Information detection method, information detection device, electronic device and readable storage medium

Abstract

The present application discloses an information detection method, an information detection device, an electronic device and a readable storage medium, which belong to the field of communication technology. The information detection method includes: searching for a set of associated terms in a preset database according to an event message, wherein the terms in the set of associated terms are associated with the semantics of the event message; determining a first semantic vector of the event message according to the semantic vector of each term in the set of associated terms; obtaining a second semantic vector of a keyword in the event message; determining the similarity between the event message and a target message in a preset database according to the first semantic vector and the second semantic vector, wherein the target message is associated with the set of associated terms.

Description

Technical Field

[0001] The present application belongs to the field of communication technology, and specifically relates to an information detection method, an information detection device, an electronic device and a readable storage medium. Background Art

[0002] In the related art, information flow services need to process a large number of message services, among which message services are filled with a large number of false messages, so it is necessary to identify and detect these false messages.

[0003] Currently, false information is generally identified and detected in massive information flow services through manual review or user reporting.

[0004] However, manual review or user reporting requires a lot of manpower and has low detection efficiency. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide an information detection method, an information detection device, an electronic device and a readable storage medium, which realizes the identification of false messages without manual work, and further improves the recognition efficiency while ensuring the recognition accuracy.

[0006] In a first aspect, an embodiment of the present application provides an information detection method, comprising: searching for a set of associated terms in a preset database based on an event message, wherein the terms in the set of associated terms are semantically associated with the event message; determining a first semantic vector of the event message based on a semantic vector of each term in the set of associated terms; obtaining a second semantic vector of a keyword in the event message; and determining the similarity between the event message and a target message in the set of associated terms based on the first semantic vector and the second semantic vector, wherein the target message is associated with the set of associated terms.

[0007] In a second aspect, an embodiment of the present application provides an information detection device, comprising: a search module, used to search for a set of associated terms in a preset database based on an event message, wherein the terms in the set of associated terms are semantically associated with the event message; a first determination module, used to determine a first semantic vector of the event message based on the semantic vector of each term in the set of associated terms; an acquisition module, used to acquire a second semantic vector of a keyword in the event message; and a second determination module, used to determine the similarity between the event message and a target message in the set of associated terms based on the first semantic vector and the second semantic vector, wherein the target message is associated with the set of associated terms.

[0008] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the information detection method of the first aspect.

[0009] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the information detection method of the first aspect are implemented.

[0010] In a fifth aspect, an embodiment of the present application provides a chip, which includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the information detection method of the first aspect.

[0011] In a sixth aspect, an embodiment of the present application provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the information detection method of the first aspect.

[0012] In the embodiment of the present application, by finding the associated word set corresponding to the event message in the preset database, and generating the first semantic vector related to the event message based on the associated word set, the deep semantics of the event message can be better understood through the first semantic vector. By comparing the first semantic vector with the second semantic vector generated according to the keywords in the event message, it is possible to accurately determine whether the event message is consistent with the corresponding target message in the preset database, thereby improving the accuracy of automatic recognition of event credibility, realizing the identification of false messages without manual work, and further improving the recognition efficiency while ensuring the recognition accuracy.

[0013] The first semantic vector in the embodiment of the present application is a semantic vector generated based on a synonym library. The first semantic vector can better understand the deep semantics behind the keywords. Therefore, applying it to false message identification can better and more accurately map various similar event keywords to a unified and standard event standard name, thereby better identifying false messages. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram showing a flow chart of an information detection method provided in an embodiment of the present application is shown;

[0015] Figure 2 A schematic diagram showing a first semantic vector calculation process provided by an embodiment of the present application is shown;

[0016] Figure 3 The structure block diagram of the information detection device provided by the embodiment of the present application is shown;

[0017] Figure 4 A structural block diagram of an electronic device provided in an embodiment of the present application is shown;

[0018] Figure 5 A schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0020] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0021] The following is combined with Figures 1 to 5 , the information detection method, information detection device, electronic device and readable storage medium provided in the embodiments of the present application are described in detail through specific embodiments and their application scenarios.

[0022] An information detection method is provided in an embodiment of the present application. Figure 1 A schematic diagram of the process of the information detection method provided in the embodiment of the present application is shown. Figure 1 As shown, the information detection method includes:

[0023] Step 102, searching for a set of related terms in a preset database according to the event message;

[0024] Wherein, the entries in the associated entry set are associated with the event message;

[0025] The preset database includes at least one of the following: a synonym of a target term, a target message, and a source of the target message.

[0026] It is worth noting that before the event message is authenticated, a preset database needs to be established. Among them, the preset database is established based on the factual message. Moreover, the preset database includes a synonym library of event words constructed based on a synonym dictionary. The vocabulary in the synonym library includes the standard name of the event and the synonyms corresponding to the standard name. For example, if the standard name is "death", the synonyms include "pass away, pass away, drive a crane to the west, and leave the world..." When establishing the synonym library, the dictionary stores the standard name of each standard name corresponding to its antonym, such as "death" corresponding to "alive". The preset database also includes a factual event library, which can be specifically established based on the encyclopedia data in the network to construct the factual event library of each person, for example: (someone, married, date of birth xx, xx, xx). The preset database also includes a white list of message sources, and the website address, author, reporter and other message sources of the authoritative message source are marked as white lists. In the process of detecting event messages, if the event message is released by a message source in the white list, it is determined that its credibility is high. The preset database also includes a person name library, which collects the aliases, nicknames, titles and other names of each person, and establishes a mapping relationship between the aliases, nicknames, titles and other names and the person's name.

[0027] It is understandable that the preset database can be stored in the cloud or configured locally on the electronic device. When the electronic device needs to authenticate the event message, it can authenticate the event message by accessing the preset database in the cloud or calling the local preset database.

