Information processing method and device, equipment and storage medium

Abstract

The invention discloses an information processing method and device, equipment and a storage medium, and the method comprises the steps: obtaining a first prompt text used for describing a first task, the first task being used for indicating to extract values of a plurality of indexes from first target data, and the first prompt text comprising description information of the first task, the first target data and names of the plurality of indexes; processing the first prompt text through a large model to obtain a first answer text; under the condition that the first answer text does not include the value of the target index, screening out second target data from the first target data, the second target data being related to the target index; a second prompt text used for describing a second task is obtained, the second task is used for indicating that a value of a target index is extracted from second target data, and the second prompt text comprises description information of the second task, the second target data and a name of the target index; and the second prompt text is processed through the large model to obtain the second answer text, so that the accuracy of data analysis is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to an information processing method, apparatus, device, and storage medium. Background Art

[0002] With the development of technology, large language models (i.e., large models) have shown impressive performance advantages in tasks such as text generation. Therefore, many enterprises use large models to analyze their financial report data and obtain analysis results.

[0003] In practical applications, financial report data often has a long content, reaching millions of words. Existing large models have weak processing performance for long texts and are difficult to accurately analyze long texts, resulting in inaccurate analysis results. Summary of the Invention

[0004] Embodiments of this application provide an information processing method, apparatus, device, and storage medium to improve the accuracy of data analysis.

[0005] In a first aspect, embodiments of this application provide an information processing method, including:

[0006] Obtain a first prompt text for describing a first task, where the first task is used to indicate extracting values of multiple metrics from first target data, and the first prompt text includes description information of the first task, the first target data, and names of multiple metrics;

[0007] Process the first prompt text through a large model to obtain a first answer text;

[0008] In a case where the first answer text does not include the value of a target metric, screen out second target data from the first target data, where the target metric is one of the multiple metrics, and the second target data is related to the target metric;

[0009] Obtain a second prompt text for describing a second task, where the second task is used to indicate extracting the value of the target metric from the second target data, and the second prompt text includes description information of the second task, the second target data, and the name of the target metric;

[0010] Process the second prompt text through the large model to obtain a second answer text, and the second answer text includes the value of the target metric.

[0011] Optionally, determining the second target data from the first target data includes:

[0012] Screen out data related to the target metric from the first target data;

[0013] Construct a knowledge base based on data related to the target metrics;

[0014] Retrieve and recall passage information related to the target metrics from the knowledge base based on the target metrics, and use it as the second target data.

[0015] Optionally, before obtaining the first prompt text, the method further includes:

[0016] Obtain financial report data;

[0017] Identify the data types of the financial report data to obtain financial report data corresponding to multiple different data types;

[0018] Based on the multiple different data types, perform parsing processing on the financial report data corresponding to the multiple different data types to obtain parsing data corresponding to the multiple different data types, and use it as the first target data.

[0019] Optionally, the data type includes a text type. Based on the multiple different data types, performing parsing processing on the financial report data corresponding to the multiple different data types includes:

[0020] When the financial report data of the text type includes multi-column text, perform splicing processing on the multi-column text based on the position information of the multi-column text to obtain parsing data corresponding to the text type.

[0021] Optionally, the data type includes an image type. Based on the multiple different data types, performing parsing processing on the financial report data corresponding to the multiple different data types includes:

[0022] Perform optical character recognition (OCR) processing on the financial report data of the image type to obtain parsing data corresponding to the image type.

[0023] Optionally, the data type includes a table type. Based on the multiple different data types, performing parsing processing on the financial report data corresponding to the multiple different data types includes:

[0024] Use the name of each column in the financial report data of the table type as the key and the value of each column as the value to obtain key-value pair data corresponding to the financial report data of the table type;

[0025] Filter out key-value pair data related to the multiple metrics from the key-value pair data and use it as parsing data corresponding to the table type.

[0026] Optionally, the method further includes:

[0027] Perform format conversion processing on the first response text and the second response text to obtain a processing result;

[0028] Output the processing result.

