CN113440856B

CN113440856B - Method and device for identifying abnormal account number in game, electronic equipment and storage medium

Info

Publication number: CN113440856B
Application number: CN202110801380.9A
Authority: CN
Inventors: 张林箭; 张聪; 冯潞潞; 吕唐杰; 陶建容; 范长杰; 胡志鹏
Original assignee: Netease Hangzhou Network Co Ltd
Current assignee: Netease Hangzhou Network Co Ltd
Priority date: 2021-07-15
Filing date: 2021-07-15
Publication date: 2024-02-02
Anticipated expiration: 2041-07-15
Also published as: CN113440856A

Abstract

The application provides a method and a device for identifying an abnormal account in a game, electronic equipment and a storage medium, and relates to the technical field of games. The method comprises the following steps: obtaining a sample dataset comprising: a game log of at least one game virtual object under the target game in a preset time period; constructing a target map according to game logs of the game virtual objects and a pre-built knowledge map framework, wherein the target map is used for representing information of the game virtual objects; determining an abnormal account corresponding to the target abnormal virtual object based on the target map; training a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the target recognition model is used for recognizing whether the game account corresponding to the virtual object to be recognized is the abnormal account. The method can improve the accuracy and efficiency of abnormal account identification.

Description

Method and device for identifying abnormal account number in game, electronic equipment and storage medium

Technical Field

The present invention relates to the field of game technologies, and in particular, to a method and apparatus for identifying an abnormal account in a game, an electronic device, and a storage medium.

Background

The washing amount behavior in the game means that some game virtual objects (game roles controlled by game players) attract other normal virtual objects to play other games in a chatting mode, so that the game revenue can be influenced, and if an abnormal account number with the washing amount behavior can be timely identified, the occurrence of the washing amount behavior can be effectively avoided, and unnecessary losses are reduced.

In the prior art, through the chat content of the obtained game virtual object, matching is carried out according to the chat content and preset keywords so as to judge whether the chat content has suspicion of washing amount, thereby further manually determining whether the account number used by the virtual object sending the chat content is an abnormal account number.

However, the keyword matching-based method is relatively one-sided in recognition and relatively poor in reliability, so that the accuracy of recognition results is relatively low.

Disclosure of Invention

The invention aims to provide a method, a device, electronic equipment and a storage medium for identifying abnormal account numbers in games, aiming at the defects in the prior art, so as to solve the problem that the accuracy of the identification result of the abnormal account numbers in games in the prior art is poor.

In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

in a first aspect, an embodiment of the present application provides a method for identifying an abnormal account in a game, including:

obtaining a sample dataset comprising: a game log of at least one game virtual object under the target game in a preset time period;

constructing a target map according to game logs of the game virtual objects and a pre-built knowledge map framework, wherein the target map is used for representing information of the game virtual objects, and the information comprises: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account;

determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object based on the target map;

training a target identification model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the target identification model is used for identifying whether the game account corresponding to the virtual object to be identified is the abnormal account.

Optionally, the object recognition model includes: a dialogue annotation classification model and an abnormal account identification model;

Training a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the training comprises the following steps:

training the dialogue annotation classification model according to the game logs of the target abnormal virtual objects, wherein the dialogue annotation classification model is used for carrying out category annotation on the game logs of the virtual objects to be identified;

training the abnormal account identification model according to the data of the abnormal account corresponding to the target abnormal virtual object and the dialogue annotation classification model, wherein the abnormal account identification model is used for identifying whether the game account corresponding to the virtual object to be identified is an abnormal account.

Optionally, the determining, based on the target map, the target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object includes:

screening suspected abnormal virtual objects from all game virtual objects based on the target map;

and determining the target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object.

Optionally, the constructing a target graph according to the game log of each game virtual object and the pre-built knowledge graph architecture includes:

According to the corresponding relation between the preset nodes and the side lengths in the knowledge graph framework, respectively converting the game logs of each game virtual object into node data and side length data;

according to the preset key information of at least one dialogue category, marking the dialogue category of each node data and the dialogue contained in each side length data to obtain a dialogue marking result corresponding to each node data and a dialogue marking result corresponding to each side length data;

determining target data of each node according to a dialogue labeling result corresponding to each node data, a dialogue labeling result corresponding to each side length data and a preset category index, wherein the target data of the node are used for representing characteristic information of a game virtual object corresponding to the node;

and constructing the target map according to the target data of each node and the game log of the game virtual object corresponding to each node.

Optionally, the game log of the game virtual object includes the following information: a session type, a session timestamp, characteristic information of a first object, the session information of the first object, characteristic information of a second object, wherein the first object characterizes a game virtual object initiating the session, the second object characterizes the game virtual object being traversed by the first object, the second object comprises at least one, the characteristic information comprises: the method comprises the steps of virtual object identification, game level of a virtual object, member level of the virtual object, creation channel of the virtual object and game account corresponding to the virtual object, wherein the dialogue type comprises: public dialogues and private dialogues.

Optionally, before the converting the game logs of each game virtual object into the node data and the side length data according to the preset correspondence between the nodes and the side lengths in the knowledge graph architecture, the method further includes:

respectively determining a character identification node of a first object in a game log of each game virtual object, wherein the character identification node is identified as the knowledge graph, a character channel node of the first object is an creation channel of the knowledge graph, and a character account node of the knowledge graph is a game account corresponding to the first object;

determining the corresponding relation between a first object and a second object in a game log of each game virtual object, the corresponding relation between the first object and a creation channel of the first object, and the corresponding relation between the first object and a game account corresponding to the first object as the side length of the knowledge graph;

and setting the corresponding relation between the nodes and the side lengths according to the determined nodes and the side lengths of the knowledge graph.

Optionally, the converting the game log of each game virtual object into node data and side length data according to the preset corresponding relationship between the node and the side length in the knowledge graph architecture includes:

according to the corresponding relation between the nodes and the side lengths, the identification of a first object in the game logs of each game virtual object, the public dialogue information corresponding to the first object, the game grade of the first object, the member grade of the first object are converted into character identification node data, the creation channel of the first object in the game logs of each game virtual object is converted into character channel node data, and the game account corresponding to the first object in the game logs of each game virtual object is converted into character account node data;

The method comprises the steps of converting identification of a first object, identification of a second object, corresponding relation between the first object and the second object and private dialogue information of the first object and the second object in a game log of each game virtual object into first side length data, converting corresponding relation between the first object and a creation channel of the first object into second side length data, converting a game account corresponding to the first object and the first object into third side length data, and obtaining the side length data according to the first side length data, the second side length data and the third side length data.

Optionally, the marking the dialogue category of each node data and the dialogue included in each side length data according to the preset key information of at least one dialogue category to obtain a dialogue marking result corresponding to each node data and a dialogue marking result corresponding to each side length data, including:

matching dialogue information contained in each node data with the key information of at least one dialogue category, and marking the dialogue information contained in each node data according to the matching result to obtain a dialogue marking result corresponding to each node data;

and matching the dialogue information contained in each side length data with the key information of at least one dialogue category, and marking the dialogue information contained in each side length data according to the matching result to obtain a dialogue marking result corresponding to each side length data.

