Risk identification method and device, electronic equipment and storage medium
By constructing the lexicon and calculating the relevance of public opinion text, the risk type of target objects is automatically identified, which solves the high cost problem caused by manual identification in the existing technology, and the automation and efficiency improvement of target objects risk identification is achieved.
Patent Information
- Application Number
- CN202411134383.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-12-31
AI Technical Summary
The existing target object risk identification methods require manual intervention, resulting in higher costs.
By obtaining the original public opinion text of the target object, building the target lexicon and risk tag lexicon, calculating the correlation between the public opinion subtext and negative keywords and risk tags, and automatically identifying the risk type of the target object.
The automation of risk identification of target objects has been achieved, reducing identification costs and improving efficiency.
Smart Images
Figure CN119226602A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of artificial intelligence technology, and in particular relates to a risk identification method, device, electronic device and storage medium. Background Art
[0002] At present, the risk identification method of some target objects (such as a brand, a company, an individual or a group, etc.) usually first obtains the public opinion text of the target object, such as news text, evaluation text, public opinion text, complaint text or report text about the target object, and then manually identifies the risk of the target object based on the obtained public opinion text. Since this method requires manual identification, the cost of risk identification of the target object is high. Summary of the invention
[0003] In view of this, the embodiments of the present application provide a risk identification method, device, electronic device and storage medium to solve the existing technical problem of high cost of risk identification for target objects.
[0004] In a first aspect, an embodiment of the present application provides a risk identification method comprising:
[0005] Obtain the original public opinion text corresponding to the target object;
[0006] For each original public opinion subtext in the original public opinion text, determining a first correlation between the original public opinion subtext and each negative keyword in the target word library;
[0007] Determining negative public opinion texts according to the first relevance of each of the original public opinion sub-texts;
[0008] For each negative public opinion subtext in the negative public opinion text, determining a second correlation degree between the negative public opinion subtext and each risk label in the risk label vocabulary;
[0009] Whether the target object has a risk and / or the risk type of the target object is determined according to the second correlation degree of each of the negative public opinion sub-texts.
[0010] Optionally, the target vocabulary is constructed in the following manner:
[0011] Obtain public opinion training text;
[0012] For each public opinion training sub-text in the public opinion training text, determining each first keyword of the public opinion training sub-text;
[0013] For each of the first keywords, determine the TF-IDF value of the first keyword, the frequency of occurrence of the first keyword in the public opinion training text, and the number of occurrences of the first keyword in the corresponding public opinion training sub-text, and determine the representativeness of the first keyword to the public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the first keyword is greater than a first preset representativeness threshold, the first keyword and the representativeness of the first keyword are associated and stored in the target vocabulary.
[0014] Optionally, the target vocabulary is updated in the following manner:
[0015] Obtaining the updated public opinion training text, and determining each second keyword in the updated public opinion training text;
[0016] For each of the second keywords, determining the solidification degree between the second keyword and the adjacent characters, if the solidification degree is greater than a first preset solidification degree threshold, determining the degree of freedom between the second keyword and the adjacent characters, if the degree of freedom is greater than the first preset degree of freedom threshold, and the target word library does not include the second keyword, then associating the second keyword with the representativeness corresponding to the second keyword and storing them in the target word library;
[0017] For each third keyword in the target vocabulary, the representativeness of the third keyword is recalculated according to the updated public opinion training text, and the representativeness of the third keyword in the target vocabulary and the third keyword is updated according to the recalculated representativeness.
[0018] Optionally, determining the first correlation between the original public opinion subtext and each negative keyword in the target vocabulary includes:
[0019] Determine each fourth keyword in the original public opinion subtext that hits the target word library, and determine the frequency of occurrence of each fourth keyword in the original public opinion subtext, the representativeness corresponding to each fourth keyword, and the number of the fourth keywords;
[0020] Determine the negative sentiment score of the original public opinion subtext, and determine the similarity between the original public opinion subtext and other original public opinion subtexts;
[0021] The first correlation of the original public opinion sub-text is determined based on the appearance frequency of each of the fourth keywords in the original public opinion sub-text, the representativeness corresponding to each of the fourth keywords, the number of the fourth keywords, the negative sentiment score of the original public opinion sub-text, and the similarity between the original public opinion sub-text and other original public opinion sub-texts.
[0022] Optionally, the risk label word library is constructed in the following manner:
[0023] Obtaining a negative public opinion training text; wherein the negative public opinion training text includes a plurality of negative public opinion training sub-texts marked with risk types;
[0024] For each of the negative public opinion training sub-texts, determining a fifth keyword in the negative public opinion training sub-text;
[0025] For each of the fifth keywords, determine the TF-IDF value of the fifth keyword, the frequency of occurrence of the fifth keyword in the negative public opinion training text, and the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text, and determine the representativeness of the fifth keyword to the negative public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the fifth keyword to the negative public opinion training text is greater than a second preset representativeness threshold, the fifth keyword, the representativeness of the fifth keyword to the negative public opinion training text, and the risk type corresponding to the fifth keyword are associated and stored in the risk label vocabulary.
[0026] Optionally, the risk label word library is updated in the following manner:
[0027] Obtaining the updated negative public opinion training text, and determining the sixth keyword in the updated negative public opinion training text;
[0028] For each of the sixth keywords, determining the solidification degree between the sixth keyword and the adjacent characters, if the solidification degree is greater than a second preset solidification degree threshold, determining the degree of freedom between the sixth keyword and the adjacent characters, if the degree of freedom is greater than the second preset degree of freedom threshold, and the risk label word library does not include the sixth keyword, then associating the sixth keyword, the representative degree corresponding to the sixth keyword, and the risk type corresponding to the sixth keyword and storing them in the risk label word library;
[0029] For each seventh keyword in the risk label vocabulary, the representativeness of the seventh keyword is recalculated according to the updated negative public opinion training text, and the representativeness of the seventh keyword and the seventh keyword in the risk label vocabulary are updated according to the recalculated representativeness.
