Financial scene searching and sorting method and device fusing category features and medium

Abstract

The first aspect of the technical scheme of the invention is to disclose a financial scene searching and sorting method fusing category features. The second aspect of the technical scheme of the invention is to disclose an electronic device. The third aspect of the technical scheme of the invention is to disclose a computer readable storage medium. A traditional keyword retrieval method is difficult to meet the requirement of a user for the financial information acquisition accuracy, and compared with the prior art, the method based on document category prediction can greatly improve the information accuracy, effectively reduce the burden of screening invalid information by the user and improve the information acquisition efficiency.

Description

Technical Field

[0001] This invention relates to a method, device, and medium for effectively searching financial documents in a financial setting by utilizing available document category features and ranking the documents that best meet the user's needs at the top. Background Technology

[0002] With the rapid development of the internet and the exponential growth of various data, users face a massive amount of information resources, making the efficient and accurate acquisition of needed information a major challenge. Search engines, as a crucial technology for solving this challenge, have become an indispensable core tool in modern society. In the financial sector, efficient information acquisition and accurate analysis are vital for decision-making, and search engines play a crucial role in this context. For example: 1) Financial professionals need to obtain the latest prices and trends of financial products such as stocks, bonds, and foreign exchange in real time, and search engines need to provide relevant data quickly and accurately to help users make immediate decisions; 2) When making investment decisions, investors often need to consult various research reports and market analysis articles. Search engines need to accurately match relevant content and provide in-depth analysis support. 3) Investment research personnel monitor specific keywords or events through search engines to promptly identify potential market risks, such as policy changes and industry crises, helping users take preventative measures in advance; 4) The financial industry is greatly affected by policies. Users need to keep up with the latest laws, regulations and policy developments, and search engines need to provide relevant interpretations and analyses quickly.

[0003] Therefore, financial users have a particularly strong demand for search engines, and optimizing search engines for use in the financial field has extremely high practical value and market prospects.

[0004] Unlike general web pages, financial documents feature more specialized categorization. For example, research reports can be divided into industry reports, company reports, and strategy reports; announcements can be categorized into financial reports, announcements of major events, and investor relations announcements; and news covers multiple subcategories such as market dynamics, policy interpretations, and company news. Furthermore, financial documents often contain additional attribute information, such as the author of the research report, the issuing institution, and the publisher of the announcement. Users' needs often explicitly or implicitly include a requirement for document categories. However, existing general search algorithms, when processing these highly specialized financial documents, often fail to consider the detailed categorization needs of users in a financial context, potentially leading to overly broad and untargeted search results that fail to effectively meet users' actual needs. Most vertical search engines, on the other hand, often rely solely on keyword matching or text similarity to retrieve relevant documents, requiring users to filter document category information by clicking different tabs, resulting in low user search efficiency. Summary of the Invention

[0005] The technical problem this invention aims to solve is that search ranking algorithms based on keyword search or text similarity matching are difficult to capture users' needs for professional information such as document categories in financial scenarios. While filtering through tab menus can help users narrow down their search scope to some extent, it is cumbersome and lacks flexibility. Users need to manually select multiple tabs for layer-by-layer filtering, which not only increases the user's operational burden but may also lead to missed or false detections due to insufficiently refined option settings or users' lack of understanding of the options.

[0006] To address the aforementioned technical problems, the first aspect of the present invention discloses a financial scenario search and ranking method that integrates category features, characterized by comprising the following steps: Step 1: Identify the document category information associated with the user's question and obtain the relevance between the current user's question and each document category; Step 2: Based on the pre-set relevance threshold and the relevance prediction results output in Step 1, select candidate document category groups with high relevance from all document categories; Step 3: Based on the candidate document category group obtained in Step 2, construct the recall number of different document categories in the candidate document category group; Step 4: Based on the recall counts for different document categories established in Step 3, construct the recall conditions for the document search API and perform document recall. Step 5: Based on the candidate document categories and relevance in the candidate document category group obtained in Step 2, use the pre-set features to perform rule-based weighted scoring on the documents recalled in Step 4. Based on the weighted scoring results, sort the recalled documents and return the number of documents requested by the front end.

[0007] Preferably, step 1 includes the following steps: Step 101: Train a BERT text multi-classification model based on classification data annotated by human experts to identify the relevance between user questions and document categories; Step 102: During online inference, use the trained BERT classification model to predict the relevance of user questions to documents of different categories.

[0008] Preferably, in step 101, when performing manual expert annotation on the categorized data: if the user's question explicitly mentions the document type required; if the user's question does not explicitly mention the document type required, then the annotation is performed based on which documents the relevant answer may appear in.

