An integrated financial services system

Abstract

The application relates to the field of financial systems, and discloses a comprehensive financial service system which comprises a knowledge query module, an information extraction module and a front-end interaction module; the knowledge query module comprises a data acquisition unit, a data storage unit and a query service unit; the information extraction module is mainly responsible for establishing an information extraction model and logically realizing an information extraction function; the information extraction module comprises a data extraction unit and a financial announcement extraction unit; the data extraction unit and the financial announcement extraction unit are arranged, which are beneficial to information extraction of listed company financial announcements by constructing a financial field knowledge graph, can effectively reduce the human cost consumed by a supervisory agency during information examination, thereby more efficiently supervising listed companies, and can form a comprehensive cognition of the present situation of a company through structured company information, so that correct decisions can be quickly made during investment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of financial systems, more particularly to a comprehensive financial service system. BACKGROUND

[0002] In the context of rapid economic development and social progress in China, with the rise of the computer industry, the demand for information and knowledge in various industries in society is also increasing. In various vertical industries, the financial sector has a particularly high demand for information. Financial practitioners can access a large amount of information related to the financial sector from the Internet. However, the large amount of data from the Internet is multi-source and heterogeneous, and needs to be carefully screened by practitioners. Most information in the financial sector is published in the form of company announcements, and key information needs to be accurately and timely obtained from the text announcements. With the development of the economy, the number of listed companies is increasing, and professionals need to spend a lot of effort reading documents. Most existing comprehensive financial service systems do not use knowledge graphs.

[0003] And most existing comprehensive financial service systems cannot perform information extraction. Information extraction is the basis for building a knowledge graph and provides data support for the entire knowledge graph. Information extraction extracts information from text and obtains structured data from unstructured data. Therefore, the existing comprehensive financial service system has caused a serious waste of human and time resources. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a comprehensive financial service system to solve the problems existing in the background art.

[0005] The present application provides the following technical solution: a comprehensive financial service system, comprising a knowledge query module, an information extraction module and a front-end interaction module, the knowledge query module comprising a data acquisition unit, a data storage unit and a query service unit, the information extraction module being mainly responsible for establishing an information extraction model and logically implementing an information extraction function, the information extraction module comprising a data extraction unit and a financial announcement extraction unit, the data extraction unit being responsible for extracting financial information such as company entities and important shareholders from the Internet, organizing the data into a form required for building a financial knowledge graph, comprising the following steps:

[0006] Step S01: first, select the data source;

[0007] Step S02: extract the page data and interface data separately;

[0008] Step S03: data cleaning is performed on the information extracted from the two;

[0009] Step S04: data cleaning and fusion by using regularization means, removing inconsistent data, and forming a reliable and complete information base;

[0010] The financial announcement extraction unit comprises a data preprocessing subunit, an event sentence positioning subunit, and an event extraction subunit. The data preprocessing subunit preprocesses the announcement text information published by listed companies through a text classification model and a named entity recognition model, filters out announcements containing rich knowledge information from a large amount of text data, and transmits the announcement data of specific categories to the event sentence positioning subunit after preprocessing the data. The event sentence positioning subunit rapidly and accurately locates the professional expression language in the announcement text after receiving the announcement data transmitted by the data preprocessing subunit, locates the central event sentence, selects entity information, trigger word information, quantity word information, and title similarity as the features of sentence classification. The entity information feature calculation formula is S entity (S i )=C(S i ), wherein C(S i ) represents the number of organization names and names contained in the sentence S i , the trigger word information feature calculation formula is: The quantity word information feature calculation formula is S number (S i )=count(S i ), wherein count(S i ) represents the number of quantity and date words contained in the sentence S i , and the title similarity feature calculation formula is: wherein w si represents a sentence vector, S cos is the cosine similarity of two sentences, and the feature calculation formula of the announcement text Sk is the linear combination of the above four feature indicators, that is: S(S k )=∑w i s i (S k ), wherein i∈{entity、keyverb、number、cos}, the event extraction subunit uses the semantic role labeling method to do semantic role labeling on the core predicate in the event sentence, identifies the agent and the patient of the predicate, establishes a semantic role labeling model, and improves the labeling accuracy. The semantic labeling model formula is Y t =f(W1x t )1≤t≤N, wherein x t is the vector representation of the tth word, W1 is the weight matrix of nonlinear transformation, and f represents the tanh function.

