Public opinion risk identification method, device, electronic device and storage medium

Abstract

The present disclosure provides a method for identifying public opinion risks, which can be applied to the fields of artificial intelligence and financial technology. The method includes: using a risk identification model to process a target identification text, identifying the risk category of the target identification text, the first risk entity, and the first category of the first risk entity; when the risk category is risky, calling the entity vocabulary and the risk keyword library, performing entity matching on the target identification text, and obtaining the second risk entity of the target identification text and the second category of the second risk entity; in the case where the first risk entity includes multiple entities, calling the entity disambiguation database, performing entity disambiguation on multiple first risk entities, and obtaining the third risk entity of the target identification text and the third category of the third risk entity; entity fusion of the second risk entity and the second category with the third risk entity and the third category, and obtaining the target risk entity and target category of the target identification text. The present disclosure also provides a device, equipment, medium, and program product.

Description

Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence and financial technology, and specifically to a method, device, electronic device, storage medium and program product for identifying public opinion risks. Background Art

[0002] Quickly and accurately identifying financial text content involving risk public opinion from massive financial news texts, and extracting the event subjects corresponding to the risk public opinion, is of great significance to the risk control departments, risk management teams and their branches of some financial institutions. It can assist staff to more effectively carry out risk prediction, risk level assessment, risk avoidance and other tasks.

[0003] At present, the method for identifying risk public opinion subjects generally first identifies the public opinion risk entity through entity recognition, and then uses classification methods to identify the risk category corresponding to the text containing the entity. However, this method is prone to accumulating errors and reducing the accuracy of intelligent recognition. Summary of the invention

[0004] In view of the above problems, the present disclosure provides a method, apparatus, device, storage medium and program product for identifying public opinion risks.

[0005] According to a first aspect of the present disclosure, a method for identifying public opinion risks is provided, including: processing a target identification text with a risk identification model to identify the risk category of the target identification text, a first risk entity in the target identification text, and a first category of the first risk entity; when the risk category is risky, calling an entity vocabulary and a risk keyword library to perform entity matching on the target identification text to obtain a second risk entity of the target identification text and a second category of the second risk entity; when the first risk entity includes multiple entities, calling an entity disambiguation database to perform entity disambiguation on the multiple first risk entities to obtain a third risk entity of the target identification text and a third category of the third risk entity; and entity merging the second risk entity and the second category with the third risk entity and the third category to obtain a target risk entity of the target identification text and a target category of the target risk entity.

[0006] According to an embodiment of the present disclosure, a target recognition text is processed using a risk identification model to identify the risk category of the target recognition text, the first risk entity in the target recognition text, and the first category of the first risk entity, including: preprocessing the target recognition text to obtain a plurality of target recognition text units for the target recognition text; performing feature extraction on the target recognition text unit data of the plurality of target recognition text units to obtain a feature vector sequence; performing entity recognition on the feature vector sequence to identify the risk category of the target recognition text, the first risk entity in the target recognition text, and the first category of the first risk entity.

[0007] According to an embodiment of the present disclosure, feature extraction is performed on target recognition text unit data of multiple target recognition text units to obtain a feature vector sequence, including: lexical feature extraction is performed on each target recognition text unit data to obtain a lexical feature vector sequence; using a position encoder, position feature extraction is performed on each target recognition text unit data to obtain a position feature vector sequence; based on the lexical feature vector sequence and the position feature vector sequence, a feature vector sequence is obtained.

[0008] According to an embodiment of the present disclosure, the risk identification model is based on the JointBERT language representation model.

[0009] According to an embodiment of the present disclosure, in the case where the risk category is risky, an entity vocabulary and a risk keyword library are called to perform entity matching on the target recognition text to obtain a second risk entity of the target recognition text and a second category of the second risk entity, including: in response to a matching request for a corresponding entity of the target recognition text, the entity vocabulary library is called to match the corresponding entity of the target recognition text to obtain the corresponding entity of the target recognition text; in response to a matching request for risk keywords of the target recognition text, the risk keyword library is called to perform risk keyword matching on the risk keywords of the target recognition text to obtain a risk event category associated with the risk keywords; according to the corresponding entity and the risk event category of the target recognition text, the second risk entity of the target recognition text and the second category of the second risk entity are obtained.

