Information summary information extraction method and device, electronic equipment and storage medium

Abstract

The invention discloses an information summary information extraction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining conference text information of target conference information, the conference text information comprising speaking text information of at least one conference speaking object; segmenting the conference text information according to conference speaking objects to obtain speaking text segments of the conference speaking objects; performing key information extraction on the speaking text segment of each conference speaking object to obtain key point information of each conference speaking object; based on a preset content category and the key point information of each conference speaking object, generating category summary information corresponding to each preset content category; and based on the category summary information corresponding to each preset content category, generating information summary information corresponding to the target conference information. The information summary information in the conference text information is automatically extracted, and the extraction efficiency of the information summary information is improved.

Description

Technical Field

[0001] The present application belongs to the field of information processing, and in particular, relates to a method, device, electronic device and storage medium for extracting news summary information. Background Art

[0002] In a telephone conference, it is usually necessary to record the content of the conference. In order to facilitate viewing, the content of the conference is generally sorted out to form information summary information. In the prior art, the information summary information of the telephone conference is often extracted by manual processing and editing. The information summary information extracted by this method is easily affected by the subjective influence of the processor, so that the generated information summary information deviates from the content expressed by the original conference speaker in the telephone conference, and the accuracy is low. In addition, it takes a long time for manual extraction of the information summary information, and the extraction efficiency of the information summary information is low. Summary of the invention

[0003] The embodiment of the present application provides an implementation scheme different from the prior art to solve the technical problem that the information summary information is manually extracted, resulting in low extraction efficiency of the information summary information.

[0004] In a first aspect, the present application provides a method for extracting information summary information, including: obtaining conference text information of target conference information, the conference text information containing speech text information of at least one conference speaker; segmenting the conference text information according to the conference speaker to obtain speech text segments of each conference speaker; extracting key information from the speech text segments of each conference speaker to obtain key point information of each conference speaker; based on preset content categories and the key point information of each conference speaker, generating category summary information corresponding to each preset content category; based on the category summary information corresponding to each preset content category, generating information summary information corresponding to the target conference information.

[0005] In a second aspect, the present application provides an information summary information extraction device, comprising: an acquisition unit, used to acquire conference text information of target conference information, wherein the conference text information includes speech text information of at least one conference speech object; a segmentation unit, used to segment the conference text information according to the conference speech object to obtain speech text segments of each conference speech object; an extraction unit, used to extract key information from the speech text segments of each of the conference speech objects respectively to obtain key point information of each of the conference speech objects; a first generation unit, used to generate category summary information corresponding to each of the preset content categories based on the preset content category and the key point information of each of the conference speech objects; and a second generation unit, used to generate information summary information corresponding to the target conference information based on the category summary information corresponding to each of the preset content categories.

[0006] In a third aspect, the present application provides an electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any method in the first aspect or any possible implementation manner of the first aspect by executing the executable instructions.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the computer program implements any method in the first aspect or any possible implementation manner of the first aspect.

[0008] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method described in the first aspect or any possible implementation manner of the first aspect.

[0009] The solution provided by the present application is to obtain the conference text information of the target conference information, wherein the conference text information contains the speech text information of at least one conference speaker; the conference text information is segmented according to the conference speaker to obtain the speech text segments of each conference speaker; the key information of the speech text segments of each conference speaker is extracted respectively to obtain the key point information of each conference speaker; based on the preset content category and the key point information of each conference speaker, the category summary information corresponding to each preset content category is generated; based on the category summary information corresponding to each preset content category, the information summary information corresponding to the target conference information is generated. The solution realizes the automatic extraction of information summary information in conference text information, and improves the extraction efficiency of information summary information. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0011] Figure 1 A flowchart of a news summary information extraction method provided in one embodiment of the present application;

[0012] Figure 2 A schematic diagram of the relationship between a database and an information library provided in one embodiment of the present application;

[0013] Figure 3 A schematic diagram of a flow chart of information summary information provided for an exemplary embodiment of the present application;

[0014] Figure 4 A schematic diagram of the structure of a news summary information extraction device provided in one embodiment of the present application;

[0015] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, but cannot be understood as limiting the present application.

