Abnormity detection method and device in game, equipment and storage medium

Abstract

The invention provides an anomaly detection method and device in a game, equipment and a storage medium, and the anomaly detection method comprises the steps: obtaining chat record information of a plurality of players in the game, and determining target chat record information containing a target keyword from the obtained chat record information; and performing classification prediction on the target chat record information, and when it is determined that a classification prediction result of the target chat record information belongs to an abnormal category, determining a detected game abnormality from the target chat record information. Therefore, the game abnormity which is not fed back by the player through an official channel can be actively detected in a mode of actively performing game abnormity detection on the chat record information of the player in the game, and the efficiency and the accuracy of abnormity detection in the game are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of game technology, and in particular to a method, device, equipment and storage medium for detecting anomalies in games. Background Art

[0002] In existing games, before the game is launched, testers can detect whether there are game bugs that need to be fixed (i.e., game anomalies) in the game through in-game testing. After the game is launched, it mainly depends on player feedback. When players report abnormal problems encountered in the game through channels such as the in-game user feedback system and the game's official website customer service, developers can detect and repair the game anomalies reported by players based on the player feedback information received. As a result, developers can only passively detect some game anomalies that may exist in the game according to the player's feedback information, which makes the detection efficiency of game anomalies low. Summary of the invention

[0003] In view of this, the present application provides a method, device, equipment and storage medium for detecting anomalies in games. By actively detecting game anomalies in the chat record information of players in the game, it is possible to actively detect game anomalies that players have not reported through official channels, thereby effectively improving the efficiency and accuracy of anomaly detection in the game.

[0004] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings.

[0005] In a first aspect, an embodiment of the present application provides a method for detecting anomalies in a game, the method comprising:

[0006] Acquire chat record information of multiple players in the game, and determine target chat record information containing target keywords from the acquired multiple chat record information; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game;

[0007] The target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

[0008] In a second aspect, an embodiment of the present application provides an anomaly detection device in a game, the anomaly detection device comprising:

[0009] A first detection module is used to obtain chat record information of multiple players in the game, and determine target chat record information containing target keywords from the multiple chat record information obtained; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game;

[0010] The second detection module is used to classify and predict the target chat record information. When it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

[0011] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for detecting anomalies in a game when executing the computer program.

[0012] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for detecting anomalies in a game are executed.

[0013] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0014] The embodiments of the present application provide a method, device, equipment and storage medium for detecting anomalies in a game, which obtains chat record information of multiple players in the game, determines target chat record information containing target keywords from the multiple chat record information obtained, and classifies and predicts the target chat record information. When it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, the detected game anomaly is determined from the target chat record information. In this way, the present application can actively detect game anomalies that players have not reported through official channels by actively detecting game anomalies in the chat record information of players in the game, effectively improving the efficiency and accuracy of anomaly detection in the game. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A schematic diagram showing a flow chart of an anomaly detection method in a game provided by an embodiment of the present application;

[0017] Figure 2 A schematic diagram showing a flow chart of a method for detecting whether a chat record information contains a target keyword provided by an embodiment of the present application;

[0018] Figure 3 A schematic diagram of a process for predicting sentiment classification of target chat record information provided by an embodiment of the present application is shown;

[0019] Figure 4 A schematic diagram of a process for predicting abnormal text classification of target chat record information provided by an embodiment of the present application is shown;

[0020] Figure 5 A schematic diagram of a process for predicting sentiment classification and abnormal text classification of target chat record information provided by an embodiment of the present application is shown;

[0021] Figure 6 A flow chart of a method for excluding non-abnormal text from target chat record information provided by an embodiment of the present application is shown;

[0022] Figure 7 A schematic diagram of the structure of an anomaly detection device in a game provided by an embodiment of the present application is shown;

[0023] Figure 8 A schematic diagram of the structure of an electronic device 800 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0025] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0026] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0027] In existing games, before the game is launched, testers can detect whether there are game bugs that need to be fixed in the game through in-game testing. After the game is launched, it mainly depends on the feedback of players. When players report abnormal problems encountered in the game through channels such as the user feedback system in the game and the customer service of the official website of the game, the developers can detect and repair the game abnormalities reported by the players based on the feedback information received from the players. As a result, the developers can only passively detect some game abnormalities that may exist in the game according to the feedback information from the players, which makes the detection efficiency of game abnormalities low.

[0028] Based on this, the embodiments of the present application provide a method, apparatus, device and storage medium for detecting anomalies in games. By actively detecting game anomalies in the chat record information of players in the game, it is possible to actively detect game anomalies that players have not reported through official channels, thereby effectively improving the efficiency and accuracy of anomaly detection in the game.

[0029] In one embodiment of the present application, an anomaly detection method in a game can be run on a terminal device or a server. The terminal device can be a local terminal device. When the anomaly detection method in the game is run on a server, the anomaly detection method can be implemented and executed based on a cloud interaction system, wherein the cloud interaction system includes a server and a client device (i.e., a terminal device).

[0030] In an optional implementation, various cloud applications, such as cloud games, can be run under the cloud interaction system. Taking cloud games as an example, cloud games refer to a game mode based on cloud computing. In the operation mode of cloud games, the operating body of the game program and the main body of the game screen presentation are separated. The storage and operation of the abnormal detection method in the game are completed on the cloud game server. The role of the client device is to receive and send data and present the game screen. For example, the client device can be a display device with data transmission function close to the user side, such as a mobile terminal, a TV, a computer, a handheld computer, etc.; but the terminal device for information processing is a cloud game server in the cloud. When playing the game, the player operates the client device to send an operation instruction to the cloud game server. The cloud game server runs the game according to the operation instruction, encodes and compresses the game screen and other data, and returns it to the client device through the network. Finally, the client device decodes and outputs the game screen.

