Public opinion clustering method, system, electronic equipment and medium in the scenario of securities firm credit rating

Abstract

The present invention provides a method, system, electronic device and medium for clustering public opinion in a brokerage credit rating scenario, the method comprising obtaining first text data from various information sources; classifying the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise; inputting the first text data into a trained fingerprint generation model and outputting an article fingerprint vector; screening out a second public opinion category that is still within the window validity period from the first public opinion category belonging to the same enterprise; respectively calculating the similarity between the article fingerprint vector of each of the second text data of the same enterprise and the central vector of each of the second public opinion categories; and classifying each second text data into the corresponding second public opinion category based on the similarity. The present invention is based on dynamic clustering of validity period, which can better capture the changing trend of public opinion over time, solves the problem that traditional clustering ignores the coherence of public opinion, and improves the continuity and accuracy of clustering results.

Description

Technical Field

[0001] The present invention relates to the technical field of public opinion analysis, and in particular to a public opinion clustering method, system, electronic equipment and medium in a brokerage firm credit rating scenario. Background Art

[0002] In the modern financial market, the credit evaluation of securities firms is crucial. Public opinion information about a company and its business activities has a significant impact on the assessment of its credit status, market performance and future potential. Traditional public opinion analysis methods mainly rely on manual classification and basic text analysis tools, which are inefficient and cannot effectively process large-scale data, resulting in insufficient accuracy and timeliness of information analysis.

[0003] At present, there are several public opinion clustering methods in the industry:

[0004] 1. Traditional text clustering method

[0005] Keyword-based clustering methods rely on the keywords that appear in the text for clustering. These methods usually use statistical features such as word frequency (TF) and inverse document frequency (IDF) to represent the text, and then apply traditional clustering algorithms such as K-Means. This method is simple to implement, but it relies on the selection of keywords, is difficult to capture semantic information, and has poor handling of the ambiguity and contextual relationships of the text.

[0006] 2. Text Clustering Based on Word Embedding

[0007] Text clustering based on word embedding is a clustering method that uses word vector representation to capture text semantic information. First, by training a word vector model (such as Word2Vec, GloVe), each word can be mapped to a high-dimensional vector representation. These vectors capture the semantic relationship between words, such as similarity and association. Then the text is preprocessed, including text cleaning, word segmentation, and stop word removal. After the text processing is completed, a batch of words will be obtained. The corresponding word embeddings can be generated using the pre-trained word vector model, and the vector representation of the text is generated by combining these word embeddings. This method can effectively capture the semantic relationship between words, but the word vector is static and cannot handle polysemous words or sentence-level semantic information.

[0008] 3. Text clustering based on deep learning

[0009] Deep learning-based text clustering generates context-sensitive text embedding vectors through deep neural network models (such as BERT, Transformer, etc.), and then clusters these vectors using clustering algorithms (such as K-Means, DBSCAN). Deep learning models can capture the complex semantic relationships of text, making the clustering results more semantically consistent. Compared with traditional methods, deep learning-based clustering can process more complex text data, but it requires higher computing resources, and the cost of training the model with private data is high.

[0010] In recent years, with the rapid development of natural language processing technology, text analysis methods based on machine learning have been widely used in public opinion analysis. However, existing technical solutions still have limitations, especially when dealing with public opinion in specific fields. On the one hand, most of the current pre-training models are trained based on general corpus, and the effect is not ideal when directly applied to the financial field. The generated vectors are difficult to accurately express the deep semantics of the text, which affects the accuracy of clustering. On the other hand, traditional clustering methods usually rely on fixed time windows and fail to fully consider the characteristics of public opinion changing dynamically over time, making it difficult to capture the coherence of public opinion. Therefore, there is an urgent need for a more intelligent and flexible public opinion clustering method that can combine the specific needs of the financial field to improve the efficiency and accuracy of public opinion analysis. Summary of the invention

[0011] The present invention provides a public opinion clustering method, system, electronic device and medium in a securities firm credit rating scenario, which can improve the accuracy of clustering.

[0012] According to one aspect of the present invention, a method for clustering public opinion in a brokerage credit rating scenario is provided, comprising:

[0013] Acquire first text data from various information sources;

[0014] Based on the relevance between the content of the first text data and the enterprise, classifying the first text data into a corresponding enterprise dimension;

[0015] Input the first text data into the trained fingerprint generation model, and output the article fingerprint vector;

[0016] Filter out the second public opinion category that is still within the window validity period from the first public opinion category belonging to the same enterprise;

[0017] respectively calculating the similarity between the article fingerprint vector of each of the second text data of the same enterprise and the central vector of each of the second public opinion categories; the same enterprise includes different public opinion categories;

[0018] Each second text data is classified into the corresponding second public opinion category based on the similarity.

