Live broadcast interaction information identification method and device, live broadcast system, equipment and medium

Abstract

The invention relates to a live broadcast interaction information identification method and device, a live broadcast system, electronic equipment and a computer readable storage medium. The method comprises the following steps: in a live broadcast process, obtaining interaction information input in a live broadcast room in real time, and extracting a target text from the interaction information; performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on a word root generated by live broadcast sensitive information and a substitute word derived from the word root; detecting whether sensitive information exists in the target text or not according to the matching result of each character; when the sensitive information exists in the target text, intercepting the interaction information in real time and preventing the interaction information from being displayed on the live broadcast room interface; according to the technical scheme, the identification coverage is wider, the efficiency is higher, the variants appearing in the live broadcast process can be intercepted more accurately and comprehensively, and the detection leakage risk identification cost is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of network live broadcasting, and in particular to a method, device, live broadcasting system, electronic device and computer-readable storage medium for identifying live broadcasting interactive information. Background Art

[0002] In the live broadcast platform ecosystem, a large number of users participate in live broadcast interactions, generating a massive amount of interactive information every day, such as user information, public screen speeches, group chats, private chats, dynamic information, broadcast titles, etc. Due to network content supervision requirements, the platform must identify these interactive information to avoid the appearance of sensitive information that violates regulations, so as to ensure that the interactive information content of the live broadcast complies with relevant regulations.

[0003] Commonly used identification technology solutions generally use the AC automaton (Aho-Corasick automaton) algorithm (a multi-mode matching algorithm) to traverse and match keywords in the text content of the live interactive information that needs to be identified to identify sensitive information; however, Internet information changes rapidly, and various new words and homophonic words emerge in an endless stream, which are extremely easy to bypass detection. When variant characters, homophonic characters, etc. appear in the user input interactive information, the identification efficiency is low. Moreover, since keywords cannot be exhaustively listed, the above technical solution is difficult to ensure comprehensive coverage, and the risk of missed detection is high, which also seriously affects the stability of the live broadcast platform operation. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, live broadcast system, electronic device and computer-readable storage medium for identifying live broadcast interaction information to improve the accuracy and efficiency of live broadcast interaction information identification.

[0005] A method for identifying live broadcast interaction information, comprising:

[0006] During the live broadcast process, the interactive information input in the live broadcast room is obtained in real time, and the target text is extracted from the interactive information;

[0007] Performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on the root word generated by the live broadcast sensitive information and the replacement words derived therefrom;

[0008] Detecting whether there is sensitive information in the target text according to the matching results of each character;

[0009] When sensitive information exists in the target text, the interactive information is intercepted in real time and prevented from being displayed on the live broadcast room interface.

[0010] In one embodiment, the method for identifying live broadcast interaction information further includes:

[0011] Detect sensitive information that appears during live broadcasts on live broadcast platforms;

[0012] Generate a root word and a replacement word derived from the root word according to the sensitive information, and generate a normalized keyword word library according to a mapping relationship between the root word and the replacement word;

[0013] A dictionary tree is constructed using the root words and replacement words of the keyword vocabulary.

[0014] In one embodiment, the interactive information input in the live broadcast room is obtained in real time, and the target text is extracted from the interactive information, including:

[0015] When interactive information is input in the live broadcast room, a screenshot image of the live broadcast room interface is obtained in real time through the live broadcast service system; text recognition is performed on the screenshot image based on OCR technology to obtain the target text.

[0016] In one embodiment, generating a root word and a replacement word derived from the root word according to the sensitive information, and generating a normalized keyword word library according to a mapping relationship between the root word and the replacement word includes:

[0017] Performing semantic analysis on the sensitive information and generating a certain number of word roots, and building a normalized word library based on the word roots;

[0018] Deriving a plurality of replacement words according to the root word, and constructing a normalized replacement library based on the mapping relationship between the replacement words and the root word;

[0019] A normalized keyword vocabulary is obtained according to the normalized vocabulary and the normalized replacement library.

[0020] In one embodiment, constructing a dictionary tree using the word roots and replacement words of the keyword word library includes:

[0021] Construct the root word and all pattern strings of the replacement word into a dictionary tree, and set the failure pointer of each node of the dictionary tree; initialize the current state to the root node.