[0010] According to an embodiment of the present disclosure, in the case where the first risk entity includes multiple entities, an entity disambiguation database is called to perform entity disambiguation on the multiple first risk entities to obtain a third risk entity of the target recognition text and a third category of the third risk entity, including: when it is determined that the multiple first risk entities are in the same position in the target recognition text, in response to a request for entity disambiguation on the multiple first risk entities, the entity disambiguation database is called; based on the data feature structure between the multiple first risk entities, a regular expression matching the data feature structure is extracted from the entity disambiguation data; based on the regular expression, the multiple first risk entities are entity disambiguated to obtain the third risk entity of the target recognition text and the third category of the third risk entity.

[0011] According to an embodiment of the present disclosure, the second risk entity and the second category are entity fused with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text, including: performing a union operation on the second risk entity and the second category with the third risk entity and the third category, so as to entity fuse the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text.

[0012] The second aspect of the present disclosure provides a public opinion risk identification device, including: an identification module, which is used to use a risk identification model to process a target identification text, identify the risk category of the target identification text, the first risk entity in the target identification text, and the first category of the first risk entity; an entity matching module, which is used to call an entity vocabulary and a risk keyword library to perform entity matching on the target identification text when the risk category is risky, so as to obtain a second risk entity of the target identification text and a second category of the second risk entity; an entity disambiguation module, which is used to call an entity disambiguation database when the first risk entity includes multiple entities, perform entity disambiguation on multiple first risk entities, and obtain a third risk entity of the target identification text and a third category of the third risk entity; an entity fusion module, which performs entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain a target risk entity of the target identification text and a target category of the target risk entity.

[0013] The third aspect of the present disclosure provides an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned public opinion risk identification method.

[0014] The fourth aspect of the present disclosure also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above-mentioned public opinion risk identification method.

[0015] The fifth aspect of the present disclosure also provides a computer program product, including a computer program, which implements the above-mentioned public opinion risk identification method when executed by a processor.

[0016] According to the embodiments of the present disclosure, by utilizing a risk identification model to simultaneously identify the risk category and entity category of the target identification text, and then combining entity matching and entity disambiguation technology to perform entity fusion analysis on the above identification results, the model cumulative error is reduced, the recognition effect is improved, and it is further helpful for business personnel to predict and avoid risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0018] Figure 1 A diagram schematically shows an application scenario of the method and device for identifying public opinion risks according to an embodiment of the present disclosure;

[0019] Figure 2 A flowchart of a method for identifying public opinion risks according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 3 A flowchart schematically shows a method for processing a target identification text by using a risk identification model to identify a risk category of the target identification text, a first risk entity, and a first category of the first risk entity according to an embodiment of the present disclosure;

[0021] Figure 4 A flowchart of a method for entity disambiguation of a plurality of first risk entities according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 5 A schematic diagram of a method for identifying public opinion risks using a risk identification model according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 6 A schematic diagram of a method for entity fusion according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 7 A schematic diagram shows a structural block diagram of a public opinion risk identification device according to an embodiment of the present disclosure; and

[0025] Figure 8 A block diagram of an electronic device suitable for implementing the public opinion risk method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0026] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0027] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0028] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0029] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0030] In the technical solution disclosed in the present invention, the acquisition, collection, storage, use, processing, transmission, provision, disclosure and application of data are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and do not violate public order and good morals.

[0031] In the technical solution of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0032] An embodiment of the present disclosure provides a method for identifying public opinion risks, including: using a risk identification model to process a target identification text, identifying the risk category of the target identification text, a first risk entity in the target identification text, and a first category of the first risk entity; when the risk category is risky, calling an entity vocabulary and a risk keyword library, and performing entity matching on the target identification text to obtain a second risk entity of the target identification text and a second category of the second risk entity; when the first risk entity includes multiple entities, calling an entity disambiguation database, and performing entity disambiguation on the multiple first risk entities to obtain a third risk entity of the target identification text and a third category of the third risk entity; and performing entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain a target risk entity of the target identification text and a target category of the target risk entity.

[0033] Figure 1 The application scenario diagram of the public opinion risk identification method and device according to the embodiment of the present disclosure is schematically shown.

[0034] like Figure 1As shown, the application scenario 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0035] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0036] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0037] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for websites browsed by users using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0038] It should be noted that the public opinion risk identification method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the public opinion risk identification device provided in the embodiment of the present disclosure can generally be set in the server 105. The public opinion risk identification method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the public opinion risk identification device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0039] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0040] The following will be based on Figure 1 The scene described by Figure 2 to Figure 6 The public opinion risk identification method of the disclosed embodiment is described in detail.

[0041] Figure 2 The flowchart of the method for identifying public opinion risks according to an embodiment of the present disclosure is schematically shown.