Optionally, the indexes of the preset category include: the method comprises the steps of accumulating times of dialogue categories, diversity of first sentence dialogues and interaction degree of dialogues, wherein the interaction degree of the dialogues is used for representing dialogue initiative of game virtual objects;

the determining the target data of each node according to the dialogue labeling result corresponding to each node data, the dialogue labeling result corresponding to each side length data and the index of the preset category comprises the following steps:

determining the accumulated times of dialogue categories corresponding to each node according to all dialogue labeling results contained in each node data and the dialogue labeling results of the first object contained in each side length data;

determining the diversity of the first sentence dialogue corresponding to each node according to the number of the first sentence dialogue information of the first object in the dialogue information of the first object and the second object contained in each side length data after duplication removal and the number of all the second objects corresponding to the first object;

determining the interaction degree of the dialogs corresponding to each node according to the number of dialogs of the first object in the dialog information of the first object and the second object contained in each side length data and the total number of dialogs corresponding to the first object, wherein the total number of dialogs corresponding to the first object comprises: the total number of conversations of the first object and the second object in the node data corresponding to the first object, and all the conversations in the side length data corresponding to the first object.

And determining target data of each node according to the data of each node, the accumulated times of dialogue categories corresponding to each node, the diversity of first sentence dialogues and the interaction degree of dialogues.

Optionally, the constructing the target map according to the target data of each node and the game log of the game virtual object corresponding to each node includes:

and constructing the target map according to the corresponding relation between the first object and the second object in the game log of the game virtual object corresponding to each node, the corresponding relation between the first object and the creation channel of the first object, the corresponding relation between the first object and the game account corresponding to the first object and the target data of each node.

Optionally, the screening the suspected abnormal virtual objects from the game virtual objects based on the target atlas includes:

and determining suspected abnormal virtual objects from game virtual objects corresponding to all the nodes according to target data of all the nodes in the target map and a preset evaluation threshold.

Optionally, the determining the target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object includes:

According to the corresponding relation between the first object and the second object in the target map, determining an associated virtual object corresponding to each suspected abnormal virtual object, wherein the associated virtual object is used for representing a game virtual object which performs dialogue with the suspected abnormal virtual object;

determining a target abnormal virtual object from each suspected abnormal virtual object according to the loss rate of the associated virtual object corresponding to each suspected abnormal virtual object;

and determining an abnormal account corresponding to the target abnormal virtual object according to the target abnormal virtual object and the corresponding relation between the first object and the game account corresponding to the first object in the target map.

Optionally, the training the dialog annotation classification model according to the game log of the target abnormal virtual object includes:

performing dialogue class labeling on dialogue information in a game log of the target abnormal virtual object to acquire first sample data;

and training and acquiring the dialogue annotation classification model by adopting the first sample data.

Optionally, the training the abnormal account identification model according to the data of the abnormal account corresponding to the target abnormal virtual object and the dialogue annotation classification model includes:

Taking target data of a first node, a second node and a third node corresponding to the abnormal account as second sample data, wherein the first node is a node where an abnormal virtual object corresponding to the abnormal account in the target map is located, the second node is a next node of the first node in the target map, and the third node is a next node of the second node in the target map;

and training and obtaining the abnormal account identification model by adopting the second sample data and the dialogue annotation classification model.

Optionally, training to obtain the abnormal account identification model by using the second sample data and the dialogue annotation classification model includes:

performing dialogue classification labeling on the second sample data by adopting the dialogue labeling classification model;

and training and acquiring the abnormal account identification model according to the marked second sample data.

In a second aspect, an embodiment of the present application further provides an apparatus for identifying an abnormal account in a game, including: the system comprises an acquisition module, a construction module, a determination module and a training module;

the acquisition module is configured to acquire a sample data set, where the sample data set includes: a game log of at least one game virtual object under the target game in a preset time period;

The building module is configured to build a target graph according to a game log of each game virtual object and a pre-built knowledge graph architecture, where the target graph is used to represent information of the game virtual object, and the information includes: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account;

the determining module is used for determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object based on the target map;

the training module is used for training a target identification model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, and the target identification model is used for identifying whether the game account corresponding to the virtual object to be identified is the abnormal account.

the training module is specifically configured to train the dialogue annotation classification model according to the game log of the target abnormal virtual object, where the dialogue annotation classification model is used to perform category annotation on the game log of the virtual object to be identified;

Optionally, the determining module is specifically configured to screen suspected abnormal virtual objects from all game virtual objects based on the target atlas;

Optionally, the building module is specifically configured to convert game logs of each game virtual object into node data and side length data according to a preset corresponding relationship between nodes and side lengths in the knowledge graph architecture;

and determining target data of each node according to the dialogue labeling result corresponding to each node data, the dialogue labeling result corresponding to each side length data and the index of the preset category, wherein the target data of the node is used for representing the characteristic information of the game virtual object corresponding to the node.

Optionally, the determining module is further configured to determine, respectively, that a character identifier node of a first object in a game log of each game virtual object is identified as the knowledge graph, a character channel node of the first object is an creation channel of the knowledge graph, and a character account node of a game account corresponding to the first object is the knowledge graph;

Optionally, the building module is specifically configured to convert, according to the correspondence between the nodes and the side lengths, an identifier of a first object in a game log of each game virtual object, common dialogue information corresponding to the first object, a game level of the first object, a member level of the first object into character identifier node data, convert a creation channel of the first object in the game log of each game virtual object into character channel node data, and convert a game account corresponding to the first object in the game log of each game virtual object into character account node data;

Optionally, the construction module is specifically configured to match dialogue information included in each node data with the key information of the at least one dialogue category, and perform dialogue category labeling on the dialogue information included in each node data according to a matching result, so as to obtain a dialogue labeling result corresponding to each node data;

optionally, the construction module is specifically configured to determine the cumulative number of times of the conversation category corresponding to each node according to all conversation annotation results included in the data of each node and conversation annotation results of the first object included in the data of each side;

Determining the interaction degree of the dialogs corresponding to each node according to the number of dialogs of the first object in the dialog information of the first object and the second object contained in each side length data and the total number of dialogs corresponding to the first object, wherein the total number of dialogs corresponding to the first object comprises: the total number of conversations of the first object and the second object in the node data corresponding to the first object and the total number of conversations in the side length data corresponding to the first object;

Optionally, the building module is specifically configured to build the target map according to a corresponding relationship between the first object and the second object in the game log of the game virtual object corresponding to each node, a corresponding relationship between the first object and a creation channel of the first object, a corresponding relationship between the first object and a game account corresponding to the first object, and target data of each node.

Optionally, the determining module is specifically configured to determine a suspected abnormal virtual object from game virtual objects corresponding to each node according to target data of each node in the target map and a preset evaluation threshold.

Optionally, the determining module is specifically configured to determine, according to a correspondence between the first object and the second object in the target map, an associated virtual object corresponding to each suspected abnormal virtual object, where the associated virtual object is used to characterize a game virtual object that performs a session with the suspected abnormal virtual object;

Optionally, the training module is specifically configured to perform dialogue class labeling on dialogue information in the game log of the target abnormal virtual object, and obtain first sample data;

Optionally, the training module is specifically configured to use target data of a first node, a second node, and a third node corresponding to the abnormal account as second sample data, where the first node is a node where an abnormal virtual object corresponding to the abnormal account in the target graph is located, the second node is a next node of the first node in the target graph, and the third node is a next node of the second node in the target graph;

Optionally, the training module is specifically configured to perform dialogue classification labeling on the second sample data by using the dialogue labeling classification model;

In a third aspect, an embodiment of the present application provides an electronic device, including: a processor, a storage medium, and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating over the bus when the electronic device is operating, the processor executing the machine-readable instructions to perform the steps of the method as provided in the first aspect when executed.