[0022] In one embodiment, performing multi-pattern matching on each character of the target text in a preset dictionary tree includes:

[0023] Traversing each character of the target text based on a multi-mode string matching algorithm, and performing state transition according to the current state and the input character;

[0024] If the current state has a corresponding transfer edge, transfer to the next state, otherwise backtrack according to the failure pointer;

[0025] If the current state is the terminal state of a pattern string, the position where the pattern string matches is recorded.

[0026] In one embodiment, the method for identifying live broadcast interaction information further includes:

[0027] If there are multiple characters at the current position, recursively traverse downwards from the current position using a new branch until the traversal is exited.

[0028] In one embodiment, recursively traversing downward from the current position using a newly opened branch includes:

[0029] If the path before the current position has been traversed in the trunk, prefix prune the path before the current position, skip the path from the root node to the current node and recursively traverse downward;

[0030] If the traversal fails from the current position, when the number of failures is greater than the number of characters at the current position, the path after the current position is suffix pruned, and the failed path of the current node is skipped and recursive traversal is continued downward.

[0031] In one embodiment, the method for identifying live interactive information is characterized by further comprising:

[0032] When the sensitive information in the target text meets the manual review conditions, the interactive information is sent to the manual review system for review; and corresponding processing operations are performed on the user according to the review results returned by the manual review system.

[0033] A device for identifying live interactive information, comprising:

[0034] An extraction module is used to obtain interactive information input in the live broadcast room in real time during the live broadcast process, and extract the target text from the interactive information;

[0035] A matching module, used for performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on the root word generated by the live broadcast sensitive information and the replacement words derived therefrom;

[0036] A detection module, used to detect whether sensitive information exists in the target text according to the matching results of each character;

[0037] The interception module is used to intercept the interactive information in real time and prevent it from being displayed on the live broadcast room interface when sensitive information exists in the target text.

[0038] A live broadcast system, comprising: a client and a live broadcast server; wherein the client comprises an anchor terminal and a viewer terminal respectively connected to the live broadcast server through a network;

[0039] The live broadcast server is used to receive the anchor video uploaded by the anchor terminal and forward it to the audience terminal, and use the live broadcast interactive information identification method to identify the interactive information uploaded by the client terminal;

[0040] The anchor terminal is used to collect anchor videos and upload them to the live broadcast server, and upload the interactive information of the anchor side to the live broadcast server for identification processing;

[0041] The audience terminal is used for audience users accessing the live broadcast room, playing the anchor video sent by the live broadcast server, and uploading the interactive information on the audience side to the live broadcast server for identification processing.

[0042] An electronic device, comprising:

[0043] one or more processors;

[0044] Memory;

[0045] One or more applications, wherein the one or more applications are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs are configured to execute the steps of the above-mentioned live interactive information identification method.

[0046] A computer-readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, wherein the at least one instruction, the at least one program, the code set or the instruction set is loaded by the processor and executes the steps of the above-mentioned method for identifying live interactive information.

[0047] The technical solution of the present application obtains interactive information input into the live broadcast room in real time during the live broadcast process, and extracts target text from the interactive information; performs multi-mode matching in a dictionary tree constructed based on word roots generated based on sensitive information of the live broadcast and replacement words derived therefrom; detects sensitive information in the target text; and intercepts interactive information containing sensitive information in real time and prevents it from being displayed on the live broadcast room interface; this technical solution has a wider identification coverage and higher efficiency, and can achieve more accurate and comprehensive interception of variant characters appearing during the live broadcast process, reducing the risk of missed detection and the cost of identification.