[0028] Specifically, after receiving the event message, the electronic device searches for a set of related terms in a preset database according to the semantics of the event message. The set of related terms is a real event in the preset database and synonyms corresponding to the event name in the real event.

[0029] Step 104, determining a first semantic vector of the event message according to the semantic vector of each term in the associated term set;

[0030] Specifically, the first semantic vector is calculated based on the semantic vector of each associated term in the associated term set, and the representation method of the first semantic vector is the "self-reducing attention" event semantic representation method. The specific calculation method is that before the attention calculation, the semantic vectors of all synonyms in the associated vocabulary are subtracted from the vocabulary vector of the event name in the associated vocabulary, only including the part that is different from the event name, and then the semantic vector of the event name is subjected to the attention operation with the semantic vector of the retained synonym, so as to obtain the first semantic vector associated with the event message.

[0031] It is worth noting that the first semantic vector of the event message is different from the semantic vector of the event name in the associated entry. For example, the semantics of the event "death" and the semantics of the keyword "death" are not exactly the same, and the meaning represented by the first semantic vector of the event message is relatively abstract.

[0032] Step 106, obtaining a second semantic vector of the keyword in the event message;

[0033] Specifically, keywords in the event message are extracted according to the semantic meaning of the event message, and the second semantic vector is the semantic vector of the keyword.

[0034] Step 108: Determine the similarity between the event message and a target message in a preset database based on the first semantic vector and the second semantic vector, and associate the target message with the associated term set.

[0035] Among them, the target message is a "real event" stored in a preset database, and multiple entries in the associated term set all correspond to the target message. For example, if the target message is that AA was born on XX year XX month XX day, then the associated term set includes "birth", "birthday", "birth" and so on. Specifically, the target message in the preset database can be found according to the semantics of the event message, and the corresponding associated term set can be found according to the semantics of the target message. In the process of judging the authenticity of the event message, the similarity between the event message and the target message can be obtained by comparing the first semantic vector and the second semantic vector. In the embodiment of the present application, the first semantic vector is a semantic vector determined by the associated term set found in the preset database, and the second semantic vector is a semantic vector of a keyword in the event message. By comparing the similarity of the first semantic vector and the second semantic vector, it is determined whether the above two semantic vectors are similar, and it is possible to judge whether the keywords in the event message belong to the synonym library in the preset database. It can be understood that the first semantic vector can reflect the event meaning corresponding to the event message in the preset database, that is, the real event meaning related to the message event. The second semantic vector can reflect the event meaning actually expressed by the event message. The similarity obtained by comparing the first semantic vector with the second semantic vector can reflect the similarity between the event message and the target message in the preset database. If the event message and the target message are detected to have a high similarity, the credibility of the event is determined to be high; if the event message and the target message are detected to have a low similarity, the credibility of the event is determined to be low.

[0036] Specifically, before searching for the associated term set in the preset database according to the event message, it also includes: extracting the name and the corresponding event in the event message through the semantics of the event message, and mapping the name to the standard name in the preset database, and the standard term of the target message in the preset database can be queried through the standard name. According to the standard term of the target message, the associated term set is queried in the synonym library of the preset database, and the associated term includes the standard term of the target message and the synonym term set of the target message. By inputting the target message and the synonym term set into BERT (Pre-training of Deep Bidirectional Transformers for Language Understanding, pre-training model), the semantic vector of each term can be obtained, and by calculating the semantic vector of each term, the first semantic vector corresponding to the event message can be obtained. Then the keywords extracted from the event message, that is, the name and the corresponding event, are input into BERT, and the second semantic vector can be obtained. By comparing the first semantic vector with the second semantic vector, the similarity between the first semantic vector and the second semantic vector can be determined, so as to determine whether the keyword of the event message exists in the preset database, that is, to determine the credibility of the event message.

[0037] In the embodiment of the present application, by finding the associated word set corresponding to the event message in the preset database, and generating the first semantic vector related to the event message based on the associated word set, the deep semantics of the event message can be better understood through the first semantic vector. By comparing the first semantic vector with the second semantic vector generated according to the keywords in the event message, it is possible to accurately determine whether the event message is consistent with the corresponding target message in the preset database, thereby improving the accuracy of automatic recognition of event credibility, realizing the identification of false messages without manual work, and further improving the recognition efficiency while ensuring the recognition accuracy.

[0038] It is worth noting that the first semantic vector in the present application is a semantic vector generated based on a synonym library. The first semantic vector can better understand the deep semantics behind the keywords. Therefore, applying it to false news identification can better and more accurately map various similar event keywords to a unified and standard event standard name, thereby better identifying false news.

[0039] In some embodiments of the present application, searching for a set of associated terms in a preset database based on an event message includes: searching for a first term in the preset database based on the event message, the first term being associated with the event message; and searching for N second terms in the preset database based on the first term, the N second terms including N synonyms of the first term, where N is a positive integer greater than 1.

[0040] In an embodiment of the present application, character events in an event message are extracted, and the character events include names of people and events corresponding to the names. A first entry can be found in a preset database based on the name of the character, and the first entry includes a standard name corresponding to the name and a real event corresponding to the name. Multiple synonyms of the first entry are queried in the synonym library of the preset database, and the synonyms of the first entry are used as second entries. The associated entry set includes the first entry and multiple second entries, so that the associated entry set includes not only entries of real events associated with the event message, but also entries with similar meanings to the real events, so that the first semantic vector obtained according to the associated entry set can represent the deep semantics of the event, thereby improving the subsequent recognition accuracy of the credibility of the message event.

[0041] The embodiment of the present application configures both the standard terms and synonymous terms of the real event associated with the event message in a set of associated terms, so that the first semantic vector obtained according to the set of associated terms can represent the deep semantics of the event, thereby improving the subsequent recognition accuracy of the credibility of the message event.