[0031] In an optional embodiment, the terminal device may be a local terminal device. Taking a game as an example, the local terminal device stores a game program and is used to present a game screen. The local terminal device is used to interact with the player through a graphical user interface, that is, the game program is downloaded and installed by an electronic device and run conventionally. The local terminal device may provide the graphical user interface to the player in a variety of ways, for example, it may be rendered and displayed on a display screen of the terminal, or provided to the player through a holographic projection. For example, the local terminal device may include a display screen and a processor, the display screen is used to present a graphical user interface, the graphical user interface includes a game screen, and the processor is used to run the game, generate a graphical user interface, and control the display of the graphical user interface on the display screen.

[0032] To facilitate understanding of the embodiments of the present application, a method, device, equipment and storage medium for detecting anomalies in a game provided by the embodiments of the present application are described in detail below.

[0033] Reference Figure 1 As shown, Figure 1 A flowchart of an anomaly detection method in a game provided by an embodiment of the present application is shown, wherein the anomaly detection method includes steps S101-S102; specifically:

[0034] S101, obtaining chat record information of multiple players in the game, and determining target chat record information containing a target keyword from the multiple chat record information obtained.

[0035] S102, performing classification prediction on the target chat record information, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, determining the detected game abnormality from the target chat record information.

[0036] The anomaly detection method in the above-mentioned game provided by the embodiment of the present application obtains chat record information of multiple players in the game, and determines the target chat record information containing the target keyword from the multiple chat record information obtained; classifies and predicts the target chat record information, and when it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, determines the detected game anomaly from the target chat record information. In this way, the present application can actively detect game anomalies that players have not reported through official channels by actively performing game anomaly detection on the chat record information of players in the game, effectively improving the efficiency and accuracy of anomaly detection in the game.

[0037] The following is an exemplary description of each step in the above-mentioned anomaly detection method in the game provided in the embodiment of the present application:

[0038] S101, obtaining chat record information of multiple players in the game, and determining target chat record information containing a target keyword from the multiple chat record information obtained.

[0039] Here, the anomaly detection method in the game provided in the embodiment of the present application can be applied to the target device; wherein, the target device can be a game server, or it can be an anomaly detection terminal device corresponding to the game tester; the embodiment of the present application does not impose any limitation on the specific device type to which the above-mentioned target device belongs.

[0040] Specifically, during the game operation process, the target device can pull the chat record information of all players in the game from the Kafka data stream in real time. Considering that the chat record information of players in the game usually contains a lot of non-text content that has no actual meaning (such as emoticons, pictures, links, punctuation marks, pure English, numbers and other non-Chinese characters), as an optional embodiment, in order to avoid the above-mentioned non-text content from interfering with the subsequent detection of game anomalies, the chat record information of each player obtained can also be pre-processed to filter out the above-mentioned non-text content that has no actual meaning from the chat record information.

[0041] In an embodiment of the present application, the chat record information obtained above may be a single chat message between the speaking player and other players (equivalent to the speech content of the speaking player). Considering that a single chat message may not be able to better judge whether the speaking player has encountered a game bug (that is, a game anomaly) in the game for game testers, as another optional embodiment, the context within the speaking player's chat record time window can also be aggregated to obtain context-related information for each speech message, so that game anomaly detection can be performed on the aggregated context-related information to more accurately analyze the reason why the player triggered the game bug.

[0042] In addition, the game scene in which the speaking player is located when speaking (including but not limited to the specific role position of the controlled virtual character controlled by the speaking player in the game scene and the game scene map where the controlled virtual character is located) and other related information can be associated with the speech content of the speaking player and the above-mentioned context-related information based on the character id (Identity document) of the speaking player and the timestamp of the speech. All the associated information is obtained as the chat record information of the above-mentioned speaking player (that is, the above-mentioned chat record information may include but is not limited to: the speech content of the speaking player, the context-related information of the speech content within the chat record time window, the character id of the speaking player, the specific role position of the controlled virtual character controlled by the speaking player in the game scene and the game scene map where the controlled virtual character is located), so that game testers can more accurately locate and reproduce the game bug (that is, game anomaly) mentioned by the speaking player in the chat record information based on the above-mentioned chat record information.

[0043] Specifically, in the underlying technical implementation, the chat record information after the above-mentioned association processing can be written into the Kafka data stream in real time, so that the target device can directly pull the chat record information after the above-mentioned association processing from the Kafka data stream in real time. The specific data format of the chat record information after the above-mentioned association processing in the Kafka data stream is as follows:

[0044] {

[0045] "role_id":"001",

[0046] "chat_content":"***",

[0047] "scene_id":"0-0-1",

[0048] "axis":"(113,109,105)",

[0049] "agg_chat_content":"******"

[0050] }

[0051] Among them, role_id is the role id of the speaking player, chat_content is the speech content of the speaking player, scene_id is the game scene map where the controlled virtual character controlled by the speaking player is located at the time of speaking, axis is the specific role position of the above-mentioned controlled virtual character in the game scene at the time of speaking, and agg_chat_content is the context-related information of the above-mentioned speech content within the chat record time window.

[0052] Here, the target keyword represents the key words involved in describing abnormal problems occurring in the game, wherein the words commonly used by players to describe abnormal problems encountered when encountering game abnormalities can be summarized to obtain a preset keyword list; and the above-mentioned target keyword can represent any keyword contained in the preset keyword list.

[0053] S102, performing classification prediction on the target chat record information, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, determining the detected game abnormality from the target chat record information.