[0029] In a second aspect, an embodiment of the present application provides an information processing device, including:

[0030] A first prompt text acquisition module, configured to acquire a first prompt text for describing a first task, where the first task is used to indicate extracting values of multiple metrics from first target data, and the first prompt text includes description information of the first task, the first target data, and names of the multiple metrics;

[0031] A first response text acquisition module, configured to process the first prompt text through a large model to obtain a first response text;

[0032] A data screening module, configured to screen out second target data from the first target data when the first response text does not include the value of a target metric, where the target metric is one of the multiple metrics, and the second target data is related to the target metric;

[0033] A second prompt text acquisition module, configured to acquire a second prompt text for describing a second task, where the second task is used to indicate extracting the value of the target metric from the second target data, and the second prompt text includes description information of the second task, the second target data, and the name of the target metric;

[0034] A second response text acquisition module, configured to process the second prompt text through the large model to obtain a second response text, where the second response text includes the value of the target metric.

[0035] In a third aspect, an embodiment of the present application provides an electronic device, where the device includes: a processor, a memory, and a system bus;

[0036] The processor and the memory are connected through the system bus;

[0037] The memory is used to store a program, where the program includes instructions, and when the instructions are executed by the processor, the processor is caused to execute any implementation step of the above information processing method.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when the instructions are run on an electronic device, the electronic device is caused to execute any implementation step of the above information processing method.

[0039] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages:

[0040] In the embodiments of the present application, first, a first prompt text for describing a first task is obtained. The first task is used to indicate extracting values of multiple metrics from first target data. Correspondingly, the first prompt text may include description information of the first task, the first target data, and the names of the metrics corresponding to the values of the multiple metrics. Then, the first prompt text is processed by a large model to obtain a first answer text. At this time, based on the analysis ability of the large model, the first target data can be initially analyzed to obtain the values of each metric. And in the case where the first answer text does not include the value of the target metric, the second target data can be further screened out from the first target data. The target metric is one of the above multiple metrics, and the second target data is related to the target metric. Furthermore, a second prompt text for describing a second task is obtained. The second task is used to indicate extracting the value of the target metric from the second target data. The second prompt text includes description information of the second task, the second target data, and the name of the target metric. In this way, the second prompt text can be further processed by the large model to obtain a second answer text. At this time, based on the analysis ability of the large model, the second target data can be analyzed for the second time to obtain the value of the target metric. It can be seen that considering that the analysis ability of the large model will decrease due to the growth of data content, two analysis processes are designed in this solution. In the second analysis, since the second target data in the second prompt text is the data with a smaller content volume and related to the target metric screened out from the first target data, when processing the second prompt text by the large model, the processing volume of the large model can be effectively reduced, which helps to analyze the second target data more accurately and efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of an information processing method provided by an embodiment of the present application;

[0042] Figure 2 It is a schematic diagram of a process of retrieving and recalling based on a knowledge base and obtaining a second Q&A text provided by an embodiment of the present application;

[0043] Figure 3 It is a schematic diagram of a second target data provided by an embodiment of the present application;

[0044] Figure 4 It is a schematic diagram of formatting conversion processing for a first Q&A text and a second Q&A text provided by an embodiment of the present application;

[0045] Figure 5 It is a schematic diagram of the structure of an information processing device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] As described above, in practical applications, financial report data often has a long content, reaching up to millions of words. Existing large models have weak processing capabilities for long texts and are difficult to accurately analyze texts with longer content.

[0047] Based on this, to solve the above problems, the embodiments of the present application provide an information processing method, which includes: First, obtain a first prompt text for describing a first task, where the first task is used to indicate extracting values of multiple indicators from first target data. Correspondingly, the first prompt text may include description information of the first task, the first target data, and the names of the indicators corresponding to the values of the multiple indicators. Then, process the first prompt text through a large model to obtain a first answer text. At this time, based on the analysis ability of the large model, the first target data can be initially analyzed to obtain the values of each indicator. In the case where the first answer text does not include the value of the target indicator, the second target data can be further screened out from the first target data, where the target indicator is one of the above multiple indicators, and the second target data is related to the target indicator. Furthermore, obtain a second prompt text for describing a second task, where the second task is used to indicate extracting the value of the target indicator from the second target data, and the second prompt text includes description information of the second task, the second target data, and the name of the target indicator. In this way, the second prompt text can be further processed through the large model to obtain a second answer text. At this time, based on the analysis ability of the large model, the second target data can be analyzed again to obtain the value of the target indicator.

[0048] It can be seen that considering that the analysis ability of the large model will weaken with the increase of data content, two analysis processes are designed in this solution. In the second analysis, since the second target data in the second prompt text is the data with a smaller content volume and related to the target indicator screened out from the first target data, when processing the second prompt text through the large model, the processing volume of the large model can be effectively reduced, which helps to analyze the second target data more accurately and efficiently.