[0042] In some embodiments of the present application, the first semantic vector of the event message is determined based on the semantic vector of each term in the set of associated terms, including: determining the third semantic vector of the first term, and N fourth semantic vectors of N second terms; subtracting the N fourth semantic vectors from the third semantic vector respectively to obtain N vector differences; and performing an attention operation on the N vector differences and the third semantic vector to obtain the first semantic vector.

[0043] In the embodiment of the present application, the first term and the plurality of second terms are input into BERT, and a third semantic vector corresponding to the first term and a plurality of fourth semantic vectors corresponding to the plurality of second terms can be obtained. The first semantic vector can be obtained by performing "self-reducing attention calculation" on the third semantic vector and the fourth semantic vector.

[0044] Specifically, the first semantic vector can be obtained by calculating the third semantic vector and the plurality of fourth semantic vectors through the following formula:

[0045]

[0046] Among them, V d is the first semantic vector, v0 is the third semantic vector, and v i is the fourth semantic vector of the ith order, v j is the jth fourth semantic vector, and n is the number of fourth semantic vectors.

[0047] The above formula is to first subtract the third semantic vector from all fourth semantic vectors before performing attention calculation, retain only the part that is different from the third semantic vector, and then perform attention operation on the third semantic vector and the retained fourth semantic vector, and finally obtain the first semantic vector of the entire event. The first semantic vector and the third semantic vector are different semantic vectors, wherein the first semantic vector includes the semantics representing the meaning of the first term, so the first semantic vector is more abstract than the third semantic vector and can better represent the deep semantics of the event.

[0048] Figure 2 A schematic diagram of a first semantic vector calculation process provided by an embodiment of the present application is shown, such as Figure 2 As shown, the first term 202 is "death", and the second term 204 includes: "passed away", "passed away", "passed away", and "passed away from the world". The first term and multiple second terms are input into BERT, and a third semantic vector 206 and a fourth semantic vector 208 can be obtained. The part 210 of the fourth semantic vector that is different from the third semantic vector can be obtained by self-reduction calculation. Then, the difference part 210 and the third semantic vector 206 are subjected to attention operation, and finally the first semantic vector 212 of the whole event is obtained.

[0049] It is understandable that the semantics of each synonym of the first term "death" is represented by its own context. Each synonym may have the same context, but due to grammar and context, each synonym must have more different contexts. Therefore, before paying attention, the third semantic vector is subtracted from the fourth semantic vector so that the final first semantic vector can pay more attention to the parts that are different from the standard name. This is equivalent to the final first semantic vector having more different contexts than the semantic vector corresponding to the specific keyword, and the semantic information represented by the first semantic vector is also richer.

[0050] The embodiment of the present application can obtain a first semantic vector representing richer semantic information by performing self-reducing attention calculation on the first term and multiple second terms. Since the first semantic vector can have more different contexts and the semantic information identified by the first semantic vector is richer, the accuracy of subsequent authentication of event messages is further improved.

[0051] In some embodiments of the present application, the similarity between an event message and a target message in a preset database is determined based on a first semantic vector and a second semantic vector, including: determining an approximate value of the first semantic vector and the second semantic vector; when the approximate value is greater than or equal to a preset threshold, determining that the event message and the target message are similar messages; when the approximate value is less than the preset threshold, determining that the event message and the target message are contradictory messages.

[0052] In the embodiment of the present application, when determining whether the event message matches the target message in the preset database, the approximate value of the two semantic vectors is calculated based on the first semantic vector and the second semantic vector. By comparing the approximate value with the preset threshold, the similarity between the first semantic vector and the second semantic vector, that is, the similarity between the target message and the event message, can be determined.

[0053] Specifically, the Euclidean distance between the first semantic vector and the second semantic vector is calculated by cosine similarity calculation, and the Euclidean distance between the two semantic vectors is used as the approximate value of the first semantic vector and the second semantic vector. When it is detected that the approximate value is greater than or equal to a preset threshold, it is determined that the similarity between the first semantic vector and the second semantic vector is high, and the event message and the target message in the preset database are determined to be similar messages. When it is detected that the approximate value is less than the preset threshold, it is determined that the similarity between the first semantic vector and the second semantic vector is low, and the event message and the target message in the preset database are determined to be contradictory messages.

[0054] The embodiment of the present application can determine the similarity between a target message corresponding to the first semantic vector and an event message corresponding to the second semantic vector by calculating the approximate value of the first semantic vector and the second semantic vector and comparing the approximate value with a preset threshold. Since the first semantic vector is the semantic vector of a set of associated terms of the event message in a preset database, the accuracy of the similarity judgment between the event message and the target message is improved.

[0055] In some embodiments, the first semantic vector includes a first sub-semantic vector and a second sub-semantic vector, wherein the first sub-semantic vector is a semantic vector of a "real event" corresponding to an event message, and the second sub-semantic vector is a semantic vector of an "antonymous event" corresponding to the event message.

[0056] In the embodiment of the present application, when it is detected that the message source is not in the whitelist in the preset database, the message event is detected according to the first sub-semantic vector and the second sub-semantic vector.

[0057] Specifically, the first approximation and the second approximation of the first sub-semantic vector and the second sub-semantic vector to the second vector are calculated respectively, wherein the first approximation can represent the similarity between the message event and the real event, and the second approximation can represent the similarity between the message event and the antonym event.

[0058] When the first approximation is detected to be greater than a preset threshold, the event message is determined to be consistent with the real event, that is, the event message is a real message. When the second approximation is detected to be greater than a preset threshold, the event message is determined to be consistent with the antonym event, that is, the event message is a false message. When the first approximation and the second approximation are detected to be less than the preset threshold, the event message is determined to be irrelevant to the real event.