[0030] Optionally, determining a second correlation between the negative public opinion subtext and each risk label in the risk label vocabulary includes:
[0031] Determine that the negative public opinion subtext hits each eighth keyword of the risk label vocabulary, and determine the frequency of occurrence of each eighth keyword in the negative public opinion subtext, the representativeness corresponding to each eighth keyword, and the number of the eighth keywords;
[0032] The second relevance of the negative public opinion sub-text is determined according to the appearance frequency of each of the eighth keywords in the negative public opinion sub-text, the representativeness corresponding to each of the eighth keywords, and the number of the eighth keywords.
[0033] In a second aspect, an embodiment of the present application provides a risk identification device, including:
[0034] An original public opinion text acquisition unit, used to acquire the original public opinion text corresponding to the target object;
[0035] A first correlation determination unit is used to determine, for each original public opinion subtext in the original public opinion text, a first correlation between the original public opinion subtext and each negative keyword in the target word library;
[0036] A negative public opinion text determination unit, configured to determine a negative public opinion text according to the first relevance of each of the original public opinion sub-texts;
[0037] A second correlation determination unit is used to determine, for each negative public opinion subtext in the negative public opinion text, a second correlation between the negative public opinion subtext and each risk label in the risk label vocabulary;
[0038] The risk determination unit is used to determine whether the target object has risks and / or the risk type of the target object according to the second correlation degree of each of the negative public opinion sub-texts.
[0039] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements each step of the risk identification method as described in any one of the first aspects above when executing the computer program.
[0040] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, each step in the risk identification method as described in any one of the first aspects above is implemented.
[0041] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product is run on an electronic device, the electronic device executes each step of the method described in the first aspect or any optional method of the first aspect.
[0042] The risk identification method, device, electronic device and storage medium provided in the embodiments of the present application have the following features:
[0043] Beneficial effects:
[0044] The risk identification method provided by the embodiment of the present application obtains the original public opinion text corresponding to the target object, and for each original public opinion sub-text in the original public opinion text, determines the first correlation between the original public opinion sub-text and each negative keyword in the target word library, and then determines the negative public opinion text according to the first correlation of each original public opinion sub-text, and then for each negative public opinion sub-text in the negative public opinion text, determines the second correlation between the negative public opinion sub-text and each risk label in the risk label word library, and finally determines whether the target object has risks and / or the risk type of the target object according to the second correlation of each negative public opinion sub-text. This method can automatically identify the risks of the target object without manual identification, thereby reducing the cost of risk identification for the target object. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0046] Figure 1 A flow chart of a risk identification method according to an embodiment of the present invention;
[0047] Figure 2 A schematic diagram of the structure of a risk identification device provided in an embodiment of the present application;
[0048] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0049] It should be noted that the terms used in the embodiments of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application. In the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two, and "at least one", "one or more" refers to one, two or more. The terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, it is defined that the "first" and "second" features can explicitly or implicitly include one or more of the features.
[0050] References to "one embodiment" or "some embodiments" etc. described in this specification mean that a particular feature, structure or characteristic described in conjunction with the embodiment is included in one or more embodiments of the present application. Thus, the phrases "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear at different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.
[0051] The execution subject of the risk identification method provided in the embodiment of the present application may be an electronic device (such as a risk identification device and a terminal device). The terminal device may be, for example, a mobile phone, a tablet computer, a laptop computer, a desktop computer, or other terminal devices.
[0052] The risk identification method provided in the embodiment of the present application can be applied to scenarios where risk identification of a target object is required. Wherein, as an example but not limitation, the target object may include a brand, a company, an individual or a group, etc. Based on this, the risk identification method provided in the embodiment of the present application can be applied to scenarios where risk identification of a brand, a company, an individual or a group is required. Specifically, when risk identification of a target object is required, each step of the risk identification method provided in the embodiment of the present application can be performed by an electronic device, so that risk identification of the target object can be automatically performed, thereby reducing the cost of risk identification of the target object.
[0053] See also Figure 1 , Figure 1 A flow chart of a risk identification method provided in an embodiment of the present application is provided. The risk identification method may include S101 to S105, which are described in detail as follows:
[0054] In S101, the original public opinion text corresponding to the target object is obtained.
[0055] In an embodiment of the present application, when a user needs to identify risks of a target object, the user can first obtain the original public opinion text corresponding to the target object through an electronic device.
[0056] For example, the target object may include but not be limited to a certain brand, a certain enterprise, a certain individual or a certain group; the original public opinion text may include but not be limited to news text, evaluation text, opinion text, complaint text or report text about the target object.
[0057] Specifically, the user can input the original public opinion text corresponding to the target object into the electronic device so that the electronic device can obtain the original public opinion text corresponding to the target object. In addition, the user can also train the electronic device in advance so that the electronic device can automatically obtain the original public opinion text corresponding to the target object.
[0058] In S102, for each original public opinion sub-text in the original public opinion text, a first correlation between the original public opinion sub-text and each negative keyword in the target vocabulary is determined.
[0059] In an embodiment of the present application, the original public opinion text may include several original public opinion sub-texts. The electronic device may determine, for each original public opinion sub-text in the original public opinion text, the first degree of correlation between the original public opinion sub-text and each negative keyword in the target vocabulary, thereby obtaining the first degree of correlation between each original public opinion sub-text and each negative keyword in the target vocabulary.
[0060] Among them, the target word library may store various negative keywords corresponding to a preset field. For example, the target object may be a certain brand, a certain enterprise, a certain live broadcaster or a certain live broadcast room, the preset field may be the live broadcast field, and the negative keywords may include "false propaganda, inconsistent with the actual situation, talking nonsense, acting and hype, counterfeit goods, fake goods, knockoff goods, passing off fake goods as genuine goods, counterfeit and shoddy goods, selling fake goods, and shoddy products".
[0061] In a possible implementation, the target vocabulary can be determined through steps a to c, as described in detail below:
[0062] In step a, public opinion training text is obtained.
[0063] In this implementation, the user can input the public opinion training text into the electronic device so that the electronic device can obtain the public opinion training text.