[0054] Here, the abnormality category represents the text category to which the game abnormality description information belongs; wherein, similar to the above-mentioned target keyword, the above-mentioned game abnormality description information also belongs to the relevant description information for the specific game abnormality encountered in the game; that is, considering that by executing the above-mentioned step S101, although the chat record information that does not contain the above-mentioned target keyword can be quickly filtered out, the efficiency of game abnormality detection is effectively improved, but only by detecting whether the chat record information contains the target keyword in the above-mentioned step S101, it may not be accurate to judge from a single dimension whether the chat record information contains game abnormality description information about the game abnormality. Therefore, the above-mentioned target chat record information that is determined to contain the target keyword can also be classified and predicted at least once to improve the accuracy of game abnormality detection.

[0055] It should be noted that the specific classification prediction method for the target chat record information in step S102 is not unique. For example, the target chat record information can be subjected to sentiment classification prediction to determine that the target chat record information with a sentiment classification prediction result of negative sentiment belongs to the above-mentioned abnormal category; the target chat record information can also be subjected to text classification prediction to determine that the target chat record information with a text classification prediction result belonging to the above-mentioned game abnormal description information belongs to the above-mentioned abnormal category; the target chat record information can also be subjected to the above-mentioned sentiment classification prediction and text classification prediction respectively (equivalent to performing multiple classification predictions) to determine that the target chat record information with a sentiment classification prediction result belonging to negative sentiment and a text classification prediction result belonging to the above-mentioned game abnormal description information belongs to the above-mentioned abnormal category; the embodiment of the present application does not impose any limitation on the specific classification prediction method adopted in the above-mentioned step S102.

[0056] Specifically, after determining that the target chat record information contains relevant description information about the game abnormality (i.e., determining that the classification prediction result of the target chat record information belongs to the above abnormality category), combined with the relevant description content of the specific information content of the chat record information in the above step S101, the method described in the following steps a1-a2 can be used to determine the detected game abnormality from the target chat record information:

[0057] Step a1: Determine the game abnormality description information corresponding to the game abnormality according to the speech content of the speaking player in the target chat record information.

[0058] Here, in combination with the relevant description of the specific information content of the chat record information in the above step S101, it can be known that the target chat record information specifically includes: the speech content of the speaking player, context-related information of the speech content within the chat record time window, the character id of the speaking player, the specific role position of the controlled virtual character controlled by the speaking player in the game scene, and the game scene map where the controlled virtual character is located; wherein the target device can determine the specific description information of the speaking player on the game abnormality encountered (i.e. the above-mentioned game abnormality description information) from the speech content of the speaking player.

[0059] Step a2: Determine the location where the game anomaly occurs in the game according to the game scene in which the speaking player in the target chat record information is speaking.

[0060] Here, in combination with the relevant explanation content of the specific information content of the chat record information in the above step S101, it can be known that in the target chat record information, the game scene where the speaking player is located when speaking may include: the specific role position of the controlled virtual character controlled by the speaking player in the game scene and the game scene map where the controlled virtual character is located; that is, the target device can determine the specific location of the game anomaly described by the player in the target chat record information from the game scene where the speaking player is located when speaking, so that the developer can accurately locate the location of the game operation anomaly and improve the repair efficiency.

[0061] The specific implementation process of each of the above steps in the embodiment of the present application is described in detail below:

[0062] Regarding the specific implementation of the above step S101, in an optional implementation, Figure 2 A flow chart of a method for detecting whether a chat record information contains a target keyword provided by an embodiment of the present application is shown as follows: Figure 2 As shown, when executing step S101, the method includes steps S201-S203, specifically:

[0063] S201, for each chat record information, extract a first keyword matching the custom vocabulary from the chat record information according to a pre-loaded custom vocabulary.

[0064] Here, the above-mentioned custom vocabulary is determined based on the game terms corresponding to the fixed indication objects in the game; wherein the above-mentioned fixed indication objects include but are not limited to: maps, gameplay, game characters, props, skills, etc. in the game; the embodiment of the present application does not impose any limitation on the specific game terms included in the above-mentioned custom vocabulary.

[0065] Specifically, before detecting the target keyword, by pre-loading the above-mentioned custom vocabulary list, it can be ensured that when any first keyword matched by the above-mentioned custom vocabulary list exists in the chat record information, the first keyword can be accurately extracted from the chat record information, thereby avoiding the risk of the above-mentioned first keyword being erroneously filtered out as an invalid word in the subsequent step S202.

[0066] S202, deleting invalid words from the chat record information to obtain the remaining second keywords.

[0067] Here, the invalid words include but are not limited to: meaningless function words (such as modal particles, conjunctions, prepositions, etc. without actual meaning) and participles whose importance in the chat record information is lower than a preset importance threshold.

[0068] Specifically, as an optional embodiment, the above-mentioned modal particles, conjunctions, prepositions and other function words without actual meaning and useless participles that often appear in historical chat data as unimportant words (i.e., participles whose importance in historical chat data is lower than a preset importance threshold) can be set as stop words in advance, and a stop word list composed of the above-mentioned stop words can be established; wherein, when executing step S202, the above-mentioned stop word list that is set in advance can be used to treat the above-mentioned stop words that appear in the stop word list as invalid words, and the above-mentioned invalid words can be deleted from the chat record information.

[0069] Specifically, as another optional embodiment, the chat record information in step S202 can also be segmented by a TF-IDF (Term Frequency-Inverse Document Frequency) algorithm, wherein the TF-IDF algorithm also annotates each obtained segmentation with the segmentation importance corresponding to the segmentation in the chat record information (TF represents the frequency of the segmentation in the chat record information, IDF represents the frequency of the segmentation in the corpus, and the importance of the segmentation is proportional to TF and inversely proportional to IDF), so that according to the above segmentation importance annotated by the TF-IDF algorithm for each segmentation, the segmentations whose segmentation importance is lower than a preset importance threshold can be deleted from the chat record information to obtain the remaining segmentations as the above second keywords.