[0059] In some embodiments of the present application, after determining the similarity between the event message and the target message in a preset database based on the first semantic vector and the second semantic vector, it also includes: obtaining the message source of the event message; when the message source is stored in the preset database, determining that the event message is in a credible state; when the message source is not stored in the preset database, and the event message and the target message are similar messages, determining that the event message is in a credible state; when the message source is not stored in the preset database, and the event message and the target message are contradictory messages, determining that the event message is in an untrustworthy state.

[0060] In an embodiment of the present application, after determining the similarity between the event message and the target message, the source of the event message is comprehensively considered to determine the credibility of the event message. After obtaining the source of the event message, the preset database is queried to see whether the source of the message is stored in the preset database. If it is detected that the source of the message is in the whitelist in the preset database, the event message is directly determined to be a true message. If it is detected that the source of the message is not in the whitelist in the preset database, the credibility of the event message is determined based on whether the event message is a similar message to the target message. Specifically, if the source of the event message is not stored in the preset database, if the event message and the target message are similar messages, the event message is determined to be in a credible state. If the event message and the target message are contradictory, the event message is determined to be in an untrustworthy state.

[0061] The embodiment of the present application judges the credibility of the event message by comprehensively considering the message source of the event message and the approximate values ​​of the first semantic vector and the second semantic vector, thereby further improving the accuracy of judging the credibility of the event message.

[0062] In some embodiments, the credibility of the event message can be obtained by weighting the message source and the approximate value of the semantic vector.

[0063] In the embodiment of the present application, in the process of detecting the credibility of the event message, the source of the event message is judged first. When the source of the message is in the preset database, the event message is directly judged to be a real message, which reduces the judgment steps of the event message and improves the efficiency of event message authentication. When the source of the message is in the preset database, the message event is authenticated according to the approximate value of the semantic vector, which can improve the accuracy of authentication.

[0064] In some embodiments of the present application, after determining the similarity between the event message and the target message in the preset database based on the first semantic vector and the second semantic vector, it also includes: when the event message and the target message in the preset database are similar messages, updating the preset database based on the event message.

[0065] In the embodiment of the present application, when it is detected that the event message is highly similar to the target message in the database, the event message is determined to be a true message, and the preset database is updated according to the event message, thereby ensuring the real-time nature of the data in the preset database.

[0066] Specifically, when it is detected that the approximation value of the semantic vector is low, but the source of the event message is stored in a preset database, the event message is determined to be a real event, and the preset database is updated through the event message, further improving the accuracy of the information in the preset database.

[0067] The information detection method provided in the embodiment of the present application can be executed by an information detection device. In the embodiment of the present application, the method of unlocking an electronic device performed by an information detection device is taken as an example to illustrate the device for unlocking an electronic device provided in the embodiment of the present application.

[0068] In some embodiments of the present application, an information detection device is provided. Figure 3 The structure diagram of the information detection device provided by the embodiment of the present application is shown as follows: Figure 3 As shown, the information detection device 300 includes:

[0069] A search module 302 is used to search a set of associated terms in a preset database according to the event message, wherein the terms in the set of associated terms are associated with the event message;

[0070] The preset database includes at least one of the following: a synonym of a target term, a target message, and a source of the target message.

[0071] It is worth noting that before the event message is authenticated, a preset database needs to be established. Among them, the preset database is established based on the factual message. Moreover, the preset database includes a synonym library of event words constructed based on a synonym dictionary. The vocabulary in the synonym library includes the standard name of the event and the synonyms corresponding to the standard name. For example, if the standard name is "death", the synonyms include "pass away, pass away, drive a crane to the west, and leave the world..." When establishing the synonym library, the dictionary stores the standard name of each standard name corresponding to its antonym, such as "death" corresponding to "alive". The preset database also includes a factual event library, which can be specifically established based on the encyclopedia data in the network to construct the factual event library of each person, for example: (someone, married, date of birth xx, xx, xx). The preset database also includes a white list of message sources, and the website address, author, reporter and other message sources of the authoritative message source are marked as white lists. In the process of detecting event messages, if the event message is released by a message source in the white list, it is determined that its credibility is high. The preset database also includes a person name library, which collects the aliases, nicknames, titles and other names of each person, and establishes a mapping relationship between the aliases, nicknames, titles and other names and the person's name.

[0072] It is understandable that the preset database can be stored in the cloud or configured locally on the electronic device. When the electronic device needs to authenticate the event message, it can authenticate the event message by accessing the preset database in the cloud or calling the local preset database.

[0073] Specifically, after receiving the event message, the electronic device searches for a set of related terms in a preset database according to the semantics of the event message. The set of related terms is a real event in the preset database and synonyms corresponding to the event name in the real event.

[0074] A first determination module 304, configured to determine a first semantic vector of the event message according to the semantic vector of each term in the associated term set;

[0075] Specifically, the first semantic vector is calculated based on the semantic vector of each associated term in the associated term set, and the representation method of the first semantic vector is the "self-reducing attention" event semantic representation method. The specific calculation method is that before the attention calculation, the semantic vectors of all synonyms in the associated vocabulary are subtracted from the vocabulary vector of the event name in the associated vocabulary, only including the part that is different from the event name, and then the semantic vector of the event name is subjected to the attention operation with the semantic vector of the retained synonym, so as to obtain the first semantic vector associated with the event message.

[0076] It is worth noting that the first semantic vector of the event message is different from the semantic vector of the event name in the associated entry. For example, the semantics of the event "death" and the semantics of the keyword "death" are not exactly the same, and the meaning represented by the first semantic vector of the event message is relatively abstract.

[0077] An acquisition module 306 is used to acquire a second semantic vector of a keyword in the event message;

[0078] Specifically, keywords in the event message are extracted according to the semantic meaning of the event message, and the second semantic vector is the semantic vector of the keyword.

[0079] The second determination module 308 is used to determine the similarity between the event message and a target message in a preset database according to the first semantic vector and the second semantic vector, wherein the target message is associated with the associated term set.