[0009] Preferably, step 3 includes the following steps: Step 301: Based on the relevance values ​​of different document categories, standardize to obtain the recall ratio of different document categories; Step 302: Calculate the recall quantity of documents in different categories based on the recall ratio of different document categories and the preset total recall quantity.

[0010] Preferably, in step 5, when performing rule-based weighted scoring, the relevance scores for different document categories are used as relevance weighting factors, and weighted together with the document relevance calculated from other features, as shown in the following formula:

[0011] In the formula, For users to ask questions, To recall the documents, For document category relevance, For text relevance, Rate the timeliness of the document. A combined score is given based on factors such as text relevance and timeliness.

[0012] The second aspect of the technical solution of the present invention discloses an electronic device, comprising: One or more processors; A storage device 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 software to implement the financial scenario search and ranking method based on the aforementioned fusion category features.

[0013] The third aspect of the technical solution of the present invention discloses a computer-readable storage medium storing executable instructions thereon, which, when executed by a processor, cause the processor to execute a financial scenario search and ranking method based on the aforementioned fusion category features.

[0014] Traditional keyword retrieval methods struggle to meet users' demands for accurate financial information retrieval. In contrast, the document category prediction method proposed in this invention can significantly improve information accuracy, effectively reduce the burden on users to filter out invalid information, and enhance information retrieval efficiency. Specifically, it has the following beneficial effects: 1) BERT text classification models based on the pre-training-fine-tuning paradigm have been widely used in various intelligent systems, and their accuracy and adaptability have been widely verified. 2) By analyzing the correlation between user questions and different document categories, irrelevant documents can be effectively filtered out, increasing the probability of hitting the correct document; 3) By using document category-weighted scoring and sorting, the most relevant documents can be effectively obtained, increasing the probability of hitting the homepage. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method disclosed in this invention. Detailed Implementation

[0016] Various aspects and features of the present invention are described herein with reference to the accompanying drawings.

[0017] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this disclosure will be apparent to those skilled in the art.

[0018] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the invention and, together with the general description of the invention given above and the detailed description of the embodiments given below, serve to explain the principles of the invention.

[0019] These and other features of the invention will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0020] It should also be understood that although the invention has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of the invention, which have the features described in the claims and are therefore all within the scope of protection defined herein.

[0021] The above and other aspects, features and advantages of the invention will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0022] Specific embodiments of the invention are described below with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of the invention, which may be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that would obscure this disclosure. Therefore, the specific structural and functional details claimed in this invention are not intended to be limiting, but are merely intended as the basis and representative basis for the claims to teach those skilled in the art to use this disclosure in various ways with substantially any suitable detailed structure.

[0023] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in still another embodiment,” all of which may refer to one or more of the same or different embodiments according to this disclosure.

[0024] To address the issue of professionalism in current financial search scenarios, this invention proposes an automated technical solution that considers users' detailed needs for document categories and attributes, which can greatly improve the efficiency and accuracy of document retrieval in financial scenarios.

[0025] Combination Figure 1 The first aspect of this invention discloses a search and ranking method for financial scenarios that integrates category features, comprising the following steps: Step 1: Identify the document category information associated with the user's question and obtain the relevance between the current user's question and each document category.

[0026] In a preferred embodiment of the present invention, step 1 further includes the following steps: Step 101: Train a BERT text multi-classification model based on classification data annotated by human experts to identify the relevance between user questions and document categories.

[0027] In another preferred embodiment of the present invention, when performing manual expert annotation on the categorized data: if the user's question explicitly mentions the required document type, such as research report, announcement, news, etc., the category is labeled as the mentioned document type; if the user's question does not explicitly mention the required document type, the annotation is made based on which documents the relevant answer may appear in.

[0028] Step 102: During online inference, the trained BERT classification model is used to predict the relevance between the user's question and documents of different categories. In a preferred embodiment of this invention, the BERT classification model outputs 0-1 to represent the relevance; for example, 1 can be used to represent relevance, and 0 is used to represent irrelevance.

[0029] Step 2: Based on the pre-set relevance threshold and the relevance prediction results output in Step 1, select candidate document category groups with high relevance from all document categories.

[0030] In a preferred embodiment of the present invention, step 2 specifically involves: predicting the relevance between different document categories and the current user's question using the BERT classification model from step 1. The obtained result can be represented as [(category 1, relevance 1), (category 2, relevance 2), ... ]. Based on a preset relevance threshold, document categories with low relevance are filtered out to obtain candidate document category groups.

[0031] Step 3: Based on the candidate document category group obtained in Step 2, construct the recall number of different document categories in the candidate document category group.

[0032] In a preferred embodiment of the present invention, step 3 further includes the following steps: Step 301: Based on the relevance values ​​of different document categories, standardize to obtain the recall ratio of different document categories; Step 302: Calculate the recall quantity of documents in different categories based on the recall ratio of different document categories and the preset total recall quantity.