[0019] Optionally, before inputting the first text data into a trained fingerprint generation model to output the article fingerprint, the method further includes:

[0020] Acquire historical text data in multiple financial fields and annotate the historical text data;

[0021] Taking the historical text data and the annotation labels corresponding to the historical text data as a sample, constructing a data set;

[0022] The fingerprint generation model is trained based on the data set to obtain the trained fingerprint generation model.

[0023] Optionally, the length of the window validity period is at least related to the amount of text data in each of the first public opinion categories, the time span of the text data in the first public opinion category in days, and the default validity period of the window.

[0024] Optionally, the calculation formula for the length v of the window validity period is:

[0025]

[0026] In the formula, d is the default validity period of the window; w is the window coefficient; n is the number of text data in the first public opinion category; t is the time span of the text data in the first public opinion category in days; MAX() is the maximum value selection formula.

[0027] Optionally, categorizing each second text data into a corresponding second public opinion category based on the similarity includes:

[0028] Obtaining a maximum similarity value from the similarities between the article fingerprint vector of the second text data and the central vectors of different second public opinion categories;

[0029] When the maximum similarity value is greater than a preset threshold, classifying the second text data into the second public opinion category corresponding to the maximum similarity value;

[0030] When the maximum similarity value is less than a preset threshold, a new public opinion category is generated, and the second text data is classified into the new public opinion category.

[0031] Optionally, also include:

[0032] After classifying the second text data into the second public opinion category corresponding to the maximum similarity value, adding the article fingerprint vector of the second text data into the vector set of the corresponding second public opinion category, and reselecting the updated center vector of the second public opinion category;

[0033] When the maximum similarity value is less than a preset threshold, a new public opinion category is generated, and the article fingerprint vector of the second text data is used as the central vector.

[0034] Optionally, after acquiring the first text data from each information source, the method further includes:

[0035] Preprocessing the first text data;

[0036] The preprocessing at least includes converting the first text data into a unified format and performing data cleaning and denoising on the first text data.

[0037] A second aspect of the present invention provides a public opinion clustering system in a brokerage credit rating scenario, comprising:

[0038] A first data acquisition unit, used to acquire first text data from various information sources;

[0039] a categorizing unit, configured to categorize the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise;

[0040] A fingerprint generation unit, used to input the first text data into a trained fingerprint generation model and output an article fingerprint vector;

[0041] A screening unit, used to screen out second public opinion categories that are still within the window validity period from the first public opinion categories belonging to the same enterprise;

[0042] A similarity calculation unit calculates the similarity between the article fingerprint vector of each of the second text data of the same enterprise and the center vector of each of the second public opinion categories; the same enterprise includes different public opinion categories;

[0043] The second classification unit is used to classify each second text data into the corresponding second public opinion category based on the similarity.

[0044] A third aspect of the present invention further provides an electronic device, comprising at least one processor; and

[0045] a memory communicatively connected to the at least one processor; wherein,

[0046] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the public opinion clustering method in the brokerage credit rating scenario.

[0047] The fourth aspect of the present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement a public opinion clustering method in a brokerage credit rating scenario when executed by a processor.

[0048] The technical solution of the embodiment of the present invention selects the article fingerprint vector of each first text data by classifying the first text data into the corresponding enterprise dimension and generating the article fingerprint vector corresponding to the first text data through the fingerprint generation model; and sets the window validity period to filter out the public opinion categories that are still within the time limit. By calculating the similarity between the article fingerprint vector of the second text data belonging to the same enterprise and the central vector of different public opinion categories, the second text data is classified into the corresponding public opinion category, and the dynamic clustering based on the validity period can better capture the changing trend of public opinion over time, solve the problem that traditional clustering ignores the coherence of public opinion, and improve the continuity and accuracy of clustering results.

