WO2021151333A1

WO2021151333A1 - Sensitive word recognition method and apparatus based on artificial intelligence, and computer device

Info

Publication number: WO2021151333A1
Application number: PCT/CN2020/124684
Authority: WO
Inventors: 吕焕焕; 姜国玮; 张冬; 李飞鹏
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-09-07
Filing date: 2020-10-29
Publication date: 2021-08-05
Also published as: CN112016317A

Abstract

A sensitive word recognition method and apparatus based on artificial intelligence, and a computer device, relating to the technical field of artificial intelligence. The method comprises: firstly, acquiring text information to be subjected to recognition (101); then carrying out recognition on a target word slot combination included in the text information (102), wherein the target word slot combination is composed of at least one preset word slot; next, determining, according to the target word slot combination and intermediate character and word information of the target word slot combination in the text information, whether the text information includes a sensitive word (103); and if it is determined that the text information includes a sensitive word, performing limitation processing on the text information (104). The method can improve the accuracy of sensitive word recognition. In addition, the method also relates to blockchain technology, and text data can be stored in a blockchain, so as to ensure data privacy and security.

Description

Artificial intelligence-based sensitive word recognition method, device and computer equipment

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 7, 2020, the application number is 202010927419.7, and the invention title is "Artificial Intelligence-based Sensitive Word Recognition Method, Apparatus, and Computer Equipment". The entire content of the application is approved The reference is incorporated in this application.

Technical field

This application relates to the field of artificial intelligence technology, and in particular to a method, device and computer equipment for identifying sensitive words based on artificial intelligence.

Background technique

With the development of social software, the communication methods between users have become more and more diversified. But at the same time, it also brings an unavoidable problem. The content of the transmitted information cannot be effectively controlled. Bad information containing sensitive words may be sent to users through various channels, causing certain adverse effects. Therefore, it is derived The need for sensitive word filtering. Perform effective quality inspection of sensitive words when users publish content to ensure that the quality of the output content meets the specifications.

The inventor realizes that currently, traditional sensitive word filtering usually uses one-to-one filtering, using simpler regular expressions to match sensitive words, or by maintaining a sensitive word class library, and performing text input in the corresponding table. Look up. For example, if you need to filter out the sensitive word A, then search for all the content that matches the given regular expression in the string, or find out whether there is a corresponding vocabulary or content in the sensitive word library, and return the corresponding if it exists the result of.

However, the inventor of this application found in his research that the traditional filtering method of sensitive words has great limitations. It can only match specific content and is easy to be exploited. The matching rules are skipped, such as adding symbols or adding symbols in the middle of sensitive words. Spaces, etc., in turn, cannot achieve the required sensitive word recognition effect, which affects the accuracy of sensitive word recognition.

technical problem

In view of this, this application provides an artificial intelligence-based method, device, and computer equipment for identifying sensitive words, the main purpose of which is to improve the current traditional filtering method of sensitive words, which causes the technical problem of low accuracy in identifying sensitive words.

Technical solutions

According to one aspect of the present application, there is provided a method for identifying sensitive words based on artificial intelligence. The method includes: acquiring text information to be recognized; identifying a target word slot combination contained in the text information, wherein the target word The slot combination is composed of at least one preset word slot; according to the target word slot combination and the intermediate word information of the target word slot combination in the text information, it is determined whether the text information contains sensitive words; If the text information contains sensitive words, the text information is restricted.

According to another aspect of the present application, there is provided an artificial intelligence-based sensitive word recognition device. The device includes: an acquisition module for acquiring text information to be recognized; a recognition module for recognizing content contained in the text information The target word slot combination, wherein the target word slot combination is composed of at least one preset word slot; the judgment module is used for combining the target word slot combination and the middle character in the text information according to the target word slot combination The word information determines whether the text information contains sensitive words; the processing module is used to perform restriction processing on the text information if it is determined that the text information contains sensitive words.

According to another aspect of the present application, there is provided a readable storage medium having computer readable instructions stored thereon, and when the computer readable instructions are executed by a processor, the following method is implemented: obtaining text information to be recognized; The target word slot combination included in the text information, wherein the target word slot combination is composed of at least one preset word slot; according to the target word slot combination and the target word slot combination in the middle of the text information The word information determines whether the text information contains sensitive words; if it is determined that the text information contains sensitive words, the text information is restricted.

According to another aspect of the present application, there is provided a computer device, including a readable storage medium, a processor, and computer readable instructions stored on the readable storage medium and executable on the processor, the processor executing all The computer-readable instruction implements the following method: acquiring text information to be recognized; identifying a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one preset word slot; according to the The target word slot combination and the intermediate word information of the target word slot combination in the text information are judged whether the text information contains sensitive words; if it is determined that the text information contains sensitive words, the text information is Restrict processing.

Beneficial effect

This application can accurately identify whether the text information contains sensitive words, which can improve the accuracy of sensitive word recognition and improve the processing efficiency of sensitive words.

The above description is only an overview of the technical solutions of this application. In order to understand the technical means of this application more clearly, it can be implemented in accordance with the content of the specification, and for the above and other purposes, features and advantages of this application to be more obvious and understandable, The following specifically cite specific implementations of this application.