In a fourth aspect, embodiments of the present application provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method as provided in the first aspect.

The beneficial effects of this application are:

the application provides a method, a device, electronic equipment and a storage medium for identifying abnormal account numbers in a game, wherein the method comprises the following steps: obtaining a sample dataset comprising: a game log of at least one game virtual object under the target game in a preset time period; constructing a target map according to game logs of the game virtual objects and a pre-built knowledge map framework, wherein the target map is used for representing information of the game virtual objects, and the information comprises: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account; determining an abnormal account corresponding to the target abnormal virtual object based on the target map; training a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the target recognition model is used for recognizing whether the game account corresponding to the virtual object to be recognized is the abnormal account. In the scheme, the knowledge graph is built based on the game log of the game virtual object, so that the abnormal virtual object and the abnormal account corresponding to the abnormal virtual object can be determined according to the knowledge graph, the interpretability and the credibility of the determined abnormal virtual object and the abnormal account corresponding to the abnormal virtual object are improved, in addition, the target recognition model is trained and obtained based on the determined abnormal virtual object and the related data of the abnormal account corresponding to the abnormal virtual object, and the automatic recognition of the abnormal account can be realized based on the target recognition model, and the accuracy and the efficiency of the recognition of the abnormal account are improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered limiting the scope, and that other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 2 is a second flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 3 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 4 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 5 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 6 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 7 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application;

Fig. 8 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 9 is a schematic diagram of a knowledge graph according to an embodiment of the present application;

fig. 10 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 11 is a flowchart of an identification method of an abnormal account in a game according to an embodiment of the present application;

fig. 12 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 13 is a flowchart illustrating a thirteenth method for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 14 is a schematic diagram of an apparatus for identifying an abnormal account in a game according to an embodiment of the present application;

fig. 15 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it should be understood that the accompanying drawings in the present application are only for the purpose of illustration and description, and are not intended to limit the protection scope of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. A flowchart, as used in this application, illustrates operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Moreover, one or more other operations may be added to the flow diagrams and one or more operations may be removed from the flow diagrams as directed by those skilled in the art.

In addition, the described embodiments are only some, but not all, of the embodiments of the present application. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, as provided in the accompanying drawings, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to be within the scope of the present application.

It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the presence of the features stated hereinafter, but not to exclude the addition of other features.

Fig. 1 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application; the method may be performed by a terminal device or a server, as shown in fig. 1, and the method may include:

s101, acquiring a sample data set, wherein the sample data set comprises: and a game log of at least one game virtual object under the target game in a preset time period.

Alternatively, the method of the present application may be applied to multiplayer online role playing class games, where there may be dialogue exchange between different game virtual objects, where a game virtual object may refer to a game role, i.e. a role played or controlled by a game player.

The game log generated by each game virtual object in the game process can be stored in a server background database and the like, wherein the game log can comprise chat data of the game virtual object, game combat data and the like. While chat logs may include, but are not limited to, private conversation information between game virtual objects and other game virtual objects, utterances of game virtual objects published on public channels, information broadcast by game virtual objects in a game lobby or game square, and the like. Optionally, a game log of at least one game virtual object under the target game in any preset time period may be obtained from the database, where the at least one game virtual object may be randomly extracted.

S102, constructing a target map according to game logs of the game virtual objects and a pre-built knowledge map framework, wherein the target map is used for representing information of the game virtual objects, and the information comprises: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account.

It should be noted that, the knowledge graph is a semantic network system with a very large scale, and its main purpose is to describe the association relationship between entities or concepts in the real world. Through a large amount of data collection, the data are arranged into a knowledge base which can be processed by a machine, and visual display is realized.

Optionally, the architectures of the knowledge maps of different types are different, and game logs of each game virtual object can be processed according to the architecture of the knowledge maps used in the application, and the target map can be built. The target atlas can visually display information of each game virtual object.

The knowledge graph architecture referred to in this embodiment may be composed of a plurality of nodes and a plurality of side lengths, and the nodes may include three types: role nodes, creation channel nodes and game account nodes, the side length can also comprise three types: based on the knowledge graph architecture, the game logs of the game virtual objects can be subjected to data processing to build a target graph.

The feature information of the game virtual object may refer to game feature information of the game virtual object, for example: identification of game avatars, ratings, and the like.

S103, determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object based on the target map.

Alternatively, based on the constructed target map, a target abnormal virtual object may be determined from the game virtual objects according to the information of the game virtual objects characterized in the target map, where the target abnormal virtual object is a virtual object with abnormal operation, for example: there are virtual objects of the wash-off behavior. In addition, according to the corresponding relation between the game virtual object and the game account in the target map, the abnormal account corresponding to the target abnormal virtual object can be determined.

S104, training a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the target recognition model is used for recognizing whether the game account corresponding to the virtual object to be recognized is the abnormal account.

In some embodiments, because the determined target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object have a certain credibility, sample data for training the target recognition model can be expanded or constructed based on the determined target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, so that the target recognition model is obtained through training, intelligent recognition on the abnormal account is realized by adopting the target recognition model, and the workload of manual recognition is reduced.

In summary, the method for identifying abnormal accounts in a game provided in this embodiment includes: obtaining a sample dataset comprising: a game log of at least one game virtual object under the target game in a preset time period; constructing a target map according to game logs of the game virtual objects and a pre-built knowledge map framework, wherein the target map is used for representing information of the game virtual objects, and the information comprises: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account; determining an abnormal account corresponding to the target abnormal virtual object based on the target map; training a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, wherein the target recognition model is used for recognizing whether the game account corresponding to the virtual object to be recognized is the abnormal account. In the scheme, the knowledge graph is built based on the game log of the game virtual object, so that the abnormal virtual object and the abnormal account corresponding to the abnormal virtual object can be determined according to the knowledge graph, the interpretability and the credibility of the determined abnormal virtual object and the abnormal account corresponding to the abnormal virtual object are improved, in addition, the target recognition model is trained and obtained based on the determined abnormal virtual object and the related data of the abnormal account corresponding to the abnormal virtual object, and the automatic recognition of the abnormal account can be realized based on the target recognition model, and the accuracy and the efficiency of the recognition of the abnormal account are improved.

Fig. 2 is a second flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S104, the object recognition model may include: a dialogue annotation classification model and an abnormal account identification model;

in step S104, training the target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object may include:

s201, training a dialogue annotation classification model according to the game logs of the target abnormal virtual objects, wherein the dialogue annotation classification model is used for carrying out category annotation on the game logs of the virtual objects to be identified.

In some embodiments, a dialog annotation corpus may be augmented from the determined game logs of all target abnormal virtual objects for training a dialog annotation classification model, wherein the dialog annotation classification model may be used to annotate the dialog class with the game logs of the virtual objects to be identified.

S202, training an abnormal account identification model according to data of an abnormal account corresponding to the target abnormal virtual object and a dialogue annotation classification model, wherein the abnormal account identification model is used for identifying whether a game account corresponding to the virtual object to be identified is an abnormal account.

Optionally, the abnormal account data may be expanded according to the data of the abnormal account corresponding to the determined target abnormal virtual object, and the abnormal account identification model may be trained based on the expanded abnormal account data in combination with the dialog annotation classification model obtained by training.