[0049] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 A public opinion clustering method in a brokerage credit rating scenario provided according to the first embodiment of the present invention;

[0052] Figure 2 A public opinion clustering method in a brokerage credit rating scenario provided according to the second embodiment of the present invention;

[0053] Figure 3 This is a system architecture diagram of a public opinion clustering system in a brokerage credit rating scenario provided according to Embodiment 3 of the present invention;

[0054] Figure 4 It is a structural schematic diagram of an electronic device for implementing the public opinion clustering method in the brokerage credit rating scenario of an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0056] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0057] Embodiment 1

[0058] Figure 1 A method flow chart of a public opinion clustering method in a brokerage credit rating scenario is provided for the first embodiment of the present invention. Figure 1 As shown, the system includes:

[0059] S101: Acquire first text data from various information sources.

[0060] The information sources include, but are not limited to, public opinion text data from news, announcements, social media, financial forums, etc. The first text may be obtained by a web crawler or other data acquisition methods; and the current first text data may be obtained from each information source in real time, or the first text data may be obtained at preset intervals, such as once a day.

[0061] In this embodiment, the first text data is mainly article data in the financial field, and specifically may be public opinion articles of various categories in the financial field.

[0062] S102: Classify the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise.

[0063] Among them, the content in the first text data may include public opinion information associated with some enterprises. Therefore, according to the association between the content and the enterprises in the multiple first text data obtained, the first text data can be divided into enterprise dimensions that have an association relationship with the content of the first text data.

[0064] Specifically, when the content of the first text data contains the situation of a certain enterprise, the first text data can be classified under the dimension of the enterprise; when the content of the first text data belongs to the business scope of a certain enterprise, the first text data can also be classified under the dimension of the enterprise; or when the content of the first text data is related to a certain enterprise executive, the first text data can be classified under the dimension of the enterprise. The association relationship has many forms, and this embodiment is only for exemplary description.

[0065] S103: Input the first text data into a trained fingerprint generation model, and output an article fingerprint vector.

[0066] The fingerprint generation model is a basic semantic model using machine learning, which can be a semantic model with a self-attention mechanism, such as a Transformer model. This type of semantic model can be used to obtain the article fingerprint vector of any text data. In this embodiment, a suitable semantic model and a loss function corresponding to the semantic model can be selected as needed. For example, TripletLoss can be selected as the loss function.

[0067] S104. Filter out the second public opinion categories that are still within the window validity period from the first public opinion categories belonging to the same enterprise.

[0068] After the first text data is classified into the corresponding enterprise dimension, it is necessary to screen the first public opinion categories contained in the same enterprise to determine the public opinion categories that are still within the window validity period. The first public opinion category is all public opinion categories belonging to the same enterprise; the second public opinion category is the public opinion category that is still within the window validity period.

[0069] Since public opinion categories have a certain timeliness, for example, public opinion from a month ago may be outdated and can no longer be a hot topic, so it is necessary to remove the first public opinion category from a month ago and only keep the public opinion categories that are still time-effective within a month. The window validity period can be set as needed, for example, it can be set to one month, two months, or one week.

[0070] S105, respectively calculating the similarity between the article fingerprint vector of each of the second text data of the same enterprise and each of the second central vectors; the same enterprise includes different public opinion categories.

[0071] S106: Classify each of the second text data into a corresponding public opinion category based on the similarity.

[0072] For the same enterprise, different public opinion categories may be included. For example, public opinion categories may include finance-related, legal-related, personnel-related, etc. This embodiment can calculate the similarity between the article fingerprint vector of each second text data and the center vector of different second public opinion categories, and determine which second public opinion category a second text data belongs to by the size of the similarity. Among them, the center vector refers to the one calculated based on the article fingerprint vectors of all text data in a second public opinion category. Among them, the second text data refers to the text data belonging to the same enterprise dimension after the first text data is classified.

[0073] The technical solution of the embodiment of the present invention selects the article fingerprint vector of each first text data by classifying the first text data into the corresponding enterprise dimension and generating the article fingerprint vector corresponding to the first text data through the fingerprint generation model; and sets the window validity period to filter out the public opinion categories that are still within the time limit. By calculating the similarity between the article fingerprint vector of the second text data belonging to the same enterprise and the central vector of different public opinion categories, the second text data is classified into the corresponding public opinion category, and the dynamic clustering based on the validity period can better capture the changing trend of public opinion over time, solve the problem that traditional clustering ignores the coherence of public opinion, and improve the continuity and accuracy of clustering results.

[0074] Embodiment 2

[0075] Figure 2 A flowchart of a public opinion clustering method in a brokerage credit rating scenario provided in the second embodiment of the present invention includes:

[0076] S201: Acquire first text data from various information sources.