[0048] Furthermore, a normalized pruning algorithm is used to traverse and match the target text. Through the processing techniques of prefix pruning and suffix pruning, it is only necessary to traverse the target string once to match all the contained root words and derived replacement words, which facilitates the rapid detection of various sensitive information during the live broadcast process and further improves the efficiency of identifying live interactive information. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a schematic diagram of an example live broadcast service application scenario;

[0050] Figure 2 is a flow chart of a method for identifying live broadcast interaction information according to an embodiment;

[0051] Figure 3 This is an example diagram of identification of live interactive information;

[0052] Figure 4 is a schematic diagram of an example matching process;

[0053] Figure 5 is a diagram of an example text "graph" traversal;

[0054] Figure 6 is a schematic diagram of the structure of a device for identifying live interactive information according to an embodiment;

[0055] Figure 7 This is a schematic diagram of the live broadcast system structure;

[0056] Figure 8 is a block diagram of an example electronic device. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] In the embodiments of the present application, the words "first", "second", etc. are used to distinguish between identical or similar items with substantially the same effects and functions. "at least one" means one or more, and "multiple" means two or more. For example, multiple objects refer to two or more objects. "Include" or "comprises" and similar words mean that the information appearing before "include" or "comprises" covers the information listed after "include" or "comprises" and its equivalents, and does not exclude other information. "And / or" mentioned in the embodiments of the present application indicates that three types of relationships may exist. The character " / " generally indicates that the objects associated before and after are in an "or" relationship.

[0059] refer to Figure 1 As shown, Figure 1 This is a schematic diagram of an example live broadcast service application scenario. The technical solution provided in the embodiment of the present application can be applied to Figure 1 In the live broadcast service application scenario, the live broadcast service application scenario includes a live broadcast server and a client; the anchor end and the audience end, the anchor end and the audience end communicate data with the live broadcast server through a communication network, so that the anchor on the anchor end and the audience users on the audience end can conduct real-time network live broadcast; among them, for the anchor end and the audience end, their terminal devices can be but not limited to various personal computers, laptops, smart phones and tablets, and the live broadcast server can be implemented with an independent server or a server cluster composed of multiple servers.

[0060] The live broadcast interactive information identification method of the present application is applied on the live broadcast server, and the interactive information (such as public screen speech information, user information modification, etc.) input by the client user in the live broadcast room is detected through various business systems to identify sensitive information and intercept it in real time to prevent sensitive information from being displayed on the screen or externally.

[0061] refer to Figure 2 As shown, Figure 2 The flowchart of the method for identifying live broadcast interaction information of an embodiment is applied to a live broadcast server, which mainly includes the following steps:

[0062] Step S10, during the live broadcast process, obtaining interactive information input in the live broadcast room in real time, and extracting target text from the interactive information;

[0063] In this step, during the live broadcast in the live broadcast room, the live broadcast server monitors the client in real time. When the user enters interactive information during the live broadcast, such as public screen speeches, data updates, etc., the interactive information entered by the user in the live broadcast room is obtained from the client in real time, and then the target text is extracted from these interactive information and uploaded for review and identification.

[0064] In one embodiment, the method of acquiring the interactive information input in the live broadcast room in real time and extracting the target text from the interactive information in step S10 includes:

[0065] When interactive information is input in the live broadcast room, a screenshot image of the live broadcast room interface is obtained in real time through the live broadcast service system; based on the OCR (Optical Character Recognition) technology, text recognition is performed on the screenshot image to obtain the target text.

[0066] For example, when a user speaks on a public screen and inputs the speech content, a screenshot of the live broadcast room interface is taken through the client to obtain a screenshot image, which is then sent to the live broadcast server for OCR text recognition to obtain the target content.

[0067] Exemplarily, before performing OCR text recognition on the screenshot image of the live broadcast room interface, the current frame screenshot image can be compared with the previous frame to obtain newly added interactive information, and OCR text recognition can be performed on the newly added interactive information to obtain the identified target content.

[0068] Step S20, performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on the root word generated by the live broadcast sensitive information and the replacement words derived therefrom.

[0069] In this step, the live broadcast server can derive replacement words based on the roots generated by the live broadcast sensitive information, so that variant characters, homophonic characters, etc. related to the roots can be covered. A dictionary tree (such as a Trie tree) is constructed based on the roots generated by the live broadcast sensitive information and the replacement words derived from them. Then, multi-mode matching is performed on each character of the target text based on a multi-mode string matching algorithm (such as an AC automaton algorithm), that is, the characters of the target text are traversed and matched one by one in the dictionary tree, and each character of the target text is traversed to match, so as to match all the contained roots and derived replacement words.