[0070] It should be noted that, when executing the above step S202, in addition to using the above stop word list alone or using the above TF-IDF algorithm alone, the above stop word list and the above TF-IDF algorithm can also be used at the same time to delete invalid words from the chat record information; the specific method of deleting invalid words in the above step S202 is not limited in any way in the embodiment of the present application.

[0071] S203: According to the keyword set consisting of the first keyword and the second keyword, when it is detected that there is at least one target keyword matching a preset keyword list in the keyword set, determining that the chat record information belongs to the target chat record information.

[0072] Here, the target keyword may represent any keyword in a preset keyword list, wherein the preset keyword list may be determined based on common descriptive words involved in describing abnormal problems occurring in a game and the game terms in the above-mentioned custom vocabulary list.

[0073] Specifically, when encountering a game bug (i.e., a game abnormality), the player's chat record information generally includes some general descriptive words that describe the game bug. The general descriptive words that appear in the historical chat record information can be counted in advance, and a public keyword list can be established based on the counted general descriptive words; among them, the general descriptive words include but are not limited to: stuck, stuck, good stuck, crash, bug, problematic, etc.; the specific content of the above-mentioned general descriptive words is not limited in any way in the embodiments of the present application.

[0074] Specifically, game testers will pay more attention to specific scenes in the game, for example, maps, gameplay, heroes, props, skills, etc. in the game (i.e., the game terms in the above-mentioned custom vocabulary). These game terms can be used as a game project keyword list, and the game project keyword list is merged with the above-mentioned public keyword list to obtain the merged result as the above-mentioned preset keyword list; at this time, when executing step S203, the above-mentioned keyword set extracted from the chat record information is matched with the preset keyword list, and when the above-mentioned keyword set has at least one target keyword that matches the preset keyword list, it is determined that the chat record information belongs to the target chat record information (equivalent to the fact that the speaking player in the chat record information has a certain possibility of triggering a game bug).

[0075] Regarding the specific implementation of the above step S102, in an optional implementation, Figure 3 A schematic diagram of a process for predicting emotion classification of target chat record information provided by an embodiment of the present application is shown, such as Figure 3 As shown, when executing step S102, the method includes steps S301-S302, specifically:

[0076] S301, inputting the target chat record information into a sentiment classification model, and outputting a sentiment classification prediction result of the target chat record information.

[0077] Here, considering that players usually have negative emotions after encountering game bugs, when players speak and complain in the chat system, their subjective emotions tend to be negative. Therefore, as an optional embodiment, the target chat record information of the players can be emotionally classified, and the target chat record information containing the target keywords can be further screened out as the chat record information with a higher possibility of encountering game anomalies, which is beneficial to improving the accuracy of game anomaly detection.

[0078] It should be noted that the above-mentioned sentiment classification model can be a pre-trained sentiment classification model, or it can be an open source sentiment classification model directly used; taking the uie-senta-base sentiment classification model as an example, the uie-senta-base model supports sentence-level sentiment analysis. After the target chat record information is input into the uie-senta-base model, the uie-senta-base model can output the classification probability that the target chat record information belongs to positive sentiment or the classification probability that it belongs to negative sentiment; the specific model structure of the above-mentioned sentiment classification model is not limited in any way in the embodiments of the present application.

[0079] S302: When the emotion classification prediction result belongs to a negative emotion, determining that the classification prediction result of the target chat record information belongs to the abnormal category.

[0080] Specifically, taking the sentiment classification model as the above-mentioned uie-senta-base model as an example, after inputting the target chat record information into the uie-senta-base model, the uie-senta-base model can output the classification probability that the target chat record information belongs to positive emotion or the classification probability that it belongs to negative emotion; wherein, when the classification probability that the target chat record information belongs to negative emotion is higher than the specified threshold, it can be determined that the classification prediction result of the target chat record information belongs to the abnormal category (equivalent to the target chat record information containing descriptive information about the game abnormality).

[0081] Regarding the specific implementation of the above step S102, in an optional implementation, Figure 4 A flow chart of a method for classifying and predicting abnormal text in target chat record information provided by an embodiment of the present application is shown. Figure 4 As shown, when executing step S102, the method includes steps S401-S402, specifically:

[0082] S401, inputting the target chat record information into a pre-trained abnormal text classification model, and outputting a text classification prediction result of the target chat record information.

[0083] Here, before executing step S401, historical chat record information of multiple players can be collected from the game in advance as training data, wherein each historical chat record information carries a real label, and the real label is used to mark whether the historical chat record information contains game abnormality description information (i.e., description information about game abnormalities). The above training data is input into the original text classification model, and text classification prediction is performed on each input historical chat record information through the original text classification model, and the text classification prediction result of each historical chat record information (i.e., the classification probability of containing game abnormality description information and the classification probability of not containing game abnormality description information in the historical chat record information) is output. According to the classification loss between the text classification prediction result of each historical chat record information and the real label of the historical chat record information, the original text classification model is trained until the original text classification model converges, and the converged original text classification model is obtained as the trained abnormal text classification model.

[0084] Specifically, by inputting the target chat information into the pre-trained abnormal text classification model, the classification probability of the target chat record information containing the game abnormality description information and the classification probability of the target chat record information not containing the game abnormality description information (i.e., the text classification prediction result of the target chat record information) can be output.