[0017] The terms "first" and "second" etc. in the specification, claims and drawings of the embodiments of the present application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0018] In a telephone conference, it is usually necessary to record the content of the conference. In order to facilitate viewing, the content of the conference is generally sorted out to form information summary information. In the prior art, the information summary information of the telephone conference is often extracted by manual processing and editing. The information summary information extracted by this method is easily affected by the subjective influence of the processor, so that the generated information summary information deviates from the content expressed by the original conference speaker in the telephone conference, and the accuracy is low. In addition, it takes a long time for manual extraction of the information summary information, and the extraction efficiency of the information summary information is low.

[0019] In addition, manual extraction is difficult to ensure the consistency and accuracy of the summary content, and is easily affected by the subjective thinking of the extractor. It is difficult to maintain consistency in the structure and style of the information summary information, and further processing is required, which wastes human resources and is inefficient. It is also difficult to meet the needs of mass production, which easily leads to untimely extraction of information summary information, making the information summary information with high requirements for information timeliness lose its function and waste resources.

[0020] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0021] Figure 1 A flowchart of a news summary information extraction method provided by an exemplary embodiment of the present application, the method at least includes the following steps S1-S5:

[0022] S1. Acquire conference text information of target conference information, wherein the conference text information includes speech text information of at least one conference speaker;

[0023] Optionally, the target conference information refers to information data consisting of text content, title information and language information extracted from conference messages such as telephone and video. The conference message may refer to a performance conference call in the financial field or a conference call in other fields, which is not limited here.

[0024] Optionally, the conference text information refers to text content extracted from the target conference information.

[0025] S2, segmenting the conference text information according to the conference speaker, to obtain speech text segments of each conference speaker;

[0026] The speech text segment refers to the speech content of each speaker in conference messages such as telephone and video.

[0027] Optionally, in the aforementioned S2, the conference text information is segmented according to the conference speech objects to obtain the speech text segments of each conference speech object, including: using an object segmentation function to segment the conference text information according to the conference speech objects to obtain the speech text segments of each conference speech object.

[0028] Optionally, the object segmentation function is used to extract speech text segments belonging to the conference speech object in the conference text information according to the conference speech object.

[0029] Specifically, the speaker list may be obtained first, and then the conference text information may be segmented according to the speaker list through the object segmentation function to obtain speech text segments corresponding to the conference speech objects of each speaker in the speaker list.

[0030] Furthermore, for each conference speaker in the speaker list, at least one speech text segment corresponding to the conference speaker can be spliced ​​to obtain a speech text segment corresponding to the conference speaker, and then obtain a speech text segment corresponding to each conference speaker in the speaker list.

[0031] S3, extracting key information from the speech text segments of each of the conference speakers to obtain key point information of each of the conference speakers;

[0032] The key point information is used to describe the main content information of the corresponding speech text segment. Specifically, the key point information may include content information related to a preset content category in the speech text segment.

[0033] Specifically, after obtaining speech text segments of different conference speakers, redundant information in the speech text segments can be deleted, and information related to a preset content type in the speech text segments can be extracted to obtain condensed text information related to the preset content type.

[0034] Furthermore, a language processing model may be determined to input speech text segments of different conference speakers into the language processing model respectively, so as to obtain key point information corresponding to the speech text ends of different conference speakers through the language processing model.

[0035] Optionally, the key point information of the conference speech object may include key point text segments under the preset content category; the preset content category refers to a category set for the conference content, for example, the preset content category includes a financial performance category and a business progress category.

[0036] The above-mentioned step S3 extracts key information from the speech text segments of each conference speaker to obtain key point information of each conference speaker, including the following steps S31-S33:

[0037] S31, for each of the conference speakers, segmenting the speech text segments of the conference speaker to obtain a plurality of segmented text information;

[0038] Optionally, in the aforementioned S31, the speech text segments of the conference speech object are segmented to obtain a plurality of segmented text information, including: using a length segmentation function to segment the speech text segments of the conference speech object to obtain a plurality of segmented text information.

