22
Also published as: CN116503879B

Abstract

The invention discloses a threat behavior identification method and device applied to an e-commerce platform, wherein the method comprises the following steps: after a certain online session is established, a monitoring server acquires monitoring authorization rights of all participating objects related to the current online session for the current online session; the monitoring server judges whether the current online conversation meets the monitoring conditions according to the monitoring authorization rights of all the participating objects for the current online conversation, monitors conversation content of the current online conversation if the current online conversation meets the monitoring conditions to obtain multi-dimensional monitoring information, judges whether threat behaviors exist in conversation behaviors of the current online conversation based on the multi-dimensional monitoring information and a pre-trained behavior recognition model, matches a processing strategy corresponding to the threat behaviors if the threat behaviors exist in the conversation behaviors of the current online conversation, and executes matched processing operation on relevant responsibility objects of the threat behaviors. Therefore, the invention can improve the recognition efficiency and recognition accuracy of the threat behaviors and can also improve the processing efficiency and processing accuracy of the threat behaviors.

Description

Threat behavior identification method and device applied to e-commerce platform

Technical Field

The invention relates to the technical field of big data analysis, in particular to a threat behavior identification and device applied to an electronic commerce platform.

Background

With the rapid development of internet technology, the intelligent functions of the e-commerce platform are more and more, so that a user can purchase required articles (entity articles or virtual articles) on the e-commerce platform, and can communicate with the e-commerce platform customer service or the merchant customer service where the e-commerce platform resides on line based on the online service function provided by the e-commerce platform, thereby being beneficial to improving the use experience of the user on the e-commerce platform.

However, it is found in practice that when two communicating parties perform a conversation through an online service function, there is a behavior that an individual with great significance deliberately finds out and threatens the other party in language, which is not beneficial to building a good environment for the online service function of the e-commerce platform. For these threat behaviors, after the communication party reports or complains, the related personnel of the electronic commerce platform can identify whether the complaint behavior is the threat behavior based on the evidence provided by the complaint party and/or the complaint party, which is not beneficial to improving the identification efficiency and the identification accuracy of the threat behavior.

It can be seen that how to improve the recognition efficiency and recognition accuracy of the threat behaviors existing in the e-commerce platform is particularly important.

Disclosure of Invention

The invention aims to solve the technical problem of providing a threat behavior recognition method and device applied to an e-commerce platform, which can improve the recognition efficiency and recognition accuracy of threat behaviors.

In order to solve the technical problem, a first aspect of the embodiment of the invention discloses a threat behavior identification method applied to an e-commerce platform, the method is applied to a monitoring server of the e-commerce platform, and the method comprises the following steps:

after a certain online conversation is established, the monitoring server acquires monitoring authorization rights of all participation objects related to a current online conversation for the current online conversation, wherein all participation objects comprise an active participation party and at least one passive participation party, and the active participation party is an initiator of the current online conversation;

the monitoring server judges whether the current online conversation meets monitoring conditions according to the monitoring authorization rights of all the participating objects for the current online conversation, and monitors conversation content of the current online conversation when the current online conversation meets the monitoring conditions to obtain multi-dimensional monitoring information;

and the monitoring server judges whether a threat behavior exists in the dialogue behavior of the current online dialogue based on the multidimensional monitoring information and a pre-trained behavior recognition model, matches a processing strategy corresponding to the threat behavior when judging that the threat behavior exists in the dialogue behavior of the current online dialogue, and executes processing operation matched with the processing strategy on related responsible objects of the threat behavior.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the monitoring server performs a processing operation matched with the processing policy on the relevant responsibility object of the threat behavior, including:

when the relevant responsibility objects of the threat behaviors comprise at least two responsibility objects, the monitoring server determines the responsibility duty ratio corresponding to each responsibility object according to the multi-dimensional monitoring information;

the monitoring server judges whether a corresponding target historical online conversation exists in the current online conversation or not based on the monitored historical online conversation, and when judging that the corresponding target historical online conversation exists in the current online conversation, the monitoring server determines a responsibility duty ratio adjustment coefficient of each responsibility object based on historical multidimensional monitoring information corresponding to the target historical online conversation;

for each responsibility object, the monitoring server performs an adjustment operation on the responsibility duty ratio of the responsibility object according to the responsibility duty ratio adjustment coefficient of the responsibility object to update the responsibility duty ratio of the responsibility object, and performs a processing operation matched with the processing policy on the responsibility object according to the updated responsibility duty ratio of the responsibility object.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the multidimensional information includes contextual dialogue text information of the current online dialogue and picture information, and the picture information includes a plurality of dialogue pictures involved in the current online dialogue, a picture type of each of the dialogue pictures, and a picture content of each of the dialogue pictures; the picture type of the dialogue picture is used for indicating whether the dialogue picture is a commodity picture or a non-commodity picture, and the picture content of the dialogue picture comprises a foreground target and additional characters in the dialogue picture; all the dialogue pictures comprise single dialogue pictures and dialogue pictures analyzed from video content;

the monitoring server judges whether a threat behavior exists in the dialogue behavior of the current online dialogue based on the multidimensional monitoring information and a pre-trained behavior recognition model, and the monitoring server comprises the following steps:

the monitoring server classifies the multi-dimensional monitoring information to obtain text monitoring information and picture monitoring information;

the monitoring server filters the text monitoring information to obtain text information to be identified, and filters the picture monitoring information to obtain picture information to be identified;

The monitoring server inputs the character information to be identified into a pre-trained character behavior identification model to obtain a character threat behavior identification result, and inputs the picture information to be identified into a pre-trained picture behavior identification model to obtain a picture threat behavior identification result;

the monitoring server judges whether a threat behavior exists in the dialogue behavior of the current online dialogue according to the character threat behavior recognition result and the picture threat behavior recognition result;

when the text threat behavior exists in the text threat behavior recognition result or when the picture threat behavior exists in the picture threat behavior recognition result, threat behavior exists in the dialogue behavior of the current online dialogue.