[0033] Step 4: Based on the recall counts of different document categories constructed in Step 3, construct the recall conditions for the document search API and perform document recall.

[0034] In one preferred embodiment of the present invention, when retrieving documents, the backend retrieval API is called for different document categories to retrieve the documents.

[0035] Step 5: Based on the candidate document categories and relevance in the candidate document category group obtained in Step 2, use other features, including text similarity and timeliness, to perform rule-based weighted scoring on the documents recalled in Step 4. Based on the weighted scoring results, sort the recalled documents and return the number of documents requested by the front end.

[0036] In a preferred embodiment of the present invention, when performing rule-based weighted scoring, the relevance scores of different document categories are used as relevance weighting factors, and the document relevance calculated with other features is weighted and scored, as shown in the following formula:

[0037] In the formula, For users to ask questions, To recall the documents, For document category relevance, For text relevance, Rate the timeliness of the document. A combined score is given based on factors such as text relevance and timeliness.

[0038] A second aspect of this invention is to provide an electronic device, including at least a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-described financial scenario search and ranking method when executing the computer program in the memory.

[0039] In some embodiments, the processor executing a computer program may be a processing device that includes one or more general-purpose processing devices, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor that runs other instruction sets, or a processor that runs a combination of instruction sets. The processor may also be one or more special-purpose processing devices, such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), system-on-a-chip (SoCs), etc.

[0040] The memory may be a read-only memory (ROM), random access memory (RAM), phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), flash drives or other forms of flash memory, cache, registers, static memory, optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape cassette or other magnetic storage devices, or any other possible non-transitory medium used to store information or instructions that can be accessed by computer equipment.

[0041] A third aspect of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned financial scenario search and ranking method.

[0042] The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. In this invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device; for example, it can be the aforementioned memory.

[0043] The computer programs of embodiments of the present invention can be organized into one or more computer-executable components or modules. Various aspects of this disclosure can be implemented with any number and combination of such components or modules. For example, aspects of this disclosure are not limited to the specific computer-executable instructions or specific components or modules shown in the drawings and described herein. Other embodiments may include different computer-executable instructions or components having more or fewer functions than those shown and described herein.

[0044] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A search and ranking method for financial scenarios that integrates category features, characterized in that, Includes the following steps: Step 1: Identify the document category information associated with the user's question and obtain the relevance between the current user's question and each document category; Step 2: Based on the relevance prediction results output in Step 1, filter out candidate document category groups with high relevance from all document categories according to the pre-set relevance threshold; Step 3: Based on the candidate document category groups obtained in Step 2, construct the recall quantity of different document categories in the candidate document category groups; Step 4: Based on the recall quantity of different document categories constructed in Step 3, construct the recall conditions for the document search API and perform document recall; Step 5: Based on the candidate document categories and relevance in the candidate document category groups obtained in Step 2, use the pre-set auxiliary features to perform rule-based weighted scoring on the documents recalled through Step 4, sort the recalled documents based on the weighted scoring results, and return the number of documents requested by the front end.

2. The financial scenario search and ranking method integrating category features as described in claim 1, characterized in that, Step 1 includes the following steps: Step 101, train a BERT text multi-classification model to identify the relevance between user questions and document categories based on classification data annotated by human experts; Step 102, during online inference, use the trained BERT classification model to predict the relevance between user questions and documents of different categories.

3. The financial scenario search and ranking method integrating category features as described in claim 2, characterized in that, In step 101, when performing manual expert annotation on the categorized data: if the user's question explicitly mentions the document type required; if the user's question does not explicitly mention the document type required, then the annotation is performed based on which documents the relevant answer might appear in.

4. The financial scenario search and ranking method integrating category features as described in claim 1, characterized in that, Step 3 includes the following steps: Step 301, standardize the relevance values ​​of different document categories to obtain the recall ratio of different document categories; Step 302, calculate the recall quantity of different document categories based on the recall ratio of different document categories and the preset total recall quantity.

5. The financial scenario search and ranking method integrating category features as described in claim 1, characterized in that, In step 5, when performing rule-based weighted scoring, the relevance scores for different document categories are used as relevance weighting factors, and weighted together with the document relevance calculated from other features, as shown in the following formula: In the formula, For users to ask questions, To recall the documents, For document category relevance, For text relevance, Rate the timeliness of the document. A combined score is given based on factors such as text relevance and timeliness.

6. An electronic device, comprising: One or more processors; A storage device 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 software to implement the financial scenario search and ranking method according to claim 1.

7. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the financial scenario search and ranking method according to claim 1.

Images