[0042] like Figure 2 As shown, the public opinion risk identification method of this embodiment includes operations S210 to S240.

[0043] In operation S210, the target recognition text is processed using a risk recognition model to identify a risk category of the target recognition text, a first risk entity in the target recognition text, and a first category of the first risk entity.

[0044] According to the embodiments of the present disclosure, the risk identification model is a recognition text model trained by a deep learning algorithm, and the model can be trained by a text classification model training method. The risk identification model predicts the risk category of each target recognition text to obtain a classification of each target recognition text, which can be a binary classification, including a risk category or a non-risk category; at the same time, entity recognition is performed on the entities in each target recognition text to obtain the first risk entity and the first category of the first risk entity in each target recognition text.

[0045] According to an embodiment of the present disclosure, the target recognition text may be a text sentence, which may be a text sentence in Chinese and may be composed of multiple characters; or a text sentence in English and may be composed of multiple words. Here, the language type of the target recognition text is not limited.

[0046] According to an embodiment of the present disclosure, the purpose of risk entity identification is to identify subjects such as names of people, places, names of organizations, and company names in the target identification text. For example, company names can be identified from the target identification text; the category of risk entities can be the category of risk entities that can be identified based on keywords in the target identification text.

[0047] In operation S220, when the risk category is risky, the entity word library and the risk keyword library are called to perform entity matching on the target recognition text to obtain the second risk entity of the target recognition text and the second category of the second risk entity.

[0048] According to an embodiment of the present disclosure, an entity word library may represent a database storing company names and basic attribute information. The company names and basic attribute information stored in the entity word library may be obtained from Internet open source channels and / or specific business channels. For example, crawler technology may be used to dynamically obtain the information from Internet open source channels. An entity word library containing company names and basic attribute information may also be formed through accumulation in the financial field.

[0049] According to an embodiment of the present disclosure, the risk keyword library may store a database representing risk event categories and corresponding keywords. For example, if the risk event category is asset abnormality, the corresponding keywords may include: loss, loss, liquidation, etc.; the corresponding keywords for breach of contract and dishonesty events include breach of contract, dishonesty, overdue, default, etc.

[0050] According to an embodiment of the present disclosure, when it is determined that the target recognition text is risky, the entity vocabulary and risk keyword library are used to match the corresponding risk entities and risk entity categories from the target recognition text to obtain the second risk entity and the second risk entity category of the target recognition text.

[0051] In operation S230 , when the first risk entity includes multiple entities, an entity disambiguation database is called to perform entity disambiguation on the multiple first risk entities to obtain a third risk entity of the target recognition text and a third category of the third risk entity.

[0052] According to an embodiment of the present disclosure, the entity disambiguation database stores regular expressions that can characterize multiple entities with the same meaning. The purpose of entity disambiguation is to eliminate ambiguity among multiple entities with the same meaning.

[0053] According to an embodiment of the present disclosure, in the process of performing risk entity recognition on the target recognition text using the risk recognition model, multiple first risk entities may be recognized for the same entity position in the target recognition text. In the case where multiple first risk entities are determined to be recognized, the entity disambiguation database is called to perform entity disambiguation on the multiple first risk entities to obtain the third risk entity of the target recognition text and the third category of the third risk entity.

[0054] In operation S240, entity fusion is performed on the second risk entity and the second category with the third risk entity and the third category to obtain a target risk entity and a target category of the target risk entity of the target recognition text.

[0055] According to an embodiment of the present disclosure, the purpose of entity fusion is to combine the second risk entity obtained by entity matching and the third risk entity obtained by entity disambiguation, as well as the corresponding second category and the corresponding third category, to obtain the target risk entity of the target recognition text and the target category of the target risk entity.

[0056] According to the embodiments of the present disclosure, by utilizing a risk identification model to simultaneously identify the risk category and entity category of the target identification text, and then combining entity matching and entity disambiguation technology to perform entity fusion analysis on the above identification results, the model cumulative error is reduced, the recognition effect is improved, and it is further helpful for business personnel to predict and avoid risks.

[0057] Figure 3A flowchart of a method for processing a target recognition text using a risk recognition model to identify a risk category of the target recognition text, a first risk entity, and a first category of the first risk entity is schematically shown according to an embodiment of the present disclosure.

[0058] like Figure 3 As shown, the method may include operations S310 to S330.

[0059] In operation S310, the target recognition text is preprocessed to obtain a plurality of target recognition text units for the target recognition text.

[0060] In operation S320, feature extraction is performed on target recognition text unit data of a plurality of target recognition text units to obtain a feature vector sequence.