Fig. 3 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S103, determining, based on the target map, the target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object may include:

s301, based on the target map, the suspected abnormal virtual objects are screened from the game virtual objects.

In one implementation manner, suspected abnormal virtual objects can be screened from the game virtual objects according to information of the game virtual objects in the target map, and rough screening can be performed through set screening rules.

S302, determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object.

Optionally, whether each suspected abnormal virtual object is determined to be the target abnormal virtual object may be further determined according to a preset evaluation rule, that is, each suspected abnormal virtual object is verified by some evaluation rules, so as to provide strong evidence from the aspect of interpretability, to prove whether each suspected abnormal virtual object is the target abnormal virtual object.

Fig. 4 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S102, constructing the target spectrum according to the game log of each game virtual object and the pre-built knowledge-graph architecture may include:

s401, according to the corresponding relation between the preset nodes and the side lengths in the knowledge graph framework, the game logs of the game virtual objects are respectively converted into node data and side length data.

Optionally, according to the knowledge graph architecture, data to be represented by each node and each side in the knowledge graph to be built can be determined, so that data processing can be performed on the game logs of each game virtual object according to the data to be represented by each node and each side, so as to convert the game logs of each game virtual object into node data and side data.

S402, marking the dialogue type of each node data and the dialogue contained in each side length data according to the preset key information of at least one dialogue type, and obtaining a dialogue marking result corresponding to each node data and a dialogue marking result corresponding to each side length data.

In some embodiments, according to preset key information of at least one dialogue type, the dialogue type label is performed on each dialogue in the obtained node data and each side length data, that is, the label processing is performed on each dialogue in each node data and each side length data. The preset at least one dialogue category may include a dialogue category corresponding to an abnormal scene to be identified, for example: and identifying the abnormal account number with the washing amount behavior, wherein the preset at least one dialogue category can comprise dialogue categories such as a bid game and the like. If the abnormal account number with the spam language is identified, the preset at least one dialogue category can comprise the dialogue category such as abuse and severe speech.

S403, determining target data of each node according to the dialogue labeling result corresponding to each node data, the dialogue labeling result corresponding to each side length data and the index of the preset category, wherein the target data of the node is used for representing the characteristic information of the game virtual object corresponding to the node.

In some embodiments, the dialog labeling result corresponding to each node data may be added to the dialog included in each node data in a suffix manner, and similarly, the dialog labeling result corresponding to each side data may be added to the corresponding dialog included in each side data in a suffix manner.

Optionally, according to the data of each node and the data of each side length, according to the indexes of the preset category, each index of the game virtual object corresponding to each node is counted, and similarly, each index corresponding to each node can be added to the dialogue contained in each node, and the addition of the labeling results is similar.

And obtaining the target data of each node based on the dialogue labeling result and the node data of the index information.

S404, constructing a target map according to the target data of each node and the game logs of the game virtual objects corresponding to each node.

Alternatively, the target data of each node may be added to the game log of the game virtual object corresponding to each node based on the target data of each node and the game log of the game virtual object corresponding to each node, so as to obtain each node data of the target map.

Optionally, the game log of the game virtual object may include the following information: the method comprises the steps of a dialogue type, a dialogue time stamp, feature information of a first object, dialogue information of the first object and feature information of a second object, wherein the first object represents a game virtual object initiating the dialogue, the second object represents the game virtual object which is interacted by the first object, the second object comprises at least one of the following information: virtual object identification, game level of the virtual object, member level of the virtual object, creation channel of the virtual object, game account corresponding to the virtual object, and dialogue type comprises: public dialogues and private dialogues.

Alternatively, the game log of each game virtual object may include a plurality of pieces, and the information included in each game log may include, but is not limited to, what has been described above.

The dialog types may include: the public dialogue and the private dialogue can determine dialogue types according to chat channels of the dialogues, and if the dialogues are private channels, chat objects exist; if a public channel, the boring object, the public channel may include, for example, a world channel, a conviction channel, etc. The session timestamp may be a recorded time of the session. For private conversations, there is a conversation initiator and a conversationed person, then the first object may refer to the conversation initiator, i.e., the game virtual object that initiated the conversation, and the second object may refer to the game virtual object that is being conversationed by the conversationed person, i.e., the game virtual object that initiated the conversation. The information of the first object may include: identification of game virtual objects that initiate a session, such as: a character ID; game level, i.e., role level, of the game virtual object initiating the session; the more the game player charges in the game, the higher the membership grade of the corresponding game virtual object; the creation channel of the game virtual object that initiates the session may refer to what channel the game player created the game virtual object through, where the creation channel may be, for example: a official network, XX region, etc.; a game account number of a game virtual object initiating a dialogue, that is, a game player creates the game virtual object in a certain creation channel through a certain game account number; similarly to the information of the first object, the information of the second object, i.e. the information of the game virtual object to which the game virtual object that initiated the session has been session, may also comprise the above information. For a common dialogue, however, the information of the second object may not be involved, since there is no corresponding second object.

The correspondence between game account numbers, creation channels, and character IDs is illustrated by a simple example as follows: for example, one game account a (account ID 1) that creates two virtual characters (character ID1 and character ID 2) may be registered with device a (character channel ID 1) and may be registered with device B (character channel ID 2); in still other cases, one device a (character channel ID 2) may log in a plurality of game accounts a (account ID 1), b (account ID 2), c (account ID 3) d (account ID 4) at the same time.

In addition, the game log of each game virtual object may also include session information for the first object, i.e., all private sessions and public sessions of the game virtual object that includes the originating session.

Fig. 5 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S401, before converting the game logs of each game virtual object into the node data and the side length data according to the preset correspondence between the nodes and the side lengths in the knowledge graph architecture, the method of the present application may further include:

s501, determining a role identification node of a first object in a game log of each game virtual object, wherein the role identification node is identified as a knowledge graph, a role channel node of the first object, wherein a creation channel of the first object is the knowledge graph, and a role account node of the first object, wherein a game account corresponding to the first object is the knowledge graph.

In one implementation, the created knowledge graph may include three classes of nodes, including: the method comprises the steps of identifying a first object in a game log of each game virtual object, creating a channel of the first object and a game account corresponding to the first object. Based on the types of the three types of nodes, the identification of the first object in the game log can be determined to be a role identification node, the creation channel of the first object is a role channel node, and the game account corresponding to the first object is a role account node.

Optionally, the node of the identification type of the first object in the game log of each game virtual object may include: all public conversations of the game virtual object, game level, member level, etc. of the game virtual object.

S502, determining the corresponding relation between the first object and the second object in the game log of each game virtual object, the corresponding relation between the first object and the creation channel of the first object, and the corresponding relation between the first object and the game account corresponding to the first object as the side length of the knowledge graph.

In addition, the created knowledge graph may include three types of side lengths, including: a dialog relationship of a first object and a second object, for example: the first object is a dialog initiator, and the side length can be pointed to the second object by the first object; the method also comprises a first object and a corresponding relation of the first object creating channel, namely a subordinate relation; the corresponding relation between the first object and the game account corresponding to the first object is also called a subordinate relation. The side length of the dialog relationship between the first object and the second object also includes information such as multi-round dialog content of private dialog of the first object and the second object.

S503, setting the corresponding relation between the nodes and the side lengths according to the determined nodes and the side lengths of the knowledge graph.