[0085] It should be noted that the above-mentioned abnormal text classification model can be a BERT (Bidirectional Encoder Representations from Transformers, a deep bidirectional language representation model based on the Transformer architecture) model, or it can be a BERT model as a basic model and fine-tuned on its basis. The specific steps are: use the fine-tune method to load the pre-trained BERT model, load the cleaned historical chat record information into the model as training data, calculate the error between the predicted value and the expected value through forward propagation, and then calculate the gradient of the loss of the network weight through back propagation. This gradient will be fed back to the optimization method (such as the gradient descent method) to update the network weight parameters to minimize the loss function; the embodiment of the present application does not impose any limitation on the specific model structure to which the above-mentioned abnormal text classification model belongs.

[0086] S402, when the text classification prediction result belongs to the abnormal text category containing game abnormality description information, determine that the classification prediction result of the target chat record information belongs to the abnormal category.

[0087] Specifically, after the target chat record information is input into the abnormal text classification model, the abnormal text classification model can output the classification probability of the target chat record information containing game abnormality description information and the classification probability of not containing game abnormality description information; wherein, when the classification probability of the target chat record information containing game abnormality description information is higher than a preset threshold, it can be determined that the classification prediction result of the target chat record information belongs to the abnormal text category containing game abnormality description information (equivalent to the target chat record information containing description information about the game abnormality).

[0088] Regarding the specific implementation of the above step S102, in an optional implementation, Figure 5 A flow chart of a method for performing sentiment classification prediction and abnormal text classification prediction on target chat record information provided by an embodiment of the present application is shown, such as Figure 5 As shown, when executing step S102, the method includes steps S501-S503, specifically:

[0089] S501, inputting the target chat record information into a sentiment classification model, and outputting a sentiment classification prediction result of the target chat record information.

[0090] Here, the specific implementation of step S501 can refer to the specific implementation of the aforementioned step S301, and the repeated parts will not be repeated here.

[0091] S502, when the emotion classification prediction result belongs to negative emotion, the target chat record information is input into a pre-trained abnormal text classification model, and the text classification prediction result of the target chat record information is output.

[0092] Here, the specific implementation of step S502 may refer to the specific implementation of the aforementioned step S302 and step S401, and the repeated parts will not be repeated here.

[0093] S503: When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, determine that the classification prediction result of the target chat record information belongs to the abnormal category.

[0094] Here, the specific implementation of step S503 may refer to the specific implementation of the aforementioned step S402, and the repeated parts will not be repeated here.

[0095] It should be noted that, compared with the classification prediction method of using sentiment classification alone as shown in the above steps S301-S302 and the classification prediction method of using abnormal text classification alone as shown in the above steps S401-S402, when the target chat record information is respectively subjected to the above sentiment classification prediction and abnormal text classification prediction according to the method shown in the above steps S501-S503 (equivalent to performing multiple classification predictions), it is beneficial to improve the accuracy of game anomaly detection.

[0096] Based on the above steps S101-S102, considering that there is still a certain probability that the target chat record information may be a false detection caused by a bad gaming experience caused by factors such as a player's network freeze or device freeze, therefore, in an optional implementation, Figure 6 A flow chart of a method for excluding non-abnormal text from target chat record information provided by an embodiment of the present application is shown, such as Figure 6 As shown, after executing step S102, the method further includes steps S601-S603, specifically:

[0097] S601, performing vector encoding on the target chat record information to obtain a target vector corresponding to the target chat record information.

[0098] Here, the vector encoding method is not unique, and the embodiment of the present application does not impose any limitation on the specific encoding method of the above-mentioned vector encoding.

[0099] Specifically, as an optional embodiment, the target chat record information can be input into the Embedding model, the target chat record information is vector-encoded through the Embedding model, and the vector encoding result of the target chat record information is output as the target vector corresponding to the target chat record information.

[0100] S602: Perform a vector search on the target vector in a preset vector database to obtain a vector search result of the target vector.

[0101] Here, a plurality of known vectors are stored in the preset vector database, wherein the text information matched by each known vector belongs to non-game abnormality description information; that is, the preset vector database stores the vector encoding results matched by the normal text information that is misjudged as game abnormality description information.

[0102] Specifically, the vector similarity between the target vector and each known vector stored in the preset vector database can be calculated respectively, and then each known vector can be sorted in order of vector similarity from high to low, and a preset number of known vectors with the highest ranking in vector similarity results can be determined as vector retrieval results for the target vector; wherein, the specific value of the above preset number can be set according to actual vector retrieval requirements, and the embodiment of the present application does not impose any limitation on the specific value of the above preset number.

[0103] It should be noted that the above-mentioned preset vector database can be an open source Milvus vector database or other types of vector databases. The embodiment of the present application does not impose any limitation on the specific database type to which the above-mentioned preset vector database belongs.

[0104] S603: When there is a known vector in the vector retrieval result whose similarity with the target vector is greater than or equal to a preset similarity threshold, it is determined that no game abnormality is detected in the target chat record information.

[0105] Here, after obtaining the vector retrieval result of the target vector, according to the similarity between each known vector in the vector retrieval result and the target vector, when there is a known vector with a similarity higher than a preset similarity threshold, it can be determined that the target chat record information is similar to the known vector, that is, based on the vector encoding result that matches the normal text information that is mistakenly judged as game abnormality description information, the target chat record information can be determined to also be normal text information that is mistakenly judged as game abnormality description information, and the target chat record information does not contain game abnormality description information (that is, it is determined that no game abnormality is detected in the target chat record information).

[0106] Specifically, if there is no known vector with a similarity higher than a preset similarity threshold in the vector retrieval result, it can be determined that the target chat information contains game abnormality description information, and thus the detected game abnormality can be determined from the target chat record information according to the method shown in step S102 above.