In an optional implementation manner, in the first aspect of the embodiment of the present invention, the text monitoring information includes a preset keyword in text content and/or target text information including preset semantics;

the monitoring server filters the text monitoring information to obtain text information to be identified, and the text information to be identified comprises the following steps:

the monitoring server determines the sending object of each target text message, and filters all the target text messages according to the predetermined filtering control conditions and by combining all the related information of the sending objects to obtain text messages to be identified;

The related information of the sending object comprises at least one of dialogue style of the sending object, emotion information when the sending object sends the target text information, initiative degree of the sending object sending the target text information, and feedback information of other participating objects received after the sending object sends the target text information aiming at the target text information.

In an optional implementation manner, in a first aspect of the embodiment of the present invention, the filtering, by the monitoring server, the picture monitoring information to obtain picture information to be identified includes:

the monitoring server filters the dialogue pictures of the expression package category from all the dialogue pictures based on the picture type and the picture content of each dialogue picture to obtain a first filtering result;

the monitoring server filters out dialogue pictures with corresponding foreground targets of a preset target type and corresponding additional characters of a preset character type from the first filtering result to obtain a second filtering result;

the monitoring server determines the picture information corresponding to all the dialogue pictures included in the second filtering result as picture information to be identified;

The additional characters of the preset character type are additional characters with preset keywords and/or preset semantics in the character content.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, after determining that the threat behavior exists in the session behavior of the current online session, the method further includes:

the monitoring server continuously monitors the follow-up dialogue content corresponding to the current online dialogue;

and the monitoring server judges the resolution degree corresponding to the threat behavior based on the follow-up dialogue content, judges whether the resolution degree indicates that the threat behavior is solved, and when the judgment result is negative, triggers the execution of the processing strategy corresponding to the threat behavior, and executes the processing operation matched with the processing strategy on the relevant responsible object of the threat behavior.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the method further includes:

after judging that the threat behavior exists in the dialogue behavior of the current online dialogue, the monitoring server monitors the severity change trend of the threat behavior along with the duration of the current online dialogue;

The monitoring server judges whether the severity trend indicates that the severity of the threat behavior is deepened along with the increase of the duration of the current online conversation, and when the severity trend indicates that the severity trend is deepened along with the increase of the duration of the current online conversation, generates a guiding conversation robot model matched with the threat behavior, virtualizes the guiding conversation robot model into a conversation robot, and adds the conversation robot as a new participation object to the current online conversation to guide the threat behavior

The second aspect of the embodiment of the invention discloses a threat behavior identification device applied to an e-commerce platform, wherein the device is applied to a monitoring server of the e-commerce platform and comprises the following components:

the system comprises an acquisition module, a monitoring and authorization module and a control module, wherein the acquisition module is used for acquiring monitoring and authorization rights of all participation objects related to a current online conversation for the current online conversation after the online conversation is established, and all the participation objects comprise an active participation party and at least one passive participation party, and the active participation party is an initiator of the current online conversation;

the judging module is used for judging whether the current online conversation meets the monitoring condition according to the monitoring authorization authority of all the participating objects for the current online conversation;

The monitoring module is used for monitoring the dialogue content of the current online dialogue to obtain multi-dimensional monitoring information when the current online dialogue meets the monitoring conditions;

the behavior recognition module is used for judging whether threat behaviors exist in the dialogue behaviors of the current online dialogue or not based on the multidimensional monitoring information and a pre-trained behavior recognition model;

and the behavior processing module is used for matching a processing strategy corresponding to the threat behavior when judging that the threat behavior exists in the dialogue behavior of the current online dialogue, and executing processing operation matched with the processing strategy on the relevant responsible object of the threat behavior.

As an optional implementation manner, in the second aspect of the embodiment of the present invention, a specific manner of the behavior processing module performing, on the relevant responsible object of the threat behavior, a processing operation matched with the processing policy includes:

when the relevant responsibility objects of the threat behaviors comprise at least two responsibility objects, determining the responsibility duty ratio corresponding to each responsibility object according to the multi-dimensional monitoring information;

judging whether the current online conversation has a corresponding target historical online conversation or not based on the monitored historical online conversation, and determining a duty ratio adjustment coefficient of each duty object based on historical multidimensional monitoring information corresponding to the target historical online conversation when judging that the current online conversation has the corresponding target historical online conversation;

As an optional implementation manner, in the second aspect of the embodiment of the present invention, the multidimensional information includes contextual dialogue text information of the current online dialogue and picture information, and the picture information includes a plurality of dialogue pictures involved in the current online dialogue, a picture type of each of the dialogue pictures, and a picture content of each of the dialogue pictures; the picture type of the dialogue picture is used for indicating whether the dialogue picture is a commodity picture or a non-commodity picture, and the picture content of the dialogue picture comprises a foreground target and additional characters in the dialogue picture; all the dialogue pictures comprise single dialogue pictures and dialogue pictures analyzed from video content;

and the behavior recognition module judges whether a threat behavior exists in the dialogue behavior of the current online dialogue based on the multidimensional monitoring information and a pre-trained behavior recognition model, wherein the specific mode comprises the following steps of:

Classifying the multi-dimensional monitoring information to obtain text monitoring information and picture monitoring information;

filtering the text monitoring information to obtain text information to be identified, and filtering the picture monitoring information to obtain picture information to be identified;

inputting the character information to be identified into a pre-trained character behavior identification model to obtain a character threat behavior identification result, and inputting the picture information to be identified into a pre-trained picture behavior identification model to obtain a picture threat behavior identification result;

judging whether a threat behavior exists in the dialogue behavior of the current online dialogue according to the character threat behavior recognition result and the picture threat behavior recognition result;

In a second aspect of the embodiment of the present invention, the text monitoring information includes a preset keyword and/or target text information including preset semantics in text content;

The specific mode of filtering the text monitoring information by the behavior recognition module to obtain the text information to be recognized comprises the following steps:

determining a sending object of each target text message, and filtering all the target text messages according to a predetermined filtering control condition and combining relevant information of all the sending objects to obtain text messages to be identified;

In a second aspect of the embodiment of the present invention, the specific manner in which the behavior recognition module filters the picture monitoring information to obtain the picture information to be recognized includes:

based on the picture type and the picture content of each dialogue picture, filtering dialogue pictures of the expression package category from all the dialogue pictures to obtain a first filtering result;

Filtering dialogue pictures with corresponding foreground targets of a preset target type and corresponding additional characters of a preset character type from the first filtering result to obtain a second filtering result;

determining picture information corresponding to all dialogue pictures included in the second filtering result as picture information to be identified;

As an optional implementation manner, in the second aspect of the embodiment of the present invention, the monitoring module is further configured to continuously monitor subsequent dialogue contents corresponding to the current online dialogue after the behavior recognition module determines that the threat behavior exists in the dialogue behavior of the current online dialogue;

and the judging module judges the resolution degree corresponding to the threat behavior based on the follow-up dialogue content, judges whether the resolution degree indicates that the threat behavior is solved, and when the judgment result is negative, triggers the behavior processing module to execute the processing strategy corresponding to the threat behavior and execute the processing operation matched with the processing strategy on the relevant responsible object of the threat behavior.

As an optional implementation manner, in the second aspect of the embodiment of the present invention, the monitoring module is further configured to monitor, after determining that the threat behavior exists in the session behavior of the current online session, a severity trend of the threat behavior with respect to a duration of the current online session;

the judging module is further configured to judge whether the severity change trend indicates that the severity of the threat behavior deepens with an increase in duration of the current online session;

and, the apparatus further comprises:

and the guiding module is used for generating a guiding dialogue robot model matched with the threat behaviors when the judging module judges that the severity degree variation trend shows that the severity degree of the threat behaviors deepens along with the increase of the duration time of the current online dialogue, virtualizing the guiding dialogue robot model into a dialogue robot, and adding the dialogue robot as a new participation object into the current online dialogue so as to guide the threat behaviors.

The third aspect of the present invention discloses another threat behavior identification apparatus applied to an e-commerce platform, the apparatus being applied to a monitoring server of the e-commerce platform, the apparatus comprising:

A memory storing executable program code;

a processor coupled to the memory;

the processor invokes the executable program code stored in the memory to execute the threat behavior identification method disclosed in the first aspect of the invention and applied to the e-commerce platform.

A fourth aspect of the invention discloses a computer storage medium storing computer instructions that, when invoked, are adapted to perform the threat behavior identification disclosed in the first aspect of the invention as applied to an e-commerce platform.

Compared with the prior art, the embodiment of the invention has the following beneficial effects:

in the embodiment of the invention, after a certain online conversation is established, a monitoring server acquires monitoring authorization rights of all participating objects related to the current online conversation for the current online conversation; the monitoring server judges whether the current online conversation meets the monitoring conditions according to the monitoring authorization rights of all the participating objects for the current online conversation, monitors conversation content of the current online conversation if the current online conversation meets the monitoring conditions to obtain multi-dimensional monitoring information, judges whether threat behaviors exist in conversation behaviors of the current online conversation based on the multi-dimensional monitoring information and a pre-trained behavior recognition model, matches a processing strategy corresponding to the threat behaviors if the threat behaviors exist in the conversation behaviors of the current online conversation, and executes matched processing operation on relevant responsibility objects of the threat behaviors. Therefore, the invention can realize the identification of the threat behaviors in the electronic commerce platform based on the multidimensional monitoring information obtained by monitoring the online dialogue, improves the identification efficiency and the identification accuracy of the threat behaviors, and can also improve the processing efficiency and the processing accuracy of the threat behaviors. In addition, when the relevant responsibility objects of the threat behaviors are processed, the responsibility duty ratio of the responsibility objects can be intelligently determined according to the multi-dimensional monitoring information, and the responsibility duty ratio is adjusted based on the historical multi-dimensional monitoring information corresponding to the corresponding historical online conversations, so that the accuracy of the responsibility duty ratio of the responsibility objects is improved, and further the processing efficiency and the processing reliability of the responsibility objects corresponding to the threat behaviors are improved. In addition, the multidimensional monitoring information not only comprises contextual dialogue text information, but also comprises picture information, the picture information not only relates to dialogue pictures, but also further relates to picture content and picture types of the dialogue pictures, the picture content of the dialogue pictures can comprise foreground targets and additional words in the dialogue, the multidimensional monitoring information is beneficial to improving the recognition comprehensiveness and recognition accuracy of threat behaviors, further, when threat behavior recognition is realized based on the multidimensional monitoring information, the text monitoring information and the picture monitoring information obtained by classification can be filtered, the recognition accuracy is improved, the recognition efficiency is further improved, the text threat behavior recognition result and the picture threat behavior recognition result can be respectively determined based on the filtering result, and the threat behavior recognition accuracy is improved. In addition, the filtering accuracy of the text monitoring information can be improved based on the related information corresponding to the sending object of the key text information and combined with the pre-set filtering control condition, and the intelligent filtering of the picture monitoring information can be realized by combining the picture type, the picture content, the picture foreground target and the additional text type in the picture, so that the filtering efficiency and the filtering accuracy of the picture monitoring information are improved. In addition, after the threat behavior of the online dialogue is identified, the degree of the threat behavior can be continuously monitored, and if the threat behavior is not solved, the corresponding processing strategy is matched, so that the reliability of the matched processing strategy is improved. In addition, when the threat behavior is identified and the severity of the threat behavior is gradually deepened, the intelligent guiding of the threat behavior is realized through the virtual generated dialogue robot, so that the intelligent guiding method of the threat behavior is provided, and the gradual deepening degree of the threat behavior is reduced.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a threat behavior identification method applied to an e-commerce platform disclosed in an embodiment of the invention;