[0064] After acquiring the public opinion training text, the electronic device can perform data cleaning on the public opinion training text. For example, the electronic device can delete meaningless text data, special symbols, emoticons, HTML tags and other non-standard format text in the public opinion training text.
[0065] In step b, for each public opinion training sub-text in the public opinion training text, each first keyword of the public opinion training sub-text is determined.
[0066] In this implementation, the electronic device can perform operations such as word segmentation and stop word removal on each public opinion training sub-text in the public opinion training text, and identify the first keywords of the public opinion training sub-text through the TF-IDF algorithm.
[0067] In step c, for each first keyword, determine the TF-IDF value of the first keyword, the frequency of occurrence of the first keyword in the public opinion training text, and the number of occurrences of the first keyword in the corresponding public opinion training sub-text, and determine the representativeness of the first keyword to the public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the first keyword is greater than a first preset representativeness threshold, the first keyword and the representativeness of the first keyword are associated and stored in the target vocabulary.
[0068] In this implementation, after determining each first keyword of each public opinion training sub-text, the electronic device can determine the TF-IDF value of the first keyword for each first keyword through the TF-IDF algorithm, and determine the frequency of occurrence of the first keyword in the public opinion training text, and determine the number of occurrences of the first keyword in the corresponding public opinion training sub-text. Afterwards, the electronic device can determine the representativeness of the first keyword to the public opinion training text based on the determined TF-IDF value of the first keyword, the frequency of occurrence of the first keyword in the public opinion training text, and the number of occurrences of the first keyword in the corresponding public opinion training sub-text.
[0069] Exemplarily, the representativeness of the first keyword to the public opinion training text may be determined by the following formula:
[0070]
[0071] in, v i is the TF-IDF value of the first keyword, n ji is the number of occurrences of the first keyword in the corresponding public opinion training subtext, f i is the frequency of occurrence of the first keyword in the public opinion training text.
[0072] After determining the representativeness of the first keyword to the public opinion training text, the electronic device can associate the first keyword with the representativeness of the first keyword and store it in the target word library when the representativeness of the first keyword is greater than the first preset representativeness threshold. The specific value of the first preset representativeness threshold can be set according to actual needs and is not limited here. For example, the first preset representativeness threshold can be 0.65.
[0073] By executing the above method for each first keyword, the first keywords that meet the requirements can be stored in the target vocabulary, and then the target vocabulary is constructed.
[0074] After the target vocabulary is constructed, the target vocabulary can be updated, that is, new keywords in the target vocabulary can be obtained by obtaining new public opinion training texts.
[0075] In a possible implementation, the target word library may be updated through steps d to f, as detailed below:
[0076] In step d, the updated public opinion training text is obtained, and each second keyword in the updated public opinion training text is determined.
[0077] In this implementation, the electronic device can obtain the updated public opinion training text, and can obtain the second keywords in the updated public opinion training text by segmenting the updated public opinion training text and removing stop words, punctuation marks, numbers and other useless information.
[0078] In step e, for each second keyword, the degree of solidification between the second keyword and the adjacent characters is determined. If the degree of solidification is greater than a first preset degree of solidification threshold, the degree of freedom between the second keyword and the adjacent characters is determined. If the degree of freedom is greater than the first preset degree of freedom threshold, and the second keyword is not included in the target vocabulary, the second keyword and the corresponding representative degree are associated and stored in the target vocabulary.
[0079] In this implementation, the electronic device may determine the solidification degree between each second keyword and adjacent characters. Exemplarily, the solidification degree may be determined by the following formula:
[0080]
[0081] Where a represents the second keyword, b represents the characters adjacent to the second keyword, and P(ba i ) indicates the fragment ba i Frequency of occurrence; P(a i b) represents fragment a i b frequency; P(b) represents the frequency of occurrence of segment b; P(a i ) represents fragment a i Frequency of occurrence; i or a i The maximum character length of b is 5; P(x) represents the frequency of occurrence of segment x; count(x) represents the number of times segment x appears in the corpus, and length(T) represents the character length of the text corpus T.
[0082] After determining the solidification degree of the second keyword and the adjacent characters, if the solidification degree is greater than the first preset solidification degree threshold, the electronic device can determine the degree of freedom of the second keyword and the adjacent characters. The first preset solidification degree threshold can be set according to actual needs and is not limited here. For example, the first preset solidification degree threshold can be 1.5.
[0083] Optionally, the degree of freedom of the second keyword and the adjacent characters may be determined by the following formula:
[0084]
[0085] in, represents the probability that the character on the left side of segment y is z1; Indicates the probability that the character on the right side of segment y is z2; the combined length of segment y and the characters on the left and right sides does not exceed 5.
[0086] After determining the degree of freedom of the second keyword and the adjacent characters, if the degree of freedom is greater than the first preset degree of freedom threshold, the electronic device can determine whether the target word library includes the second keyword. If the target word library does not include the second keyword, the second keyword and the degree of representativeness corresponding to the second keyword are associated and stored in the target word library. The first preset degree of freedom threshold can be set according to actual needs and is not limited here. For example, the first preset degree of freedom threshold can be 3.
[0087] In step f, for each third keyword in the target vocabulary, the representativeness of the third keyword is recalculated according to the updated public opinion training text, and the representativeness of the third keyword in the target vocabulary and the third keyword is updated according to the recalculated representativeness.
[0088] In this implementation, after the target vocabulary is preliminarily updated (a part of the second keywords are added), the electronic device can recalculate the representativeness of each third keyword in the target vocabulary according to the updated public opinion training text, and update the third keyword in the target vocabulary and the representativeness of the third keyword according to the recalculated representativeness.
[0089] Specifically, for each third keyword, if the representativeness of the third keyword is less than the first preset representativeness threshold, the third keyword can be removed from the target word library. In addition, the recalculated representativeness can replace the original representativeness for storage.
[0090] The above describes the method for constructing and updating the target vocabulary, and the following describes the method for calculating the first relevance.