[0061] In operation S330, entity recognition is performed on the feature vector sequence to obtain the risk category of the target recognition text, the first risk entity in the target recognition text, and the first category of the first risk entity.

[0062] According to an embodiment of the present disclosure, preprocessing the target recognition text may include: segmenting the acquired target recognition text, and segmenting the target recognition text into a plurality of target recognition text units. The target recognition text unit is the smallest unit processed by the risk recognition model. For example, for a target recognition text in Chinese, the target recognition text unit may be a character; for a target recognition text in English, the target recognition text unit may be a word.

[0063] According to an embodiment of the present disclosure, feature extraction of the obtained multiple target recognition text units may include: during the feature extraction process, the risk recognition model may add additional tags to each target recognition text unit, for example, adding [CLS] to represent the classification tag, [SEP] to represent the short sentence tag, O to represent that it does not belong to any risk entity, B to represent the initial position of the risk entity, I to represent the middle or end position of the risk entity, etc. Feature extraction is then performed on each target recognition unit to obtain a feature vector sequence of the target recognition text unit.

[0064] According to an embodiment of the present disclosure, entity recognition of a feature vector sequence may include: extracting sequence position information of the feature vector sequence based on a conversion model of a risk identification model, and performing, for example, conditional random field analysis on the feature vector sequence according to the extraction result to identify the risk category of the target identification text, the first risk entity and the first category of the first risk entity.

[0065] According to an embodiment of the present disclosure, for example, taking the Chinese text of the target recognition text "Company A applies for bankruptcy liquidation [CLS]" as an example, the target recognition text units are "A", "public", "company", "apply", "bankruptcy", "production", "liquidation", "calculation", "CLS", and feature extraction is performed on each target recognition text unit to obtain a feature vector sequence, and then entity recognition is performed on the feature vector to obtain an entity recognition result. For example, "A" is the first letter of the name of Company A, and its labeled category is B-asset abnormal event subject, and "public" is the middle position of the name of Company A, and its labeled category is I-asset abnormal event subject, where "asset abnormal event subject" is characterized as the risk entity category corresponding to the risk entity. The risk category of this target recognition text is "risky", and the labeling category corresponding to "CLS" is "risk_news".

[0066] According to an embodiment of the present disclosure, feature extraction is performed on target recognition text unit data of multiple target recognition text units to obtain a feature vector sequence, including: performing lexical feature extraction on each target recognition text unit data to obtain a lexical feature vector sequence; using a position encoder to perform position feature extraction on each target recognition text unit data to obtain a position feature vector sequence; and obtaining a feature vector sequence based on the lexical feature vector sequence and the position feature vector sequence.

[0067] According to an embodiment of the present disclosure, a lexical feature vector sequence can represent the corresponding entity and entity category corresponding to each target recognition text unit; a position feature vector sequence can represent the relative or absolute position of each target recognition text unit in the target recognition text. The feature vector sequence can include the entity in the target recognition text, entity category information, and the position information of the entity in the target recognition text.

[0068] According to the embodiments of the present disclosure, a risk identification model is used to simultaneously identify risk entities and entity categories of a target identification text, without the need to first perform entity identification and then perform entity category identification, thereby reducing the cumulative error of the model and improving the recognition accuracy.

[0069] According to an embodiment of the present disclosure, the risk identification model is based on the JointBERT language representation model.

[0070] According to an embodiment of the present disclosure, the JointBERT model is a model for training language representation, which can perform word segmentation on the input recognition text, process the word segmentation results as units, and obtain feature extraction results corresponding to each word segmentation result.

[0071] According to an embodiment of the present disclosure, the risk identification model is based on the JointBERT language representation model and can be trained by the following method: preparing training recognition text data; inputting sentences in the training recognition text data into the risk identification training model, and training to obtain the risk identification model.

[0072] According to an embodiment of the present disclosure, training recognition text data can use a web crawler to capture as much financial news and public opinion as possible from the source of news and public opinion risk data in the financial field, and store the financial public opinion news in the form of text.

[0073] According to an embodiment of the present disclosure, sources of news and public opinion risk data in the financial field may include corporate news and financial news sections of major portal websites and various large, medium and small websites related to finance, business, etc.