[0070] In one embodiment, the method for constructing a dictionary tree may include the following steps:

[0071] (1) Detect sensitive information that appears during live broadcasts on live broadcast platforms.

[0072] Specifically, the live broadcast server can record the sensitive information that appears during the live broadcast process of the live broadcast platform and store it in a sensitive information database.

[0073] (2) Generate a root word and a replacement word derived from the root word according to the sensitive information, and generate a normalized keyword thesaurus according to a mapping relationship between the root word and the replacement word.

[0074] Specifically, replacement words are derived from word roots. Matching word roots can be generated based on sensitive information. Several replacement words can be derived from synonyms, near-synonyms, homophones, near-phonetic characters or near-images. Then, based on the derived replacement words, they are associated with the mapping relationship between the word roots and the replacement words. A normalized keyword thesaurus can be established based on the word roots and their mapping relationship.

[0075] In one embodiment, a method for generating a normalized keyword vocabulary may include the following:

[0076] a. Perform semantic analysis based on the sensitive information and generate a certain number of word roots, and construct a normalized word library based on the word roots; specifically, a small number of word roots can be managed and maintained in the background of the live broadcast server, and the word roots can be normalized to construct a normalized word library.

[0077] b. Deriving a number of replacement words according to the root, and constructing a normalized replacement library based on the mapping relationship between the replacement words and the root; specifically, based on the pre-stored root through the mapping relationship between the root and the replacement word, a root can derive a number of derived words, and then construct a normalized replacement library based on the mapping relationship between the root and the replacement word.

[0078] c. A normalized keyword vocabulary is obtained based on the normalized vocabulary and the normalized replacement library; specifically, in the normalized keyword vocabulary constructed based on the normalized vocabulary and the normalized replacement library, the root word and all the derivative words can be matched and intercepted, thereby expanding the coverage of the vocabulary and achieving the effect of covering a surface with one point.

[0079] In the scheme of the above embodiment, the live broadcast server generates a normalized keyword thesaurus based on the root word and the replacement word derived therefrom, through the mapping relationship between the root word and the replacement word, and can expand the coverage of various variant situations by adding replacement words, thereby improving the coverage of the thesaurus, so that it can be used for matching and intercepting various variant words. The generated normalized keyword thesaurus can greatly reduce the keyword maintenance cost and difficulty in target text identification, the accuracy will be better, and it can quickly respond to the interception of various sensitive information.

[0080] (3) Construct a dictionary tree using the root words and replacement words of the keyword vocabulary.

[0081] In this step, the root word and the replacement word text string are used to construct a dictionary tree, which is a tree structure used to quickly query character strings or character prefixes.

[0082] In one embodiment, all pattern strings of the root word and the replacement word can be constructed into a dictionary tree, and a failure pointer of each node of the dictionary tree is set; and the current state is initialized to the root node.

[0083] Specifically, all pattern strings are constructed into a dictionary tree, and the failure pointer failurelink of each node is set; the dictionary tree is initialized, and all sensitive information consisting of roots and replacement words are inserted into the dictionary tree as strings.

[0084] The scheme of the above-mentioned embodiment constructs a dictionary tree through word roots and their replacement words. Compared with the dictionary tree constructed by the conventional AC automaton algorithm, it has a wider coverage. Even if new sensitive information is detected during the live broadcast, the interception coverage can be improved by simply adding word roots. The target string content of various variant words input by the user can be covered through the replacement words.

[0085] In one embodiment, the method of performing multi-pattern matching on each character of the target text in a preset dictionary tree in step S20 may include the following:

[0086] Based on a multi-pattern string matching algorithm (such as the AC automaton algorithm), each character of the target text is traversed, and the state is transferred according to the current state and the input character; if the current state has a corresponding transfer edge, it is transferred to the next state, otherwise it is backtracked according to the failure pointer; if the current state is the terminal state of a pattern string, the position of the pattern string match is recorded.