[0049] It should be noted that the embodiments of the present application do not limit the execution subject of the information processing method. For example, the information processing method of the embodiments of the present application can be applied to information processing devices such as terminal devices or servers. Among them, the terminal device can be an electronic device such as a smart phone, a computer, a personal digital assistant (PDA), or a tablet computer. The server can be an independent server, a cluster server, or a cloud server.

[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part rather than all of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0051] Figure 1 This is a flowchart of an information processing method provided for an embodiment of this application. In combination with Figure 1 As shown, for the information processing method provided in the embodiment of this application, the corresponding information processing device is used as the execution subject to describe the specific implementation of the solution. This information processing method may include the following steps S101 to step S105.

[0052] S101: Obtain a first prompt text for describing a first task. The first task is used to indicate extracting values of multiple metrics from first target data. The first prompt text includes description information of the first task, the first target data, and the names of the metrics corresponding to the values of the multiple metrics respectively.

[0053] In the embodiment of this application, taking the financial report analysis scenario as an example, when obtaining the first prompt text, the financial report data can be preprocessed to obtain the first target data. For ease of understanding, the following will give an exemplary illustration in combination with a possible implementation manner.

[0054] As a possible implementation manner, first, the financial report data can be obtained.

[0055] Next, the data types of the financial report data can be identified to obtain the financial report data corresponding to multiple different data types. In practical applications, the data types of the financial report data may include text type, image type, and table type. Here, the embodiment of this application does not specifically limit the implementation manner of identifying the data types of the financial report data, and any existing or future possible algorithm that can identify the data types in a document can be used for implementation.

[0056] Then, based on multiple different data types, the financial report data corresponding to multiple different data types can be parsed and processed to obtain the parsed data corresponding to multiple different data types, which is used as the first target data. For ease of understanding, the following will separately describe the corresponding parsing and processing processes in combination with different data types.

[0057] As an example, for financial report data of text type, in the case where the financial report data of text type includes multi-column text, the multi-column text can be spliced based on the position information of the multi-column text to obtain the analysis data corresponding to the text type. That is to say, the front and back splicing can be performed according to the position of each column of text in the document, so as to obtain the analysis data corresponding to the text type.

[0058] As another example, for financial report data of image type, the financial report data of image type can be subjected to optical character recognition (OCR) processing to obtain the analysis data corresponding to the image type. Here, the embodiments of the present application do not specifically limit the implementation manner of performing OCR processing on the financial report data of image type, and any existing or future possible OCR algorithm can be used for implementation.

[0059] As yet another example, for financial report data of table type, the name of each column in the financial report data of table type can be used as the key, and the value of each column can be used as the value to obtain the key-value pair data corresponding to the financial report data of table type.

[0060] For example, the financial report data of table type is first represented by Table 1 below.

[0061] Table 1

[0062]

[0063] Then, the column name A can be used as the key, and the index names 1 and 2 can be used as the values respectively. The column name B can be used as the key, and the numerical values 1 and 2 can be used as the values respectively. The column name C can be used as the key, and the strings 1 and 2 can be used as the values respectively. That is, column name A: index name 1 \t column name B: numerical value 1 \t column name C: string 1 \t\n column name A: index name 2 \t column name B: numerical value 2 \t column name C: string 2 \t\n.

[0064] Since the financial report data not only has the characteristic of long content, but also may contain many tables, and most of the values of the indicators appear in the tables, which makes the analysis difficult. Therefore, representing the financial report data of table type in the form of key-value pairs is convenient for the large model to better understand the relationship between the table header and the corresponding content, thus helping to better analyze the financial report data.

[0065] Then, the key-value pair data related to multiple indicators can be screened out from the key-value pair data and used as the analysis data corresponding to the table type.

[0066] Specifically, the index information library can be obtained in advance, and the index information library includes multiple indexes. Then, for each of the multiple indexes, the index is matched with the key-value pair data to determine whether the index exists in the key-value pair data, and the key-value pair data that does not contain the index is filtered, so as to screen out the key-value pair data related to the multiple indexes.

[0067] In this way, by parsing and processing the financial report data according to different data types respectively, the first target data with a format convenient for large models to process can be obtained, which helps to accurately analyze the financial report data.

[0068] In addition, in practical applications, the first prompt text can also include additional constraint information, which can be used to explain the relevant regulations of the financial index data, so as to facilitate the large language model to process the first prompt text more accurately. For the convenience of understanding, the first prompt text will be introduced below in combination with Table 2.