Description of the drawings

The drawings described here are used to provide a further understanding of the application and constitute a part of the application. The exemplary embodiments and descriptions of the application are used to explain the application, and do not constitute an improper limitation of the local application.

FIG. 1 shows a schematic flowchart of an artificial intelligence-based sensitive word recognition method provided by an embodiment of the present application.

FIG. 2 shows a schematic flowchart of another method for identifying sensitive words based on artificial intelligence according to an embodiment of the present application.

Fig. 3 shows a schematic structural diagram of an artificial intelligence-based sensitive word recognition device provided by an embodiment of the present application.

Embodiments of the present invention

Hereinafter, the present application will be described in detail with reference to the drawings and in conjunction with the embodiments. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other if there is no conflict.

The technical solution of this application can be applied to the fields of artificial intelligence, blockchain and/or big data technology to realize sensitive word recognition. Optionally, the data involved in this application, such as text information, can be stored in a database, or can be stored in a blockchain, which is not limited in this application.

In view of the technical problem that the improvement of the current traditional sensitive word filtering method will cause the low accuracy of sensitive word recognition, this embodiment provides an artificial intelligence-based sensitive word recognition method. As shown in FIG. 1, the method includes the following steps.

101. Acquire text information to be recognized.

The text information to be recognized can be the text information of the communication message to be published, such as the message sending text in the instant messaging software, the online communication text between the platform customer service staff and the user, and the message publishing text of the public platform (such as the text published by the web comment , The text of the product evaluation, the text sent by the video barrage, etc.) etc. In addition, the text information to be recognized can also be text within a specified range (such as a specified range text in a publicly published electronic book, a specified range text in a publicly issued notification message, etc.), etc.

The execution subject of this embodiment may be a device or equipment for sensitive word recognition and processing, which may be deployed on the client or server, etc., which can improve the accuracy of sensitive word recognition.

102. Identify the target word slot combination contained in the text information.

Among them, the target word slot combination is composed of at least one preset word slot. In this embodiment, word slots can be set in advance. These word slots can be determined according to different sensitive words. Specifically, they can include word slots of sensitive words (such as "reduction of principal", "reduction of rent", "personal loan", etc., as well as Bank card number, ID number, account password format, a series of digital symbols and other word slots), non-sensitive word slots (such as "no", "must" and other word slots, as well as single numbers, single words and other word slots ), the word slot of the sensitive word synonyms (such as the word slot that is essentially synonymous with the sensitive word but does not belong to the scope of the sensitive word), and can also contain each participle obtained by splitting the sensitive word (such as for the sensitive word "Go to your unit to investigate "The three word slots of "Fuck You", "Unit" and "Investigation" obtained by splitting). Then these word slots are combined and matched according to the corresponding sensitive word recognition to obtain the word slot combination.

For this embodiment, the pre-statistical word slot combination can be stored in a predetermined storage location (such as a database, a mapping table, etc.). When subsequently identifying the word slot combination contained in the text information, each word segmentation in the text information can be combined with a predetermined storage location. Each word slot combination in the storage location is matched, and a matching word slot combination is found as the target word slot combination contained in the text information.

103. Determine whether the text information contains sensitive words according to the target word slot combination contained in the text information and the intermediate word information of the target word slot combination in the text information.

The intermediate word information may be word information that appears between each word slot included in the word slot combination in the text information. For example, the text message is "XX finds someone to go to your unit, and finds out XX after doing a background investigation on you", where XX stands for words that are omitted from the text message, and the target word slot combination contained in the text message is "Go You" + "unit" + "investigation", and the "de" between "fuck you" and "unit", and "being a background for you" between "unit" and "investigation" are the middle words .

In this embodiment, the word slot combination corresponding to the sensitive word has the same meaning as that of the sensitive word to a certain extent. It can be a word slot combination composed of the sensitive word itself; or it is not a sensitive word, but Combination of word slots with the meaning of sensitive words, etc. In specific application scenarios, sometimes the published text that actually contains sensitive words will be mixed with spaces, symbols, or added some words, or the same semantic rewriting through other texts, etc., which will affect whether there is sensitive in the text information. The accuracy of word discrimination. However, this embodiment not only judges the word slot combination and the intermediate word information of the word slot combination in the text information, it can accurately identify whether the text information contains sensitive words in these cases, which can improve the sensitivity. Accuracy of word recognition.

104. If it is determined that the text information contains sensitive words, perform restriction processing on the text information.

For example, when it is determined that the text information contains sensitive words, the text information can be marked and reminded to inform the existence of sensitive word information, such as highlighting the text part containing the target word slot combination in the text information (such as highlighting, bolding, adding Underscore, etc.), or restrict the sending of communication messages containing the text information, etc.

Through the artificial intelligence-based sensitive word recognition method in this embodiment, the target word slot combination contained in the text information can be identified. The word slot combination is composed of at least one preset word slot, and then according to the target word slot combination and target The word slot combines the intermediate word information in the text information to determine whether the text information contains sensitive words. Compared with the current existing traditional filtering method for sensitive words, this embodiment uses the method of discriminating intermediate words between the word slot combination + word slot combination, even if symbols or spaces are added in the text sensitive words, or some words are added, Or the same semantic rewriting through other texts, etc., can accurately identify whether the text information contains sensitive words, which can improve the accuracy of sensitive word recognition. If it is determined that the text information contains sensitive words, the text information can also be restricted and processed in time. The entire process of sensitive word recognition + restriction processing can be automated, which improves the efficiency of sensitive word processing.

Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully explain the specific implementation process in this embodiment, another artificial intelligence-based sensitive word recognition method is provided. As shown in FIG. 2, this method Including the following steps.

201. Acquire text information to be recognized.

Before the sensitive word recognition is performed on the text information to be recognized, in order to ensure the security and privacy of the text information, optionally, the text information can be pre-stored in the blockchain. Accordingly, step 201 can specifically include: Obtain the text information to be recognized from the block chain. For example, the text information to be recognized can be obtained from the target node of the blockchain, and then sensitive word recognition can be performed on the text information. It should be noted that the blockchain referred to in this embodiment is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

202. Clear character spaces and preset special symbols in the text information.

Among them, the preset special matches can be symbols such as "@", "#", "￥", "\", "/", and "*". For this embodiment, before the word slot combination recognition is performed on the text information, the character spaces and preset special symbols in the text information are cleared, which can effectively reduce noise interference and improve the accurate matching of word slot combinations and corresponding detection rules. .

Further, in addition to removing character spaces and preset special symbols in the text information, it can also remove continuously repeated words and symbols in the text information, and can also remove rare words in the text information that will cause matching interference. , Which can further reduce noise interference.

203. Identify the target word slot combination contained in the text information after clearing character spaces and preset special symbols.

In this embodiment, the configuration of the sensitive word recognition rule can be performed first, and the rule configuration can be divided into three layers: slot, rule, and model. The word slot contains some preset keywords such as sensitive words, non-sensitive words, sensitive word synonyms, etc. The rule is a combination of word slots (equivalent to a preset verification rule, that is, the text information meets the criterion for the presence of sensitive words) , The model is a combination of rules (equivalent to a combination of multiple verification rules). After the word slot is established, rules and models can be freely combined, and a sensitive word filtering strategy that meets the requirements can be formulated according to the business scenario. For example, after establishing word slots, rules, and models, each participle in the text information after clearing character spaces and preset special symbols is matched with the word slot combination in the rule, and then the target word slot combination contained in it is found.

It should be noted that after recognition and matching, there may be at least one target word slot combination in the text information, that is, there are multiple sets of different target word slot combinations, and then comprehensive judgments are made based on these target word slot combinations, and steps 204 to 206a are specifically executed. , 206b shows the process.

204. Obtain a target verification rule combination according to at least one target verification rule corresponding to each target word slot combination.

In this embodiment, a single word slot combination may correspond to at least one target verification rule, and each verification rule is equivalent to a preset criterion for containing sensitive words. When a single word slot combination corresponds to at least two verification rules, it is equivalent to a combination of verification rules. It should be noted that this embodiment can pre-determine whether the word slot combination corresponds to a single check rule or a check rule combination containing at least two check rules according to actual needs. That is, the qualifier slot combination can be used in the rule layer or the model. Floor. In this way, the range of sensitive words can be limited, that is, the content of detection. The scope of action is the rule layer or the model layer. The detection is performed in the specified range, and then the verification rules can be flexibly used for sensitive word recognition. In the case of, a variety of verification rules can be used to make accurate judgments from different angles, which can improve the accuracy of sensitive word recognition.

In this embodiment, at least one target verification rule corresponding to each target word slot combination is combined to obtain a target verification rule combination, and the target verification rule combination includes at least one preset sensitive word determination criterion.

205. According to the word slot arrangement information of the target word slot combination in the text information and the intermediate word information between the word slots, respectively determine whether the text information meets the multiple preset sensitive word determination criteria in the target check rule combination.

In the specific judgment process, use each preset sensitive word judgment standard in the target test rule combination, and judge the text according to the word slot arrangement information of the target word slot combination in the text information and the intermediate word information between the word slots Whether the information meets one or more of these sensitive word criteria.

Illustratively, given several examples of criteria for determining sensitive words, step 205 may specifically include: if the criteria for determining sensitive words is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words is limited If the range is to meet the criteria of the judgment, then the arrangement information of the word slots in the judgment (such as the sequence in which the word slots appear in the text) conforms to the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is less than or equal to the predetermined word slot sequence. When the number threshold is set, it is determined that the text information contains sensitive words; if the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range does not meet the judgment standard, When it is determined that the word slot arrangement information matches the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is greater than or equal to the preset number threshold, it is determined that the text information contains sensitive words.

Among them, because sometimes the sequence of different word collocations is different, the semantics obtained will not be different, so each word slot combination has its own corresponding preset word slot sequence, which is used to determine whether it has the meaning of a sensitive word. The word slot combination can correspond to at least one preset word slot sequence according to actual conditions. The middle word part is used to modify the semantics of the word slot combination, and the preset number threshold is used to determine whether the language-modified word slot combination still has the meaning of a sensitive word. The threshold size can be preset according to the actual situation.