[0011] The front-end interaction module is responsible for providing a system interface to a user and realizing interaction of user data and background data, and is an interface for communication between the system and the user.

[0012] Further, the data acquisition unit selects a financial related information source from the Internet, crawls network data, cleans the data obtained from the information source to obtain structured applicable data, the data storage unit first stores the extracted data by using a relational database, then constructs a triple according to a designed financial knowledge graph, and finally stores the triple in a graph database for persistent processing, and the query service unit is responsible for realizing a query function of the background, connecting a database to realize query on user input data and returning a query result.

[0013] Further, the text classification model is a CNN+LSTM model, and the named entity recognition model regards the entity recognition problem as sequence labeling, uses a conditional random field (CRF) to learn the relationship between labels, and solves the sequence labeling problem.

[0014] Further, the query step of the query service unit is as follows:

[0015] Step S11: a user inputs query information through a front-end interaction interface;

[0016] Step S12: the query service unit transmits the query information to different query functions, generates corresponding query statements, connects the data acquisition unit to perform information query, and acquires query data;

[0017] Step S13: returns the query result to the front end for display and output.

[0018] Further, the data storage unit uses a Neo4j graph database as a knowledge graph triple storage tool.

[0019] Further, the functions of the front-end interaction module include system function display, information transmission function, data rendering function and knowledge graph visualization function.

[0020] Further, the step S02 of collecting information from page data by using a crawler technology is as follows:

[0021] Step S21: acquires a URL of a target web page, and sends an HTTP request to the URL;

[0022] Step S22: parses a returned HTTP response, and judges whether a crawler stopping condition is met;

[0023] Step S23: stores a parsing result.

[0024] The technical effects and advantages of the present application are as follows:

[0025] 1. The present application is advantageous for information extraction of listed company financial announcements by constructing a financial field knowledge graph through the data extraction unit and the financial announcement extraction unit, and through text classification, event sentence positioning and fine-grained extraction to extract information from the information disclosure announcement in the financial field. Financial practitioners can directly obtain the key content of the financial entities and relationships and financial announcements they want to understand, effectively reduce the human cost consumed by regulatory agencies when reviewing information, and more efficiently supervise listed companies. At the same time, the structured company information can form a comprehensive understanding of the current situation of the company, so that correct decisions can be made quickly when investing. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The figure is a structural diagram of the comprehensive financial service system of the present application.

[0027] Figure 2 The figure is a step flow chart of the data extraction unit of the present application.

[0028] Figure 3 The figure is a flow chart of the financial announcement extraction unit of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the present application will be described clearly and completely in combination with the drawings in the present application. In addition, the forms of each structure described in the following embodiments are only examples, and the comprehensive financial service system involved in the present application is not limited to each structure described in the following embodiments. All other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of the present application.

[0030] The terms "module", "system" and the like used in the present application are intended to include computer-related entities such as, but not limited to, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a module can be, but is not limited to, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and / or a computer. For example, an application running on a computing device and the computing device can both be modules. One or more modules can be located within a process and / or thread of execution, and a module can also be located on a computer and / or distributed between two or more computers.

[0031] This invention provides a comprehensive financial service system, including a knowledge query module, an information extraction module, and a front-end interaction module. The knowledge query module includes a data acquisition unit, a data storage unit, and a query business unit. The information extraction module is mainly responsible for establishing the information extraction model and implementing the logic of the information extraction function. The information extraction module includes a data extraction unit and a financial announcement extraction unit. The data extraction unit is responsible for extracting financial information from the Internet, such as information on company entities and major shareholders, and organizing the data into the form required to construct a financial knowledge graph, including the following steps:

[0032] Step S01: First, select the data source;

[0033] Step S02: Extract page-type data and interface-type data separately. Page-type data uses web crawling and web page parsing techniques to obtain target information, while interface-type data is extracted by connecting to and calling the corresponding interfaces.