[0069] Table 2

[0070]

[0071] S102: Process the first prompt text through a large model to obtain the first answer text.

[0072] For the convenience of understanding, still taking the example provided in Table 2 above, the first answer text will be introduced in combination with Table 3.

[0073] Table 3

[0074]

[0075]

[0076] S103: In the case that the first answer text does not include the value of the target index, screen out the second target data from the first target data, where the target index is one of the multiple indexes, and the second target data is related to the target index.

[0077] In practical applications, as private data of enterprises, financial report data generally does not meet the conditions for being uploaded to large models for training, and there are data security issues. Moreover, the knowledge of current large models themselves comes from their training data, which is basically constructed from publicly available data on the network. Therefore, it is often impossible to obtain and use the above non-public financial report data for training, resulting in large models not having the relevant knowledge in financial report data. In this way, the limitations of large models in terms of knowledge can lead to inaccurate analysis of financial report data. Correspondingly, there may also be a certain hallucination problem in the analysis results output by large models, that is, some seemingly correct but actually wrong results, which require users to have knowledge in the relevant field to be able to distinguish the errors in the analysis results.

[0078] Based on this, in the case where the first answer text does not include the value of the target indicator, as shown in Table 3 above, in the first answer text, the values of some indicators may not be listed, but are represented by the words "not listed separately". In this way, the second target data can be further screened for secondary analysis.

[0079] In specific implementation, for the target indicator, first, the data related to the target indicator can be screened out from the first target data. Specifically, the first target data can be matched with the target indicator to determine whether the target indicator exists in the first target data, and the data without the target indicator can be filtered out, so as to screen out the data related to the target indicator.

[0080] Next, combined with Figure 2 As shown, a knowledge base can be constructed based on the data related to the target indicator. Here, the embodiments of the present application do not specifically limit the implementation manner of constructing the knowledge base, and any existing or possibly emerging algorithm for constructing the knowledge base in the future can be used for implementation. In this way, by constructing the knowledge base, more knowledge can be provided for the large model and learned locally, thereby solving the problems of knowledge limitations and hallucinations existing in the large model, and avoiding data leakage and causing data security problems through local learning.

[0081] Then, based on the target indicator, the paragraph information related to the target indicator can be retrieved and recalled from the knowledge base and used as the second target data. Among them, the second target data retrieved and recalled from the knowledge base can be as Figure 3 shown. Here, the embodiments of the present application do not specifically limit the implementation manner of retrieving and recalling the paragraph information related to the target indicator, and any existing or possibly emerging algorithm for retrieving and recalling can be used for implementation.

[0082] In this way, the second target data is obtained by means of retrieval and recall. The second target data is the data with a smaller content volume and related to the target indicator selected from the first target data. Therefore, when processing the second prompt text through the large model, the processing volume of the large model can be effectively reduced, which helps to analyze the second target data more accurately and efficiently.

[0083] S104: Obtain a second prompt text for describing the second task. The second task is used to indicate extracting the value of the target indicator from the second target data. The second prompt text includes the description information of the second task, the second target data, and the name of the target indicator.

[0084] In practical applications, the second prompt text may also include additional constraint information, which can be used to explain the relevant regulations of financial indicator data, so as to facilitate the large language model to process the second prompt text more accurately. For the convenience of understanding, the second prompt text will be introduced below in combination with Table 4.

[0085] Table 4

[0086]

[0087]

[0088] S105: Process the second prompt text through the large model to obtain a second answer text, and the second answer text includes the value of the target indicator.

[0089] For the convenience of understanding, still taking the example provided in Table 3 above, the first answer text will be introduced in combination with Table 5.

[0090] Table 5

[0091]

[0092] Furthermore, after obtaining the first answer text and the second answer text, format conversion processing can also be performed on the first answer text and the second answer text to obtain a processing result and output the processing result.

[0093] For example, since the value format of financial indicators is special. For example, the value of the indicator uses parentheses to represent the negative sign, that is, "(100) million" represents "-100 million". Therefore, during the format conversion processing, the parentheses of the values of the indicators in the first answer text and the second answer text can be converted into negative signs, such as converting "(100) million" into "-100 million".

[0094] Another example is that only the values of the indicators can be retained in the first answer text and the second answer text, without additional other descriptions, as Figure 4 shown.