For example, a word slot combination composed of word slots "Go to you", "unit", and "investigation" can be matched to get the corresponding preset verification rule as the judgment standard of [and], and editable among the three word slots is allowed The number of characters in is 8. If the message sent by the user contains these three words at the same time and there are less than 8 words in the middle, it will be judged to meet the criterion of [and], that is, it is determined that the message sent by the user contains sensitive words. And if there are more than 8 edited characters among the three sensitive words, it is judged that it does not meet the criterion of [and], that is, it is determined that the information sent by the user does not contain sensitive words. [And] It means that multiple word slots before and after this rule can be hit, and the middle word can be set. For example, for the text message of "Look for someone to go to your unit, do a background check on you.", the middle word of the word slot combination in this text is the "between" "Go to you" and "Unit". ", and "Be a background for you" between "unit" and "investigation" are the middle words. If the sum of the number of middle words is equal to 8, it is considered to hit the corresponding verification rule standard. Then determine that this text contains sensitive words.

For another example, the word slot combination consisting of the word slot of sensitive words "reduction of principal" and the word slot of non-sensitive words "will not", after matching, the corresponding preset verification rule is [Non] judgment standard, There are 3 editable characters between the two sensitive words. If the information sent by the user contains these two words at the same time, and there are less than 3 words in the middle, it will be judged to meet the criterion of [Non], that is, it is determined that the information sent by the user does not contain sensitive words. [Non] The verification rule is to set a word slot of a sensitive word and a word slot of a non-sensitive word. If two words appear together, they will not be hit, and the middle word can be set. For example, for the text message of "no principal reduction", the number of words in the middle between the two word slots of "no" and "principal reduction" is 0, which is considered to have not hit the corresponding verification rule Standards, and then determine that this text does not contain sensitive words. As for "no, you can rest assured, the principal will be reduced", the middle word between the two word slots of "no" and "principal reduction" ", you can rest assured, will definitely" the number is greater than 3, it is considered that the corresponding verification rule standard is hit, and then it is determined that the text contains sensitive words.

In practical applications, the combination of verification rules may include at least two verification rules, which is equivalent to identifying sensitive words through the model in step 203. For example, the combination of verification rules contains three verification rules. The first verification rule is ID verification, the second verification rule is sensitive word + and verification, and the third verification rule is sensitive word + non-verification. . Among them, when using verification rule 1 to identify sensitive words, it can be recognized whether the text information (after removing character spaces, preset special symbols, rare words and other noisy texts) contains a string of numeric word slots, if it contains this type of word slot The word slot can determine whether the string of numbers corresponding to the word slot conforms to the ID card format. If it meets the ID card format, it is determined that it meets the first verification rule, and it can be considered that there are sensitive words in the text information. The above two examples can be referred to when using verification rule 2 and verification rule 3 to identify sensitive words, which will not be repeated here.

In the actual verification process, the target verification rule combination may contain a large number of verification rules. If the verification judgment is performed one by one, it will affect the problem of time efficiency. Therefore, in order to improve the efficiency of sensitive word recognition, optionally, step 205 may further specifically include: if the target inspection rule combination includes at least one preset sensitive word determination criterion with different execution priorities, then according to each target inspection rule combination. The execution priority of the sensitive word judgment standard is from high to low, and the text information is judged in turn; in the process of sequential judgment, if it is determined that there is a sensitive word judgment standard that the text information meets, then the subsequent judgment on the text information is stopped , And use the currently obtained judgment result as the result of judging the text information by using the target inspection rule combination.

For example, the target check rule combination contains five check rules, and the five check rules are preset with execution priority in the check rule combination (for example, the preset priority is based on the sensitive word recognition success rate from high to low) , The order of execution priority from high to low is: verification rule one>verification rule three>verification rule four>verification rule five>verification rule two, and then follow the order of this sort and use the corresponding verification in turn The rule judges the text information. If it is judged that there are sensitive words in the text information through verification rule three, the subsequent verification process of verification rule four, verification rule five, and verification rule two will be stopped. Through this optional method, there is no need to check sensitive words one by one, and the judgment result can be obtained as quickly as possible, which can improve the efficiency of sensitive word recognition.

In the specific application implementation of this embodiment, the calculation priority range can be limited. Similar to the four arithmetic operations, the verification rules appearing in the priority range are executed first. If there are specific regular symbols in the verification rules that represent different meanings, the verification rules that match the priority range can be placed in brackets "()". When the verification rules are executed, the content of the rules in the brackets will be executed first, and then the others will be executed. The verification rules.

206a. If the text information meets at least one set of sensitive word determination criteria in the target inspection rule combination, it is determined that the text information contains sensitive words.

In this embodiment, a set of sensitive word judgment standards may include at least one sensitive word judgment standard, that is, one, or two, or more sensitive word judgment standards, which may be determined according to actual sensitive word judgment accuracy requirements. .

In step 206b, which is parallel to step 206a, if none of the text information meets the sensitive word determination criteria in the target inspection rule combination, it is determined that the text information does not contain sensitive words.

In this embodiment, a plurality of preset sensitive word judgment criteria in the combination of inspection rules may be used to perform sensitive word recognition judgment on the text information respectively. If the text information is judged to meet at least one of the preset sensitive word judgment criteria, then It can be determined that the text information contains sensitive words, which can improve the accuracy of sensitive word recognition.