[0039] Optionally, the length segmentation function may segment the speech text segment based on a preset length to obtain segmented text information that does not exceed the preset length.

[0040] Specifically, the length segmentation function segments the input speech text segment, and divides the speech text segment into several parts of segmented text information, wherein the length of each part of the segmented text information in the several parts does not exceed a preset length.

[0041] Optionally, the preset length is the longest text length that can be accepted by the large language model.

[0042] The length segmentation function is used to segment the speech text into segments of a length suitable for input into the large language model, laying the foundation for the subsequent work of using the large language model to generate summaries. By segmenting the speech text segments and inputting them into the large language model, the integrity of the speeches of the conference speakers is preserved, and the large language model is convenient for generating summaries, ensuring the smoothness and accuracy of the summary generation process.

[0043] S32, based on the preset content category and each of the segmented text information in the plurality of segmented text information, obtaining segmented summary information corresponding to each of the segmented text information under each of the preset content categories;

[0044] Optionally, the preset content categories may include financial performance categories and business progress categories, wherein the financial performance category is used to identify that the text information is information related to finance, for example, information related to finance includes financial data and financial factor information that affects changes in financial data, wherein financial data includes but is not limited to revenue, net profit, gross profit, earnings per share, etc., and financial factor information includes but is not limited to the content listed here, which can be set specifically according to actual needs. The business progress category is used to identify that the text information is information related to the company's development, such as operating results, business planning, prospects, product research and development, market expansion, strategic cooperation, and brand building.

[0045] Optionally, the acquiring, based on the preset content category and each of the plurality of segmented text information, the segmented summary information corresponding to each of the segmented text information under each of the preset content categories comprises the following steps:

[0046] For each of the multiple segmented text information, a pre-trained large language model can be used to perform summary extraction on the segmented text information based on a preset content category to obtain the segmented summary information corresponding to the segmented text information under different preset content types, and then obtain the segmented summary information corresponding to each segmented text information under different preset content types.

[0047] The pre-trained large language model is used to extract summary content information from the input text. Specifically, the category label corresponding to the preset content category can be obtained, and then a prompt (such as a third prompt) is generated based on the category label and the segmented text information. By inputting the prompt into the pre-trained large language model for extracting summary content information, the large language model can obtain the corresponding segmented summary information from the segmented text information according to the preset content category indicated by the category label. The large language model can be obtained by performing prompt learning training on the LLM (Large Language Model) model. The LLM model is a text processing model obtained by pre-training with a large amount of corpus, which can understand and generate natural language text. The LLM model is trained to generate summary content information through prompt learning, which can improve the accuracy of summary content information generation. Prompt refers to instructions and context information used to guide the large language model to generate summary content information. Specifically, the prompt may include system templates and user templates.

[0048] For example, the preset content categories include financial performance categories and business progress categories. After the speech text segments of the conference speaker are segmented to obtain segmented text information A and segmented text information B, for the segmented text information A, corresponding prompts can be generated based on the category label of the financial performance category, the category label of the business progress category, and the segmented text information A, and then the prompts are input into the above-mentioned pre-trained large language model to extract summary content information of the segmented text information, and obtain the segmented summary information A of the segmented text information under the financial performance category and the segmented summary information B of the segmented text information under the business progress category. Summary information B. At the same time, for the segmented text information B, a corresponding prompt can be generated based on the category label of the financial performance category, the category label of the business progress category and the segmented text information A, and then the prompt is input into the large language model to realize the extraction of summary content information of the segmented text information, and obtain the segmented summary information C of the segmented text information B under the financial performance category and the segmented summary information D under the business progress category, and finally obtain the segmented summary information A and segmented summary information C under the financial performance category, and the segmented summary information B and segmented summary D under the business progress category.