[0091] In a possible implementation manner, the first degree of association may be determined through steps g to i.
[0092] Details are as follows:
[0093] In step g, each fourth keyword that hits the target vocabulary in the original public opinion subtext is determined, and the occurrence frequency of each fourth keyword in the original public opinion subtext, the representativeness corresponding to each fourth keyword and the number of fourth keywords are determined.
[0094] In this implementation, the electronic device can determine, for each original public opinion sub-text in the original public opinion text, each fourth keyword in the original public opinion sub-text that hits the target vocabulary, and determine the frequency of occurrence of each fourth keyword in the original public opinion sub-text, the representativeness corresponding to each fourth keyword, and the number of fourth keywords.
[0095] In step h, the negative sentiment score of the original public opinion subtext is determined, and the similarity between the original public opinion subtext and other original public opinion subtexts is determined.
[0096] In this implementation, the electronic device can determine the negative sentiment score of the original public opinion sub-text through a text sentiment classifier, and can convert the text feature vector of the original public opinion sub-text into a semantic vector through a semantic matching model, and calculate the similarity between the original public opinion sub-text and other original public opinion sub-texts through cosine similarity.
[0097] In step i, the first correlation of the original public opinion sub-text is determined based on the occurrence frequency of each fourth keyword in the original public opinion sub-text, the representativeness of each fourth keyword, the number of fourth keywords, the negative sentiment score of the original public opinion sub-text, and the similarity between the original public opinion sub-text and other original public opinion sub-texts.
[0098] In this implementation, after determining the negative sentiment score of the original public opinion subtext and the similarity between the original public opinion subtext and other original public opinion subtexts, the electronic device can calculate the negative attention of the original public opinion subtext by the following formula:
[0099]
[0100] Among them, e i Indicates the negative sentiment score of the original public opinion sub-text; s ij represents the similarity between the original public opinion subtext and other original public opinion subtexts; α, φ, It is a hyperparameter with default values of 0.6, 0.7, and 5.
[0101] Afterwards, the electronic device can determine the first relevance of the original public opinion sub-text based on the appearance frequency of each fourth keyword in the original public opinion sub-text, the representativeness corresponding to each fourth keyword, the number of fourth keywords and the negative attention of the original public opinion sub-text.
[0102] Exemplarily, the first relevance of the original public opinion subtext may be determined by the following formula:
[0103]
[0104] Among them, w ig represents the gth fourth keyword in the original public opinion subtext; r ig represents the representativeness corresponding to the g-th fourth keyword; f ig represents the frequency of occurrence of the gth fourth keyword in the original public opinion subtext; n i Indicates the number of fourth keywords in the original public opinion sub-text that hit the target vocabulary.
[0105] The electronic device can determine the first relevance of the original public opinion sub-text through the above method. Based on this, the electronic device can determine the first relevance of each original public opinion sub-text.
[0106] In S103, negative public opinion texts are determined according to the first relevance of each original public opinion sub-text.
[0107] In an embodiment of the present application, after determining the first relevance of each original public opinion sub-text, the electronic device can determine the negative public opinion text according to the first relevance of each original public opinion sub-text.
[0108] Exemplarily, the electronic device may determine an original public opinion sub-text having a first correlation degree greater than a first correlation degree threshold as a negative public opinion sub-text in the negative public opinion text, thereby obtaining a negative public opinion text including a plurality of negative public opinion sub-texts.
[0109] Among them, the first correlation threshold can be set according to actual needs and is not limited here. For example, the first correlation threshold can be 0.85, that is, the electronic device can determine the original public opinion sub-text with a first correlation greater than 0.85 as a negative public opinion sub-text, thereby obtaining a negative public opinion text.
[0110] In S104, for each negative public opinion subtext in the negative public opinion text, a second correlation degree between the negative public opinion subtext and each risk label in the risk label vocabulary is determined.
[0111] In an embodiment of the present application, after determining the negative public opinion text, the electronic device can determine, for each negative public opinion sub-text in the negative public opinion text, a second degree of correlation between the negative public opinion sub-text and each risk label in the risk label vocabulary, thereby obtaining a second degree of correlation between each negative public opinion sub-text in the negative public opinion text and each risk label in the risk label vocabulary.
[0112] Among them, the risk label word library may store risk labels related to the preset field corresponding to the target word library. Exemplarily, the preset field may be the field of live streaming with goods, and the risk labels in the risk label word library may include "false propaganda for live streaming with goods", "live streaming with counterfeit goods", "live streaming with inferior products", etc.
[0113] In a possible implementation, the risk label word library can be determined through steps j to l, as described in detail as follows:
[0114] In step j, negative public opinion training text is obtained.
[0115] Among them, the negative public opinion training text includes several negative public opinion training sub-texts marked with risk types;
[0116] In this implementation, a user can input a negative public opinion training text including a plurality of negative public opinion training subtexts marked with risk types into an electronic device, so that the electronic device can obtain the negative public opinion training text. The risk types of the negative public opinion training subtexts can be obtained by manual marking.
[0117] After obtaining the negative public opinion training text, the electronic device can perform data cleaning on the negative public opinion training text. For example, the electronic device can delete meaningless text data, special symbols, emoticons, HTML tags and other non-standard format text in the negative public opinion training text.
[0118] In step k, for each negative public opinion training sub-text, the fifth keyword in the negative public opinion training sub-text is determined.
[0119] In this implementation, after obtaining the negative public opinion training text, the electronic device can perform word segmentation, remove stop words, and other operations on each negative public opinion training sub-text in the negative public opinion training text, and identify the fifth keywords of the negative public opinion training sub-text through the TF-IDF algorithm.
[0120] In step 1, for each fifth keyword, the TF-IDF value of the fifth keyword, the frequency of occurrence of the fifth keyword in the negative public opinion training text, and the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text are determined, and based on the TF-IDF value, the frequency of occurrence and the number of occurrences, the representativeness of the fifth keyword to the negative public opinion training text is determined; if the representativeness of the fifth keyword to the negative public opinion training text is greater than the second preset representativeness threshold, the fifth keyword, the representativeness of the fifth keyword to the negative public opinion training text, and the risk type corresponding to the fifth keyword are associated and stored in the risk label vocabulary.