[0107] Here, as an optional embodiment, when it is determined that no game abnormality is detected in the target chat record information, the target chat record information can be stored in a preset vector database so that the known vectors stored in the preset vector database can be updated according to the above target chat record information detected in real time, which is beneficial to improve the subsequent detection accuracy of game abnormalities.

[0108] In addition, as another optional embodiment, when it is detected that the number of known vectors stored in the preset vector database is greater than or equal to the preset number threshold, the chat record information matched by each known vector in the preset vector database can be added to the model training data of the above-mentioned abnormal text classification model as the text information with the real label of "does not contain game abnormality description information", and the abnormal text classification model can be re-trained so that the abnormal text classification model can better understand the difference between the chat record information that does not contain game abnormality description information and the chat record information that contains game abnormality description information, which is conducive to improving the accuracy of the model output results of the abnormal text classification model.

[0109] Based on the above anomaly detection method provided by the embodiment of the present application, the chat record information of multiple players in the game is obtained, and the target chat record information containing the target keyword is determined from the multiple chat record information obtained; the target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, the detected game anomaly is determined from the target chat record information. In this way, the present application can actively detect game anomalies that players have not reported through official channels by actively performing game anomaly detection on the chat record information of players in the game, effectively improving the efficiency and accuracy of anomaly detection in the game.

[0110] Based on the same inventive concept, the present application also provides an anomaly detection device corresponding to the anomaly detection method in the above-mentioned game. Since the principle of solving the problem by the anomaly detection device in the embodiment of the present application is similar to the anomaly detection method in the above-mentioned game in the embodiment of the present application, the implementation of the anomaly detection device can refer to the implementation of the above-mentioned anomaly detection method, and the repeated parts will not be repeated.

[0111] Reference Figure 7 As shown, Figure 7 A schematic diagram of the structure of an anomaly detection device in a game provided in an embodiment of the present application is shown, wherein the anomaly detection device comprises:

[0112] The first detection module 701 is used to obtain chat record information of multiple players in the game, and determine target chat record information containing target keywords from the multiple chat record information obtained; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game;

[0113] The second detection module 702 is used to classify and predict the target chat record information. When it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

[0114] In an optional implementation, when determining the target chat record information containing the target keyword from the acquired multiple chat record information, the first detection module 701 is used to:

[0115] For each chat record information, extracting a first keyword matching the custom vocabulary from the chat record information according to a pre-loaded custom vocabulary; wherein the custom vocabulary is determined according to game terms corresponding to fixed indicating objects in the game;

[0116] Deleting invalid words from the chat record information to obtain the remaining second keywords; wherein the invalid words include: meaningless function words and participles in the chat record information whose participle importance is lower than a preset importance threshold;

[0117] According to a keyword set consisting of the first keyword and the second keyword, when it is detected that there is at least one target keyword matching a preset keyword list in the keyword set, it is determined that the chat record information belongs to the target chat record information; wherein, the preset keyword list is determined based on the general descriptive vocabulary involved in describing abnormal problems occurring in the game and the game terms.

[0118] In an optional implementation, when the target chat record information is classified and predicted, the second detection module 702 is used to:

[0119] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0120] When the emotion classification prediction result belongs to a negative emotion, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0121] In an optional implementation, when the target chat record information is classified and predicted, the second detection module 702 is used to:

[0122] Input the target chat record information into a pre-trained abnormal text classification model, and output a text classification prediction result of the target chat record information;

[0123] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0124] In an optional implementation, when the target chat record information is classified and predicted, the second detection module 702 is used to:

[0125] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0126] When the sentiment classification prediction result belongs to negative sentiment, the target chat record information is input into a pre-trained abnormal text classification model, and the text classification prediction result of the target chat record information is output;

[0127] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0128] In an optional implementation, when the detected game anomaly is determined from the target chat record information, the second detection module 702 is used to:

[0129] Determine the game abnormality description information corresponding to the game abnormality according to the speech content of the speaking player in the target chat record information;

[0130] The occurrence position of the game anomaly in the game is determined according to the game scene in which the speaking player in the target chat record information is located when speaking.

[0131] In an optional implementation, after determining that the classification prediction result of the target chat record information belongs to an abnormal category, the second detection module 702 is further configured to:

[0132] Performing vector encoding on the target chat record information to obtain a target vector corresponding to the target chat record information;

[0133] In a preset vector database, a vector search is performed on the target vector to obtain a vector search result of the target vector; wherein the preset vector database stores a plurality of known vectors, and the text information matched by the known vectors belongs to non-game abnormality description information;

[0134] When there is a known vector in the vector retrieval result whose similarity with the target vector is greater than or equal to a preset similarity threshold, it is determined that no game abnormality is detected in the target chat record information.

[0135] In an optional implementation, the second detection module 702 is further configured to:

[0136] When it is determined that no game abnormality is detected in the target chat record information, the target chat record information is stored in the preset vector database.

[0137] Based on the anomaly detection device in the above game provided by the embodiment of the present application, the chat record information of multiple players in the game is obtained, and the target chat record information containing the target keyword is determined from the multiple chat record information obtained; the target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, the detected game anomaly is determined from the target chat record information. In this way, the present application can actively detect game anomalies that players have not reported through official channels by actively performing game anomaly detection on the chat record information of players in the game, effectively improving the efficiency and accuracy of anomaly detection in the game.

[0138] Based on the same inventive concept, the present application also provides an electronic device corresponding to the anomaly detection method in the above-mentioned game. Since the principle of solving the problem by the electronic device in the embodiment of the present application is similar to the anomaly detection method in the above-mentioned game in the embodiment of the present application, the implementation of the electronic device can refer to the implementation of the above-mentioned anomaly detection method, and the repeated parts will not be repeated.