[0034] Step S03: Perform data cleaning on the information extracted from both sources;

[0035] Step S04: Use rule-based methods to clean and integrate the data, remove inconsistent data, and form a reliable and complete information database;

[0036] The financial announcement extraction unit includes a data preprocessing subunit, an event sentence localization subunit, and an event extraction subunit. The data preprocessing subunit preprocesses the text information of announcements released by listed companies using a text classification model and a named entity recognition model, filtering out announcements containing rich knowledge information from a large amount of text data. After preprocessing, the data preprocessing subunit selects several specific categories of announcement data and transmits them to the event sentence localization subunit. Upon receiving the announcement data from the data preprocessing subunit, the event sentence localization subunit quickly and accurately locates the professional descriptive language in the announcement text, pinpointing the central event sentence. It selects four features—entity information, trigger word information, quantifier information, and similarity to the title—as sentence classification features. The formula for calculating the entity information feature is: S... entity (S i )=C(S i ), where C(S i ) represents sentence S i The number of organization names and personal names contained therein, and the formula for calculating the trigger word information feature is as follows: The formula for calculating the quantifier information feature is: S number (S i ) = count(S i ), where count(S) i ) represents sentence Si The title similarity feature calculation formula is as follows: Wherein, w si represents a sentence vector, S cos is the cosine similarity of two sentences, and the feature calculation formula of the sentence Sk in the announcement text is a linear combination of the above four feature indexes, namely: S(S k )=∑w i s i (S k ), wherein i∈{entity, keyverb, number, cos}, the event extraction subunit performs more fine-grained information extraction on the event sentence after obtaining the event sentence, mainly extracts event elements, and the event elements include entities, event triggers, event arguments, argument roles, etc. The event extraction subunit performs semantic role labeling on the core predicate in the event sentence by using the semantic role labeling method, identifies the agent and the patient of the predicate, establishes a semantic role labeling model, and improves the labeling accuracy. The formula of the semantic labeling model is: Y t =f(W1x t )1≤t≤N, wherein x t is the vector representation of the tth word, W1 is the weight matrix of the nonlinear transformation, and f represents the tanh function.

[0037] The front-end interaction module is responsible for providing a system interface to the user and realizing the interaction between user data and background data, and is an interface for the system and the user to communicate.

[0038] In a preferred embodiment, the data acquisition unit selects a financial related information source from the Internet, crawls network data, cleanses the data obtained from the information source to obtain structured applicable data, the data storage unit first stores the extracted data by using a relational database, then constructs triples according to a designed financial knowledge graph, and finally stores the triples in a graph database for persistent processing, and the query business unit is responsible for realizing the query function of the background, connecting the database to realize the query of the user input data and returning the query result.

[0039] In a preferred embodiment, the text classification model is a CNN+LSTM model, the named entity recognition model regards the entity recognition problem as sequence labeling, uses a conditional random field (CRF) to learn the relationship between labels, and solves the sequence labeling problem.

[0040] In a preferred embodiment, the query step of the query business unit is as follows:

[0041] Step S11: The user inputs query information through the front-end interaction interface.

[0042] Step S12: the query service unit passes the query information to different query functions to generate corresponding query statements, connects with the data acquisition unit to query information, and acquires query data;

[0043] Step S13: the result after the query is returned to the front end for display and output;

[0044] In a preferred embodiment, the data storage unit uses a Neo4j graph database as a knowledge graph triple storage tool.

[0045] In a preferred embodiment, the functions of the front-end interaction module include system function display, information transmission function, data rendering function, and knowledge graph visualization function, realizing the interactive interface for the user to query knowledge and the interactive interface for the user to upload files to extract announcement information, and returning the query information of the user visually.

[0046] In a preferred embodiment, the step of collecting information from page data by using the crawler technology in step S02 is as follows:

[0047] Step S21: obtaining the URL of the target webpage, and sending an HTTP request to the URL;

[0048] Step S22: parsing the returned HTTP response to determine whether the crawler stopping condition is met;

[0049] Step S23: storing the parsed result.

[0050] In the present embodiment, it needs to be specifically pointed out that the difference between the present embodiment and the prior art mainly lies in that the present embodiment has a data extraction unit and a financial announcement extraction unit, constructs a financial field knowledge graph to extract information from the financial announcements of listed companies, extracts information from the information disclosure announcements in the financial field through text classification, event sentence positioning, and fine-grained extraction, financial practitioners can directly obtain the financial entities and relationships and the key content of the financial announcements that they want to understand, can effectively reduce the human cost consumed by the regulatory agencies when checking information, thereby more efficiently supervising the listed companies, and at the same time, can form a comprehensive understanding of the current situation of the company through the structured company information, thereby quickly making correct decisions when investing.