207. If it is determined that the text information contains sensitive words, perform restriction processing on the text information.

Optionally, restricting the text information may specifically include: preventing the publication of the text information; or, using preset characters (such as "*", "-" and other characters in the text part containing the target word slot combination in the text information, (It has the effect of desensitization) before publishing after replacement; or, sending the text information to the review module for review, and publishing if the review is passed. For example, after determining that the text information contains sensitive words, the system will prevent the user from publishing sensitive words, or directly delete the content containing sensitive words sent by the user. For some less sensitive words, they will not be deleted immediately after they are sent out, and the reviewers need to conduct a second manual review.

Further optionally, after step 207, the method of this embodiment may further include: recording the text part of the target word slot combination in the text information as sample data; and then periodically analyzing each recorded sample data to make statistics on each sample data Word combinations that appear more frequently than the preset frequency threshold and are different from the existing word slot combinations; calculate the semantic similarity between the word combinations obtained by statistics and the preset sensitive words and/or preset sensitive sentences; The target word combination whose semantic similarity is greater than the preset similarity threshold is used as a new word slot combination, and the verification rule corresponding to the new word slot combination is updated according to the sample data containing the new word slot combination; it can be used later The new word slot combination and corresponding inspection rules determine whether other text information contains sensitive words. Through this method of automatically updating the word slot combination and its corresponding inspection rules, the automatic update of the sensitive word recognition system can be realized, so as to further improve the accuracy of subsequent sensitive word recognition. The entire sensitive word recognition system is equivalent to having the function of machine learning, which can realize the accurate recognition of sensitive words by artificial intelligence.

For example, for an article on a sensitive topic, it usually not only includes a set of word slots with the meaning of a sensitive word, but also uses a variety of different words to express the sensitive topic. Therefore, using existing word slot combinations and their corresponding verification rules, text data determined to contain sensitive words may sometimes include other word combinations with the meaning of sensitive words. This embodiment collects these text data as sample data; regularly analyzes the sample data, finds word combinations that appear more than a certain threshold and is different from the existing word slot combinations, and compares them with the preset sensitivity Words and/or preset sensitive sentences are calculated for semantic similarity, and then find those new word slot combinations that have not been discovered before and also have the meaning of sensitive words, and formulate their corresponding verification rules. In this way, the new word slot combination and corresponding inspection rules can be used to determine whether other text information contains sensitive words, so as to find more text data that actually has the meaning of sensitive words.

The method of this embodiment can also be applied to a system for intelligent sensitive word quality inspection. Algorithms can be used to match entries. Specific rules and strategies can be set to reduce noise interference. It can span text and perform accurate sensitive words. filter. After constructing the sensitive vocabulary, the algorithm is used to traverse the text and match with the sensitive word tree to achieve the function of identifying and filtering sensitive vocabulary. Intelligent strategies can be customized according to customer needs to efficiently filter prohibited messages, malicious promotion, vulgar abuse, low-quality irrigation and other sensitive words and prohibited variants. The intelligent quality inspection system has a high accuracy of content review and recognition, which can quickly process text, greatly reduce the workload of manual review, eliminate online risks, improve content output quality, purify the network environment, and ensure a good user experience.

Further, as a specific implementation of the method shown in FIG. 1 and FIG. 2, this embodiment provides an artificial intelligence-based sensitive word recognition device. As shown in FIG. 3, the device includes: an acquisition module 31, a recognition module 32, The judgment module 33 and the processing module 34.

The obtaining module 31 is used to obtain the text information to be recognized.

The recognition module 32 is configured to recognize a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one predetermined word slot.

The judgment module 33 is configured to judge whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information.

The processing module 34 is configured to perform filtering processing on the text information if it is determined that the text information contains sensitive words.

In a specific application scenario, the judgment module 33 is specifically configured to obtain a target verification rule combination according to at least one target verification rule corresponding to each target word slot combination; according to the target word slot combination in the text information The word slot arrangement information and the intermediate word information between the word slots in, respectively determine whether the text information meets the multiple preset sensitive word determination criteria in the target inspection rule combination; if the text information meets the At least one set of sensitive word determination criteria in the target inspection rule combination is determined to include the sensitive words; if none of the text information meets the determination criteria of each sensitive word in the target inspection rule combination, then the text information is determined Does not contain sensitive words.

In a specific application scenario, the judging module 33 is specifically used to determine if the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words is within a limited range. The criterion for determination is to determine the text information when it is determined that the word slot arrangement information matches the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is less than or equal to the preset number threshold. Contains sensitive words; if the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words does not meet the judgment criterion within a limited range, then the word is judged When the slot arrangement information conforms to the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is greater than or equal to the preset number threshold, it is determined that the text information contains sensitive words.

In a specific application scenario, the judging module 33 is specifically further configured to, if the target inspection rule combination includes at least one preset sensitive word determination criterion with different execution priorities, then according to each sensitive word in the target inspection rule combination. The execution priority of the word judgment standard is from high to low, and the text information is judged in sequence; in the process of sequential judgment, if it is determined that there is a sensitive word judgment standard that the text information meets, then the subsequent review of the text information is stopped. The text information is judged, and the currently obtained judgment result is used as the result of judging the text information by using the target inspection rule combination.