[0121] In this implementation, after determining each fifth keyword of each negative public opinion training sub-text, the electronic device can determine the TF-IDF value of the fifth keyword through the TF-IDF algorithm for each fifth keyword, and determine the frequency of occurrence of the fifth keyword in the negative public opinion training text, and determine the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text. Afterwards, the electronic device can determine the representativeness of the fifth keyword to the negative public opinion training text based on the determined TF-IDF value of the fifth keyword, the frequency of occurrence of the fifth keyword in the negative public opinion training text, and the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text.
[0122] Exemplarily, the representativeness of the fifth keyword to the public opinion training text may be determined by the following formula:
[0123]
[0124] in, v i is the TF-IDF value of the fifth keyword, n ji is the number of occurrences of the fifth keyword in the corresponding negative public opinion training subtext, f i is the frequency of occurrence of the fifth keyword in the negative public opinion training text.
[0125] After determining the representativeness of the fifth keyword to the negative public opinion training text, the electronic device can associate and store the fifth keyword, the representativeness of the fifth keyword to the negative public opinion training text, and the risk type corresponding to the fifth keyword in the risk label word library when the representativeness of the fifth keyword is greater than the second preset representativeness threshold. The specific value of the second preset representativeness threshold can be set according to actual needs and is not limited here. For example, the second preset representativeness threshold can be 0.65.
[0126] By executing the above method for each fifth keyword, the fifth keyword that meets the requirements can be stored in the risk label vocabulary, and then the risk label vocabulary is constructed.
[0127] After the risk label vocabulary is constructed, the risk label vocabulary can be updated, that is, new keywords in the risk label vocabulary are obtained by obtaining new negative public opinion training texts.
[0128] In a possible implementation, the target word library may be updated through steps m to o, as described in detail as follows:
[0129] In step m, the updated negative public opinion training text is obtained, and the sixth keyword in the updated negative public opinion training text is determined.
[0130] In this implementation, the electronic device can obtain the updated negative public opinion training text, and can obtain each sixth keyword in the updated negative public opinion training text by segmenting the updated negative public opinion training text and removing stop words, punctuation marks, numbers and other useless information.
[0131] In step n, for each sixth keyword, the solidification degree of the sixth keyword and the adjacent characters is determined. If the solidification degree is greater than the second preset solidification degree threshold, the degree of freedom of the sixth keyword and the adjacent characters is determined. If the degree of freedom is greater than the second preset degree of freedom threshold, and the sixth keyword is not included in the risk label vocabulary, the sixth keyword, the representative degree corresponding to the sixth keyword, and the risk type corresponding to the sixth keyword are associated and stored in the risk label vocabulary.
[0132] In this implementation, the electronic device may determine the solidification degree between each sixth keyword and adjacent characters. Exemplarily, the solidification degree may be determined by the following formula:
[0133]
[0134] Wherein, a represents the sixth keyword, b represents the character adjacent to the sixth keyword, and P(ba i ) indicates the fragment ba i Frequency of occurrence; P(a i b) represents fragment a i b frequency; P(b) represents the frequency of occurrence of segment b; P(a i ) represents fragment a i Frequency of occurrence; i or a i The maximum character length of b is 5; P(x) represents the frequency of occurrence of segment x; count(x) represents the number of times segment x appears in the corpus, and length(T) represents the character length of the text corpus T.
[0135] After determining the freezing degree of the sixth keyword and the adjacent characters, if the freezing degree is greater than the second preset freezing degree threshold, the electronic device can determine the degree of freedom of the sixth keyword and the adjacent characters. The second preset freezing degree threshold can be set according to actual needs and is not limited here. For example, the second preset freezing degree threshold can be 1.5.
[0136] Optionally, the degree of freedom of the sixth keyword and the adjacent characters may be determined by the following formula:
[0137]
[0138] in, represents the probability that the character on the left side of segment y is z1; Indicates the probability that the character on the right side of segment y is z2; the combined length of segment y and the characters on the left and right sides does not exceed 5.
[0139] After determining the degrees of freedom of the sixth keyword and the adjacent characters, if the degrees of freedom are greater than the second preset degrees of freedom threshold, the electronic device can determine whether the risk label word library includes the sixth keyword. If the risk label word library does not include the sixth keyword, the sixth keyword, the representative degree corresponding to the sixth keyword, and the risk type corresponding to the sixth keyword are associated and stored in the risk label word library. The second preset degrees of freedom threshold can be set according to actual needs and is not limited here. For example, the second preset degrees of freedom threshold can be 3.
[0140] In step o, for each seventh keyword in the risk label vocabulary, the representativeness of the seventh keyword is recalculated according to the updated negative public opinion training text, and the representativeness of the seventh keyword and the seventh keyword in the risk label vocabulary are updated according to the recalculated representativeness.
[0141] In this implementation, after the risk label vocabulary is preliminarily updated (a part of the sixth keyword is added), the electronic device can recalculate the representativeness of each seventh keyword in the risk label vocabulary according to the updated negative public opinion training text, and update the representativeness of the seventh keyword and the sixth keyword in the risk label vocabulary according to the recalculated representativeness.
[0142] Specifically, for each seventh keyword, if the representativeness of the seventh keyword is less than the second preset representativeness threshold, the seventh keyword can be removed from the risk tag word library. In addition, the recalculated representativeness can replace the original representativeness for storage.
[0143] The above describes the construction and update method of the risk label word library. The following describes the method for calculating the second correlation degree.
[0144] In a possible implementation manner, the first degree of association may be determined through steps p to q.
[0145] Details are as follows:
[0146] In step p, determine each eighth keyword of the risk label vocabulary that the negative public opinion sub-text hits, and determine the occurrence frequency of each eighth keyword in the negative public opinion sub-text, the representative degree corresponding to each eighth keyword, and the number of eighth keywords.