[0095] As can be seen from the relevant content of the above steps S101 - S105, in the embodiment of the present application, first, a first prompt text for describing a first task is obtained. The first task is used to indicate extracting values of multiple indicators from first target data. Correspondingly, the first prompt text may include description information of the first task, the first target data, and the names of the indicators corresponding to the values of the multiple indicators. Then, the first prompt text is processed by a large model to obtain a first answer text. At this time, based on the analysis ability of the large model, the first target data can be preliminarily analyzed to obtain the values of each indicator. And in the case where the first answer text does not include the value of the target indicator, the second target data can be further screened out from the first target data. The target indicator is one of the above - mentioned multiple indicators, and the second target data is related to the target indicator. Furthermore, a second prompt text for describing a second task is obtained. The second task is used to indicate extracting the value of the target indicator from the second target data. The second prompt text includes description information of the second task, the second target data, and the name of the target indicator. In this way, the second prompt text can be further processed by the large model to obtain a second answer text. At this time, based on the analysis ability of the large model, the second target data can be analyzed again to obtain the value of the target indicator. Thus, considering that the analysis ability of the large model will weaken due to the growth of data content, two analysis processes are designed in this solution. In the second analysis, since the second target data in the second prompt text is the data with a smaller content volume and related to the target indicator screened out from the first target data, when processing the second prompt text by the large model, the processing volume of the large model can be effectively reduced, which helps to analyze the second target data more accurately and efficiently.

[0096] Furthermore, based on the information processing method provided in the above - mentioned embodiment, the embodiment of the present application can also provide an information processing device. The following will describe this information processing device in combination with the embodiment and the drawings respectively.

[0097] Figure 5 The following is a schematic structural diagram of an information processing device provided in the embodiment of the present application. In combination with Figure 5 As shown, the information processing device 500 provided in the embodiment of the present application may include:

[0098] A first prompt text acquisition module 501, configured to acquire a first prompt text for describing a first task, where the first task is used to indicate extracting values of multiple indicators from first target data, and the first prompt text includes description information of the first task, the first target data, and the names of the multiple indicators;

[0099] A first answer text acquisition module 502, configured to process the first prompt text through a large model to obtain a first answer text;

[0100] A data screening module 503, configured to screen out second target data from the first target data when the first answer text does not include the value of a target indicator, where the target indicator is one of the multiple indicators, and the second target data is related to the target indicator;

[0101] A second prompt text acquisition module 504, configured to acquire a second prompt text for describing a second task, where the second task is used to indicate extracting the value of the target indicator from the second target data, and the second prompt text includes a description information of the second task, the second target data, and the name of the target indicator;

[0102] A second answer text acquisition module 505, configured to process the second prompt text through the large model to obtain a second answer text, where the second answer text includes the value of the target indicator.

[0103] Optionally, the data screening module 503 is specifically configured to:

[0104] Screen out data related to the target indicator from the first target data;

[0105] Build a knowledge base based on the data related to the target indicator;

[0106] Retrieve and recall passage information related to the target indicator from the knowledge base based on the target indicator, and use it as the second target data.

[0107] Optionally, the information processing device 500 further includes:

[0108] A data acquisition module, configured to acquire financial report data;

[0109] A data identification module, configured to identify the data type of the financial report data to obtain financial report data corresponding to multiple different data types;

[0110] A data parsing module, configured to perform parsing processing on the financial report data corresponding to the multiple different data types based on the multiple different data types to obtain parsed data corresponding to the multiple different data types, and use it as the first target data.

[0111] Optionally, the data type includes a text type, and the data parsing module is specifically configured to:

[0112] When the financial report data of the text type includes multi-column text, perform splicing processing on the multi-column text based on the position information of the multi-column text to obtain the parsed data corresponding to the text type.

[0113] Optionally, the data type includes an image type, and the data parsing module is specifically configured to:

[0114] Perform optical character recognition (OCR) processing on the financial report data of the image type to obtain the parsed data corresponding to the image type.

[0115] Optionally, the data type includes a table type, and the data parsing module is specifically configured to:

[0116] Use the name of each column in the financial report data of the table type as the key and the value of each column as the value to obtain the key-value pair data corresponding to the financial report data of the table type;

[0117] Filter out the key-value pair data related to the multiple metrics from the key-value pair data and use it as the parsed data corresponding to the table type.

[0118] Optionally, the information processing device 500 further includes:

[0119] A format conversion module, configured to perform format conversion processing on the first answer text and the second answer text to obtain a processing result;

[0120] A data output module, configured to output the processing result.