[0074] According to an embodiment of the present disclosure, when the risk category is risky, an entity vocabulary and a risk keyword library are called to perform entity matching on the target recognition text to obtain a second risk entity of the target recognition text and a second category of the second risk entity, including: in response to a matching request for a corresponding entity of the target recognition text, the entity vocabulary library is called to match the corresponding entity of the target recognition text to obtain the corresponding entity of the target recognition text; in response to a matching request for risk keywords of the target recognition text, the risk keyword library is called to perform risk keyword matching on the risk keywords of the target recognition text to obtain a risk event category associated with the risk keywords; according to the corresponding entity and the risk event category of the target recognition text, the second risk entity of the target recognition text and the second category of the second risk entity are obtained.

[0075] According to an embodiment of the present disclosure, when it is determined that the risk category of the target recognition text is risky, the entity word library is first called to associate the target recognition text with the entity word library, and the entity word library is used to match the entity in the target recognition text with the entity word library. When it is determined that the entity word library contains the corresponding entity of the target recognition text, the risk keyword library is called again to match the keywords in the target recognition text with the keywords in the risk keyword library to obtain the risk event category corresponding to the keyword, and the corresponding entity obtained by using the entity word library and the risk event category corresponding to the keyword obtained by using the risk keyword library are combined to obtain the second risk entity of the target recognition text and the second category of the second risk entity.

[0076] According to an embodiment of the present disclosure, taking the target identification text "Company A applies for bankruptcy liquidation" as an example, the entity vocabulary is first called to match the entity corresponding to the risk entity "Company A" of the target identification text in the entity vocabulary. When it is determined that the entity vocabulary contains the corresponding entity "Company A" of the target identification text, the risk keyword library is called to match "bankruptcy liquidation" in the target identification text as a keyword with the keywords in the risk keyword library. Keywords with the same or similar semantics as "bankruptcy liquidation" are matched in the risk keyword library, and the risk event category corresponding to the keyword is "abnormal asset event". The risk entity "Company A" is combined with the risk event category "abnormal asset event" to obtain the second risk entity of the target identification text as "Company A", and the second category of the second risk entity is "abnormal asset event".

[0077] According to an embodiment of the present disclosure, taking the target recognition text "Company A applies for bankruptcy liquidation" as an example, the entity vocabulary is first called to match the entity corresponding to the risk entity "Company A" of the target recognition text in the entity vocabulary. If it is determined that the corresponding entity is not matched in the entity vocabulary, the risk keyword library is not called for matching.

[0078] Figure 4 The flowchart of a method for entity disambiguation of multiple first risk entities according to an embodiment of the present disclosure is schematically shown.

[0079] like Figure 4 As shown, the method may include operations S410 to S430.

[0080] In operation S410 , when it is determined that a plurality of first risk entities are at the same position in a target recognition text, in response to a request for entity disambiguation of the plurality of first risk entities, an entity disambiguation database is called.

[0081] In operation S420 , according to the data feature structure between the plurality of first risk entities, a regular expression matching the data feature structure is extracted from the entity disambiguation data.

[0082] In operation S430, entity disambiguation is performed on the plurality of first risk entities according to the regular expression to obtain a third risk entity of the target recognition text and a third category of the third risk entity.

[0083] According to an embodiment of the present disclosure, in the process of identifying the first risk entity of the target recognition text using the risk recognition model, the risk recognition model may identify multiple first risk entities at the same position where each target recognition text unit appears in the target recognition text. When multiple first risk entities are determined to be at the same position in the target recognition text, the entity disambiguation database is called to perform entity disambiguation on the multiple first risk entities to eliminate ambiguity.

[0084] According to the embodiments of the present disclosure, the regular expressions stored in the entity disambiguation database reflect the expression relationship of eliminating ambiguity of multiple entities with the same meaning and turning them into one entity. For example, "B" and "B Company" represent two entities, and the regular expression for entity disambiguation of the two entities of the data feature structure can be "B+B Company=B Company"; "C Company" and "Original C Company" represent two entities, and the regular expression for entity disambiguation of the two entities of the data feature structure can be "C Company+Original C Company=C Company". It should be noted that the regular expressions stored in the entity disambiguation database are not limited to the above examples, and the risk entity is not limited to between two. The entity disambiguation rule expressions between more than two risk entities are also stored in the entity disambiguation database.

[0085] According to an embodiment of the present disclosure, a risk identification model is used to identify a target identification text. For example, the two first risk entities at the same position in "Company A applies for bankruptcy liquidation" are "Company A" and "Original Company A", and the corresponding first risk first category is "Abnormal asset events". According to the data feature structure of the two first risk entities, a regular expression matching the data feature structure is extracted from the entity disambiguation data. For example, "Company C + Original Company C = Company C", then the two first risk entities are entity disambiguated according to the regular expression of the data feature structure in the entity disambiguation data, and the third risk entity is obtained as "Company A" and the third category of the third risk entity is "Abnormal asset events".