[0077] S202: pre-process the first text data.

[0078] The preprocessing at least includes converting the first text data into a unified format and performing data cleaning and denoising on the first text data.

[0079] The acquired first text data can be processed according to a preset format, for example, the first text data can be stored in the format of text title, text body, text source, and text release time. Data cleaning and data denoising can be processed using conventional data cleaning and data denoising methods. Of course, appropriate data cleaning and data denoising methods can also be selected as needed. For example, in this embodiment, invalid information such as HTML tags and special symbols in the first text data can be removed, and the article in a unified format can be retained.

[0080] S203: Classify the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise.

[0081] S204: Acquire historical text data in multiple financial fields, and annotate the historical text data.

[0082] The data acquired in this embodiment is data in the financial field. When training the fingerprint generation model, a large amount of historical text data in the financial field can be acquired and annotated. Using the annotated historical text data to train the fingerprint generation model can optimize the adaptability of the model in financial scenarios, solve the problem that the traditional pre-training model has poor understanding of text in the financial field, and significantly improve the semantic accuracy of public opinion clustering.

[0083] In one embodiment, preprocessing of the historical text data at least includes converting the historical text data into a unified format, and performing data cleaning and denoising on the historical text data.

[0084] The acquired historical text data can be processed according to a preset format, for example, the historical text data can be stored in the format of text title, text body, text source, and text release time. Similarly, before generating the article fingerprint vector of the first text data, the first text data will be processed in the same preprocessing manner to ensure that the generated vector can accurately express the deep semantics of the text, thereby improving the accuracy of clustering.

[0085] S205: Take the historical text data and the annotation labels corresponding to the historical text data as a sample to construct a data set.

[0086] S206: Train the fingerprint generation model based on the data set to obtain the trained fingerprint generation model.

[0087] A historical text data and the annotated label corresponding to the historical text data can be regarded as a sample, and then multiple historical text data can constitute multiple sample data, thereby forming a sample set. The sample set can be divided into a training set, a validation set, and a test set in proportion, and the training set can be annotated to train the fingerprint generation model, so as to gradually fine-tune the parameters of the fingerprint generation model to obtain a trained fingerprint generation model.

[0088] The fingerprint generation model can be a semantic model with a self-attention mechanism, such as a Transformer model. This type of semantic model can be used to obtain the article fingerprint vector of any text data. In this embodiment, a suitable semantic model and a loss function corresponding to the semantic model can be selected as needed. For example, TripletLoss can be selected as the loss function.

[0089] In one embodiment, only the title of a historical text data and the annotated label of the historical text data may be used as a sample to form a sample set to simplify the training of the model.

[0090] S207: Input the first text data into a trained fingerprint generation model, and output an article fingerprint vector.

[0091] In one embodiment, only the title data of the first text data may be input into the trained fingerprint generation model to output the article fingerprint vector, so as to speed up the operation.

[0092] In one embodiment, the length of the window validity period is at least related to the amount of text data in each of the first public opinion categories, the time span of the text data in the first public opinion category in days, and the default validity period of the window.

[0093] Since the public opinion category has a certain timeliness, its timeliness can be related to the number of text data of the first public opinion category belonging to the same enterprise and the time span of the text data of the first public opinion category, that is, the more texts in the current public opinion category, the higher the popularity of the public opinion category, and the longer the span time allowed for the public opinion information of the enterprise. In addition, the timeliness of the public opinion category is also related to the default validity period of the window, that is, a default window validity period can be set in the initial state.

[0094] Specifically, the calculation formula for the length v of the window validity period is:

[0095]

[0096] In the formula, d is the default validity period of the window; w is the window coefficient; n is the number of text data in the first public opinion category; t is the time span of the text data in the first public opinion category in days; MAX() is the maximum value selection formula.

[0097] S208. Filter out the second public opinion categories that are still within the window validity period from the first public opinion categories belonging to the same enterprise.

[0098] S209, respectively calculating the similarity between the article fingerprint vector of each of the second text data of the same enterprise and the central vector of each of the second public opinion categories; the same enterprise includes different public opinion categories.

[0099] S210: Obtain the maximum similarity value from the similarities between the article fingerprint vector of the second text data and the central vectors of different second public opinion categories.