Alternatively, based on the determined types of the nodes and the side lengths in the knowledge graph and the information respectively contained in the determined types, the correspondence between the nodes and the side lengths in the knowledge graph to be created can be designed.

Fig. 6 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S401, according to a preset correspondence between nodes and side lengths in the knowledge graph architecture, converting game logs of each game virtual object into node data and side length data respectively may include:

s601, according to the corresponding relation between the nodes and the side lengths, the identification of the first object in the game logs of each game virtual object, the public dialogue information corresponding to the first object, the game grade of the first object, the member grade of the first object are converted into character identification node data, the creation channel of the first object in the game logs of each game virtual object is converted into character channel node data, and the game account corresponding to the first object in the game logs of each game virtual object is converted into character account node data.

Alternatively, the data related to the node included in the game log of each game virtual object may be converted into the node data based on the correspondence between the node and the side length set as described above, for example: the role identification node data of a certain game virtual object may include { "identification of virtual object": "12345", "game level": 40, "membership grade": "4", "public dialogue information": [ "I have preferential recharging channel, do you need it", "do you play this game for the first time" ] }.

S602, converting the identification of a first object, the identification of a second object, the corresponding relation between the first object and the second object and the private dialogue information of the first object and the second object in the game log of each game virtual object into first side length data, converting the corresponding relation between the first object and the creation channel of the first object into second side length data, converting the game account corresponding to the first object and the first object into third side length data, and obtaining the side length data according to the first side length data, the second side length data and the third side length data.

For the correspondence between the first object and the second object in the game log of each game virtual object, the side length from the first object to the second object may be established, for example, the information about the side length from a certain first object to the second object includes: { "first object": "12345", "second object": "12346", "private dialog content": [ "1_hello" hand over friends bar "," 2_good "hand over", "1_is now available", "1_we do a game task together", "2_is available" ] }. Where the dialog content is preceded by a prefix of 1 and 2 to distinguish which virtual object is said, 1 representing the first object and 2 representing the second object. For the first object and the creation channel of the first object, the creation channel of the first object can be established to the side length of the first object, and similarly, the side length of the game account corresponding to the first object is established, so that the processing of three side lengths in the knowledge graph is completed for the game log of one game virtual object.

Fig. 7 is a flowchart of a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S402, according to preset key information of at least one dialogue type, dialogue type labeling is performed on each node data and the dialogues included in each side length data, so as to obtain a dialogue labeling result corresponding to each node data and a dialogue labeling result corresponding to each side length data, which may include:

and S701, matching dialogue information contained in each node data with key information of at least one dialogue category, and marking the dialogue information contained in each node data according to the matching result to obtain a dialogue marking result corresponding to each node data.

In this embodiment, taking the identification of an abnormal account number with respect to the existence of a wash-out behavior in a game as an example, the at least one dialogue category set correspondingly may include the following seven categories: 1, 2 welfare recharging, 3 slot games, 4 fixed play, 5 sets of nearly, 6 shouting together play and 7 request contact ways. For each type of feature, some rules (regular expressions or keywords) may be prepared in advance, such as key information corresponding to the benefit recharge category: "give benefit", "recharge preferential", etc.

Each piece of dialogue information contained in the node data can be respectively matched with the key information of each dialogue category, and when any piece of dialogue information is matched with a certain key information, the dialogue category corresponding to the key information can be marked on the piece of dialogue information, for example: when the key information such as 'welfare' and 'preferential recharging' is matched, a label of 'welfare recharging' is marked for the dialogue information.

Optionally, the seven types of identification can be performed on each piece of dialogue information contained in the data of each node, so as to obtain a labeling result of each piece of dialogue information, and the labeling result is recorded on the node where the dialogue information is located. Such as { "identification of game virtual object": "12345", "game level": 40, "membership grade": "4", "dialogue information": [ "I have preferential recharge channel, do you need to do so 0100000", "do you play this game for the first time" do you get it to play it for the first time }. Wherein, the 7-class labeling result is used as a suffix to be placed on each piece of dialogue information, if a certain dialogue class is successfully matched, the number of the corresponding position is marked with 1, and the result of the normal corpus is_0000000. For example, 0100000 indicates that the type of dialogue information is welfare refill, 0010000 indicates that the type of dialogue information is slot game, and the like.

S702, matching dialogue information contained in each side length data with key information of at least one dialogue category, and marking the dialogue information contained in each side length data according to the matching result to obtain a dialogue marking result corresponding to each side length data.

Similarly, the same labeling process is performed for each piece of dialogue information included in each side length data, and the labeling result is recorded on the corresponding side length.

For example: { "identification of first object": "12345", "identification of second object": "12346", "dialogue information": [ "1_hello" and "friend bar_ 0000100", "2_well" 0000000"," 1_currently available "mock_0000000", "1_we do a game task together" 0000000"," 2_possible "0000000" ] }.

FIG. 8 is a flowchart illustrating a method for identifying abnormal accounts in a game according to an embodiment of the present disclosure; optionally, the indexes of the preset category may include: the cumulative number of dialogue categories, the diversity of first sentence dialogues, and the degree of interaction of the dialogues, which are used to characterize the dialogue initiative of the game virtual object.

The cumulative number of conversational categories may refer to the cumulative number of all conversations of the game virtual object corresponding to the node on the 7-class feature, wherein only conversations beginning with 1 are considered for the conversational information on the edge, that is, only conversations of the first object are counted. The diversity of the first sentence conversation refers to the diversity of the first sentence spoken in the conversation process of the first object and the second object. The interaction degree of the dialogs may refer to the proportion of the number of all dialogs spoken by the first object to the number of all dialogs in the private dialogs with the second object.

Optionally, in step S403, determining the target data of each node according to the session labeling result corresponding to each node data, the session labeling result corresponding to each side length data, and the index of the preset category may include:

s801, determining the accumulated times of the dialogue categories corresponding to each node according to all dialogue labeling results contained in the data of each node and the dialogue labeling results of the first object contained in the data of each side length.

Optionally, for any game virtual object, statistics of preset category indexes can be performed according to dialogue labeling results of each piece of dialogue information contained in each piece of node data and each piece of side length data of the game virtual object, and the statistics can be used as feature information of the game virtual object.

Optionally, for the node where any game virtual object is located, the cumulative number of times of each dialogue information in the node data on the 7-class feature and the cumulative number of times of the dialogue information of the first object in the side length data on the 7-class feature may be counted, and the total cumulative number of times is taken as the cumulative number of times of the dialogue class corresponding to the node.

S802, determining diversity of first sentence conversations corresponding to each node according to the number of the first sentence conversations of the first object after duplication elimination and the number of all the second objects corresponding to the first object in the conversation information of the first object and the second object contained in each side length data.

Alternatively, the number of first sentence dialogs that are spoken by the first object and the number of second objects corresponding to the first object (i.e., all second objects that are set up by the first object) may be counted in the dialog information of the first object and the second object, and typically, one game virtual object may actively converse with a plurality of second objects in the game. The ratio of the number of the first sentence dialogs, which are spoken by the first object, to the number of all the second objects corresponding to the first object in the dialog information of the first object and the second object can be calculated, so that the diversity of the first sentence dialogs corresponding to the nodes is obtained.

Wherein the ratio is more toward 0, the worse the diversity of the first sentence dialogue, the more likely it is that the first object will have wash-out behavior to use the same technique to speak with a different second object.

