Method for automatically generating broadcast video of enterprise financial news based on AIGC

Abstract

The invention discloses a method for automatically generating a broadcast video for enterprise financial news on the basis of AIGC (Artificial Intelligence Generation Content), and relates to a method for automatically generating a broadcast video for enterprise financial news on the basis of AIGC (Artificial Intelligence Generation Content). According to the method, key information in enterprise financial news is extracted through the AIGC technology, a data video is automatically generated, and the method aims at improving the understanding ability of green investors for complex financial information. On the basis, the invention further develops an interactive interface, the controllability of the AI aided design process is enhanced, and a user can flexibly control the video generation process and adjust the video content generated by the AI. Compared with the prior art, the problem of simplifying and transmitting complex news narration containing various financial information is effectively solved. Core information is extracted from enterprise financial news, a video script is generated in combination with a large language model (LLM) based on a knowledge base, and the generative model is used for generating news pictures. Meanwhile, the system automatically retrieves financial data of related enterprises and generates a visual chart easy to understand, and a user does not need to manually retrieve related information. Finally, the visual elements are automatically arranged and organized into a coherent news broadcast video. According to the method, a content creator can be helped to quickly generate a high-quality enterprise financial news video, comprehensive and in-depth enterprise financial news interpretation is provided for green investors, and the method has high practical value and wide development prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic product generation methods, and in particular to an AIGC-based method for automatically generating broadcast videos of corporate financial news. Background Art

[0002] Corporate financial news can provide investors with insights into the company's overall financial health and growth prospects, and is an important channel for obtaining company dynamics and financial information. However, traditional corporate financial news is usually long, information-intensive, and has complex financial concepts, which increases the difficulty for novice investors to understand. Data videos significantly improve the ability of non-professionals to understand financial information, and improve the efficiency and effectiveness of information retrieval and insight understanding.

[0003] There are already some tools for automatically generating information visualizations in financial news to reduce the workload of news creators and lower the threshold for information understanding of financial news. However, most of these automated tools are limited to generating a single financial information visualization and cannot cover the complete news narrative. For example, generating a visualization of a company's stock market trend. In practical applications, the financial information contained in corporate financial news not only involves company stocks, but also unique attributes such as company shareholders, products, and business activities. In addition, creating corporate financial news broadcast videos is challenging and requires news creators to have both storytelling and data video production capabilities. Existing tools that assist designers in creating data videos mainly focus on coordinating narratives and visualization animations to guide users to easily create universal data videos. However, these methods lack support for handling financial narratives and coordinating the design of visual elements, and still require authors to have high professional skills and use independent tools. At present, there is no method that fully considers the diversity of financial information contained in corporate financial news and automatically generates broadcast videos containing relevant financial data visualizations. The emergence of AIGCs provides a new way to automatically create such videos because of their ease of use and powerful generation capabilities. In view of this, we propose a method for automatically generating broadcast videos for corporate financial news based on AIGCs. Summary of the invention

[0004] The purpose of the present invention is to propose a method for automatically generating broadcast videos of corporate financial news based on AIGC to solve the problems in the background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for automatically generating broadcast videos of corporate financial news based on AIGC, comprising the following contents:

[0007] S1. Set the data time range of the broadcast video and input the style of the news picture through natural language, and input one or more corporate financial news texts;

[0008] S2, a large language model (LLM) based on the corporate financial news knowledge base, using prompt word engineering to generate prompt word instructions to reproduce key information in the news;

[0009] S3, formulating a series of instructions based on the prompt word engineering described in S2 to guide a large language model (LLM) to generate prompt words for a text-to-image generation model;

[0010] S4. Retrieve and match relevant data facts based on the financial attributes of each chapter;

[0011] S5. Divide the video clips into narrative clips and visual clips according to the visual elements that appear. Apply the element layout algorithm based on saliency detection to layout the visual elements in the video. Use the BASNet model to extract the saliency map from the original chart and use it as the basis for generating the visual element layout. The visual elements after layout will match the corresponding animation template.

[0012] S6. Generate multiple broadcast video clips, automatically splice and output the corporate financial news broadcast video.