[0087] Specifically, traverse each character of the target text and perform state transition according to the current state and the input character; if the current state has a corresponding transition edge, transfer to the next state; if there is no transition edge, backtrack according to the failure pointer; if the current state is a terminal state of a pattern string, it means that a match for the pattern string is found in the target text, and the matching position is recorded or related operations are performed.

[0088] In one embodiment, in the process of traversing the target text, in order to quickly match all the included root words and derived replacement words, thereby achieving the purpose of quickly detecting various sensitive information, this embodiment provides a solution for traversing and matching the target text using a normalized pruning algorithm, including the following:

[0089] S201, if there are multiple characters at the current position, recursively traverse downward from the current position using a newly opened branch until the traversal is exited.

[0090] Specifically, since the target text is normalized into one for traversal, there may be multiple characters corresponding to the same position. Therefore, during the traversal process, it is determined whether there are multiple characters at the current position. If so, a new branch is opened from the current position to recursively traverse downward until the traversal is exited.

[0091] S202: If the path before the current position has been traversed in the trunk, prefix pruning is performed on the path before the current position, and the path from the root node to the current node is skipped and recursively traversed downward.

[0092] Specifically, if the path before the current position has been traversed in the trunk, then there is no need to start from the root node when opening a new branch. The path from the root node to the current node has been traversed and does not need to be traversed again. Based on this, the path from the root node to the current node can be directly skipped and recursively traversed downward, thereby reducing the traversal path and improving traversal efficiency.

[0093] S203, if the traversal from the current position fails, when the number of failures is greater than the number of characters at the current position, the path after the current position is suffix pruned, and the failed path of the current node is skipped and the traversal is recursively performed downward.

[0094] Specifically, if the traversal from the current position fails, the number of failures at the current position is recorded. When the number of failures is greater than the number of characters at the current position, the new branch does not need to start from the root node. The path from the root node to the current node has been traversed and does not need to be traversed again. The path after the current position is suffix pruned, and the failed path of the current node is directly skipped and recursive traversal is performed downward, thereby reducing the traversal path and improving traversal efficiency.

[0095] As in the solution of the above-mentioned embodiment, during the pattern matching process, a normalized pruning algorithm is used to perform prefix pruning and suffix pruning. It is only necessary to traverse the target text string once to match all the included roots and derived replacement words, which facilitates the rapid detection of various sensitive information and improves the detection efficiency of interactive information.

[0096] Step S30, detecting whether there is sensitive information in the target text according to the matching results of each character.

[0097] In this step, the traversal and matching process is repeated by traversing and matching each character of the target text one by one until the entire target text is traversed. Combining the matching results of each character, it is possible to detect whether there is sensitive information in the target text.

[0098] For example, when determining whether there is sensitive information in the target text, the semantics of the target text can be analyzed based on the character matching root or replacement word of the target text, and the corresponding strategy can be used to determine whether there is sensitive information; thereby providing important technical support for the subsequent interception processing of the live broadcast platform.

[0099] Step S40, when there is sensitive information in the target text, the interactive information is intercepted in real time and prevented from being displayed on the live broadcast room interface.

[0100] Specifically, in combination with the sensitive information detection results, when it is found that the interactive information input by the user contains sensitive information, the interactive information is intercepted and processed, and the interactive information is prevented from being displayed on the live broadcast room interface. For example, if the public screen speech contains sensitive information, the speech is prohibited from being displayed on the public screen. If it is a user information modification, the modification is refused.

[0101] In one embodiment, different processing can be performed according to the degree of interception of interactive information. When the sensitive information in the target text meets the manual review conditions, the interactive information is sent to the manual review system for review; and corresponding processing operations are performed on the user based on the review results returned by the manual review system.

[0102] For example, if the sensitive information meets the conditions for manual review, the interactive information will be sent to the manual review system for review. If the interactive information is found to be in violation of the rules, corresponding penalties may be imposed based on the severity of the violation, such as clearing the information, prohibiting modification of the information, or blocking the account.

[0103] In order to make the technical solution of this application more clear, several application examples are described below.