FIG. 2 is a schematic structural diagram of a threat behavior recognition apparatus for an e-commerce platform according to an embodiment of the present invention;

FIG. 3 is a schematic structural diagram of another threat behavior identification apparatus for an e-commerce platform according to an embodiment of the invention;

fig. 4 is a schematic structural diagram of a threat behavior recognition apparatus applied to an e-commerce platform according to another embodiment of the present invention.

Description of the embodiments

In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The terms first, second and the like in the description and in the claims and in the above-described figures are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, article, or device that comprises a list of steps or elements is not limited to the list of steps or elements but may, in the alternative, include other steps or elements not expressly listed or inherent to such process, method, article, or device.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.

The invention discloses a threat behavior identification method and device applied to an e-commerce platform, which can realize the identification of threat behaviors in the e-commerce platform based on multidimensional monitoring information obtained by monitoring online conversations, improve the identification efficiency and the identification accuracy of the threat behaviors, and also improve the processing efficiency and the processing accuracy of the threat behaviors. The following will describe in detail.

Examples

Referring to fig. 1, fig. 1 is a flow chart of a threat behavior recognition method applied to an e-commerce platform according to an embodiment of the present invention. The method described in fig. 1 may be applied to a monitoring server of an e-commerce platform, where the monitoring server may be a local server or a cloud server, and the embodiment of the present invention is not limited. As shown in fig. 1, the threat behavior identification method applied to the e-commerce platform may include the following steps:

101. after a certain online session is established, the monitoring server acquires monitoring authorization rights of all participating objects related to the current online session for the current online session.

All the participation objects comprise an active participant and at least one passive participant, the active participant is an initiator of the current online conversation, the passive participant comprises a participant invited by the active participant and can also comprise other participants invited by the required passive participant, and the embodiment of the invention is not limited.

102. And the monitoring server judges whether the current online conversation meets the monitoring conditions according to the monitoring authorization rights of all the participating objects for the current online conversation, and monitors the conversation content of the current online conversation when the current online conversation meets the monitoring conditions to obtain multi-dimensional monitoring information.

In the embodiment of the invention, the monitoring server judges whether the current online conversation meets the monitoring condition according to the monitoring authority of all the participated objects for the current online conversation, and the method comprises the following steps:

the monitoring server monitors whether the monitoring authority of the active participant for the current online conversation is the authority, if yes, the current online conversation is determined to meet the monitoring condition, if not, the number proportion of the passive participants, which are aiming at the monitoring authority of the current online conversation and are the authority, is monitored, whether the number proportion is larger than or equal to a preset proportion threshold value is judged, and if yes, the current online conversation is determined to meet the monitoring condition.

Therefore, before monitoring the current online conversation, the embodiment of the invention judges the monitoring authorization authority, which is beneficial to improving the monitoring legitimacy and the monitoring effectiveness of the current online conversation.

103. The monitoring server judges whether threat behaviors exist in the dialogue behaviors of the current online dialogue or not based on the multidimensional monitoring information and a pre-trained behavior recognition model.

The behavior recognition model is obtained by training an initial behavior recognition model to be converged based on a large number of sample multidimensional monitoring information and corresponding threat behavior labeling results, and the pre-trained behavior recognition model is used for recognizing at least one threat behavior existing in a current online dialogue, and further, at least one of the types of the existing threat behaviors, the probability of the threat behaviors, the severity of the threat behaviors and the like can be further included.

Optionally, the multidimensional information includes contextual dialogue text information of the current online dialogue and picture information including a plurality of dialogue pictures involved in the current online dialogue, a picture type of each dialogue picture, and a picture content of each dialogue picture. Further, the picture type of the dialogue picture is used for indicating whether the dialogue picture is a commodity picture or a non-commodity picture, and the picture content of the dialogue picture comprises a foreground target and additional text in the dialogue picture. Still further, all of the conversation pictures include a single conversation picture and conversation pictures parsed from the video content. It should be noted that, the commodity picture may be a physical commodity picture or a virtual commodity picture, where the virtual commodity picture corresponds to a service class commodity, and the virtual commodity picture may specifically represent service standard information and/or service result information corresponding to the service class commodity.