[0051] Finally, the above only describes the preferred embodiments of the present application and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0052] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0053] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0054] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0055] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0056] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0057] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A comprehensive financial service system, comprising a knowledge query module, an information extraction module, and a front-end interaction module, characterized in that: The knowledge query module includes a data acquisition unit, a data storage unit, and a query business unit. The information extraction module is mainly responsible for establishing the information extraction model and implementing the logic of the information extraction function. The information extraction module includes a data extraction unit and a financial announcement extraction unit. The data extraction unit is responsible for extracting financial information from the Internet and organizing the data into the form required to construct a financial knowledge graph, including the following steps: Step S01: First, select the data source; Step S02: Extract page-type data and interface-type data separately; Step S03: Perform data cleaning on the information extracted from both sources; Step S04: Use rule-based methods to clean and integrate the data, remove inconsistent data, and form a reliable and complete information database; The financial announcement extraction unit includes a data preprocessing subunit, an event sentence localization subunit, and an event extraction subunit. The data preprocessing subunit preprocesses the text information of announcements released by listed companies using a text classification model and a named entity recognition model, filtering out announcements containing rich knowledge information from a large amount of text data. After preprocessing, the data preprocessing subunit selects several specific categories of announcement data and transmits them to the event sentence localization subunit. Upon receiving the announcement data from the data preprocessing subunit, the event sentence localization subunit quickly and accurately locates the professional descriptive language in the announcement text, pinpointing the central event sentence. It selects four features—entity information, trigger word information, quantifier information, and similarity to the title—as sentence classification features. The formula for calculating the entity information feature is: S... entity (S i )=C(S i ), where C(S i ) represents sentence S i The number of organization names and personal names contained therein, and the formula for calculating the trigger word information feature is as follows: The formula for calculating the quantifier information feature is: S number (S i ) = count(S i ), where count(S) i ) represents sentence S i The formula for calculating the title similarity feature is as follows: (Including the number of representations and the number of date words) Among them, w si S represents the sentence vector. c os is the cosine similarity between two sentences. The feature calculation formula for the statement Sk in the announcement text is a linear combination of the above four feature indicators, namely: S(S k )=∑w i s i (S k ), where i∈{entity、keyverb、number、cos}, the event extraction subunit uses semantic role labeling to perform semantic role labeling on the core predicates in the event sentence, identifies the agent and patient of the predicate, and establishes a semantic role labeling model. The formula for the semantic role labeling model is: Y t =f(W1x t )1≤t≤N, where x t W1 is the vector representation of the t-th word segment, W1 is the weight matrix of the nonlinear transformation, and f represents the tanh function. The front-end interaction module is responsible for providing the system interface to the user and realizing the interaction between user data and back-end data. It is the interface for communication between the system and the user.

2. The integrated financial service system according to claim 1, characterized in that: The data acquisition unit selects financial information sources from the Internet, crawls the network data, cleans the data obtained from the information sources to obtain structured and applicable data, and the data storage unit first stores the extracted data using a relational database, then constructs triples based on the designed financial knowledge graph, and finally stores it in a graph database for persistent processing. The query business unit is responsible for implementing the query function in the background, connecting to the database to query the user input data and returning the query results.

3. The integrated financial service system according to claim 1, characterized in that: The text classification model is a CNN+LSTM model, and the named entity recognition model treats the entity recognition problem as sequence labeling, using a conditional random field (CRF) to learn the relationship between labels and solve the sequence labeling problem.

4. The integrated financial service system according to claim 2, characterized in that: The query steps for the query business unit are as follows: Step S11: The user enters query information through the front-end interactive interface; Step S12: The query business unit transmits the query information to different query functions, generates corresponding query statements, connects with the data acquisition unit to perform information query, and obtains query data; Step S13: Return the query results to the front end for display and output.

5. A comprehensive financial service system according to claim 2, characterized in that: The data storage unit uses Neo4j graph database as the knowledge graph triple storage tool.

6. The integrated financial service system according to claim 1, characterized in that: The functions of the front-end interaction module include system function display, information transmission, data rendering, and knowledge graph visualization.

7. The integrated financial service system according to claim 1, characterized in that: The steps in step S02 for collecting information from page-type data using web crawling technology are as follows: Step S21: Obtain the URL of the target webpage and send an HTTP request to that URL; Step S22: Parse the returned HTTP response and determine whether the crawler stops. Step S23: Store the parsing results.

Images