S803, determining the interaction degree of the dialogs corresponding to each node according to the dialog number of the first object and the dialog total number corresponding to the first object in the dialog information of the first object and the second object contained in each side length data, wherein the dialog total number corresponding to the first object comprises: the total number of conversations of the first object and the second object in the node data corresponding to the first object, and all the conversations in the side length data corresponding to the first object.

Alternatively, the number of all dialogs spoken by the first object may be counted in the dialog information of the first object and all second objects, and the ratio of the number of all dialogs spoken by the first object to the number of all dialogs may be calculated. In general, this value tends to be about 1, indicating that the higher the dialog initiative of the first object, the more likely it is that the "wash-out" behaved virtual object is ever harassing other virtual objects, and is more suspicious.

S804, determining target data of each node according to the data of each node, the accumulated times of dialogue categories corresponding to each node, the diversity of first sentence dialogues and the interaction degree of dialogues.

Alternatively, through the above processing, the obtained target data of a certain node may be expressed as { "identification of game virtual object": "12345", "game level": 40, "membership grade": "4", "dialogue information": [ "I have preferential recharging channel, do you need to do so with the name of" do you play this game for the first time "do you play" with the name of "bid game": 0, "welfare top-up": 1, "slot game": 0, "fixed on white": 0, "sleeve is nearly": 1, "megaphone together play": 0, "claim contact: 0, "diversity of first sentence dialogue": 0.2, "interaction degree of dialog": 0.5}.

Fig. 9 is a schematic diagram of a knowledge graph according to an embodiment of the present application. Optionally, in step S404, constructing the target map according to the target data of each node and the game log of the game virtual object corresponding to each node may include: and constructing a target map according to the corresponding relation between the first object and the second object in the game log of the game virtual object corresponding to each node, the corresponding relation between the first object and the creation channel of the first object, the corresponding relation between the first object and the game account corresponding to the first object and the target data of each node.

In one implementation manner, the node data and the side length data obtained by the processing can be uploaded to a graph database based on the architecture of the designed knowledge graph, and the target graph is constructed. In this embodiment, the open source graph database system may be a hugegraph.

Optionally, the target graph obtained by building may be shown in fig. 9, where the graph only schematically shows a relationship graph within one hop of any game virtual object, that is, only the first object is used as a main body, and shows a relationship between the first object and the second object, and in practical application, the second object may also be used as a main body, that is, as a session initiator, and may show a relationship between the second object and other objects corresponding to the second object, that is, shows a relationship between the second object and other objects that are being interacted with by the second object, so that a complete knowledge graph that is built by all the extracted game virtual objects together may be obtained.

Optionally, based on a knowledge graph mode, a dialogue network of each game virtual object is constructed, so that the interpretive that the determined abnormal account has abnormality is increased to a certain extent.

Optionally, in step S301, screening the suspected abnormal virtual objects from the game virtual objects based on the target atlas may include: and determining suspected abnormal virtual objects from game virtual objects corresponding to the nodes according to target data of the nodes in the target map and a preset evaluation threshold.

In one implementation, multiple evaluation thresholds may be preset, such as: the game level threshold value, the membership level threshold value, the diversity threshold value of the first sentence dialogue, the interaction degree threshold value of the dialogue and the like of the preset game virtual object can be used for inquiring each node in the knowledge graph based on each threshold value, for example: all nodes with game grades smaller than 40 and member grades smaller than 3, diversity of first sentence conversations smaller than 0.3 and interaction degree of conversations higher than 0.6 are queried, so that virtual objects corresponding to queried nodes can be determined to be suspected abnormal virtual objects.

Optionally, based on the determined virtual object with suspected abnormality, a virtual object having the same creation channel or the same game account number as the virtual object with suspected abnormality may be found as the virtual object with suspected abnormality according to the corresponding relationship between the first object and the second object displayed in the target map.

Fig. 10 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S302, determining the target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object may include:

s1001, according to the corresponding relation between the first object and the second object in the target map, determining an associated virtual object corresponding to each suspected abnormal virtual object, wherein the associated virtual object is used for representing a game virtual object which is in dialogue with the suspected abnormal virtual object.

Optionally, based on the determined suspected abnormal virtual objects, an associated virtual object corresponding to each suspected abnormal virtual object may be searched in the knowledge graph, where the associated virtual object is a virtual object that is disturbed by the suspected abnormal virtual object, that is, a virtual object that is interacted by the suspected abnormal virtual object.

S1002, determining a target abnormal virtual object from the suspected abnormal virtual objects according to the loss rate of the associated virtual object corresponding to the suspected abnormal virtual objects.

For any suspected abnormal virtual object, the loss rate of all the associated virtual objects corresponding to the suspected abnormal virtual object can be counted, when the loss rate is larger than the set average loss rate, the suspected abnormal virtual object can be objectively determined that the suspected abnormal virtual object actually causes the loss of part of the virtual objects, the washing amount behavior exists, and the suspected abnormal virtual object can be determined as the target abnormal virtual object.

Optionally, the determined abnormal virtual object can be further confirmed manually and penalized correspondingly.

S1003, determining an abnormal account corresponding to the target abnormal virtual object according to the target abnormal virtual object and the corresponding relation between the first object and the game account corresponding to the first object in the target map.

Optionally, based on the correspondence between the first object and the game account corresponding to the first object shown in the knowledge graph, an abnormal account corresponding to each target abnormal virtual object may be determined, that is, the game account creating the target abnormal virtual object is taken as the abnormal account corresponding to the target abnormal virtual object.

Fig. 11 is a flowchart of an identification method of an abnormal account in a game according to an embodiment of the present application; optionally, in step S201, training a dialogue annotation classification model according to the game log of the target abnormal virtual object may include:

s1101, labeling dialogue types of dialogue information in a game log of the target abnormal virtual object, and acquiring first sample data.

In one implementation manner, all the dialogue information of all the determined target abnormal virtual objects can be obtained, and each dialogue information can be labeled in a manual labeling manner to expand rules or be classified into the seven types of features to expand labeling corpus, so as to obtain first sample data.

S1102, training and obtaining a dialogue annotation classification model by adopting first sample data.

Alternatively, based on the first sample data, a dialogue labeling classification model can be trained and acquired in combination with dialogue type labels labeled by each sample data.

In some embodiments, the method provided in the present application may be performed repeatedly, and after a certain number of labeling corpuses are accumulated, a machine learning or deep learning method may be used to train a dialogue labeling classification model, for example, CNN (Convolutional Neural Networks, convolutional neural network), LSTM (Long Short-Term Memory network), transducer (machine translation model), etc., and specific training processes will not be repeated here. After the initial dialogue labeling classification model is trained, dialogue information can be labeled simultaneously by combining a rule and a model, then manual labeling is continued, an iterative model is trained on line, and the recognition accuracy of the model is continuously improved.

Fig. 12 is a flowchart illustrating a method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S202, training the abnormal account identification model according to the data of the abnormal account corresponding to the target abnormal virtual object and the dialogue annotation classification model may include:

And S1201, taking target data of a first node, a second node and a third node corresponding to the abnormal account as second sample data, wherein the first node is a node where an abnormal virtual object corresponding to the abnormal account in the target map is located, the second node is a next node of the first node in the target map, and the third node is a next node of the second node in the target map.