[0147] In this implementation, the electronic device can determine, for each negative public opinion sub-text in the negative public opinion text, each eighth keyword in the risk label vocabulary that the negative public opinion sub-text hits, and determine the frequency of occurrence of each eighth keyword in the negative public opinion sub-text, the representativeness corresponding to each eighth keyword, and the number of eighth keywords.
[0148] In step q, the second relevance of the negative public opinion subtext is determined according to the occurrence frequency of each eighth keyword in the negative public opinion subtext, the representativeness corresponding to each eighth keyword and the number of eighth keywords.
[0149] In this implementation, after the electronic device determines the occurrence frequency of each eighth keyword in the negative public opinion sub-text, the representativeness corresponding to each eighth keyword and the number of eighth keywords, it can calculate the second correlation degree of the negative public opinion sub-text by the following formula.
[0150]
[0151] in, Indicates the gth eighth keyword in the negative public opinion subtext; represents the representativeness corresponding to the g-th eighth keyword; represents the frequency of occurrence of the g-th eighth keyword in the negative public opinion subtext; Indicates the number of the eighth keyword.
[0152] The electronic device can determine the second degree of association between the negative public opinion sub-text and each risk label in the risk label vocabulary through the above method. Based on this, the electronic device can determine the second degree of association between each negative public opinion sub-text and each risk label in the risk label vocabulary.
[0153] In S105, whether the target object has risk and / or the risk type of the target object is determined according to the second correlation degree of each negative public opinion subtext.
[0154] In an embodiment of the present application, after determining the second degree of association between each negative public opinion sub-text and each risk label in the risk label vocabulary, the electronic device can determine whether the target object is at risk and / or the risk type of the target object based on the second degree of association between each negative public opinion sub-text and each risk label in the risk label vocabulary.
[0155] Exemplarily, the electronic device can determine whether there is any target negative public opinion sub-text whose second correlation degree is greater than the second correlation degree threshold. If so, the electronic device can determine that there is a risk in the target object. In addition, the electronic device can also determine the risk label corresponding to the second correlation degree greater than the second correlation degree threshold as the target risk label, and can determine the risk type of the target object as the target risk label.
[0156] Exemplarily, if the second correlation between the negative public opinion sub-text A and the risk label a in the risk label vocabulary is greater than the second correlation threshold, the second correlation between the negative public opinion sub-text A and the risk label b in the risk label vocabulary is less than the second correlation threshold, and the second correlation between the negative public opinion sub-text A and the risk label c in the risk label vocabulary is less than the second correlation threshold, it can be determined that there is a risk in the target object, and the risk type of the target object can be determined to be risk label a.
[0157] Exemplarily, if the second correlation between the negative public opinion sub-text A and the risk label d in the risk label vocabulary is less than the second correlation threshold, the second correlation between the negative public opinion sub-text A and the risk label e in the risk label vocabulary is less than the second correlation threshold, and the second correlation between the negative public opinion sub-text A and the risk label f in the risk label vocabulary is less than the second correlation threshold, it can be determined that there is no risk for the target object.
[0158] The second correlation threshold may be set according to actual needs and is not limited here. Exemplarily, the second correlation threshold may be 0.85.
[0159] In a possible implementation, after determining whether the target object has a risk and / or the risk type of the target object, the electronic device may output to the user the result of whether the target object has a risk and / or the risk type of the target object.
[0160] From the above, it can be seen that the risk identification method provided by the embodiment of the present application obtains the original public opinion text corresponding to the target object, and for each original public opinion sub-text in the original public opinion text, determines the first correlation between the original public opinion sub-text and each negative keyword in the target vocabulary, and then determines the negative public opinion text based on the first correlation of each original public opinion sub-text, and then for each negative public opinion sub-text in the negative public opinion text, determines the second correlation between the negative public opinion sub-text and each risk label in the risk label vocabulary, and finally determines whether the target object has risks and / or the risk type of the target object based on the second correlation of each negative public opinion sub-text. This method can automatically identify the risks of the target object without manual identification, thereby reducing the cost of risk identification for the target object.
[0161] Based on the risk identification method provided in the above embodiment, the present application further provides a risk identification device for implementing the above method embodiment. Figure 2 , Figure 2 This is a schematic diagram of the structure of a risk identification device provided in an embodiment of the present application. Figure 2 As shown, the risk identification device 20 may include an original public opinion text acquisition unit 21, a first relevance determination unit 22, a negative public opinion text determination unit 23, a second relevance determination unit 24 and a risk determination unit 25. Among them:
[0162] The original public opinion text acquisition unit 21 is used to acquire the original public opinion text corresponding to the target object.
[0163] The first relevance determination unit 22 is used to determine, for each original public opinion sub-text in the original public opinion text, a first relevance between the original public opinion sub-text and each negative keyword in the target vocabulary.
[0164] The negative public opinion text determination unit 23 is used to determine the negative public opinion text according to the first relevance of each original public opinion sub-text.
[0165] The second correlation determination unit 24 is used to determine, for each negative public opinion subtext in the negative public opinion text, a second correlation between the negative public opinion subtext and each risk label in the risk label vocabulary.
[0166] The risk determination unit 25 is used to determine whether the target object has a risk and / or the risk type of the target object according to the second relevance of each negative public opinion subtext.
[0167] Optionally, the risk identification device 20 may further include a target word library construction unit.
[0168] The target vocabulary building unit is specifically used for:
[0169] Obtain public opinion training text;
[0170] For each public opinion training sub-text in the public opinion training text, determining each first keyword of the public opinion training sub-text;
[0171] For each first keyword, determine the TF-IDF value of the first keyword, the frequency of occurrence of the first keyword in the public opinion training text, and the number of occurrences of the first keyword in the corresponding public opinion training sub-text, and determine the representativeness of the first keyword to the public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the first keyword is greater than a first preset representativeness threshold, the first keyword and the representativeness of the first keyword are associated and stored in the target vocabulary.