[0104] refer to Figure 3 As shown, Figure 3It is an example diagram for identifying live interactive information. Assume that on the live server side, the root words "same city service" and some inappropriate words like "silly", "stupid" are managed and maintained. The normalized keyword library constructed based on the replacement words derived from the root words includes: same (child, tong), city (journey, cheng), service (government office, fu), service (thing, wu), silly (what, sha), coin (ratio, bi); when the client user inputs the interactive information "child government office thing, come... big silly", through normalization conversion, it becomes "child [same] government office [service] thing [service], come... big what [silly] ratio [coin]". Pass multiple characters of the target text to the trie tree for traversal and matching, and use the normalization pruning algorithm to traverse and match the keyword library once to obtain the detection results: "same city service, silly, stupid", and its matching process is as Figure 4 shown Figure 4 It is a schematic diagram of an example matching process; the algorithm benefit comparison results of the normalization pruning algorithm are as follows in the table:

[0105] Pruning path income Remark Full path 504 0 2*1*2*2*1*1*3*3*7＝504 Prefix pruning 206 298 Suffix pruning 27 271

[0106] As can be seen from the above table, for the technical solution of this embodiment, after prefix pruning and suffix pruning processing, higher benefits can be obtained, thereby greatly improving the efficiency of traversal and matching.

[0107] As Figure 4 shown, the final hit is "same city service... silly", from which it can be identified that the input interactive information contains sensitive information, and thus the interactive information is intercepted and prevented from being displayed on the live broadcast interface.

[0108] For the traversal process of the target text, since the target text is extended from one-dimensional to two-dimensional, forming a data structure of the text "graph", sequential traversal is no longer used during traversal, and the text "graph" traversal method is adopted;示例性, depth-first search (Depth First Search, abbreviated as DFS) or breadth-first search (Breath First Search, abbreviated as BFS) can be used, and the goal of traversal is to traverse all paths once; preferably, the depth-first traversal scheme is adopted. During the traversal process, as Figure 5 shown Figure 5 It is a schematic diagram of an example text "graph" traversal. Assume that the input text "graph": "child [same] government office [service] thing [service], come... big what [sha, silly] ratio [stupid, coin]" is traversed, and when traversing and backtracking, the trie tree status bits of the AC automaton are backtracked together.

[0109] Next, an embodiment of the apparatus for identifying live interactive information is described.

[0110] Refer to Figure 6 shown Figure 6The structure diagram of a device for identifying live interactive information according to an embodiment includes:

[0111] The extraction module 10 is used to obtain the interactive information input in the live broadcast room in real time during the live broadcast process, and extract the target text from the interactive information;

[0112] A matching module 20, for performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on a root word generated from live broadcast sensitive information and its derived replacement words;

[0113] A detection module 30, used to detect whether there is sensitive information in the target text according to the matching results of each character;

[0114] The interception module 40 is used to intercept the interactive information in real time and prevent it from being displayed on the live broadcast room interface when sensitive information exists in the target text.

[0115] The live broadcast interaction information identification device of this embodiment can execute a live broadcast interaction information identification method provided by the embodiment of the present application, and its implementation principle is similar. The actions performed by each module in the live broadcast interaction information identification device in each embodiment of the present application correspond to the steps in the live broadcast interaction information identification method in each embodiment of the present application. For the detailed functional description of each module in the live broadcast interaction information identification device, please refer to the description in the corresponding live broadcast interaction information identification method shown in the previous text, which will not be repeated here.

[0116] An embodiment of the live broadcast system is described below.

[0117] The live broadcast system of this embodiment refers to Figure 7 As shown, Figure 7 This is an example of a live broadcast system structure diagram. The system framework may include a client and a live broadcast server. The client includes an anchor terminal and an audience terminal. The anchor terminal and multiple audience terminals respectively establish communication connections with the live broadcast server through a wired network or a wireless network. The anchor user broadcasts live through the anchor terminal, such as performing a real talent show, and the audience user chooses to enter a virtual live broadcast room through the audience terminal to watch the anchor user broadcast live. The audience terminal and the anchor terminal can enter the live broadcast platform through the client installed on the electronic device. For example, the anchor terminal and the audience terminal can be computer devices, such as PDA, smart phone, tablet computer, desktop computer or laptop computer, etc., which is not limited to this. It can also be a software module of an application. The live broadcast server includes a background server for providing background services for electronic devices, which can be implemented with an independent server or a server cluster composed of multiple servers.