Optionally, for the determined abnormal account, information of the abnormal account may be: the method comprises the steps of obtaining second sample data by taking node data corresponding to an abnormal account (a node where a game virtual object corresponding to the abnormal account is located) and sub-image data within two hops taking the node as a center as data of the abnormal account. As fig. 9 only shows the sub-graph data within one hop, in practical application, the sub-graph data within two hops can be obtained from the complete knowledge graph. Of course, the data of the abnormal account number can be based on sub-image data within three hops or within four hops, and can be obtained according to actual requirements.

S1202, training to obtain an abnormal account identification model by adopting the second sample data and the dialogue annotation classification model.

Optionally, the abnormal account identification model may be obtained through training based on the second sample data obtained through training in combination with the dialogue annotation classification model.

Fig. 13 is a flowchart illustrating a thirteenth method for identifying an abnormal account in a game according to an embodiment of the present application; optionally, in step S1202, training to obtain the abnormal account identification model using the second sample data and the dialogue annotation classification model may include:

s1301, performing dialogue classification annotation on the second sample data by adopting a dialogue annotation classification model.

Optionally, a dialogue labeling classification model obtained through training may be used to label dialogue information contained in the second sample data, so as to obtain labeled second sample data.

S1302, training and obtaining an abnormal account identification model according to the second sample data after labeling.

Optionally, the second sample data after labeling and the abnormal account number are used as the input of the model, and the abnormal account number identification model is obtained through training. Because the accuracy of the labeling result of the second sample data is higher, and the abnormal account number is determined through objective evaluation, the accuracy of the abnormal account number identification model obtained through training is higher.

The following describes a device, an electronic device, a storage medium, etc. for executing the method for identifying an abnormal account in a game provided by the present application, and specific implementation processes and technical effects of the method are referred to above, which are not described in detail below.

Fig. 14 is a schematic diagram of an apparatus for identifying an abnormal account in a game according to an embodiment of the present application, where a function implemented by the apparatus for identifying an abnormal account in a game corresponds to a step executed by the method. The apparatus may be understood as a terminal device or a server as described above, or a processor of a server, or may be understood as a component, which is independent from the server or the processor and performs the functions of the present application under the control of the server, and optionally the apparatus may include: an acquisition module 140, a construction module 141, a determination module 142, and a training module 142;

an acquisition module 140 for acquiring a sample data set comprising: a game log of at least one game virtual object under the target game in a preset time period;

the construction module 141 is configured to construct a target graph according to a game log of each game virtual object and a pre-constructed knowledge graph architecture, where the target graph is used to represent information of the game virtual object, and the information includes: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account;

The determining module 142 is configured to determine, based on the target map, a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object;

the training module 143 is configured to train a target recognition model according to the game log of the target abnormal virtual object and the data of the abnormal account corresponding to the target abnormal virtual object, where the target recognition model is used to recognize whether the game account corresponding to the virtual object to be recognized is the abnormal account.

the training module 143 is specifically configured to train a dialogue annotation classification model according to the game log of the target abnormal virtual object, where the dialogue annotation classification model is used to perform category annotation on the game log of the virtual object to be identified;

training an abnormal account identification model according to the data of the abnormal account corresponding to the target abnormal virtual object and the dialogue annotation classification model, wherein the abnormal account identification model is used for identifying whether the game account corresponding to the virtual object to be identified is an abnormal account.

Optionally, the determining module 142 is specifically configured to screen suspected abnormal virtual objects from the game virtual objects based on the target atlas;

And determining an abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object.

Optionally, the construction module 141 is specifically configured to convert the game logs of each game virtual object into node data and side length data according to a preset corresponding relationship between nodes and side lengths in the knowledge graph architecture;

And constructing a target map according to the target data of each node and the game log of the game virtual object corresponding to each node.

Optionally, the game log of the game virtual object includes the following information: the method comprises the steps of a dialogue type, a dialogue time stamp, feature information of a first object, dialogue information of the first object and feature information of a second object, wherein the first object represents a game virtual object initiating the dialogue, the second object represents the game virtual object being intersected by the first object, the second object comprises at least one, and the feature information comprises: virtual object identification, game level of the virtual object, member level of the virtual object, creation channel of the virtual object, game account corresponding to the virtual object, and dialogue type comprises: public dialogues and private dialogues.

Optionally, the determining module 142 is further configured to determine a role identifier node of the first object in the game log of each game virtual object, where the identifier is a knowledge graph, a role channel node of the first object, where a creation channel of the first object is a knowledge graph, and a role account node of the first object, where a game account corresponding to the first object is a knowledge graph;

determining the corresponding relation between a first object and a second object in a game log of each game virtual object, the corresponding relation between the first object and a creation channel of the first object, and the corresponding relation between the first object and a game account corresponding to the first object as the side length of a knowledge graph;

Optionally, the construction module 141 is specifically configured to convert, according to a correspondence between nodes and side lengths, an identifier of a first object in a game log of each game virtual object, common dialogue information corresponding to the first object, a game level of the first object, a member level of the first object into character identifier node data, a creation channel of the first object in the game log of each game virtual object into character channel node data, and a game account corresponding to the first object in the game log of each game virtual object into character account node data;

Optionally, the construction module 141 is specifically configured to match dialogue information included in each node data with key information of at least one dialogue category, and label the dialogue information included in each node data according to the matching result, so as to obtain a dialogue labeling result corresponding to each node data;

Optionally, the indexes of the preset category include: the accumulated times of dialogue categories, the diversity of first sentence dialogues and the interaction degree of dialogues are used for representing the dialogue initiative of the game virtual objects;

Optionally, the construction module 141 is specifically configured to determine the cumulative number of times of the session category corresponding to each node according to all session labeling results included in the data of each node and the session labeling results of the first object included in the data of each side;

according to the conversation quantity of the first object in the conversation information of the first object and the second object contained in each side length data and the conversation total quantity corresponding to the first object, determining the interaction degree of the conversations corresponding to each node, wherein the conversation total quantity corresponding to the first object comprises: the total number of conversations of the first object and the second object in the node data corresponding to the first object and the total number of conversations in the side length data corresponding to the first object;

Optionally, the construction module 141 is specifically configured to construct a target graph according to a corresponding relationship between the first object and the second object in the game log of the game virtual object corresponding to each node, a corresponding relationship between the first object and a creation channel of the first object, a corresponding relationship between the first object and a game account corresponding to the first object, and target data of each node.

Optionally, the determining module 142 is specifically configured to determine a suspected abnormal virtual object from game virtual objects corresponding to each node according to target data of each node in the target map and a preset evaluation threshold.

Optionally, the determining module 142 is specifically configured to determine, according to a correspondence between the first object and the second object in the target map, an associated virtual object corresponding to each suspected abnormal virtual object, where the associated virtual object is used to characterize a game virtual object that performs a session with the suspected abnormal virtual object;

determining a target abnormal virtual object from the suspected abnormal virtual objects according to the loss rate of the associated virtual object corresponding to the suspected abnormal virtual objects;

Optionally, the training module 143 is specifically configured to perform dialogue class labeling on dialogue information in the game log of the target abnormal virtual object, and obtain first sample data;

and training and obtaining a dialogue annotation classification model by adopting the first sample data.

Optionally, the training module 143 is specifically configured to use target data of a first node, a second node, and a third node corresponding to the abnormal account as second sample data, where the first node is a node where an abnormal virtual object corresponding to the abnormal account in the target graph is located, the second node is a next node of the first node in the target graph, and the third node is a next node of the second node in the target graph;

And training to obtain an abnormal account identification model by adopting the second sample data and the dialogue annotation classification model.