[0121] Furthermore, an embodiment of the present application further provides an electronic device, including: a processor, a memory, and a system bus;

[0122] The processor and the memory are connected through the system bus;

[0123] The memory is used to store one or more programs, and the one or more programs include instructions that, when executed by the processor, cause the processor to execute any implementation step of the above information processing method.

[0124] Furthermore, an embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored, and when the instructions run on an electronic device, any implementation step of the above information processing method is enabled.

[0125] As can be seen from the description of the above embodiments, those skilled in the art can clearly understand that all or part of the steps in the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in each embodiment or some parts of the embodiments of the present application. It should be noted that the various embodiments in this specification are described in a progressive manner, and the key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0126] For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For the relevant parts, reference can be made to the description in the method part.

[0127] It should also be noted that in this text, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0128] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An information processing method, characterized in that: include: Acquire a first prompt text for describing a first task, where the first task is used to instruct to extract values ​​of multiple indicators from first target data, and the first prompt text includes description information of the first task, the first target data, and names of multiple indicators; Processing the first prompt text by using a large model to obtain a first answer text; In the case where the first answer text does not include a value of a target indicator, filtering out second target data from the first target data, the target indicator being one of the multiple indicators, and the second target data being related to the target indicator; Acquire a second prompt text for describing a second task, where the second task is used to instruct to extract a value of the target indicator from the second target data, and the second prompt text includes description information of the second task, the second target data, and a name of the target indicator; The second prompt text is processed by the large model to obtain a second answer text, wherein the second answer text includes the value of the target indicator.

2. The information processing method according to claim 1, characterized in that: The determining the second target data from the first target data comprises: Filtering data related to the target indicator from the first target data; Building a knowledge base based on data related to the target indicator; Based on the target indicator, paragraph information related to the target indicator is retrieved from the knowledge base and used as the second target data.

3. The information processing method according to claim 1, characterized in that: Before obtaining the first prompt text, the method further includes: Obtain financial report data; Identify the data type of the financial report data to obtain financial report data corresponding to a plurality of different data types; Based on the multiple different data types, the financial report data corresponding to the multiple different data types are parsed and processed to obtain parsed data corresponding to the multiple different data types and use them as the first target data.

4. The information processing method according to claim 3, characterized in that: The data type includes a text type, and the parsing and processing of the financial report data corresponding to the multiple different data types respectively based on the multiple different data types includes: In the case where the financial report data of the text type includes multiple columns of text, the multiple columns of text are spliced ​​based on position information of the multiple columns of text to obtain parsed data corresponding to the text type.

5. The information processing method according to claim 3, characterized in that: The data type includes an image type, and the parsing and processing of the financial report data corresponding to the multiple different data types respectively based on the multiple different data types includes: Optical character recognition (OCR) processing is performed on the financial report data of the image type to obtain parsed data corresponding to the image type.

6. The information processing method according to claim 3, characterized in that: The data type includes a table type, and the parsing and processing of the financial report data corresponding to the multiple different data types respectively based on the multiple different data types includes: Using the name of each column in the financial report data of the table type as a key and the value of each column as a value, key-value pair data corresponding to the financial report data of the table type is obtained; The key-value pair data related to the multiple indicators are screened out from the key-value pair data and used as parsed data corresponding to the table type.

7. The information processing method according to any one of claims 1 to 6, characterized in that: The method further comprises: Performing format conversion processing on the first answer text and the second answer text to obtain a processing result; The processing result is outputted.

8. An information processing device, characterized in that: include: A first prompt text acquisition module, used to acquire a first prompt text for describing a first task, wherein the first task is used to instruct to extract values ​​of multiple indicators from first target data, and the first prompt text includes description information of the first task, the first target data, and names of multiple indicators; A first answer text acquisition module, used for processing the first prompt text through a large model to obtain a first answer text; a data screening module, configured to screen out second target data from the first target data when the first answer text does not include a value of the target indicator, the target indicator being one of the multiple indicators, and the second target data being related to the target indicator; A second prompt text acquisition module, used to acquire a second prompt text for describing a second task, wherein the second task is used to instruct to extract the value of the target indicator from the second target data, and the second prompt text includes description information of the second task, the second target data, and the name of the target indicator; The second answer text acquisition module is used to process the second prompt text through the large model to obtain a second answer text, and the second answer text includes the value of the target indicator.

9. An electronic device, characterized in that: The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes the steps of the information processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and when the computer program is executed by a terminal device, the steps of the information processing method according to any one of claims 1 to 7 are implemented.

Images