Still further, the trained behavior recognition model is continuously optimized according to the updating of the sample, so that the accuracy of threat behaviors of the behavior recognition model in the online dialogue recognition is improved.

104. When judging that the threat behavior exists in the dialogue behavior of the current online dialogue, the monitoring server matches the processing strategy corresponding to the threat behavior and executes the processing operation matched with the processing strategy on the relevant responsibility object of the threat behavior.

In the embodiment of the invention, the monitoring server can store different processing strategies corresponding to different threat behaviors. Further, the monitoring server matches a processing policy corresponding to the threat behavior may include:

when a plurality of processing strategies exist in a certain threat behavior, the monitoring server can further calculate the matching degree of the current threat behavior and each processing strategy according to relevant information (such as responsible party information, behavior probability information of the threat behavior, current severity of the threat behavior, severity change trend information of the threat behavior and the like) corresponding to the threat behavior, and screen the processing strategies (such as warning, forbidden language processing, alarm processing, dialogue content covering processing and the like) with the matching degree meeting the preset matching degree conditions (such as highest matching degree conditions) so as to be beneficial to improving the matching degree of the matched processing strategies and the threat behavior existing in the current online dialogue, and further beneficial to improving the processing efficiency and processing accuracy of the threat behavior.

Therefore, the method described by the embodiment of the invention can realize the identification of the threat behaviors in the e-commerce platform based on the multidimensional monitoring information obtained by monitoring the online conversation, improves the identification efficiency and the identification accuracy of the threat behaviors, and can also improve the processing efficiency and the processing accuracy of the threat behaviors.

In an alternative embodiment, the monitoring server performs processing operations matching the processing policy on relevant responsible objects of the threat activity, comprising:

when the related responsibility objects of the threat behaviors comprise at least two responsibility objects, the monitoring server determines the responsibility duty ratio corresponding to each responsibility object according to the multi-dimensional monitoring information;

the monitoring server judges whether the current online conversation has a corresponding target historical online conversation or not based on the monitored historical online conversation, and when judging that the current online conversation has the corresponding target historical online conversation, the monitoring server determines a duty ratio adjustment coefficient of each responsible object based on historical multidimensional monitoring information corresponding to the target historical online conversation;

for each responsibility object, the monitoring server performs an adjustment operation on the responsibility ratio of the responsibility object according to the responsibility ratio adjustment coefficient of the responsibility object to update the responsibility ratio of the responsibility object, and performs a processing operation matched with the processing policy on the responsibility object according to the updated responsibility ratio of the responsibility object.

Therefore, when the relevant responsibility objects of the threat behaviors are processed, the responsibility duty ratio of the responsibility objects can be intelligently determined according to the multi-dimensional monitoring information, and the responsibility duty ratio is adjusted based on the historical multi-dimensional monitoring information corresponding to the conversations on the corresponding historical lines, so that the accuracy of the responsibility duty ratio of the responsibility objects is improved, and further the processing efficiency and the processing reliability of the processing of the responsibility objects corresponding to the threat behaviors are improved.

In another alternative embodiment, the monitoring server determines whether a threat behavior exists in the dialogue behavior of the current online dialogue based on the multidimensional monitoring information and the pre-trained behavior recognition model, and may include:

and the monitoring server judges whether the threat behaviors exist in the dialogue behaviors of the current online dialogue according to the character threat behavior recognition result and the picture threat behavior recognition result.

When the character threat behavior exists in the character threat behavior recognition result or when the picture threat behavior exists in the picture threat behavior recognition result, threat behavior exists in the dialogue behavior of the current online dialogue.

It can be seen that this optional embodiment can be favorable to improving threat behavior's discernment comprehensiveness and discernment accuracy through aforementioned multidimensional monitoring information, further, when realizing threat behavior discernment based on multidimensional monitoring information, can filter the text monitoring information and the picture monitoring information that obtain by classification, be favorable to further improving discernment efficiency when improving discernment accuracy, and can also confirm text threat behavior discernment result and picture threat behavior discernment result respectively based on filtering result, be favorable to improving threat behavior discernment accuracy's discernment comprehensiveness simultaneously.

In yet another alternative embodiment, the text monitoring information includes a first preset keyword in the text content and/or a target text information including a first preset semantic. And the monitoring server filters the text monitoring information to obtain text information to be identified, and the method comprises the following steps:

the monitoring server determines the sending object of each target text message, and filters all the target text messages according to the predetermined filtering control conditions and by combining the related information of all the sending objects, so as to obtain the text message to be identified. For example, this alternative embodiment can filter out threatening textual information that is laughing.

It can be seen that the filtering accuracy of the text monitoring information can be improved by combining the preset filtering control conditions based on the related information corresponding to the sending object of the key text information.

In yet another optional embodiment, the monitoring server filters the picture monitoring information to obtain the picture information to be identified, including:

the monitoring server filters dialogue pictures of the expression package category from all dialogue pictures based on the picture type and the picture content of each dialogue picture to obtain a first filtering result;

the monitoring server filters out dialogue pictures (such as dialogue pictures related to the filtered commodity) with corresponding foreground targets as preset target types and corresponding additional characters as preset character types from the first filtering result to obtain a second filtering result;

And the monitoring server determines the picture information corresponding to all the dialogue pictures included in the second filtering result as the picture information to be identified.