[0100] There are many ways to calculate similarity, such as Euclidean distance, Manhattan distance, Chebyshev distance, cosine similarity, correlation coefficient, etc. In this embodiment, cosine similarity can be used to calculate the similarity between the article fingerprint vector of the second text data and the central vectors of different public opinion categories, and find the maximum similarity value and the second public opinion category corresponding to the maximum similarity value.

[0101] S211, when the maximum similarity value is greater than a preset threshold, classifying the second text data into the second public opinion category corresponding to the maximum similarity value; and adding the article fingerprint vector of the second text data into the vector set of the corresponding second public opinion category, and reselecting the updated center vector of the second public opinion category;

[0102] A preset k can be set in advance. When the maximum similarity value is greater than k, the second text data is divided into the second public opinion category corresponding to the maximum similarity value, and the article fingerprint vector is added to the vector set of the corresponding second public opinion category, and the central vector of the second public opinion category is reselected based on the updated vector set.

[0103] S212: When the maximum similarity value is less than a preset threshold, a new public opinion category is generated, and the second text data is classified into the new public opinion category, and the article fingerprint vector of the second text data is used as the central vector.

[0104] When the maximum similarity value is less than k, the second text data is classified into a new public opinion category, and a corresponding article fingerprint vector is used as the central vector of the newly generated public opinion category.

[0105] The embodiments of the present invention optimize the adaptability of the model in financial scenarios by annotating and training the financial data, solve the problem that the traditional pre-trained model has a poor understanding of the text in the financial field, and significantly improve the semantic accuracy of public opinion clustering. By dynamically adjusting the length of the window validity period, the window period can be dynamically calculated in real time, so that new public opinion data can be quickly aggregated with existing public opinion categories, effectively solving the problem of slow response of traditional clustering methods and ensuring the timeliness of public opinion analysis. Through dynamic clustering based on the window validity period, the changing trend of public opinion over time can be better captured, solving the problem that traditional clustering ignores the coherence of public opinion, and improving the continuity and accuracy of clustering results.

[0106] Embodiment 3

[0107] Figure 3 This is a schematic diagram of the structure of a public opinion clustering system in a brokerage credit rating scenario provided by the third embodiment of the present invention. Figure 3 As shown, the device comprises:

[0108] A first data acquisition unit 301 is used to acquire first text data from various information sources;

[0109] A classification unit 302, configured to classify the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise;

[0110] A fingerprint generating unit 303 is used to input the first text data into a trained fingerprint generating model and output an article fingerprint vector;

[0111] A screening unit 304 is used to screen out second public opinion categories that are still within the window validity period from the first public opinion categories belonging to the same enterprise;

[0112] A similarity calculation unit 305 is used to respectively calculate the similarity between the article fingerprint vector of each second text data of the same enterprise and the central vector of each second public opinion category; the same enterprise includes different public opinion categories;

[0113] The second classification unit 306 is used to classify each second text data into the corresponding second public opinion category based on the similarity.

[0114] In one embodiment, a preprocessing unit is further included, which is used to preprocess the first text data; the preprocessing at least includes converting the first text data into a unified format, and performing data cleaning and denoising on the first text data.

[0115] It also includes a training unit, which is used to obtain historical text data in multiple financial fields and annotate the historical text data; use the historical text data and the annotated labels corresponding to the historical text data as a sample to construct a data set; train the fingerprint generation model based on the data set to obtain the trained fingerprint generation model.

[0116] In one embodiment, the second classification unit 306 is also used to obtain the maximum similarity from the similarity between the article fingerprint vector of the second text data and the center vectors of different second public opinion categories; when the maximum similarity is greater than a preset threshold, the second text data is classified into the second public opinion category corresponding to the maximum similarity; when the maximum similarity is less than the preset threshold, a new public opinion category is generated, and the second text data is classified into the new public opinion category.

[0117] In one embodiment, the second classification unit 306 is also used to add the article fingerprint vector of the second text data to the corresponding vector set of the second public opinion category after classifying the second text data into the second public opinion category corresponding to the maximum similarity value, and reselect the center vector of the updated second public opinion category; when the maximum similarity value is less than a preset threshold, a new public opinion category is generated, and the article fingerprint vector of the second text data is used as the center vector.

[0118] The message data processing resource scheduling device provided in the embodiment of the present invention can execute the message data processing resource scheduling device method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0119] Embodiment 4

[0120] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0121] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0122] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0123] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a public opinion clustering method in a brokerage credit rating scenario.