[0080] Among them, the target message is a "real event" stored in a preset database, and multiple entries in the associated entry set all correspond to the target message. For example, if the target message is that AA was born on XX / XX / XX, then the associated entry set includes "birth", "birthday", "death", etc. Specifically, the target message in the preset database can be found according to the semantics of the event message, and the corresponding associated entry set can be found according to the semantics of the target message. In the process of judging the authenticity of the event message, the similarity between the event message and the target message can be obtained by comparing the first semantic vector and the second semantic vector.

[0081] In an embodiment of the present application, the first semantic vector is a semantic vector determined by a set of associated terms found in a preset database, and the second semantic vector is a semantic vector of a keyword in an event message. By comparing the similarity of the first semantic vector and the second semantic vector, it is determined whether the two semantic vectors are similar, and it is possible to determine whether the keywords in the event message belong to the synonym library in the preset database. It is understandable that the first semantic vector can reflect the event meaning corresponding to the event message in the preset database, that is, the real event meaning related to the message event. The second semantic vector can reflect the event meaning actually expressed by the event message. The similarity obtained by comparing the first semantic vector with the second semantic vector can reflect the similarity between the event message and the target message in the preset database. When it is detected that the similarity between the event message and the target message is high, it is determined that the credibility of the event is high, and when it is detected that the similarity between the event message and the target message is low, it is determined that the credibility of the event is low.

[0082] Specifically, before searching for the associated term set in the preset database according to the event message, it also includes: extracting the name and the corresponding event in the event message through the semantics of the event message, and mapping the name to the standard name in the preset database, and the standard term of the target message in the preset database can be queried through the standard name. According to the standard term of the target message, the associated term set is queried in the synonym library of the preset database, and the associated term includes the standard term of the target message and the synonym term set of the target message. By inputting the target message and the synonym term set into BERT (Pre-training of Deep Bidirectional Transformers for Language Understanding, pre-training model), the semantic vector of each term can be obtained, and by calculating the semantic vector of each term, the first semantic vector corresponding to the event message can be obtained. Then the keywords extracted from the event message, that is, the name and the corresponding event, are input into BERT, and the second semantic vector can be obtained. By comparing the first semantic vector with the second semantic vector, the similarity between the first semantic vector and the second semantic vector can be determined, so as to determine whether the keyword of the event message exists in the preset database, that is, to determine the credibility of the event message.

[0083] In the embodiment of the present application, by finding the associated word set corresponding to the event message in the preset database, and generating the first semantic vector related to the event message based on the associated word set, the deep semantics of the event message can be better understood through the first semantic vector. By comparing the first semantic vector with the second semantic vector generated according to the keywords in the event message, it is possible to accurately determine whether the event message is consistent with the corresponding target message in the preset database, thereby improving the accuracy of automatic recognition of event credibility, realizing the identification of false messages without manual work, and further improving the recognition efficiency while ensuring the recognition accuracy.

[0084] It is worth noting that the first semantic vector in the present application is a semantic vector generated based on a synonym library. The first semantic vector can better understand the deep semantics behind the keywords. Therefore, applying it to false news identification can better and more accurately map various similar event keywords to a unified and standard event standard name, thereby better identifying false news.

[0085] In some embodiments of the present application, the search module 302 is further used to search for a first entry in a preset database according to the event message, the first entry being associated with the event message;

[0086] The search module 302 is further configured to search for N second terms in a preset database, where the N second terms include N synonyms of the first term, and N is a positive integer greater than 1.

[0087] In an embodiment of the present application, character events in an event message are extracted, and the character events include names of people and events corresponding to the names. A first entry can be found in a preset database based on the name of the character, and the first entry includes a standard name corresponding to the name and a real event corresponding to the name. Multiple synonyms of the first entry are queried in the synonym library of the preset database, and the synonyms of the first entry are used as second entries. The associated entry set includes the first entry and multiple second entries, so that the associated entry set includes not only entries of real events associated with the event message, but also entries with similar meanings to the real events, so that the first semantic vector obtained according to the associated entry set can represent the deep semantics of the event, thereby improving the subsequent recognition accuracy of the credibility of the message event.

[0088] The embodiment of the present application configures both the standard terms and synonymous terms of the real event associated with the event message in a set of associated terms, so that the first semantic vector obtained according to the set of associated terms can represent the deep semantics of the event, thereby improving the subsequent recognition accuracy of the credibility of the message event.

[0089] In some embodiments of the present application, the first determination module 304 is further configured to determine a third semantic vector of the first term and N fourth semantic vectors of the N second terms;

[0090] The information detection device further comprises:

[0091] A calculation module, used for performing subtraction calculations on the N fourth semantic vectors from the third semantic vector respectively to obtain N vector differences;

[0092] A processing module is used to perform an attention operation on the N vector differences and the third semantic vector to obtain a first semantic vector.

[0093] In the embodiment of the present application, the first term and the plurality of second terms are input into BERT, and a third semantic vector corresponding to the first term and a plurality of fourth semantic vectors corresponding to the plurality of second terms can be obtained. The first semantic vector can be obtained by performing "self-reducing attention calculation" on the third semantic vector and the fourth semantic vector.

[0094] Specifically, the first semantic vector can be obtained by calculating the third semantic vector and the plurality of fourth semantic vectors through the following formula:

[0095]

[0096] Among them, V d is the first semantic vector, v0 is the third semantic vector, and v i is the fourth semantic vector of the ith order, v jis the jth fourth semantic vector, and n is the number of fourth semantic vectors.

[0097] The above formula is to first subtract the third semantic vector from all fourth semantic vectors before performing attention calculation, retain only the part that is different from the third semantic vector, and then perform attention operation on the third semantic vector and the retained fourth semantic vector, and finally obtain the first semantic vector of the entire event. The first semantic vector and the third semantic vector are different semantic vectors, wherein the first semantic vector includes the semantics representing the meaning of the first term, so the first semantic vector is more abstract than the third semantic vector and can better represent the deep semantics of the event.