[0086] According to an embodiment of the present disclosure, the second risk entity and the second category are entity-fused with the third risk entity and the third category to obtain a target risk entity and a target category of the target risk entity of the target recognition text, including:

[0087] The second risk entity and the second category are unioned with the third risk entity and the third category so as to perform entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text.

[0088] According to an embodiment of the present disclosure, the second risk entity and the second category of the second risk entity of the target recognition text obtained through entity matching are respectively unioned with the third risk entity and the third category of the third risk entity of the target recognition text obtained through entity disambiguation to obtain the target risk entity of the target recognition text and the target category of the target risk entity.

[0089] According to an embodiment of the present disclosure, for example, the second risk entity and the second category of the second risk entity are "Company A-Abnormal Asset Event" and "E-Default and Breach of Trust Event" respectively; the third risk entity and the third category of the third risk entity are "Company A-Abnormal Asset Event", and the above-mentioned second risk entity and the second category are merged with the third risk entity and the third category to achieve entity fusion, and the risk entity output is obtained as "Company A-Abnormal Asset Event" and "E-Default and Breach of Trust Event" to supplement and confirm the identification results of the first risk entity and the category of the first risk entity obtained by using the risk identification model, and finally the target risk entity of the target identification text and the target category of the target risk entity are obtained as "Company A-Abnormal Asset Event".

[0090] According to an embodiment of the present disclosure, entity fusion analysis is implemented on the recognition results of the first risk entity and the first risk entity category obtained by recognizing the target recognition text using the risk identification model through the logical rules of entity matching and entity disambiguation, as a further supplement to the recognition results, thereby improving the recognition effect.

[0091] Figure 5 A schematic diagram of a method for identifying public opinion risks using a risk identification model according to an embodiment of the present disclosure is schematically shown.

[0092] Figure 6 A schematic diagram of a method for entity fusion according to an embodiment of the present disclosure is schematically shown.

[0093] like Figure 5 As shown, the target identification text 501 is input into the risk identification model 502 to identify the risk entity of the target identification text, and the risk category 503 of the target identification text, the first risk entity of the target identification text and the first category 504 of the first risk entity are obtained. The risk category 503 is a binary classification, that is, risk events and non-risk events.

[0094] like Figure 6 As shown, in the case where it is determined that the target identification text is a risk event, the target identification text 501 is entity matched with the entity vocabulary and the risk keyword library 601 to obtain the second risk entity and the second category 602 of the second risk entity; in the case where there are multiple first risk entities 504, the multiple first risk entities and the first category 504 of the first risk entity are matched with the entity disambiguation rule expression 603 to perform entity disambiguation to obtain the third risk entity and the third category 604 of the third risk entity, and the first risk entity and the second category 602 of the second risk entity are entity fused with the third risk entity and the third category 604 of the third risk entity to obtain the target risk entity of the target identification text and the target category 605 of the target risk entity.

[0095] Based on the above public opinion risk identification method, the present disclosure also provides a public opinion risk identification device. Figure 7 The device is described in detail.

[0096] Figure 7 The structural block diagram of the public opinion risk identification device according to an embodiment of the present disclosure is schematically shown.

[0097] like Figure 7 As shown, the public opinion risk identification device 700 of this embodiment may include an identification module 710, an entity matching module 720, an entity disambiguation module 730 and an entity fusion module 740.

[0098] The identification module 710 is used to process the target identification text using the risk identification model to identify the risk category of the target identification text, the first risk entity in the target identification text, and the first category of the first risk entity. In one embodiment, the identification module 710 can be used to perform the operation S210 described above, which will not be repeated here.

[0099] The entity matching module 720 is used to call the entity word library and the risk keyword library to perform entity matching on the target recognition text when the risk category is risky, so as to obtain the second risk entity of the target recognition text and the second category of the second risk entity. In one embodiment, the entity matching module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0100] The entity disambiguation module 730 is used to call the entity disambiguation database when the first risk entity includes multiple entities, perform entity disambiguation on the multiple first risk entities, and obtain the third risk entity of the target recognition text and the third category of the third risk entity. In one embodiment, the entity disambiguation module 730 can be used to perform the operation S230 described above, which will not be repeated here.

[0101] The entity fusion module 740 performs entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text. In one embodiment, the entity fusion module 740 can be used to perform the operation S240 described above, which will not be repeated here.