[0013] Preferably, the enterprise financial news knowledge base in S2 includes knowledge on financial attributes and inverted pyramid news paradigm and chapter-based organizational structure;

[0014] The financial attributes are eight privately defined attributes that are closely related to the financial information of an enterprise; the definition of the financial attributes includes attribute name, attribute description, financial data facts and news instances;

[0015] The inverted pyramid news paradigm and chapter-based organizational structure refer to relevant academic papers. Long texts are segmented into semantically meaningful financial knowledge fragments. Text embedding is used to convert text data into continuous vector representations, and a local knowledge base is established to perform searches in large data sets.

[0016] Preferably, the prompt word instruction in S2 includes four subtasks, specifically including the following contents:

[0017] S2.1. Divide the text T into n paragraphs p according to the context information i , where i = 1,…,n;

[0018] S2.2. For each paragraph p i , extract its corresponding financial attribute a i ;

[0019] S2.3. Paragraphs with the same financial attributes {pi ,p j |i,j∈[1,n],a i =a j}Reorganized into the same section S;

[0020] S2.4. Summarize each chapter S and generate m concise subtitle sentences.

[0021] Preferably, the instruction in S3 requires a large language model (LLM) to play the role of an illustrator to constitute basic information of image style, company and industry;

[0022] The tasks of the large language model (LLM) are: to expand the subtitle narration into a specific visual scene, refine the scene to generate appropriate prompt words as additional information; to generate the final prompt for each image by retaining the original basic information and modifying the additional information, and to input the generated prompt word text into the image generation model in turn to produce a set of images with consistent style.

[0023] Preferably, the process of metal property retrieval in S4 is described as follows:

[0024] Retrieve(C,A,T)={(a,f a )|a∈A,f a ∈S(C,T)}

[0025] Where C represents the formula; A represents the financial attribute set; T represents the time range; a represents the specific attribute; f a Indicates data facts; S indicates data source;

[0026] Each attribute a in the financial attribute set A corresponds to a set of data facts f a , the data fact f a Retrieved from data source S, relevant to company C within time frame T; structured data is converted into visualization icons through encoding templates, and narrative icon annotations are generated using large language models (LLM).

[0027] Preferably, the S5 specifically includes the following contents:

[0028] S5.1. Use single-layer spatial pyramid pooling technology to reduce the search space, adjust the saliency map to the size of the sub-window, and perform mean pooling operation on each sub-window; from the pooled saliency map, select the most and least significant pixels as the basis for subsequent processing;

[0029] S5.2. For narrative editing, a template matching algorithm is used to maximize the presentation of salient information in the image. After identifying the salient area, its spatial distribution characteristics are analyzed. Based on the spatial distribution characteristics, the system classifies the salient area and selects a suitable horizontal or vertical template. For salient areas covering the image, the system automatically switches to a full-screen template and applies masking technology to ensure the clarity and visual effect of the title when displayed in full screen.

[0030] S5.3. For visual clipping, the layout of its visual elements (charts and text) follows the three rules of non-overlapping, readability, and proximity;

[0031] The non-overlapping means that the text annotation box does not interfere with the key elements of the chart, and the significance threshold is determined based on the highest significance value among the least significant pixels to ensure that the candidate area does not exceed the threshold;

[0032] The readability means that the size of the text annotation box is appropriate and easy to read; each least significant pixel point is expanded horizontally and vertically to become a candidate region until a pixel point exceeding the significance threshold is encountered, and the region that meets the size standard is retained, the coordinates are mapped back to the original scale, and non-maximum suppression is applied to remove redundant candidate regions; the selection of the optimal text annotation box is based on horizontal alignment, close to the golden ratio and maximum size. After calculating the ratio index, the annotation box with the smallest ratio is selected as the optimal annotation box;

[0033] The proximity refers to placing text annotation boxes near important areas to enhance visibility and relevance, and identifying the four nearest salient pixels in each corner to fine-tune the position of the optimal annotation box.

[0034] Compared with the prior art, the present invention provides a method for automatically generating broadcast videos of corporate financial news based on AIGC, which has the following beneficial effects:

[0035] This paper proposes a method for automatically generating broadcast videos of corporate financial news based on AIGC, aiming to reduce the workload of news creators and lower the threshold for understanding financial news information. This paper creatively proposes eight key financial attributes in corporate financial news, cleverly uses LLM's ability to process natural language to extract key insights from news, generates news pictures through a text-to-image generative model, automatically retrieves data facts, creates visualizations containing financial information, and organizes them into news broadcast videos. This automated method not only allows users to create high-quality corporate financial news broadcast videos more quickly, but also allows viewers to understand the company's financial status more intuitively and comprehensively, filling the gap in existing research in this field. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1This is a schematic diagram of the process of automatically generating a corporate financial news broadcast video mentioned in Example 1 of the present invention;