[0098] The embodiment of the present application can obtain a first semantic vector representing richer semantic information by performing self-reducing attention calculation on the first term and multiple second terms. Since the first semantic vector can have more different contexts and the semantic information identified by the first semantic vector is richer, the accuracy of subsequent authentication of event messages is further improved.

[0099] In some embodiments of the present application, the second determination module 308 is further configured to determine approximate values ​​of the first semantic vector and the second semantic vector;

[0100] The second determination module 308 is further configured to determine that the event message and the target message are similar messages when the approximation value is greater than or equal to a preset threshold;

[0101] The second determination module 308 is further configured to determine that the event message and the target message are contradictory messages when the approximation value is less than a preset threshold.

[0102] In the embodiment of the present application, when determining whether the event message matches the target message in the preset database, the approximate value of the two semantic vectors is calculated based on the first semantic vector and the second semantic vector. By comparing the approximate value with the preset threshold, the similarity between the first semantic vector and the second semantic vector, that is, the similarity between the target message and the event message, can be determined.

[0103] Specifically, the Euclidean distance between the first semantic vector and the second semantic vector is calculated by cosine similarity calculation, and the Euclidean distance between the two semantic vectors is used as the approximate value of the first semantic vector and the second semantic vector. When it is detected that the approximate value is greater than or equal to a preset threshold, it is determined that the similarity between the first semantic vector and the second semantic vector is high, and the event message and the target message in the preset database are determined to be similar messages. When it is detected that the approximate value is less than the preset threshold, it is determined that the similarity between the first semantic vector and the second semantic vector is low, and the event message and the target message in the preset database are determined to be contradictory messages.

[0104] The embodiment of the present application can determine the similarity between a target message corresponding to the first semantic vector and an event message corresponding to the second semantic vector by calculating the approximate value of the first semantic vector and the second semantic vector and comparing the approximate value with a preset threshold. Since the first semantic vector is the semantic vector of a set of associated terms of the event message in a preset database, the accuracy of the similarity judgment between the event message and the target message is improved.

[0105] In some embodiments, the first semantic vector includes a first sub-semantic vector and a second sub-semantic vector, wherein the first sub-semantic vector is a semantic vector of a "real event" corresponding to an event message, and the second sub-semantic vector is a semantic vector of an "antonymous event" corresponding to the event message.

[0106] In the embodiment of the present application, when it is detected that the message source is not in the whitelist in the preset database, the message event is detected according to the first sub-semantic vector and the second sub-semantic vector.

[0107] Specifically, the first approximation and the second approximation of the first sub-semantic vector and the second sub-semantic vector to the second vector are calculated respectively, wherein the first approximation can represent the similarity between the message event and the real event, and the second approximation can represent the similarity between the message event and the antonym event.

[0108] When the first approximation is detected to be greater than a preset threshold, the event message is determined to be consistent with the real event, that is, the event message is a real message. When the second approximation is detected to be greater than a preset threshold, the event message is determined to be consistent with the antonym event, that is, the event message is a false message. When the first approximation and the second approximation are detected to be less than the preset threshold, the event message is determined to be irrelevant to the real event.

[0109] In some embodiments of the present application, the acquisition module 306 is further used to acquire the message source of the event message;

[0110] The information detection device also includes:

[0111] A third determination module is used to determine that the event message is in a credible state when the message source is stored in a preset database;

[0112] The third determination module is further used to determine that the event message is in a credible state when there is no message source stored in the preset database and the event message and the target message are similar messages;

[0113] The third determination module is further used to determine that the event message is in an untrustworthy state when the message source is not stored in the preset database and the event message and the target message are contradictory messages.

[0114] In an embodiment of the present application, after determining the similarity between the event message and the target message, the source of the event message is comprehensively considered to determine the credibility of the event message. After obtaining the source of the event message, the source of the message is queried in the preset database to see if it is stored in the preset database. When it is detected that the source of the message is in the whitelist in the preset database, the event message is directly determined to be a true message. When it is detected that the source of the message is not in the whitelist in the preset database, the credibility of the event message is judged based on whether the event message is a similar message to the target message. Specifically, in the case where the source of the event message is not stored in the preset database, if the event message and the target message are similar messages, the event message is determined to be in a credible state, and if the event message and the target message are contradictory details, the event message is determined to be in an untrustworthy state. The embodiment of the present application further improves the accuracy of the credibility judgment of the event message by comprehensively considering the source of the event message, as well as the approximate value of the first semantic vector and the second semantic vector, to judge the credibility of the event message.

[0115] In some embodiments, the credibility of the event message can be obtained by weighting the message source and the approximate value of the semantic vector.

[0116] In the embodiment of the present application, in the process of detecting the credibility of the event message, the source of the event message is judged first. When the source of the message is in the preset database, the event message is directly judged to be a real message, which reduces the judgment steps of the event message and improves the efficiency of event message authentication. When the source of the message is in the preset database, the message event is authenticated according to the approximate value of the semantic vector, which can improve the accuracy of authentication.

[0117] In some embodiments of the present application, the information detection device further includes:

[0118] The updating module is used to update the preset database according to the event message when the event message and the target message in the preset database are similar messages.

[0119] In the embodiment of the present application, when it is detected that the event message is highly similar to the target message in the database, the event message is determined to be a true message, and the preset database is updated according to the event message, thereby ensuring the real-time nature of the data in the preset database.

[0120] Specifically, when it is detected that the approximation value of the semantic vector is low, but the source of the event message is stored in a preset database, the event message is determined to be a real event, and the preset database is updated through the event message, further improving the accuracy of the information in the preset database.