The additional characters of the preset character type are additional characters with second preset keywords and/or second preset semantics in the character content.

Therefore, the optional embodiment can also be combined with the picture type, the picture content, the picture foreground target and the additional text type in the picture to realize intelligent filtering of the picture monitoring information, thereby being beneficial to improving the filtering efficiency and the filtering accuracy of the picture monitoring information.

In yet another alternative embodiment, after determining that there is a threat activity in the conversational activity of the current online conversation, the method further comprises:

and the monitoring server judges the resolution degree corresponding to the threat behavior based on the subsequent dialogue content, judges whether the resolution degree indicates that the threat behavior is solved, and when the judgment result is negative, triggers and executes the processing strategy matched with the threat behavior and executes the processing operation matched with the processing strategy on the relevant responsible object of the threat behavior.

It can be seen that, after the on-line dialogue is identified that the threat behavior exists, the alternative embodiment can also continuously monitor the resolution of the threat behavior, and if the threat behavior is not resolved, the alternative embodiment matches the corresponding processing policy, thereby being beneficial to improving the reliability of the matching processing policy.

In yet another alternative embodiment, the method may further comprise the operations of:

after judging that the threat behavior exists in the dialogue behavior of the current online dialogue, the monitoring server monitors the severity change trend of the threat behavior along with the duration time of the current online dialogue;

the monitoring server judges whether the severity change trend indicates that the severity of the threat behavior deepens along with the increase of the duration of the current online conversation, and when the judgment result is yes, a guiding conversation robot model matched with the threat behavior is generated, the guiding conversation robot model is virtualized into a conversation robot, and the conversation robot is used as a new participation object to be added into the current online conversation so as to guide the threat behavior.

In an alternative embodiment, the generating a guided dialogue robot model that matches the threat behavior may include:

the monitoring server analyzes a plurality of influence factors with deepened threat behaviors based on multidimensional monitoring information of the current upward dialogue, determines relation objects with deepened threat behaviors based on the influence factors, and generates a guiding dialogue robot model matched with the threat behaviors based on one or more of participation positions of each relation object, influence factors corresponding to each relation object, triggering factors of the influence factors corresponding to each relation object, dialogue styles of each relation object, dialogue participation degree of each relation object, responsibility ratio of each relation object and the like when the relation objects are multiple.

Further, if there are multiple participation positions of all the relationship objects, the monitoring server may split the current online conversation into multiple sub-online conversations according to the participation position, generate sub-guidance conversation robot models of each sub-online conversation based on the relationship objects involved in each sub-online conversation for threat actions, virtualize each sub-guidance conversation robot model into sub-conversation robots, and add the sub-conversation robot as a new participation object to the corresponding sub-online conversation to guide the threat actions. Still further, when the threat behavior is guided, the monitoring server monitors the consensus sub-conversation results of the sub-conversation robots and the corresponding relation objects in each sub-line conversation, synchronizes the consensus sub-conversation results to other sub-conversation robots so that the other sub-conversation robots guide new consensus sub-conversation results in the sub-line conversation where the sub-conversation robots are located, synchronizes the new consensus sub-conversation results to the other sub-conversation robots, and monitors and synchronizes the promotion sub-conversation results until the main consensus results which can be commonly used in all sub-line conversations are monitored, and synchronizes the main consensus results to the current online conversation so as to guide the threat behavior.

It can be seen that the alternative embodiment can also realize intelligent guiding of the threat behaviors through the virtually generated dialogue robot when the threat behaviors are recognized and the severity of the threat behaviors is gradually increased, so that an intelligent guiding method of the threat behaviors is provided, and the gradual increase of the threat behaviors is reduced.

Examples

Referring to fig. 2, fig. 2 is a schematic structural diagram of a threat behavior recognition apparatus applied to an e-commerce platform according to an embodiment of the invention. The device of fig. 2 may be applied to a monitoring server of an e-commerce platform, where the monitoring server may be a local server or a cloud server, and embodiments of the present invention are not limited. As shown in fig. 2, the threat behavior identification apparatus applied to the e-commerce platform may include:

the obtaining module 201 is configured to obtain, after a certain online session is established, monitoring authorization rights of all participating objects related to the current online session for the current online session.

The judging module 202 is configured to judge whether the current online session meets the monitoring condition according to the monitoring authorization rights of all the participating objects for the current online session.

The determining module 202 determines whether the current online session meets the monitoring condition according to the monitoring authorization rights of all the participating objects for the current online session, which may include:

the monitoring active participants determine whether the monitoring authorization authority of the current online conversation is the authorization authority, if yes, the current online conversation is determined to meet the monitoring condition, if not, the number proportion of the passive participants, which are aiming at the monitoring authorization authority of the current online conversation and are the authorization authority, in all the passive participants is monitored, and whether the number proportion is larger than or equal to a preset proportion threshold value is determined, if yes, the current online conversation is determined to meet the monitoring condition.

And the monitoring module 203 is configured to monitor the session content of the current online session to obtain multi-dimensional monitoring information when the judging module 202 judges that the current online session meets the monitoring condition.

The behavior recognition module 204 is configured to determine whether a threat behavior exists in the session behavior of the current online session based on the multidimensional monitoring information and a pre-trained behavior recognition model.

The behavior processing module 205 is configured to, when the behavior recognition module 204 determines that a threat behavior exists in the session behavior of the current online session, match a processing policy corresponding to the threat behavior, and perform a processing operation matched with the processing policy on an associated responsible object of the threat behavior.