In specific application scenarios, the device also includes: a recording module and an analysis module.

The recording module is configured to record the text part of the text information containing the target word slot combination as sample data after the restriction processing is performed on the text information.

The analysis module is used to periodically analyze the recorded sample data, and count the word combinations in each sample data that appear more frequently than the preset frequency threshold and are different from the existing word slot combinations; Word combinations are calculated for semantic similarity with preset sensitive words and/or preset sensitive sentences; target word combinations with semantic similarity greater than the preset similarity threshold are used as a new word slot combination, and based on the new word slot combination The sample data of the word slot combination is updated with the verification rule corresponding to the new word slot combination; the new word slot combination and the corresponding verification rule are used to determine whether other text information contains sensitive words.

In a specific application scenario, the processing module 34 is specifically configured to prevent the publication of the text information; or, replace the text part of the text information containing the target word slot combination with preset characters before publishing; or , Send the text information to the review module for review, and publish it if the review is passed.

In a specific application scenario, optionally, the text information is pre-stored in the blockchain; correspondingly, the obtaining module 31 is specifically configured to obtain the text information from the blockchain; the recognition module 32, Specifically, it is used to clear the character spaces and preset special symbols in the text information; identify the target word slot combination included in the text information after removing the character spaces and the preset special symbols.

It should be noted that, for other corresponding descriptions of the functional units involved in the artificial intelligence-based sensitive word recognition device provided in this embodiment, reference may be made to the corresponding descriptions in FIG. 1 and FIG. 2, and details are not repeated here.

Based on the above-mentioned methods shown in Figures 1 and 2, correspondingly, this embodiment also provides a readable storage medium on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the foregoing Figure 1 and Figure 2 show the artificial intelligence-based sensitive word recognition method.

Optionally, the readable storage medium involved in this application may be a computer readable storage medium. Further optionally, the storage medium involved in this application, such as a readable storage medium, may be non-volatile, such as a non-volatile readable storage medium, or may be volatile, such as a volatile readable storage medium.

Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, U disk, mobile hard disk, etc.), including several The instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute the methods in each implementation scenario of the present application.

Based on the above-mentioned method shown in Fig. 1 and Fig. 2 and the virtual device embodiment shown in Fig. 3, in order to achieve the above-mentioned purpose, this embodiment also provides a computer device, which may be a personal computer, a notebook computer, or a server. , Network equipment, etc., the physical equipment includes a storage medium and a processor; the storage medium is used to store computer-readable instructions; the processor is used to execute computer-readable instructions to implement the aforementioned manual-based Intelligent method of identifying sensitive words. Optionally, the storage medium may be a readable storage medium, such as a nonvolatile readable storage medium or a volatile readable storage medium.

Optionally, the computer device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, and so on. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), etc., and the optional user interface may also include a USB interface, a card reader interface, and the like. The optional network interface can include standard wired interface, wireless interface (such as Bluetooth interface, WI-FI interface), etc.

Those skilled in the art can understand that the computer device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or arrange different components.

The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical devices, and supports the operation of information processing programs and other software and/or programs. The network communication module is used to realize the communication between the various components in the storage medium and the communication with other hardware and software in the physical device.

Through the description of the above implementation manners, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform, or can be implemented by hardware. By applying the technical solution of this embodiment, the target word slot combination contained in the text information can be identified. The word slot combination is composed of at least one preset word slot, and then the target word slot combination and the target word slot combination are used in the text information In the middle word information, judge whether the text information contains sensitive words. Compared with the current existing traditional filtering method for sensitive words, this embodiment uses the method of discriminating intermediate words between the word slot combination + word slot combination, even if symbols or spaces are added in the text sensitive words, or some words are added, Or the same semantic rewriting through other texts, etc., can accurately identify whether the text information contains sensitive words, which can improve the accuracy of sensitive word recognition. If it is determined that the text information contains sensitive words, the text information can also be restricted and processed in time. The entire process of sensitive word recognition + restriction processing can be automated, which improves the efficiency of sensitive word processing.

Those skilled in the art can understand that the accompanying drawings are only schematic diagrams of preferred implementation scenarios, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing this application. Those skilled in the art can understand that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the description of the implementation scenario, or can be changed to be located in one or more devices different from the implementation scenario. The modules of the above implementation scenarios can be combined into one module or further divided into multiple sub-modules.

The above serial number of this application is for description only, and does not represent the pros and cons of implementation scenarios. What has been disclosed above are only a few specific implementation scenarios of this application, but this application is not limited to these, and any changes that can be thought of by those skilled in the art should fall into the protection scope of this application.