[0121] The information detection device in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices other than the terminal. Exemplary, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a palmtop computer, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), augmented reality (augmented reality, AR) / virtual reality (virtualreality, VR) equipment, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personalcomputer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personalcomputer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not specifically limited.

[0122] The information detection device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0123] The information detection device provided in the embodiment of the present application can implement each process implemented in the above method embodiment, and will not be described again here to avoid repetition.

[0124] Optionally, an embodiment of the present application further provides an electronic device, Figure 4 A structural block diagram of an electronic device according to an embodiment of the present application is shown. Figure 4 As shown, the electronic device 400 includes a processor 402 and a memory 404. The memory 404 stores programs or instructions that can be executed on the processor 402. When the program or instructions are executed by the processor 402, the various steps of the above-mentioned method embodiment are implemented and the same technical effect can be achieved. To avoid repetition, they are not repeated here.

[0125] It should be noted that the electronic devices in the embodiments of the present application include the above-mentioned mobile electronic devices and non-mobile electronic devices.

[0126] Figure 5 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of the present application.

[0127] The electronic device 500 includes but is not limited to: a radio frequency unit 501, a network module 502, an audio output unit 503, an input unit 504, a sensor 505, a display unit 506, a user input unit 507, an interface unit 508, a memory 509, and a processor 510.

[0128] Those skilled in the art will appreciate that the electronic device 500 may also include a power source (such as a battery) for supplying power to various components, and the power source may be logically connected to the processor 510 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Figure 5 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0129] The processor 510 is used to search for a set of associated terms in a preset database according to the event message, wherein the terms in the set of associated terms are associated with the event message;

[0130] The preset database includes at least one of the following: a synonym of a target term, a target message, and a source of the target message.

[0131] Processor 510, configured to determine a first semantic vector of the event message according to the semantic vector of each term in the associated term set;

[0132] Processor 510, configured to obtain a second semantic vector of a keyword in the event message;

[0133] The processor 510 is used to determine the similarity between the event message and a target message in a preset database according to the first semantic vector and the second semantic vector, and the target message is associated with the associated term set.

[0134] In the embodiment of the present application, by finding the associated word set corresponding to the event message in the preset database, and generating the first semantic vector related to the event message based on the associated word set, the deep semantics of the event message can be better understood through the first semantic vector. By comparing the first semantic vector with the second semantic vector generated according to the keywords in the event message, it is possible to accurately determine whether the event message is consistent with the corresponding target message in the preset database, thereby improving the accuracy of automatic recognition of event credibility, realizing the identification of false messages without manual work, and further improving the recognition efficiency while ensuring the recognition accuracy.

[0135] It is worth noting that the first semantic vector in the present application is a semantic vector generated based on a synonym library. The first semantic vector can better understand the deep semantics behind the keywords. Therefore, applying it to false news identification can better and more accurately map various similar event keywords to a unified and standard event standard name, thereby better identifying false news.

[0136] Further, the processor 510 is configured to search a first term in a preset database according to the event message, the first term being associated with the event message;

[0137] The processor 510 is configured to search for N second terms in a preset database, where the N second terms include N synonyms of the first term.

[0138] The embodiment of the present application configures both the standard terms and synonymous terms of the real event associated with the event message in a set of associated terms, so that the first semantic vector obtained according to the set of associated terms can represent the deep semantics of the event, thereby improving the subsequent recognition accuracy of the credibility of the message event.

[0139] Further, the processor 510 is configured to determine a third semantic vector of the first term and N fourth semantic vectors of the N second terms;

[0140] The processor 510 is configured to perform subtraction calculations on the N fourth semantic vectors from the third semantic vector respectively to obtain N vector differences;

[0141] The processor 510 is used to perform an attention operation on the N vector differences and the third semantic vector to obtain a first semantic vector.

[0142] The embodiment of the present application can obtain a first semantic vector representing richer semantic information by performing self-reducing attention calculation on the first term and multiple second terms. Since the first semantic vector can have more different contexts and the semantic information identified by the first semantic vector is richer, the accuracy of subsequent authentication of event messages is further improved.

[0143] Further, the processor 510 is configured to determine an approximate value of the first semantic vector and the second semantic vector;

[0144] The processor 510 is configured to determine that the event message and the target message are similar messages when the approximation value is greater than or equal to a preset threshold;

[0145] The processor 510 is configured to determine that the event message and the target message are conflicting messages when the approximation value is less than a preset threshold.

[0146] The embodiment of the present application can determine the similarity between a target message corresponding to the first semantic vector and an event message corresponding to the second semantic vector by calculating the approximate value of the first semantic vector and the second semantic vector and comparing the approximate value with a preset threshold. Since the first semantic vector is the semantic vector of a set of associated terms of the event message in a preset database, the accuracy of the similarity judgment between the event message and the target message is improved.

[0147] Further, the processor 510 is used to obtain a message source of the event message;

[0148] Processor 510, for determining that the event message is in a credible state when the message source is stored in a preset database;

[0149] The processor 510 is used to determine that the event message is in a credible state when the message source is not stored in the preset database and the event message and the target message are similar messages;

[0150] The processor 510 is configured to determine that the event message is in an untrustworthy state when no message source is stored in a preset database and the event message and the target message are contradictory messages.

[0151] The embodiment of the present application judges the credibility of the event message by comprehensively considering the message source of the event message and the approximate values ​​of the first semantic vector and the second semantic vector, thereby further improving the accuracy of judging the credibility of the event message.

[0152] Further, the processor 510 is configured to update the preset database according to the event message when the event message and the target message in the preset database are similar messages.

[0153] In the embodiment of the present application, when it is detected that the event message is highly similar to the message in the database, the event message is determined to be a true message, and the preset database is updated according to the event message, thereby ensuring the real-time nature of the data in the preset database.