In the embodiment of the invention, the monitoring server can store different processing strategies corresponding to different threat behaviors. Further, the behavior processing module 205 matches a processing policy corresponding to the threat behavior, which may include:

when a plurality of processing strategies exist in a certain threat behavior, the matching degree of the current threat behavior and each processing strategy is calculated according to relevant information (such as responsible party information, behavior probability information of the threat behavior, current severity of the threat behavior, severity change trend information of the threat behavior and the like) corresponding to the threat behavior, and the processing strategies (such as warning sending, forbidden speaking processing, alarming processing, conversation content covering processing and the like) with the matching degree meeting the preset matching degree conditions (such as highest matching degree conditions) are screened, so that the matching degree of the matched processing strategies and the threat behavior existing in the current online conversation is improved, and further the processing efficiency and the processing accuracy of the threat behavior are improved.

Therefore, the device described by the embodiment of the invention can realize the identification of the threat behaviors in the e-commerce platform based on the multidimensional monitoring information obtained by monitoring the online conversation, improves the identification efficiency and the identification accuracy of the threat behaviors, and can also improve the processing efficiency and the processing accuracy of the threat behaviors.

In an alternative embodiment, the specific manner in which the behavior processing module 205 performs processing operations that match processing policies on relevant responsible objects that threaten behavior includes:

when the related responsibility objects of the threat behaviors comprise at least two responsibility objects, determining the responsibility duty ratio corresponding to each responsibility object according to the multi-dimensional monitoring information;

judging whether a corresponding target historical online conversation exists in the current online conversation or not based on the monitored historical online conversation, and determining a duty ratio adjustment coefficient of each responsible object based on historical multi-dimensional monitoring information corresponding to the target historical online conversation when judging that the corresponding target historical online conversation exists in the current online conversation;

for each responsibility object, performing an adjustment operation on the responsibility ratio of the responsibility object according to the responsibility ratio adjustment coefficient of the responsibility object to update the responsibility ratio of the responsibility object, and performing a processing operation matched with the processing policy on the responsibility object according to the updated responsibility ratio of the responsibility object.

In another alternative embodiment, the specific manner of determining whether the threat behavior exists in the dialogue behavior of the current online dialogue based on the multidimensional monitoring information and the pre-trained behavior recognition model by the behavior recognition module 204 includes:

inputting character information to be identified into a pre-trained character behavior identification model to obtain character threat behavior identification results, and inputting picture information to be identified into a pre-trained picture behavior identification model to obtain picture threat behavior identification results;

And judging whether threat behaviors exist in the dialogue behaviors of the current online dialogue according to the character threat behavior recognition result and the picture threat behavior recognition result.

In yet another alternative embodiment, the text monitoring information includes a first preset keyword in the text content and/or a target text information including a first preset semantic. And, the behavior recognition module 204 filters the text monitoring information, and the specific ways of obtaining the text information to be recognized include:

And determining a sending object of each target text message, and filtering all the target text messages according to the predetermined filtering control conditions and by combining the related information of all the sending objects to obtain the text message to be identified. For example, this alternative embodiment can filter out threatening textual information that is laughing.

In yet another alternative embodiment, the specific manner in which the behavior recognition module 204 filters the picture monitoring information to obtain the picture information to be recognized includes:

Filtering dialogue pictures (such as filtering dialogue pictures related to commodities) with corresponding foreground targets of a preset target type and corresponding additional characters of a preset character type from the first filtering result to obtain a second filtering result;

and determining the picture information corresponding to all the dialogue pictures included in the second filtering result as the picture information to be identified.

In yet another alternative embodiment, the monitoring module 203 is further configured to continuously monitor subsequent dialog content corresponding to the current online dialog after the behavior recognition module 204 determines that a threat behavior exists in the dialog behavior of the current online dialog;

the judging module 202 is further configured to judge a resolution corresponding to the threat behavior based on the subsequent dialog content, and judge whether the resolution indicates that the threat behavior has been resolved, and when the judgment result is negative, trigger the behavior processing module 205 to execute the processing policy corresponding to the threat behavior and execute a processing operation matching the processing policy on the relevant responsible object of the threat behavior.

In yet another alternative embodiment, the monitoring module 203 is further configured to monitor a severity trend of the threat behavior with a duration of the current online session after the behavior recognition module 204 determines that the threat behavior exists in the session behavior of the current online session;

the determining module 202 is further configured to determine whether the severity trend indicates that the severity of the threat behavior increases with the duration of the current online session.

And, as shown in fig. 4, the apparatus further includes:

a guiding module 206, configured to generate a guiding dialogue robot model that matches the threat behavior when the determining module 202 determines that the severity trend indicates that the severity of the threat behavior increases with the duration of the current online dialogue, virtualize the guiding dialogue robot model as a dialogue robot, and add the dialogue robot as a new participant to the current online dialogue to guide the threat behavior.

Optionally, the specific manner in which the guidance module 206 generates the guidance dialog robot model that matches the threat behavior is:

and when a plurality of relationship objects exist, generating a guided conversation robot model matched with the threat behaviors based on one or more of participation positions of each relationship object, influence factors corresponding to each relationship object, triggering factors of the influence factors corresponding to each relationship object, conversation style of each relationship object, conversation participation degree of each relationship object, responsibility ratio of each relationship object and the like.