Claims

A method for identifying sensitive words based on artificial intelligence, which includes:

Obtain the text information to be recognized;

Identifying a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one preset word slot;

Judging whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information;

If it is determined that the text information contains sensitive words, restrict processing is performed on the text information.
The method according to claim 1, wherein said determining whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information, specifically include:

Obtaining a target verification rule combination according to at least one target verification rule corresponding to each of the target word slot combinations;

According to the word slot arrangement information of the target word slot combination in the text information and the intermediate word information between word slots, it is determined whether the text information meets the multiple preset sensitivity in the target check rule combination. Word criterion;

If the text information meets at least one set of sensitive word determination criteria in the target inspection rule combination, it is determined that the text information contains sensitive words;

If none of the text information meets the criteria for determining each sensitive word in the target inspection rule combination, it is determined that the text information does not contain sensitive words.
2. The method according to claim 2, wherein the word slot arrangement information and the intermediate word information between the word slots combined in the text information according to the target word slot respectively determine whether the text information conforms to The multiple preset criteria for judging sensitive words in the target inspection rule combination specifically include:

If the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range meets the judgment standard, then it is determined that the word slot arrangement information meets all When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is less than or equal to the preset number threshold, it is determined that the text information contains sensitive words;

If the sensitive word determination criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range does not meet the criterion, then it is determined that the word slot arrangement information matches When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is greater than or equal to the preset number threshold, it is determined that the text information contains sensitive words.
2. The method according to claim 2, wherein if the target check rule combination includes at least one preset sensitive word determination criterion with different execution priorities, the combination according to the target word slot is used in the text information The word slot arrangement information and the intermediate word information between the word slots in, respectively determine whether the text information meets the multiple preset sensitive word determination criteria in the target inspection rule combination, which specifically includes:

Judging the text information in turn according to the order of execution priority of each sensitive word judgment criterion in the target inspection rule combination from high to low;

In the process of sequential judgment, if it is determined that there is a sensitive word judgment standard that the text information meets, then the subsequent judgment on the text information is stopped, and the currently obtained judgment result is used as the combination of the target inspection rules for the The result of the judgment on the text information.
The method according to claim 2, wherein, after the restricting processing of the text information, the method further comprises:

Recording the text part containing the target word slot combination in the text information as sample data;

Regularly analyze the recorded sample data, and count the word combinations that appear more frequently than the preset frequency threshold in each sample data and are different from the existing word slot combinations;

Calculating the semantic similarity between the word combination obtained by statistics and the preset sensitive word and/or the preset sensitive sentence;

The target word combination with semantic similarity greater than the preset similarity threshold is used as a new word slot combination, and the check corresponding to the new word slot combination is updated according to the sample data containing the new word slot combination rule;

Using the new word slot combination and corresponding inspection rules to determine whether other text information contains sensitive words.
The method according to claim 1, wherein the restricting processing of the text information specifically includes:

Prevent the publication of the text information; or,

Publish after replacing the text part of the text information containing the target word slot combination with preset characters; or,

The text information is sent to the review module for review, and if the review is passed, it will be released.
The method according to claim 1, wherein the text information is stored in a blockchain in advance;

The obtaining of the text information to be recognized specifically includes:

Obtain the text information from the blockchain;

The recognizing the target word slot combination contained in the text information specifically includes:

Clear character spaces and preset special symbols in the text information;

Identify the target word slot combination contained in the text information after clearing character spaces and preset special symbols.
An artificial intelligence-based sensitive word recognition device, which includes:

The obtaining module is used to obtain the text information to be recognized;

A recognition module, configured to recognize a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one preset word slot;

A judging module, configured to judge whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information;

The processing module is configured to perform restriction processing on the text information if it is determined that the text information contains sensitive words.
A readable storage medium having computer readable instructions stored thereon, wherein, when the computer readable instructions are executed by a processor, the following method is implemented:

Obtain the text information to be recognized;

Identifying a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one preset word slot;

Judging whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information;

If it is determined that the text information contains sensitive words, restrict processing is performed on the text information.
The readable storage medium according to claim 9, wherein said determining whether the text information contains sensitive information based on the target word slot combination and the intermediate word information of the target word slot combination in the text information When writing words, the concrete realization:

Obtaining a target verification rule combination according to at least one target verification rule corresponding to each of the target word slot combinations;

According to the word slot arrangement information of the target word slot combination in the text information and the intermediate word information between word slots, it is determined whether the text information meets the multiple preset sensitivity in the target check rule combination. Word criterion;

If the text information meets at least one set of sensitive word determination criteria in the target inspection rule combination, it is determined that the text information contains sensitive words;

If none of the text information meets the criteria for determining each sensitive word in the target inspection rule combination, it is determined that the text information does not contain sensitive words.
10. The readable storage medium according to claim 10, wherein the text is determined according to the word slot arrangement information and the intermediate word information between the word slots in the text information according to the target word slot combination. When the information meets the multiple preset sensitive word judgment criteria in the target inspection rule combination, the specific realization is as follows:

If the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range meets the judgment standard, then it is determined that the word slot arrangement information meets all When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is less than or equal to the preset number threshold, it is determined that the text information contains sensitive words;

If the sensitive word determination criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range does not meet the criterion, then it is determined that the word slot arrangement information matches When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is greater than or equal to the preset number threshold, it is determined that the text information contains sensitive words.
The readable storage medium according to claim 10, wherein if the target check rule combination includes at least one preset sensitive word determination criterion with different execution priorities, the combination according to the target word slot is When the word slot arrangement information in the text information and the intermediate word information between the word slots are used to determine whether the text information meets the multiple preset sensitive word determination criteria in the target inspection rule combination, the specific implementation is implemented:

Judging the text information in turn according to the order of execution priority of each sensitive word judgment criterion in the target inspection rule combination from high to low;