[0139] Figure 8 A structural diagram of an electronic device 800 provided in an embodiment of the present application includes: a processor 801, a memory 802 and a bus 803, wherein the memory 802 stores machine-readable instructions executable by the processor 801. When the electronic device runs an abnormality detection method in a game in the embodiment, the processor 801 communicates with the memory 802 via the bus 803, and the processor 801 executes the machine-readable instructions. When the processor 801 executes the machine-readable instructions, the following steps are implemented, specifically:

[0140] Acquire chat record information of multiple players in the game, and determine target chat record information containing target keywords from the acquired multiple chat record information; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game;

[0141] The target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

[0142] In an optional implementation, when determining the target chat record information containing the target keyword from the plurality of chat record information obtained, the processor 801 is configured to:

[0143] For each chat record information, extracting a first keyword matching the custom vocabulary from the chat record information according to a pre-loaded custom vocabulary; wherein the custom vocabulary is determined according to game terms corresponding to fixed indicating objects in the game;

[0144] Deleting invalid words from the chat record information to obtain the remaining second keywords; wherein the invalid words include: meaningless function words and participles in the chat record information whose participle importance is lower than a preset importance threshold;

[0145] According to a keyword set consisting of the first keyword and the second keyword, when it is detected that there is at least one target keyword matching a preset keyword list in the keyword set, it is determined that the chat record information belongs to the target chat record information; wherein, the preset keyword list is determined based on the general descriptive vocabulary involved in describing abnormal problems occurring in the game and the game terms.

[0146] In an optional implementation, when performing classification prediction on the target chat record information, the processor 801 is configured to:

[0147] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0148] When the emotion classification prediction result belongs to a negative emotion, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0149] In an optional implementation, when performing classification prediction on the target chat record information, the processor 801 is configured to:

[0150] Input the target chat record information into a pre-trained abnormal text classification model, and output a text classification prediction result of the target chat record information;

[0151] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0152] In an optional implementation, when performing classification prediction on the target chat record information, the processor 801 is configured to:

[0153] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0154] When the sentiment classification prediction result belongs to negative sentiment, the target chat record information is input into a pre-trained abnormal text classification model, and the text classification prediction result of the target chat record information is output;

[0155] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0156] In an optional implementation, when determining the detected game abnormality from the target chat record information, the processor 801 is configured to:

[0157] Determine the game abnormality description information corresponding to the game abnormality according to the speech content of the speaking player in the target chat record information;

[0158] The occurrence position of the game anomaly in the game is determined according to the game scene in which the speaking player in the target chat record information is located when speaking.

[0159] In an optional implementation, after determining that the classification prediction result of the target chat record information belongs to an abnormal category, the processor 801 is further configured to:

[0160] Performing vector encoding on the target chat record information to obtain a target vector corresponding to the target chat record information;

[0161] In a preset vector database, a vector search is performed on the target vector to obtain a vector search result of the target vector; wherein the preset vector database stores a plurality of known vectors, and the text information matched by the known vectors belongs to non-game abnormality description information;

[0162] When there is a known vector in the vector retrieval result whose similarity with the target vector is greater than or equal to a preset similarity threshold, it is determined that no game abnormality is detected in the target chat record information.

[0163] In an optional implementation, the processor 801 is further configured to:

[0164] When it is determined that no game abnormality is detected in the target chat record information, the target chat record information is stored in the preset vector database.

[0165] Through the electronic device provided by the embodiment of the present application, the chat record information of multiple players in the game is obtained, and the target chat record information containing the target keyword is determined from the multiple chat record information obtained; the target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, the detected game abnormality is determined from the target chat record information. In this way, the present application can actively detect game abnormalities that players have not reported through official channels by actively detecting game abnormalities in the chat record information of players in the game, effectively improving the efficiency and accuracy of abnormality detection in the game.

[0166] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. The computer program is executed when a processor is running, and the processor performs the following steps:

[0167] Acquire chat record information of multiple players in the game, and determine target chat record information containing target keywords from the acquired multiple chat record information; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game;

[0168] The target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

[0169] In an optional implementation, when determining the target chat record information containing the target keyword from the plurality of chat record information obtained, the processor is configured to:

[0170] For each chat record information, extracting a first keyword matching the custom vocabulary from the chat record information according to a pre-loaded custom vocabulary; wherein the custom vocabulary is determined according to game terms corresponding to fixed indicating objects in the game;

[0171] Deleting invalid words from the chat record information to obtain the remaining second keywords; wherein the invalid words include: meaningless function words and participles in the chat record information whose participle importance is lower than a preset importance threshold;

[0172] According to a keyword set consisting of the first keyword and the second keyword, when it is detected that there is at least one target keyword matching a preset keyword list in the keyword set, it is determined that the chat record information belongs to the target chat record information; wherein, the preset keyword list is determined based on the general descriptive vocabulary involved in describing abnormal problems occurring in the game and the game terms.

[0173] In an optional implementation, when performing classification prediction on the target chat record information, the processor is configured to:

[0174] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0175] When the emotion classification prediction result belongs to a negative emotion, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0176] In an optional implementation, when performing classification prediction on the target chat record information, the processor is configured to:

[0177] Input the target chat record information into a pre-trained abnormal text classification model, and output a text classification prediction result of the target chat record information;

[0178] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0179] In an optional implementation, when performing classification prediction on the target chat record information, the processor is configured to:

[0180] Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information;

[0181] When the sentiment classification prediction result belongs to negative sentiment, the target chat record information is input into a pre-trained abnormal text classification model, and the text classification prediction result of the target chat record information is output;

[0182] When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

[0183] In an optional implementation, when determining the detected game abnormality from the target chat record information, the processor is configured to:

[0184] Determine the game abnormality description information corresponding to the game abnormality according to the speech content of the speaking player in the target chat record information;

[0185] The occurrence position of the game anomaly in the game is determined according to the game scene in which the speaking player in the target chat record information is located when speaking.