[0124] In some embodiments, a method for clustering public opinion in a brokerage credit rating scenario may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for clustering public opinion in the brokerage credit rating scenario described above may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute a method for clustering public opinion in a brokerage credit rating scenario in any other appropriate manner (e.g., by means of firmware).

[0125] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0126] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0127] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0128] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0129] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0130] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0131] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0132] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A public opinion clustering method in the scenario of securities firm credit rating, characterized by: include: Acquire first text data from various information sources; Based on the relevance between the content of the first text data and the enterprise, classifying the first text data into a corresponding enterprise dimension; Input the first text data into the trained fingerprint generation model, and output the article fingerprint vector; Filter out the second public opinion category that is still within the window validity period from the first public opinion category belonging to the same enterprise; The length of the window validity period is at least related to the number of text data in each of the first public opinion categories, the time span of the text data in the first public opinion category, and the default validity period of the window; Calculate the similarity between the article fingerprint vector of each second text data of the same enterprise and the central vector of each second public opinion category respectively; the same enterprise includes different public opinion categories; the second text data refers to the text data belonging to the same enterprise dimension after the first text data is classified; Each second text data is classified into the corresponding second public opinion category based on the similarity.

2. The public opinion clustering method in the brokerage credit rating scenario according to claim 1 is characterized in that: Before inputting the first text data into the trained fingerprint generation model and outputting the article fingerprint, the method further includes: Acquire historical text data in multiple financial fields and annotate the historical text data; Taking the historical text data and the annotation labels corresponding to the historical text data as a sample, constructing a data set; The fingerprint generation model is trained based on the data set to obtain the trained fingerprint generation model.

3. The public opinion clustering method in the brokerage credit rating scenario according to claim 1 is characterized in that: The calculation formula for the length v of the window validity period is: In the formula, d is the default validity period of the window; w is the window coefficient; n is the number of text data in the first public opinion category; t is the time span of the text data in the first public opinion category in days; MAX() is the maximum value selection formula.

4. The public opinion clustering method in the brokerage credit rating scenario according to claim 1 is characterized in that: The classifying each of the second text data into the corresponding second public opinion category based on the similarity includes: Obtaining a maximum similarity value from the similarities between the article fingerprint vector of the second text data and the central vectors of different second public opinion categories; When the maximum similarity value is greater than a preset threshold, classifying the second text data into the second public opinion category corresponding to the maximum similarity value; When the maximum similarity value is less than a preset threshold, a new public opinion category is regenerated, and the second text data is classified into the new public opinion category.

5. The public opinion clustering method in the brokerage credit rating scenario according to claim 4 is characterized in that: Also includes: After classifying the second text data into the second public opinion category corresponding to the maximum similarity value, adding the article fingerprint vector of the second text data into the vector set of the corresponding second public opinion category, and reselecting the updated center vector of the second public opinion category; When the maximum similarity value is less than a preset threshold, a new public opinion category is generated, and the article fingerprint vector of the second text data is used as the central vector.

6. The public opinion clustering method in the brokerage credit rating scenario according to claim 4 is characterized in that: After obtaining the first text data from each information source, the method further includes: Preprocessing the first text data; The preprocessing at least includes converting the first text data into a unified format and performing data cleaning and denoising on the first text data.

7. A public opinion clustering system in the scenario of securities firm credit rating, characterized by: include: A first data acquisition unit, used to acquire first text data from various information sources; a categorizing unit, configured to categorize the first text data into a corresponding enterprise dimension based on the relevance between the content of the first text data and the enterprise; A fingerprint generation unit, used to input the first text data into a trained fingerprint generation model and output an article fingerprint vector; A screening unit, used to screen out second public opinion categories that are still within a window validity period from first public opinion categories belonging to the same enterprise; the length of the window validity period is at least related to the amount of text data in each of the first public opinion categories, the number of days of the time span of the text data in the first public opinion category, and the length of the default window validity period; a similarity calculation unit, for respectively calculating the similarity between the article fingerprint vector of each second text data of the same enterprise and the central vector of each second public opinion category; the same enterprise includes different public opinion categories; the second text data refers to the text data belonging to the same enterprise dimension after the first text data is classified; The second classification unit is used to classify each second text data into the corresponding second public opinion category based on the similarity.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the public opinion clustering method in the brokerage credit rating scenario described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the public opinion clustering method in the brokerage credit rating scenario according to any one of claims 1 to 6 when executed.

Images