[0037] Figure 2 The eight financial attributes and their corresponding financial data facts mentioned in Embodiment 3 of the present invention;

[0038] Figure 3 This is a schematic diagram of an example of generating a broadcast video from corporate financial news mentioned in Example 3 of the present invention. DETAILED DESCRIPTION

[0039] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0040] Embodiment 1:

[0041] The present invention proposes a framework NewsVis (news visualization), which uses artificial intelligence generated content (AIGC) technology to extract key information from corporate financial news and automatically matches relevant financial data visualization to generate news broadcast videos. The framework includes three modules: news reproduction, visualization recommendation, and visual element organization. The news reproduction module aims to automatically identify financial attributes and reorganize narratives, while generating news images using a text-to-image generative model. The visualization recommendation module can automatically retrieve relevant industry and company earnings data based on the extracted financial attributes and generate expressive annotated visualizations. The visual element organization module focuses on heterogeneous visual element layout algorithms based on saliency detection, and uses animation templates to organize elements into coherent data videos. On this basis, the present invention also develops an interactive interface that enables users to adjust and improve the results generated by AI.

[0042] See also Figure 1 The present invention automatically generates broadcast videos of corporate financial news according to the following steps:

[0043] Step 1: Set the time range and image style, and enter the corporate financial news text;

[0044] Step 2: Divide the task into four subtasks to reproduce the key information in the news;

[0045] Step 3: Apply the text-to-image generative model to generate news pictures for key information;

[0046] Step 4: Retrieve and match relevant data facts to generate annotated data visualizations;

[0047] Step 5: Apply layout algorithms to visual elements and match animation templates;

[0048] Step 6: Generate multiple video clips and output the spliced ​​broadcast video.

[0049] Embodiment 2:

[0050] Based on Example 1, but different from Example 1, the present invention proposes a method for automatically generating a broadcast video of corporate financial news based on AIGC, which specifically includes the following contents:

[0051] Step a:

[0052] Set the data time range of the broadcast video and input the style of the news picture through natural language, and input one or more corporate financial news texts;

[0053] Step b:

[0054] With the support of the large language model (LLM) based on the corporate financial news knowledge base, the key information in the news is reproduced using prompt word engineering. The corporate financial news knowledge base is characterized by covering the following knowledge: financial attributes, inverted pyramid news paradigm, and chapter-based organizational structure. The financial attributes are eight privately defined attributes that are closely related to corporate financial information (such as Figure 2 As shown in Figure 1). The definition of each financial attribute includes attribute name, attribute description, financial data facts, and news examples; the knowledge of the inverted pyramid news paradigm and chapter-based organizational structure is a related academic paper. Long texts are segmented into semantically meaningful financial knowledge fragments, and text embedding is used to convert text data into continuous vector representations to establish a local knowledge base that can efficiently perform searches in large data sets. The prompt word instruction includes four subtasks, and the specific steps are as follows:

[0055] First, the text T is divided into n paragraphs p according to the context information i , where i = 1,…,n; for each paragraph p i , extract its corresponding financial attribute a i Next, the paragraphs with the same financial attributes {p i ,p j |i,j∈[1,n],a i =a j} are reorganized into the same chapter S. Finally, each chapter S is summarized to generate m concise subtitle sentences.

[0056] Step c:

[0057] A series of instructions are formulated to guide the LLM in generating prompt words for the text-to-image generative model. The instructions require the LLM to play the role of an illustrator, focusing on the basic information that constitutes the image style, company, and industry. The LLM's task is to first expand the caption narrative into a specific visual scene, and then refine the scene to generate appropriate prompt words as additional information. The final prompt for each image is generated by retaining the original basic information and modifying the additional information. The generated prompt words are sequentially input into the text-to-image generative model to produce a set of images with consistent style.

[0058] Step d:

[0059] According to the financial attributes of each chapter, relevant data facts are retrieved and matched. The retrieval process can be described as Retrieve(C,A,T)={(a,f a )|a∈A,f a ∈S(C,T)}, where each attribute a in the financial attribute set A corresponds to a set of data facts f a , these data facts are retrieved from data source S and are related to companies (C) in time frame (T). The structured data are converted into visual diagrams through a carefully designed coding template, and narrative diagram annotations are generated using LLM.