Optionally, the training module 143 is specifically configured to perform dialogue classification labeling on the second sample data by using a dialogue labeling classification model;

and training to obtain an abnormal account identification model according to the marked second sample data.

The foregoing apparatus is used for executing the method provided in the foregoing embodiment, and its implementation principle and technical effects are similar, and are not described herein again.

The above modules may be one or more integrated circuits configured to implement the above methods, for example: one or more application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as ASIC), or one or more microprocessors (digital singnal processor, abbreviated as DSP), or one or more field programmable gate arrays (Field Programmable Gate Array, abbreviated as FPGA), or the like. For another example, when a module above is implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processor that may invoke the program code. For another example, the modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

The modules may be connected or communicate with each other via wired or wireless connections. The wired connection may include a metal cable, optical cable, hybrid cable, or the like, or any combination thereof. The wireless connection may include a connection through a LAN, WAN, bluetooth, zigBee, or NFC, or any combination thereof. Two or more modules may be combined into a single module, and any one module may be divided into two or more units. It will be clearly understood by those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the method embodiments, which are not described in detail in this application.

It should be noted that these above modules may be one or more integrated circuits configured to implement the above methods, for example: one or more application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as ASIC), or one or more microprocessors (Digital Singnal Processor, abbreviated as DSP), or one or more field programmable gate arrays (Field Programmable Gate Array, abbreviated as FPGA), or the like. For another example, when a module above is implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processor that may invoke the program code. For another example, the modules may be integrated together and implemented in the form of a System-on-a-chip (SOC).

Fig. 15 is a schematic structural diagram of an electronic device according to an embodiment of the present application, where the device may include: a processor 801, and a memory 802.

The memory 802 is used for storing a program, and the processor 801 calls the program stored in the memory 802 to execute the above-described method embodiment. The specific implementation manner and the technical effect are similar, and are not repeated here.

Therein, the memory 802 stores program code that, when executed by the processor 801, causes the processor 801 to perform various steps in the method of identifying an in-game abnormal account according to various exemplary embodiments of the present application described in the section of the "exemplary method" described above in the present specification.

The processor 801 may be a general purpose processor such as a Central Processing Unit (CPU), digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, and may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present application. The general purpose processor may be a microprocessor or any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present application may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in the processor for execution.

Memory 802, as a non-volatile computer-readable storage medium, may be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. The Memory may include at least one type of storage medium, which may include, for example, flash Memory, hard disk, multimedia card, card Memory, random access Memory (Random Access Memory, RAM), static random access Memory (Static Random Access Memory, SRAM), programmable Read-Only Memory (Programmable Read Only Memory, PROM), read-Only Memory (ROM), charged erasable programmable Read-Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), magnetic Memory, magnetic disk, optical disk, and the like. The memory is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to such. The memory 802 in the embodiments of the present application may also be circuitry or any other device capable of implementing a memory function for storing program instructions and/or data.

Optionally, the present application also provides a program product, such as a computer readable storage medium, comprising a program for performing the above-described method embodiments when being executed by a processor.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.

The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (english: processor) to perform part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.

Claims

1. The method for identifying the abnormal account number in the game is characterized by comprising the following steps of:

according to the corresponding relation between the preset nodes and the side lengths in the knowledge graph framework, converting the game logs of each game virtual object into node data and side length data respectively;

constructing a target map according to the corresponding relation between a first object and a second object in a game log of a game virtual object corresponding to each node, the corresponding relation between the first object and a creation channel of the first object, the corresponding relation between the first object and a game account corresponding to the first object and target data of each node; the target atlas is used for representing information of a game virtual object, and the information comprises: characteristic information of the game virtual object, dialogue relation between the game virtual object and other game virtual objects, and corresponding relation between the game virtual object and a game account;

determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object;

2. The method of claim 1, wherein the object recognition model comprises: a dialogue annotation classification model and an abnormal account identification model;

3. The method of claim 1, wherein the game log of the game virtual object includes information as follows: a session type, a session timestamp, characteristic information of a first object, the session information of the first object, characteristic information of a second object, wherein the first object characterizes a game virtual object initiating the session, the second object characterizes the game virtual object being traversed by the first object, the second object comprises at least one, the characteristic information comprises: the method comprises the steps of virtual object identification, game level of a virtual object, member level of the virtual object, creation channel of the virtual object and game account corresponding to the virtual object, wherein the dialogue type comprises: public dialogues and private dialogues.

4. The method of claim 3, wherein before converting the game logs of each game virtual object into node data and side length data according to the preset correspondence between nodes and side lengths in the knowledge graph architecture, the method further comprises:

respectively determining a character identification node of a first object in a game log of each game virtual object, wherein the character identification node is identified as a knowledge graph, a character channel node of the first object, wherein a creation channel of the first object is the knowledge graph, and a character account node of the first object, wherein a game account corresponding to the first object is the knowledge graph;

5. The method of claim 4, wherein the converting the game log of each game virtual object into node data and side length data according to the preset correspondence between nodes and side lengths in the knowledge graph architecture comprises:

identifying a first object, a second object in a game log of each game virtual object,

The corresponding relation between the first object and the second object and the private dialogue information of the first object and the second object are converted into first side length data, the corresponding relation between the first object and the creation channel of the first object is converted into second side length data, the game account corresponding to the first object and the first object is converted into third side length data, and the side length data are obtained according to the first side length data, the second side length data and the third side length data.

6. The method according to claim 5, wherein the marking the dialogue type of each node data and the dialogue included in each side data according to the preset key information of at least one dialogue type to obtain a dialogue marking result corresponding to each node data and a dialogue marking result corresponding to each side data includes:

7. The method of claim 6, wherein the indicators of the predetermined category comprise: the method comprises the steps of accumulating times of dialogue categories, diversity of first sentence dialogues and interaction degree of dialogues, wherein the interaction degree of the dialogues is used for representing dialogue initiative of game virtual objects;

8. The method of claim 1, wherein the screening suspected abnormal virtual objects from the game virtual objects based on the target atlas comprises:

9. The method according to claim 8, wherein the determining, from the suspected abnormal virtual objects, the target abnormal virtual object and the abnormal account corresponding to the target abnormal virtual object includes:

10. The method of claim 2, wherein the training the dialog annotation classification model from the game log of the target abnormal virtual object comprises:

11. The method according to claim 10, wherein training the abnormal account identification model according to the data of the abnormal account corresponding to the target abnormal virtual object and the dialogue annotation classification model comprises:

12. The method of claim 11, wherein training the abnormal account identification model using the second sample data and the dialog annotation classification model comprises:

13. An apparatus for identifying an abnormal account in a game, comprising: the system comprises an acquisition module, a construction module, a determination module and a training module;

the building module is used for respectively converting the game logs of each game virtual object into node data and side length data according to the corresponding relation between the preset nodes and the side length in the knowledge graph framework;

the determining module is used for screening suspected abnormal virtual objects from all game virtual objects based on the target map; determining a target abnormal virtual object and an abnormal account corresponding to the target abnormal virtual object from the suspected abnormal virtual object;

14. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing program instructions executable by the processor, the processor and the storage medium communicating over the bus when the electronic device is running, the processor executing the program instructions to perform the steps of the method according to any one of claims 1 to 12 when executed.

15. A computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the method according to any of claims 1 to 12.