[0118] For the live broadcast server, it is used to receive the anchor video uploaded by the anchor end and forward it to the audience end, and use the live broadcast interactive information identification method of any of the above embodiments to identify the interactive information uploaded by the client, including various interactive information uploaded by the audience side and the anchor side; for the anchor end, it is used to access the anchor user, collect the anchor video and upload it to the live broadcast server, and also upload the interactive information of the anchor side to the live broadcast server for identification; for the audience end, it is used to access the audience user in the live broadcast room, play the anchor video sent by the live broadcast server, and also upload the interactive information of the audience side to the live broadcast server for identification.

[0119] Exemplarily, on the client side, the anchor side and the audience side are responsible for reporting the interactive information input by the user; on the live broadcast server side, each business system actively reports the interactive information, including public screen speech information, user information modification, etc., to the content security review system through the interface. The content security review system obtains the text information to obtain the target text, and uses the present application to perform real-time, fast and intelligent filtering on each character of these target texts, including variant characters, homophonic characters, near-shaped characters, synonymous characters, etc. If the intercepted keywords can be determined to contain sensitive information, they will be returned to the business system for real-time interception to prevent the sensitive information from being displayed on the screen or externally; in addition, if the sensitive information meets the conditions for manual review, the interactive information will be transferred to the manual review system for review. If the manual review finds a violation, the live broadcast server can take corresponding measures according to the severity of the violation, such as clearing the information, prohibiting modification of the information, banning the account, etc.

[0120] As in the live broadcast system of the above-mentioned embodiment, based on the matching algorithm of word roots and their derived replacement words, the live broadcast server can realize more comprehensive matching interception, achieving the matching effect of covering a surface with one point. After discovering new sensitive information, it only needs to simply add word roots to improve the coverage and interception of various variant characters. The interception coverage is wider and the accuracy is better, which reduces the risk of missing sensitive information and reduces the cost and difficulty of keyword maintenance. Through the normalized pruning algorithm to traverse and match the target text, through the processing technology of prefix pruning and suffix pruning, it is only necessary to traverse the target string once to match all the contained word roots and derived replacement words, which is convenient for quickly detecting various sensitive information and improving the detection efficiency of interactive information.

[0121] Embodiments of electronic devices and computer-readable storage media are described below.

[0122] The present application provides a technical solution for an electronic device to implement functions related to a method for identifying live interactive information. The electronic device includes: one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured for the method for identifying live interactive information in any embodiment.

[0123] like Figure 8 As shown, Figure 8 It is a block diagram of an example electronic device, which includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for obtaining a steady-state value of a delay estimation is implemented. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covered on the display screen, or a button, a trackball or a touch pad set on the housing of the electronic device, or an external keyboard, touch pad or mouse, etc.

[0124] Those skilled in the art will understand that Figure 8 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0125] The present application provides a technical solution of a computer-readable storage medium to implement functions related to a method for identifying live interactive information. The computer-readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded by a processor and executes the method for identifying live interactive information of any embodiment. In an exemplary embodiment, the computer-readable storage medium may be a non-temporary computer-readable storage medium including instructions, such as a memory including instructions. For example, the non-temporary computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0126] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for identifying live interactive information, characterized in that: include: During the live broadcast process, the interactive information input in the live broadcast room is obtained in real time, and the target text is extracted from the interactive information; Performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on the root word generated by the live broadcast sensitive information and the replacement words derived therefrom; Detecting whether there is sensitive information in the target text according to the matching results of each character; When sensitive information exists in the target text, the interactive information is intercepted in real time and prevented from being displayed on the live broadcast room interface.

2. The method for identifying live interactive information according to claim 1, characterized in that: Also includes: Detect sensitive information that appears during live broadcasts on live broadcast platforms; Generate a root word and a replacement word derived from the root word according to the sensitive information, and generate a normalized keyword word library according to a mapping relationship between the root word and the replacement word; A dictionary tree is constructed using the root words and replacement words of the keyword vocabulary.