Still further alternatively, if there are multiple participation positions of all the relationship objects, the guidance module 206 may split the current online conversation into multiple sub-online conversations according to the participation position, generate sub-guidance conversation robot models for each sub-online conversation based on the relationship objects involved in each sub-online conversation for threat behavior, virtualize each sub-guidance conversation robot model into sub-conversation robots, and add the sub-conversation robots as a new participation object to the corresponding sub-online conversation to guide threat behavior. Still further, when guiding the threat behavior, the guiding module 206 monitors the consensus sub-session results of the sub-session robots and the corresponding relationship objects in each sub-line session, synchronizes the consensus sub-session results to other sub-session robots so that the other sub-session robots guide new consensus sub-session results in the sub-line session where they are located, synchronizes the new consensus sub-session results to other sub-session robots, and monitors and synchronizes the propulsion sub-session results until it is monitored that all sub-line sessions can share the main consensus results, and synchronizes the main consensus results to the current online session, so as to realize the guiding of the threat behavior.

Examples

Referring to fig. 3, fig. 3 is a schematic structural diagram of a threat behavior recognition apparatus applied to an e-commerce platform according to an embodiment of the invention. The device of fig. 3 may be applied to a monitoring server of an e-commerce platform, where the monitoring server may be a local server or a cloud server, and the embodiment of the present invention is not limited. As shown in fig. 3, the threat behavior identification apparatus applied to the e-commerce platform may include:

a 301 memory storing executable program code;

a processor 302 coupled with the memory;

the processor 302 invokes executable program code stored in the memory 301 to perform the steps in the threat behavior identification method described in any of the embodiments one applied to the e-commerce platform.

Examples

The embodiment of the invention discloses a computer storage medium which stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps in the threat behavior identification method applied to an e-commerce platform.

Examples

An embodiment of the present invention discloses a computer program product comprising a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the threat behavior identification method described in any of the embodiments applied to an e-commerce platform.

The apparatus embodiments described above are merely illustrative, wherein the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above detailed description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course by means of hardware. Based on such understanding, the foregoing technical solutions may be embodied essentially or in part in the form of a software product that may be stored in a computer-readable storage medium including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disc Memory, magnetic disc Memory, tape Memory, or any other medium that can be used for computer-readable carrying or storing data.

Finally, it should be noted that: the embodiment of the invention discloses a threat behavior identification method and a threat behavior identification device applied to an e-commerce platform, which are disclosed by the embodiment of the invention only as a preferred embodiment of the invention, and are only used for illustrating the technical scheme of the invention, but not limiting the technical scheme; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme recorded in the various embodiments can be modified or part of technical features in the technical scheme can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.

Claims

1. The threat behavior identification method applied to the e-commerce platform is characterized by being applied to a monitoring server of the e-commerce platform, and comprises the following steps:

2. The threat behavior recognition method for an e-commerce platform of claim 1, wherein the monitoring server performs processing operations on relevant responsible objects of the threat behavior that match the processing policy, comprising:

for each responsibility object, performing an adjustment operation on the responsibility duty ratio of the responsibility object according to the responsibility duty ratio adjustment coefficient of the responsibility object to update the responsibility duty ratio of the responsibility object, and performing a processing operation matched with the processing policy on the responsibility object according to the updated responsibility duty ratio of the responsibility object.

3. The threat behavior recognition method applied to an e-commerce platform of claim 2, wherein the multidimensional information comprises contextual dialogue text information of the current online dialogue and picture information, the picture information comprising a plurality of dialogue pictures involved in the current online dialogue, a picture type of each of the dialogue pictures, and a picture content of each of the dialogue pictures; the picture type of the dialogue picture is used for indicating whether the dialogue picture is a commodity picture or a non-commodity picture, and the picture content of the dialogue picture comprises a foreground target and additional characters in the dialogue picture; all the dialogue pictures comprise single dialogue pictures and dialogue pictures analyzed from video content;

4. The threat behavior recognition method applied to an e-commerce platform of claim 3, wherein the text monitoring information comprises a target text information with preset keywords and/or preset semantics in text content;

5. The threat behavior identification method applied to an e-commerce platform according to claim 3, wherein the monitoring server filters the picture monitoring information to obtain picture information to be identified, and the method comprises the steps of:

6. The threat behavior recognition method for an e-commerce platform of claim 5, wherein after determining that the threat behavior exists in the session behavior of the current online session, the method further comprises:

7. The threat behavior identification method of claim 6 applied to an e-commerce platform, the method further comprising:

the monitoring server judges whether the severity change trend indicates that the severity of the threat behavior deepens along with the increase of the duration of the current online conversation, and when the judgment result is yes, a guiding conversation robot model matched with the threat behavior is generated, the guiding conversation robot model is virtualized into a conversation robot, and the conversation robot is used as a new participation object to join the current online conversation to guide the threat behavior.

8. Threat behavior recognition device applied to an e-commerce platform, wherein the device is applied to a monitoring server of the e-commerce platform, and the device comprises:

9. Threat behavior recognition device applied to an e-commerce platform, wherein the device is applied to a monitoring server of the e-commerce platform, and the device comprises:

a memory storing executable program code;

a processor coupled to the memory;

the processor invokes the executable program code stored in the memory to perform the threat behavior identification method of any of claims 1-7 applied to an e-commerce platform.

10. A computer storage medium storing computer instructions which, when invoked, are operable to perform the threat behavior identification method of any of claims 1-7 applied to an e-commerce platform.