[0186] In an optional implementation, after determining that the classification prediction result of the target chat record information belongs to an abnormal category, the processor is further configured to:

[0187] Performing vector encoding on the target chat record information to obtain a target vector corresponding to the target chat record information;

[0188] In a preset vector database, a vector search is performed on the target vector to obtain a vector search result of the target vector; wherein the preset vector database stores a plurality of known vectors, and the text information matched by the known vectors belongs to non-game abnormality description information;

[0189] When there is a known vector in the vector retrieval result whose similarity with the target vector is greater than or equal to a preset similarity threshold, it is determined that no game abnormality is detected in the target chat record information.

[0190] In an optional implementation, the processor is further configured to:

[0191] When it is determined that no game abnormality is detected in the target chat record information, the target chat record information is stored in the preset vector database.

[0192] Through the above-mentioned computer-readable storage medium provided by the embodiment of the present application, the chat record information of multiple players in the game is obtained, and the target chat record information containing the target keyword is determined from the multiple chat record information obtained; the target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to the abnormal category, the detected game abnormality is determined from the target chat record information. In this way, the present application can actively detect game abnormalities that players have not reported through official channels by actively detecting game abnormalities in the chat record information of players in the game, effectively improving the efficiency and accuracy of abnormality detection in the game.

[0193] In an embodiment of the present application, the computer-readable storage medium can also execute other machine-readable instructions when run by the processor to execute the anomaly detection method in the game as described in other embodiments. For the specific steps and principles of the anomaly detection method, please refer to the description of the method side embodiment, which will not be repeated here.

[0194] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the system or unit can be electrical, mechanical or other forms.

[0195] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple 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.

[0196] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0197] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0198] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0199] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for detecting anomalies in a game, characterized in that: The anomaly detection method comprises: Acquire chat record information of multiple players in the game, and determine target chat record information containing target keywords from the acquired multiple chat record information; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game; The target chat record information is classified and predicted, and when it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

2. The anomaly detection method according to claim 1, characterized in that: The step of determining target chat record information containing a target keyword from the acquired plurality of chat record information includes: For each chat record information, extracting a first keyword matching the custom vocabulary from the chat record information according to a pre-loaded custom vocabulary; wherein the custom vocabulary is determined according to game terms corresponding to fixed indicating objects in the game; Deleting invalid words from the chat record information to obtain the remaining second keywords; wherein the invalid words include: meaningless function words and participles in the chat record information whose participle importance is lower than a preset importance threshold; According to a keyword set consisting of the first keyword and the second keyword, when it is detected that there is at least one target keyword matching a preset keyword list in the keyword set, it is determined that the chat record information belongs to the target chat record information; wherein, the preset keyword list is determined based on the general descriptive words involved in describing abnormal problems occurring in the game and the game terms.

3. The anomaly detection method according to claim 1, characterized in that: The classifying and predicting the target chat record information includes: Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information; When the emotion classification prediction result belongs to a negative emotion, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

4. The anomaly detection method according to claim 1, characterized in that: The classifying and predicting the target chat record information includes: Input the target chat record information into a pre-trained abnormal text classification model, and output a text classification prediction result of the target chat record information; When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

5. The anomaly detection method according to claim 1, characterized in that: The classifying and predicting the target chat record information includes: Input the target chat record information into a sentiment classification model, and output a sentiment classification prediction result of the target chat record information; When the sentiment classification prediction result belongs to negative sentiment, the target chat record information is input into a pre-trained abnormal text classification model, and the text classification prediction result of the target chat record information is output; When the text classification prediction result belongs to the abnormal text category containing game abnormality description information, it is determined that the classification prediction result of the target chat record information belongs to the abnormal category.

6. The anomaly detection method according to claim 1, characterized in that: Determining the detected game anomaly from the target chat record information includes: Determine the game abnormality description information corresponding to the game abnormality according to the speech content of the speaking player in the target chat record information; The occurrence position of the game anomaly in the game is determined according to the game scene in which the speaking player in the target chat record information is located when speaking.

7. The anomaly detection method according to claim 1, characterized in that: After determining that the classification prediction result of the target chat record information belongs to an abnormal category, the abnormality detection method further includes: Performing vector encoding on the target chat record information to obtain a target vector corresponding to the target chat record information; In a preset vector database, a vector search is performed on the target vector to obtain a vector search result of the target vector; wherein the preset vector database stores a plurality of known vectors, and the text information matched by the known vectors belongs to non-game abnormality description information; When there is a known vector in the vector retrieval result whose similarity with the target vector is greater than or equal to a preset similarity threshold, it is determined that no game abnormality is detected in the target chat record information.

8. The abnormality detection method according to claim 7, characterized in that: The anomaly detection method further includes: When it is determined that no game abnormality is detected in the target chat record information, the target chat record information is stored in the preset vector database.

9. A device for detecting anomalies in a game, characterized in that: The abnormality detection device comprises: A first detection module is used to obtain chat record information of multiple players in the game, and determine target chat record information containing target keywords from the multiple chat record information obtained; wherein the target keyword represents a key word involved in describing an abnormal problem occurring in the game; The second detection module is used to classify and predict the target chat record information. When it is determined that the classification prediction result of the target chat record information belongs to an abnormal category, the detected game abnormality is determined from the target chat record information; wherein the abnormal category represents the text category to which the game abnormality description information belongs.

10. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the abnormality detection method as described in any one of claims 1 to 8 are performed.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the abnormality detection method according to any one of claims 1 to 8 are executed.

Images