3. The method for identifying live interactive information according to claim 1, characterized in that: Acquire interactive information input in the live broadcast room in real time, and extract target text from the interactive information, including: When interactive information is input in the live broadcast room, a screenshot image of the live broadcast room interface is obtained in real time through the live broadcast service system; text recognition is performed on the screenshot image based on OCR technology to obtain the target text.

4. The method for identifying live interactive information according to claim 2, characterized in that: Generating a root word and a replacement word derived from the root word according to the sensitive information, and generating a normalized keyword word library according to a mapping relationship between the root word and the replacement word, including: Performing semantic analysis on the sensitive information and generating a certain number of word roots, and building a normalized word library based on the word roots; Deriving a plurality of replacement words according to the root word, and constructing a normalized replacement library based on the mapping relationship between the replacement words and the root word; A normalized keyword vocabulary is obtained according to the normalized vocabulary and the normalized replacement library.

5. The method for identifying live interactive information according to claim 2, characterized in that: The word roots and replacement words of the keyword word library are used to construct a dictionary tree, including: Construct the root word and all pattern strings of the replacement word into a dictionary tree, and set the failure pointer of each node of the dictionary tree; initialize the current state to the root node.

6. The method for identifying live interactive information according to claim 5, characterized in that: Performing multi-pattern matching on each character of the target text in a preset dictionary tree includes: Traversing each character of the target text based on a multi-mode string matching algorithm, and performing state transition according to the current state and the input character; If the current state has a corresponding transfer edge, transfer to the next state, otherwise backtrack according to the failure pointer; If the current state is the terminal state of a pattern string, the position where the pattern string matches is recorded.

7. The method for identifying live interactive information according to claim 6, characterized in that: Also includes: If there are multiple characters at the current position, recursively traverse downwards from the current position using a new branch until the traversal is exited.

8. The method for identifying live interactive information according to claim 7, characterized in that: From the current position, recursively traverse down by starting a new branch, including: If the path before the current position has been traversed in the trunk, prefix prune the path before the current position, skip the path from the root node to the current node and traverse recursively downward; If the traversal fails from the current position, when the number of failures is greater than the number of characters at the current position, the path after the current position is suffix pruned, and the failed path of the current node is skipped and recursive traversal is continued downward.

9. The method for identifying live interactive information according to claim 1, characterized in that: Also includes: When the sensitive information in the target text meets the manual review conditions, the interactive information is sent to the manual review system for review; Perform corresponding processing operations on the user according to the review results returned by the manual review system.

10. A device for identifying live interactive information, characterized in that: include: An extraction module is used to obtain interactive information input in the live broadcast room in real time during the live broadcast process, and extract target text from the interactive information; A matching module, used for performing multi-mode matching on each character of the target text in a preset dictionary tree; wherein the dictionary tree is constructed based on the root word generated by the live broadcast sensitive information and the replacement words derived therefrom; A detection module, used to detect whether sensitive information exists in the target text according to the matching results of each character; The interception module is used to intercept the interactive information in real time and prevent it from being displayed on the live broadcast room interface when sensitive information exists in the target text.

11. A live broadcast system, characterized in that: include: Client and live broadcast server; wherein the client includes an anchor terminal and a viewer terminal respectively connected to the live broadcast server through a network; The live broadcast server is used to receive the anchor video uploaded by the anchor end and forward it to the audience end, and use the live broadcast interactive information identification method described in any one of claims 1 to 9 to identify the interactive information uploaded by the client; The anchor terminal is used to collect anchor videos and upload them to the live broadcast server, and upload the interactive information of the anchor side to the live broadcast server for identification processing; The audience end is used for audience users accessing the live broadcast room, playing the anchor video sent by the live broadcast server, and uploading the interactive information on the audience side to the live broadcast server for identification processing.

12. An electronic device, characterized in that: The electronic device comprises: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the steps of the method for identifying live interactive information as described in any one of claims 1-9.

13. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded by the processor and executes the steps of the method for identifying live interactive information according to any one of claims 1-9.

Images