In the process of sequential judgment, if it is determined that there is a sensitive word judgment standard that the text information meets, then the subsequent judgment on the text information is stopped, and the currently obtained judgment result is used as the combination of the target inspection rules for the The result of the judgment on the text information.
10. The readable storage medium according to claim 10, wherein, after the restriction processing is performed on the text information, the computer readable instruction is further used to implement when being executed by a processor:

Recording the text part containing the target word slot combination in the text information as sample data;

Regularly analyze the recorded sample data, and count the word combinations that appear more frequently than the preset frequency threshold in each sample data and are different from the existing word slot combinations;

Calculating the semantic similarity between the word combination obtained by statistics and the preset sensitive word and/or the preset sensitive sentence;

The target word combination with semantic similarity greater than the preset similarity threshold is used as a new word slot combination, and the check corresponding to the new word slot combination is updated according to the sample data containing the new word slot combination rule;

Using the new word slot combination and corresponding inspection rules to determine whether other text information contains sensitive words.
The readable storage medium according to claim 9, wherein the text information is stored in a blockchain in advance;

When the text information to be recognized is obtained, the specific implementation is as follows:

Obtain the text information from the blockchain;

When recognizing the target word slot combination contained in the text information, it is specifically realized:

Clear character spaces and preset special symbols in the text information;

Identify the target word slot combination contained in the text information after clearing character spaces and preset special symbols.
A computer device, including a readable storage medium, a processor, and computer readable instructions stored on the readable storage medium and running on the processor, wherein the processor executes the computer readable instructions to achieve the following method:

Obtain the text information to be recognized;

Identifying a target word slot combination contained in the text information, wherein the target word slot combination is composed of at least one preset word slot;

Judging whether the text information contains sensitive words according to the target word slot combination and the intermediate word information of the target word slot combination in the text information;

If it is determined that the text information contains sensitive words, restrict processing is performed on the text information.
15. The computer device according to claim 15, wherein when determining whether the text information contains sensitive words based on the target word slot combination and the intermediate word information of the target word slot combination in the text information ,Implementation:

Obtaining a target verification rule combination according to at least one target verification rule corresponding to each of the target word slot combinations;

According to the word slot arrangement information of the target word slot combination in the text information and the intermediate word information between word slots, it is determined whether the text information meets the multiple preset sensitivity in the target check rule combination. Word criterion;

If the text information meets at least one set of sensitive word determination criteria in the target inspection rule combination, it is determined that the text information contains sensitive words;

If none of the text information meets the criteria for determining each sensitive word in the target inspection rule combination, it is determined that the text information does not contain sensitive words.
16. The computer device according to claim 16, wherein the word slot arrangement information combined in the text information according to the target word slot and the intermediate word information between the word slots determine whether the text information is When it meets the multiple preset criteria for determining sensitive words in the target inspection rule combination, the specific realization is as follows:

If the sensitive word judgment criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range meets the judgment standard, then it is determined that the word slot arrangement information meets all When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is less than or equal to the preset number threshold, it is determined that the text information contains sensitive words;

If the sensitive word determination criterion is that each word slot in the target word slot combination appears in the text information, and the number of intermediate words within a limited range does not meet the criterion, then it is determined that the word slot arrangement information matches When the preset word slot sequence corresponding to the target word slot combination, and the number of intermediate words is greater than or equal to the preset number threshold, it is determined that the text information contains sensitive words.
16. The computer device according to claim 16, wherein if the target check rule combination includes at least one preset sensitive word determination criterion with different execution priorities, then the combination according to the target word slot is used in the text When the word slot arrangement information and the intermediate word information between the word slots in the information respectively determine whether the text information meets the multiple preset sensitive word determination criteria in the target inspection rule combination, the specific realization is achieved:

Judging the text information in turn according to the order of execution priority of each sensitive word judgment criterion in the target inspection rule combination from high to low;

In the process of sequential judgment, if it is determined that there is a sensitive word judgment standard that the text information meets, then the subsequent judgment on the text information is stopped, and the currently obtained judgment result is used as the combination of the target inspection rules for the The result of the judgment on the text information.
The computer device according to claim 16, wherein, after the restriction processing is performed on the text information, when the processor executes the computer readable instruction, it is further configured to implement:

Recording the text part containing the target word slot combination in the text information as sample data;

Regularly analyze the recorded sample data, and count the word combinations that appear more frequently than the preset frequency threshold in each sample data and are different from the existing word slot combinations;

Calculating the semantic similarity between the word combination obtained by statistics and the preset sensitive word and/or the preset sensitive sentence;

The target word combination with semantic similarity greater than the preset similarity threshold is used as a new word slot combination, and the check corresponding to the new word slot combination is updated according to the sample data containing the new word slot combination rule;

Using the new word slot combination and corresponding inspection rules to determine whether other text information contains sensitive words.
The computer device according to claim 15, wherein the text information is stored in a blockchain in advance;

When the text information to be recognized is obtained, the specific implementation is as follows:

Obtain the text information from the blockchain;

When recognizing the target word slot combination contained in the text information, it is specifically realized:

Clear character spaces and preset special symbols in the text information;

Identify the target word slot combination contained in the text information after clearing character spaces and preset special symbols.