[0102] According to the embodiments of the present disclosure, by utilizing a risk identification model to simultaneously identify the risk category and entity category of the target identification text, and then combining entity matching and entity disambiguation technology to perform entity fusion analysis on the above identification results, the model cumulative error is reduced, the recognition effect is improved, and it is further helpful for business personnel to predict and avoid risks.

[0103] According to an embodiment of the present disclosure, the identification module 710 may include: a first obtaining submodule, a first extracting submodule, and an identification submodule.

[0104] The first obtaining submodule is used to preprocess the target recognition text to obtain a plurality of target recognition text units for the target recognition text.

[0105] The first extraction submodule is used to perform feature extraction on target recognition text unit data of multiple target recognition text units to obtain a feature vector sequence.

[0106] The recognition submodule is used to perform entity recognition on the feature vector sequence to identify the risk category of the target recognition text, the first risk entity in the target recognition text, and the first category of the first risk entity.

[0107] According to an embodiment of the present disclosure, the extraction submodule may include: a first obtaining unit, a second obtaining unit, and a third obtaining unit.

[0108] The first obtaining unit is used to extract lexical features from each target recognition text unit data to obtain a lexical feature vector sequence.

[0109] The second obtaining unit is used to use a position encoder to extract position features from each target recognition text unit data to obtain a position feature vector sequence.

[0110] The third obtaining unit is used to obtain a feature vector sequence according to the lexical feature vector sequence and the position feature vector sequence.

[0111] According to an embodiment of the present disclosure, the risk identification model is based on the JointBERT language representation model.

[0112] According to an embodiment of the present disclosure, the entity matching module 720 may include: a first calling submodule, a second calling submodule, and a second obtaining submodule.

[0113] The first calling submodule is used to respond to a matching request for a corresponding entity of a target recognition text, call an entity word library, match the corresponding entity of the target recognition text, and obtain the corresponding entity of the target recognition text.

[0114] The second calling submodule is used to call the risk keyword library in response to a matching request for the risk keyword of the target recognition text, perform risk keyword matching on the risk keyword of the target recognition text, and obtain the risk event category associated with the risk keyword.

[0115] The second obtaining submodule is used to obtain the second risk entity of the target recognition text and the second category of the second risk entity according to the corresponding entity and risk event category of the target recognition text.

[0116] According to an embodiment of the present disclosure, the entity disambiguation module 730 may include: a third calling submodule, a second extracting submodule, and a third obtaining submodule.

[0117] A third calling submodule is used to call the entity disambiguation database in response to a request for entity disambiguation of the plurality of first risk entities when it is determined that the plurality of first risk entities are at the same position in the target recognition text;

[0118] A second extraction submodule is used to extract a regular expression matching the data feature structure from the entity disambiguation data according to the data feature structure between the plurality of first risk entities;

[0119] The third obtaining submodule is used to perform entity disambiguation on multiple first risk entities according to the regular expression to obtain the third risk entity of the target recognition text and the third category of the third risk entity.

[0120] According to an embodiment of the present disclosure, the entity fusion module 740 may include: a union submodule.

[0121] The union submodule is used to perform a union operation on the second risk entity and the second category with the third risk entity and the third category, so as to perform entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text.

[0122] According to an embodiment of the present disclosure, any multiple modules of the recognition module 710, the entity matching module 720, the entity disambiguation module 730, and the entity fusion module 740 can be combined in one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the recognition module 710, the entity matching module 720, the entity disambiguation module 730, and the entity fusion module 740 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or by any one of the three implementation methods of software, hardware, and firmware, or by a suitable combination of any of them. Alternatively, at least one of the recognition module 710, the entity matching module 720, the entity disambiguation module 730, and the entity fusion module 740 may be at least partially implemented as a computer program module, which may perform a corresponding function when executed.

[0123] Figure 8 A block diagram of an electronic device suitable for implementing the public opinion risk method according to an embodiment of the present disclosure is schematically shown.

[0124] like Figure 8 As shown, the electronic device 800 according to an embodiment of the present disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include an onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0125] In RAM 803, various programs and data required for the operation of electronic device 800 are stored. Processor 801, ROM 802 and RAM 803 are connected to each other via bus 804. Processor 801 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 802 and / or RAM 803. It should be noted that the program can also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in one or more memories.

[0126] According to an embodiment of the present disclosure, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to the bus 804. The electronic device 800 may further include one or more of the following components connected to the I / O interface 805: an input portion 806 including a keyboard, a mouse, etc.; an output portion 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 808 including a hard disk, etc.; and a communication portion 809 including a network interface card such as a LAN card, a modem, etc. The communication portion 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage portion 808 as needed.