[0172] Optionally, the risk identification device 20 may further include a target word library updating unit.
[0173] The target vocabulary updating unit is specifically used for:
[0174] Obtaining an updated public opinion training text, and determining each second keyword in the updated public opinion training text;
[0175] For each second keyword, determine the solidification degree of the second keyword and the adjacent characters, if the solidification degree is greater than a first preset solidification degree threshold, determine the degree of freedom of the second keyword and the adjacent characters, if the degree of freedom is greater than the first preset degree of freedom threshold, and the target word library does not include the second keyword, associate the second keyword with the representative degree corresponding to the second keyword and store them in the target word library;
[0176] For each third keyword in the target vocabulary, the representativeness of the third keyword is recalculated according to the updated public opinion training text, and the third keyword in the target vocabulary and the representativeness of the third keyword are updated according to the recalculated representativeness.
[0177] Optionally, the first correlation determination unit 22 is specifically configured to:
[0178] Determine each fourth keyword in the original public opinion subtext that hits the target word library, and determine the frequency of occurrence of each fourth keyword in the original public opinion subtext, the representativeness corresponding to each fourth keyword, and the number of fourth keywords;
[0179] Determine the negative sentiment score of the original public opinion subtext, and determine the similarity between the original public opinion subtext and other original public opinion subtexts;
[0180] The first correlation of the original public opinion sub-text is determined based on the appearance frequency of each fourth keyword in the original public opinion sub-text, the representativeness corresponding to each fourth keyword, the number of fourth keywords, the negative sentiment score of the original public opinion sub-text, and the similarity between the original public opinion sub-text and other original public opinion sub-texts.
[0181] Optionally, the risk identification device 20 may further include a risk label word library construction unit.
[0182] The risk label word library construction unit is specifically used for:
[0183] Obtaining a negative public opinion training text; wherein the negative public opinion training text includes a plurality of negative public opinion training sub-texts marked with risk types;
[0184] For each negative public opinion training sub-text, determining the fifth keyword in the negative public opinion training sub-text;
[0185] For each fifth keyword, determine the TF-IDF value of the fifth keyword, the frequency of occurrence of the fifth keyword in the negative public opinion training text, and the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text, and determine the representativeness of the fifth keyword to the negative public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the fifth keyword to the negative public opinion training text is greater than the second preset representativeness threshold, the fifth keyword, the representativeness of the fifth keyword to the negative public opinion training text, and the risk type corresponding to the fifth keyword are associated and stored in the risk label vocabulary.
[0186] Optionally, the risk identification device 20 may further include a risk label word library updating unit.
[0187] The risk label word library update unit is specifically used for:
[0188] Obtaining an updated negative public opinion training text, and determining a sixth keyword in the updated negative public opinion training text;
[0189] For each sixth keyword, determine the solidification degree of the sixth keyword and the adjacent characters, if the solidification degree is greater than a second preset solidification degree threshold, determine the degree of freedom of the sixth keyword and the adjacent characters, if the degree of freedom is greater than the second preset degree of freedom threshold, and the risk label word library does not include the sixth keyword, then associate the sixth keyword, the representative degree corresponding to the sixth keyword, and the risk type corresponding to the sixth keyword and store them in the risk label word library;
[0190] For each seventh keyword in the risk label vocabulary, the representativeness of the seventh keyword is recalculated according to the updated negative public opinion training text, and the seventh keyword in the risk label vocabulary and the representativeness of the seventh keyword are updated according to the recalculated representativeness.
[0191] Optionally, the second relevance determination unit 24 is specifically configured to:
[0192] Determine each eighth keyword of the risk label vocabulary that the negative public opinion subtext hits, and determine the frequency of occurrence of each eighth keyword in the negative public opinion subtext, the representativeness corresponding to each eighth keyword, and the number of eighth keywords;
[0193] The second relevance of the negative public opinion sub-text is determined according to the appearance frequency of each eighth keyword in the negative public opinion sub-text, the representativeness corresponding to each eighth keyword and the number of eighth keywords.
[0194] It should be noted that the information interaction, execution process and other contents between the above-mentioned units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be specifically referred to the method embodiment part and will not be repeated here.
[0195] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 3 As shown, the electronic device 3 provided in this embodiment may include: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 80. For example, a program corresponding to risk identification. When the processor 30 executes the computer program 32, the steps applied in the risk identification embodiment are implemented, for example Figure 1 Alternatively, when the processor 30 executes the computer program 32, the functions of each module / unit in the embodiment corresponding to the risk identification device 20 are realized, for example Figure 2 The functions of the units 21 to 26 are shown.
[0196] Exemplarily, the computer program 32 may be divided into one or more modules / units, one or more modules / units are stored in the memory 31 and executed by the processor 30 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program 32 in the electronic device 3. For example, the computer program 32 may be divided into an original public opinion text acquisition unit 21, a first relevance determination unit 22, a negative public opinion text determination unit 23, a second relevance determination unit 24, and a risk determination unit 25. For the specific functions of each unit, please refer to Figure 2 The relevant descriptions in the corresponding embodiments are not repeated here.
[0197] Those skilled in the art will understand that Figure 3 This is only an example of the electronic device 3 and does not constitute a limitation on the electronic device 3 , which may include more or less components than those shown in the figure, or a combination of certain components, or different components.
[0198] The processor 30 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.
[0199] The memory 31 may be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. The memory 31 may also be an external storage device of the electronic device 3, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, or a flash card, etc., equipped on the electronic device 3. Further, the memory 31 may also include both an internal storage unit and an external storage device of the electronic device 3. The memory 31 is used to store computer programs and other programs and data required by the electronic device. The memory 31 may also be used to temporarily store data that has been output or is to be output.
[0200] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional units as needed, that is, the internal structure of the risk identification device can be divided into different functional units to complete all or part of the functions described above. The functional units in the embodiment can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.
[0201] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.
[0202] An embodiment of the present application provides a computer program product. When the computer program product is executed on an electronic device, the electronic device implements the steps in the above-mentioned method embodiments.