[0060] Steps:

[0061] The video clips are divided into narrative clips and visual clips according to the visual elements that appear. The element layout algorithm based on saliency detection is applied to layout the visual elements in the video. The BASNet model is used to extract the saliency map from the original chart and use it as the basis for generating the visual element layout, which includes the following contents:

[0062] (1) Use single-layer spatial pyramid pooling technology to reduce the search space, adjust the saliency map to the sub-window size, and perform mean pooling on each sub-window. From the pooled saliency map, select the most and least salient pixels as the basis for subsequent processing.

[0063] (2) Narrative editing uses a precise template matching algorithm to maximize the presentation of salient information in the image. After identifying the salient areas, their spatial distribution characteristics, including aspect ratio and area, are analyzed. Based on these characteristics, the system classifies the salient areas and selects appropriate horizontal or vertical templates. For salient areas that cover most of the image area, the system automatically switches to a full-screen template and applies masking technology to ensure the clarity and visual effect of the title when displayed in full screen.

[0064] (3) The layout of visual elements (charts and text) in the visualization clip follows the following three rules: no overlap, readability, and proximity. No overlap means that the text annotation box should not interfere with the key elements of the chart. The saliency threshold is determined based on the highest saliency value among the least salient pixels, ensuring that the candidate region does not exceed this threshold; the readability is that the size of the text annotation box should be appropriate, preferably horizontal, to ensure easy reading. Each least salient pixel point becomes a candidate region by horizontally and vertically expanding until it stops at a pixel point that exceeds the saliency threshold. The region that meets the size criteria is retained and the coordinates are mapped back to the original scale. Non-maximum suppression is applied to remove redundant candidate regions and retain sufficiently large regions. The selection of the optimal text annotation box is based on factors such as horizontal alignment, proximity to the golden ratio, and maximum size. After calculating the ratio index, the annotation box with the smallest ratio is selected as the optimal annotation box; the proximity is that the text annotation box should be placed near the important area to enhance visibility and relevance. The four nearest salient pixels in each corner are identified to fine-tune the position of the optimal annotation box, making the box closer to the salient area and making the annotation layout more beautiful. The visual elements after layout will match the corresponding animation template.

[0065] Step f:

[0066] Generate multiple broadcast video clips, automatically splice and edit multiple video clips and output corporate financial news broadcast videos.

[0067] Embodiment 3:

[0068] Based on Example 1-2, but with the difference that, the method for automatically generating a broadcast video of a certain enterprise financial news based on AIGC proposed by the present invention is described below by taking a certain enterprise financial news broadcast video as an example, and the specific content is as follows:

[0069] Step 1: Enter a financial news N of company C in TXT text (such as Figure 3 left part), and set the time range and image style;

[0070] Step 2: With the support of LLM based on the corporate financial news knowledge base, the key information in N is reproduced using the prompt word engineering. The reproduction process includes: paragraph division, attribute extraction, paragraph merging, and summary subtitles;

[0071] Step 3: Input the generated subtitles and image styles into LLM to generate prompt words for image generation, and use the text-to-image generative model to generate news images;

[0072] Step 4: Based on the definition of financial attributes (such as Figure 2As shown in the figure, relevant financial data facts of enterprise C within the set time range are retrieved from the data source and converted into structured data, visual charts are generated through coding templates, and narrative chart annotations are generated using LLM;

[0073] Step 5: Divide the video clips into narrative clips and visual clips according to the visual elements that appear. Apply the element layout algorithm based on saliency detection to layout the visual elements in the video and match the corresponding animation templates;

[0074] Step 6: Generate multiple video clips and output the spliced ​​broadcast video generated by the financial news N of enterprise C (such as Figure 3 right part).

[0075] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for automatically generating broadcast videos of corporate financial news based on AIGC, characterized in that: The method specifically comprises the following steps: S1. Set the data time range of the broadcast video and input the style of the news picture through natural language, and input one or more corporate financial news texts; S2, a large language model based on the corporate financial news knowledge base, using prompt word engineering to generate prompt word instructions to reproduce key information in the news; S3, formulating a series of instructions based on the prompt word engineering described in S2 to guide the large language model to generate prompt words for the text-to-image generation model; S4. Retrieve and match relevant data facts based on the financial attributes of each chapter; S5. Divide the video clips into narrative clips and visual clips according to the visual elements that appear, apply the element layout algorithm based on saliency detection to layout the visual elements in the video, and use the BASNet model to extract the saliency map from the original chart as the basis for generating the visual element layout; S6. Generate multiple broadcast video clips, automatically splice and output the corporate financial news broadcast video.