[0127] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0128] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0129] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the public opinion risk identification method provided by the embodiment of the present disclosure.

[0130] The above functions defined in the system / device of the embodiment of the present disclosure are executed when the computer program is executed by the processor 801. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0131] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 809, and / or installed from a removable medium 811. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0132] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0133] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0134] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0135] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0136] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. The scope of the present disclosure is defined by the attached claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A method for identifying public opinion risks, comprising: Processing the target identification text by using the risk identification model to identify the risk category of the target identification text, the first risk entity in the target identification text, and the first category of the first risk entity; In the case where the risk category is risky, calling an entity word library and a risk keyword library to perform entity matching on the target recognition text to obtain a second risk entity of the target recognition text and a second category of the second risk entity; In the case where the first risk entity includes a plurality of entities, calling an entity disambiguation database to perform entity disambiguation on the plurality of the first risk entities to obtain a third risk entity of the target recognition text and a third category of the third risk entity; The second risk entity and the second category are entity-fused with the third risk entity and the third category to obtain the target risk entity of the target recognition text and the target category of the target risk entity.

2. The method according to claim 1, wherein: Processing the target identification text by using the risk identification model to identify the risk category of the target identification text, the first risk entity in the target identification text, and the first category of the first risk entity includes: Preprocessing the target recognition text to obtain a plurality of target recognition text units for the target recognition text; Extracting features from target recognition text unit data of the plurality of target recognition text units to obtain a feature vector sequence; Entity recognition is performed on the feature vector sequence to obtain the risk category of the target recognition text, the first risk entity in the target recognition text, and the first category of the first risk entity.

3. The method according to claim 2, wherein: The step of extracting features from the target recognition text unit data of the plurality of target recognition text units to obtain a feature vector sequence comprises: Extracting lexical features from each target recognition text unit data to obtain a lexical feature vector sequence; Using a position encoder, extracting position features from each target recognition text unit data to obtain a position feature vector sequence; A feature vector sequence is obtained according to the lexical feature vector sequence and the position feature vector sequence.

4. The method according to claim 1, wherein: The risk identification model is based on the JointBERT language representation model.

5. The method according to claim 1, wherein: When the risk category is risky, calling the entity word library and the risk keyword library to perform entity matching on the target recognition text to obtain the second risk entity of the target recognition text and the second category of the second risk entity includes: In response to a matching request for a corresponding entity of the target recognition text, calling the entity word library, matching the corresponding entity of the target recognition text, and obtaining the corresponding entity of the target recognition text; In response to a matching request for the risk keywords of the target recognition text, calling the risk keyword library, performing risk keyword matching on the risk keywords of the target recognition text, and obtaining a risk event category associated with the risk keywords; According to the corresponding entity of the target identification text and the risk event category, a second risk entity of the target identification text and a second category of the second risk entity are obtained.

6. The method according to claim 1, wherein: In the case where the first risk entity includes a plurality of entities, calling an entity disambiguation database to perform entity disambiguation on the plurality of first risk entities to obtain a third risk entity of the target recognition text and a third category of the third risk entity, including: In a case where a plurality of the first risk entities are determined to be at the same position in the target recognition text, in response to a request for entity disambiguation of the plurality of the first risk entities, calling the entity disambiguation database; According to the data feature structure between the plurality of first risk entities, extracting a regular expression matching the data feature structure from the entity disambiguation data; According to the regular expression, entity disambiguation is performed on the plurality of the first risk entities to obtain a third risk entity of the target recognition text and a third category of the third risk entity.

7. The method according to claim 1, wherein: The step of performing entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text includes: The second risk entity and the second category are unioned with the third risk entity and the third category to perform entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text.

8. A public opinion risk identification device, comprising: An identification module, used for processing a target identification text by using a risk identification model, identifying a risk category of the target identification text, a first risk entity in the target identification text, and a first category of the first risk entity; An entity matching module, for calling an entity word library and a risk keyword library to perform entity matching on the target recognition text when the risk category is risky, so as to obtain a second risk entity of the target recognition text and a second category of the second risk entity; An entity disambiguation module, configured to, when the first risk entity includes a plurality of entities, call an entity disambiguation database to perform entity disambiguation on the plurality of first risk entities, and obtain a third risk entity of the target recognition text and a third category of the third risk entity; The entity fusion module performs entity fusion on the second risk entity and the second category with the third risk entity and the third category to obtain the target risk entity and the target category of the target risk entity of the target recognition text.

9. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images