[0203] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0204] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0205] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A risk identification method, characterized in that: include: Obtain the original public opinion text corresponding to the target object; For each original public opinion subtext in the original public opinion text, determining a first correlation between the original public opinion subtext and each negative keyword in the target word library; Determining negative public opinion texts according to the first relevance of each of the original public opinion sub-texts; For each negative public opinion subtext in the negative public opinion text, determining a second correlation degree between the negative public opinion subtext and each risk label in the risk label vocabulary; Whether the target object has a risk and / or the risk type of the target object is determined according to the second correlation degree of each of the negative public opinion sub-texts.
2. The method according to claim 1, characterized in that: The target vocabulary is constructed in the following way: Obtain public opinion training text; For each public opinion training sub-text in the public opinion training text, determining each first keyword of the public opinion training sub-text; For each of the first keywords, determine the TF-IDF value of the first keyword, the frequency of occurrence of the first keyword in the public opinion training text, and the number of occurrences of the first keyword in the corresponding public opinion training sub-text, and determine the representativeness of the first keyword to the public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the first keyword is greater than a first preset representativeness threshold, the first keyword and the representativeness of the first keyword are associated and stored in the target vocabulary.
3. The method according to claim 2, characterized in that The target word library is updated in the following way: Obtaining the updated public opinion training text, and determining each second keyword in the updated public opinion training text; For each of the second keywords, determining the solidification degree between the second keyword and the adjacent characters, if the solidification degree is greater than a first preset solidification degree threshold, determining the degree of freedom between the second keyword and the adjacent characters, if the degree of freedom is greater than the first preset degree of freedom threshold, and the target word library does not include the second keyword, then associating the second keyword with the representativeness corresponding to the second keyword and storing them in the target word library; For each third keyword in the target vocabulary, the representativeness of the third keyword is recalculated according to the updated public opinion training text, and the representativeness of the third keyword in the target vocabulary and the third keyword is updated according to the recalculated representativeness.
4. The method according to claim 2 or 3, characterized in that: The step of determining the first correlation between the original public opinion subtext and each negative keyword in the target word library includes: Determine each fourth keyword in the original public opinion subtext that hits the target word library, and determine the frequency of occurrence of each fourth keyword in the original public opinion subtext, the representativeness corresponding to each fourth keyword, and the number of the fourth keywords; Determine the negative sentiment score of the original public opinion subtext, and determine the similarity between the original public opinion subtext and other original public opinion subtexts; The first correlation of the original public opinion sub-text is determined based on the appearance frequency of each of the fourth keywords in the original public opinion sub-text, the representativeness corresponding to each of the fourth keywords, the number of the fourth keywords, the negative sentiment score of the original public opinion sub-text, and the similarity between the original public opinion sub-text and other original public opinion sub-texts.
5. The method according to claim 1, characterized in that The risk label word library is constructed in the following way: Obtaining a negative public opinion training text; wherein the negative public opinion training text includes a plurality of negative public opinion training sub-texts marked with risk types; For each of the negative public opinion training sub-texts, determining a fifth keyword in the negative public opinion training sub-text; For each of the fifth keywords, determine the TF-IDF value of the fifth keyword, the frequency of occurrence of the fifth keyword in the negative public opinion training text, and the number of occurrences of the fifth keyword in the corresponding negative public opinion training sub-text, and determine the representativeness of the fifth keyword to the negative public opinion training text based on the TF-IDF value, the frequency of occurrence, and the number of occurrences. If the representativeness of the fifth keyword to the negative public opinion training text is greater than a second preset representativeness threshold, the fifth keyword, the representativeness of the fifth keyword to the negative public opinion training text, and the risk type corresponding to the fifth keyword are associated and stored in the risk label vocabulary.
6. The method according to claim 5, characterized in that The risk label word library is updated in the following way: Obtaining the updated negative public opinion training text, and determining the sixth keyword in the updated negative public opinion training text; For each of the sixth keywords, determining the solidification degree between the sixth keyword and the adjacent characters, if the solidification degree is greater than a second preset solidification degree threshold, determining the degree of freedom between the sixth keyword and the adjacent characters, if the degree of freedom is greater than the second preset degree of freedom threshold, and the risk label word library does not include the sixth keyword, then associating the sixth keyword, the representative degree corresponding to the sixth keyword, and the risk type corresponding to the sixth keyword and storing them in the risk label word library; For each seventh keyword in the risk label vocabulary, the representativeness of the seventh keyword is recalculated according to the updated negative public opinion training text, and the representativeness of the seventh keyword and the seventh keyword in the risk label vocabulary are updated according to the recalculated representativeness.
7. The method according to claim 5 or 6, characterized in that: The determining of the second correlation between the negative public opinion subtext and each risk label in the risk label vocabulary includes: Determine that the negative public opinion subtext hits each eighth keyword of the risk label vocabulary, and determine the frequency of occurrence of each eighth keyword in the negative public opinion subtext, the representativeness corresponding to each eighth keyword, and the number of the eighth keywords; The second relevance of the negative public opinion sub-text is determined according to the appearance frequency of each of the eighth keywords in the negative public opinion sub-text, the representativeness corresponding to each of the eighth keywords, and the number of the eighth keywords.
8. A risk identification device, characterized in that: include: An original public opinion text acquisition unit, used to acquire the original public opinion text corresponding to the target object; A first correlation determination unit is used to determine, for each original public opinion subtext in the original public opinion text, a first correlation between the original public opinion subtext and each negative keyword in the target word library; A negative public opinion text determination unit, configured to determine a negative public opinion text according to the first relevance of each of the original public opinion sub-texts; A second correlation determination unit is used to determine, for each negative public opinion subtext in the negative public opinion text, a second correlation between the negative public opinion subtext and each risk label in the risk label vocabulary; The risk determination unit is used to determine whether the target object has risks and / or the risk type of the target object according to the second correlation degree of each of the negative public opinion sub-texts.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the risk identification method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, each step of the risk identification method according to any one of claims 1 to 7 is implemented.