[0154] It should be understood that in the embodiment of the present application, the input unit 504 may include a graphics processor (GPU) 5041 and a microphone 5042, and the graphics processor 5041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 506 may include a display panel 5061, and the display panel 5061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 507 includes a touch panel 5071 and at least one of other input devices 5072. The touch panel 5071 is also called a touch screen. The touch panel 5071 may include two parts: a touch detection device and a touch controller. Other input devices 5072 may include, but are not limited to, a physical keyboard, function keys (such as a volume control button, a switch button, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0155] The memory 509 can be used to store software programs and various data. The memory 509 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory 509 may include a volatile memory or a non-volatile memory, or the memory 509 may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM). The memory 509 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0156] The processor 510 may include one or more processing units; optionally, the processor 510 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 510.

[0157] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of the above-mentioned information detection method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0158] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0159] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned information detection method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0160] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0161] An embodiment of the present application provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the above-mentioned information detection method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0162] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0163] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0164] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. An information detection method, characterized in that: include: Searching for a set of associated terms in a preset database according to the event message, wherein terms in the set of associated terms are associated with the event message; Determining a first semantic vector of the event message according to the semantic vector of each term in the associated term set; Obtaining a second semantic vector of a keyword in the event message; Determining, based on the first semantic vector and the second semantic vector, the similarity between the event message and a target message in the preset database, wherein the target message is associated with the set of associated terms; Among them, the first semantic vector can reflect the event meaning corresponding to the event message in the preset database, and the second semantic vector can reflect the event meaning actually expressed by the event message. The similarity obtained by comparing the first semantic vector with the second semantic vector can reflect the similarity between the event message and the target message in the preset database.

2. The information detection method according to claim 1, characterized in that: The step of searching for a set of related terms in a preset database according to the event message includes: searching the preset database for a first entry according to the event message, where the first entry is associated with the event message; According to the first term, N second terms are searched in a preset database, wherein the N second terms include N synonyms of the first term, where N is a positive integer greater than 1.

3. The information detection method according to claim 2, characterized in that: The step of determining the first semantic vector of the event message according to the semantic vector of each term in the associated term set includes: Determining a third semantic vector of the first term and N fourth semantic vectors of the N second terms; Subtracting the N fourth semantic vectors from the third semantic vector respectively to obtain N vector differences; Perform an attention operation on the N vector differences and the third semantic vector to obtain the first semantic vector.

4. The information detection method according to any one of claims 1 to 3, characterized in that: Determining the similarity between the event message and the target message in the preset database according to the first semantic vector and the second semantic vector includes: Determining approximate values ​​of the first semantic vector and the second semantic vector; When the approximation value is greater than or equal to a preset threshold, determining that the event message and the target message are similar messages; When the approximate value is smaller than the preset threshold, it is determined that the event message and the target message are conflicting messages.

5. The information detection method according to claim 4, characterized in that: After determining the similarity between the event message and the target message in the preset database according to the first semantic vector and the second semantic vector, the method further includes: Obtaining the message source of the event message: if the message source is stored in the preset database, determining that the event message is in a credible state; In the case that the message source is not stored in the preset database, and the event message and the target message are similar messages, determining that the event message is in the credible state; In the case that the message source is not stored in the preset database, and the event message and the target message are contradictory messages, it is determined that the event message is in an untrustworthy state.

6. An information detection device, characterized in that: include: A search module, used to search a set of associated terms in a preset database according to an event message, wherein terms in the set of associated terms are associated with the event message; A first determination module, configured to determine a first semantic vector of the event message according to the semantic vector of each term in the associated term set; An acquisition module, used for acquiring a second semantic vector of a keyword in the event message; A second determination module, configured to determine, based on the first semantic vector and the second semantic vector, the similarity between the event message and a target message in the preset database, wherein the target message is associated with the associated term set; Among them, the first semantic vector can reflect the event meaning corresponding to the event message in the preset database, and the second semantic vector can reflect the event meaning actually expressed by the event message. The similarity obtained by comparing the first semantic vector with the second semantic vector can reflect the similarity between the event message and the target message in the preset database.

7. The information detection device according to claim 6, characterized in that: The search module is further used to search for a first entry in the preset database according to the event message, wherein the first entry is associated with the event message; The search module is further used to search for N second terms in a preset database, where the N second terms include N synonyms of the first term, and N is a positive integer greater than 1.

8. The information detection device according to claim 7, characterized in that: The first determination module is further used to determine a third semantic vector of the first term and N fourth semantic vectors of the N second terms; The information detection device further includes: A calculation module, used for performing subtraction calculation on the N fourth semantic vectors from the third semantic vector respectively to obtain N vector difference values; A processing module is used to perform an attention operation on the multiple vector differences and the third semantic vector to obtain the first semantic vector.

9. The information detection device according to any one of claims 6 to 8, characterized in that: The second determination module is further used to determine approximate values ​​of the first semantic vector and the second semantic vector; The second determination module is further configured to determine that the event message and the target message are similar messages when the approximate value is greater than or equal to a preset threshold; The second determination module is further configured to determine that the event message and the target message are conflicting messages when the approximate value is less than the preset threshold.

10. The information detection device according to claim 9, characterized in that: The acquisition module is further used to acquire the message source of the event message; The information detection device also includes: A third determination module, configured to determine that the event message is in a credible state if the message source is stored in the preset database; The third determination module is further configured to determine that the event message is in the credible state when the message source is not stored in the preset database and the event message and the target message are similar messages; The third determination module is further configured to determine that the event message is in an untrustworthy state when the message source is not stored in the preset database and the event message and the target message are contradictory messages.

11. An electronic device, characterized in that: The method comprises a processor, a memory and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the information detection method as claimed in any one of claims 1 to 5.

12. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the information detection method as described in any one of claims 1 to 5 are implemented.

Images