2. According to claim 1, a method for automatically generating broadcast videos of corporate financial news based on AIGC is characterized in that: The knowledge base of corporate financial news described in S2 includes knowledge on financial attributes as well as the inverted pyramid news paradigm and chapter-based organizational structure; The financial attributes are eight privately defined attributes that are closely related to the financial information of an enterprise; the definition of financial attributes includes attribute name, attribute description, financial data facts and news instances; The inverted pyramid news paradigm and chapter-based organizational structure refer to relevant academic papers. Long texts are segmented into semantically meaningful financial knowledge fragments. Text embedding is used to convert text data into continuous vector representations, and a local knowledge base is established to perform searches in large data sets.

3. The method for automatically generating broadcast videos of corporate financial news based on AIGC according to claim 2 is characterized in that: The prompt word instruction in S2 includes four subtasks, specifically including the following contents: S2.

1. Divide the text T into n paragraphs p according to the context information i , where i = 1,…,n; S2.

2. For each paragraph p i , extract its corresponding financial attribute a i ; S2.

3. Paragraphs with the same financial attributes {p i ,p j |i,j∈[1,n],a i =a j }Reorganized into the same section S; S2.

4. Summarize each chapter S and generate m concise subtitle sentences.

4. The method for automatically generating broadcast videos of corporate financial news based on AIGC according to claim 3 is characterized in that: The instructions described in S3 require a large language model to play the role of an illustrator to compose basic information about the image style, company, and industry; The task of the large language model is to expand the subtitle description into a specific visual scene, extract the scene and generate appropriate prompt words as additional information; The final prompt for each image is generated by retaining the original basic information and modifying the additional information. The generated prompt word text is input into the image generation model in sequence to produce a set of images with consistent style.

5. The method for automatically generating broadcast videos of corporate financial news based on AIGC according to claim 4 is characterized in that: The process of metal property retrieval described in S4 is described as: Retrieve(C,A,T)={(a,f a )|a∈A,f a ∈S(C,T)} Where C represents the formula; A represents the financial attribute set; T represents the time range; a represents the specific attribute; f a Indicates data facts; S indicates data source; Each attribute a in the financial attribute set A corresponds to a set of data facts f a , the data fact f a Retrieved from data source S, relevant to company C within time frame T; structured data is converted into visual icons through encoding templates, and narrative icon annotations are generated using large language models.

6. The method for automatically generating broadcast videos of corporate financial news based on AIGC according to claim 5 is characterized in that: The S5 specifically includes the following contents: S5.

1. Use single-layer spatial pyramid pooling technology to reduce the search space, adjust the saliency map to the size of the sub-window, and perform mean pooling operation on each sub-window; from the pooled saliency map, select the most and least significant pixels as the basis for subsequent processing; S5.

2. For narrative editing, a template matching algorithm is used to maximize the presentation of salient information in the image. After identifying the salient area, its spatial distribution characteristics are analyzed. Based on the spatial distribution characteristics, the system classifies the salient area and selects a suitable horizontal or vertical template. For salient areas covering the image, the system automatically switches to a full-screen template and applies masking technology to ensure the clarity and visual effect of the title when displayed in full screen. S5.

3. For visual editing, the layout of visual elements follows the three rules of non-overlapping, readability, and proximity. The non-overlapping means that the text annotation box does not interfere with the key elements of the chart, and the significance threshold is determined based on the highest significance value among the least significant pixels to ensure that the candidate area does not exceed the threshold; The readability means that the text annotation box is of appropriate size and easy to read; each least significant pixel point is expanded horizontally and vertically to become a candidate region until a pixel point exceeding the significance threshold is encountered, and the region that meets the size standard is retained, the coordinates are mapped back to the original scale, and non-maximum suppression is applied to remove redundant candidate regions; The selection of the optimal text annotation box is based on horizontal alignment, proximity to the golden ratio, and maximum size. After calculating the ratio index, the annotation box with the smallest ratio is selected as the optimal annotation box; The proximity refers to placing text annotation boxes near important areas to enhance visibility and relevance, and identifying the four nearest salient pixels in each corner to fine-tune the position of the optimal annotation box.

Images