[0049] S33: for a target content category under any preset content category, generate a key point text segment of the conference speech subject under the target content category according to the segment summary information corresponding to all the segment text information under the target content category.

[0050] Optionally, the target content category refers to a financial performance category or a business progress category under a preset content category.

[0051] In some optional embodiments, in the aforementioned S33, for the target content category under any preset content category, based on the segmented summary information corresponding to all the segmented text information under the target content category, a key point text segment of the conference speech object under the target content category is generated, including: for the target content category under any preset content category, the segmented summary information corresponding to all the segmented text information under the target content category is spliced ​​to obtain the key point text segment of the conference speech object under the target content category, and then obtain the key point text segments corresponding to all the target content categories under the preset content category.

[0052] Continuing with the example that the above preset content categories include the financial performance category and the business progress category, and the speech text segment of the conference speaker includes segmented text information A and segmented text B, the segmented summary information A and the segmented summary information C can be directly spliced ​​to obtain the key point text segment of the conference speaker under the financial performance category. At the same time, the segmented summary information B and the segmented summary information D can be spliced ​​to obtain the key point text segment of the conference speaker under the business progress category.

[0053] In some other optional embodiments, in the aforementioned S33, generating the key point text segment of the conference speech object under the target content category according to the segment summary information corresponding to all the segment text information under the target content category includes the following steps S331-S333:

[0054] S331, concatenating the segment summary information corresponding to each segment text information under the target content category to obtain category content information under the target content category;

[0055] S332, generating a first prompt according to the category content information under the preset content category;

[0056] Optionally, generating a first prompt based on the category content information under the preset content category includes the following steps: using a system template for indicating the relationship between the input content and the output content of the large language model, and a user template containing the category content information under the preset content category as the first prompt.

[0057] Optionally, a prompt refers to instructions and context information used to guide the large language model to generate summary content information, wherein the prompt includes a system template and a user template.

[0058] Optionally, the system template (system_template) includes configuration instructions and preprocessing instructions, wherein the configuration instructions are used to indicate format information of input content and output content of the large prediction model, and the preprocessing instructions are used to trigger the execution of the large prediction model.

[0059] Optionally, the user template (user_template) includes a category label corresponding to a preset content category and category content information under the preset content category.

[0060] For example, taking the application scenario of the financial field as an example, the prompts are as follows: As a skilled financial analyst, refine the {input content} into the {output content} corresponding to the {category label} that is concise and easy to understand; among them, the {input content} is arranged in format A, and the {output content} is arranged in format B.

[0061] It can be understood that the prompt is used to constrain the generation result of the key point text segment (i.e., the summary content information). The first prompt can be used to input into the above-mentioned large language model to control the large language model to extract the summary content information according to the category label, so as to obtain the key point text segment corresponding to the category label.

[0062] S333: extracting a key point text segment corresponding to the target content category from the category content information under the target content category by using the first prompt and the pre-trained large language model.

[0063] Optionally, after extracting the key text segment corresponding to the target content category from the category content information under the target content category by using the first prompt and the pre-trained large language model in the aforementioned S333, the following steps S020-S022 are included:

[0064] S020, obtaining the number of at least one key point text segment corresponding to the target content category, and the length of each key point text segment in the at least one key point text segment;

[0065] S021. For the target content category, if there is a key point text segment whose length is greater than a preset length in at least one key point text segment corresponding to the target content category, the at least one key point text segment is concatenated to obtain new category content information under the target content category, and the new category content information is used as the category content information to execute the above-mentioned step S332 until there is no key point text segment whose length is greater than the preset length in at least one key point text segment corresponding to the target content category determined most recently;

[0066] Optionally, the preset length can be set to 40, 50, 65, etc., and can be set based on actual needs, and no excessive restrictions are made here.

[0067] Optionally, if there is a description about a picture in the key point text segment, the character length of the description about the picture is limited to 15 characters, or if it exceeds 15 characters, it is reduced by 30%.

[0068] S022. For the target content category, if the number of at least one key point text segment corresponding to the target content category is greater than the preset number, the at least one key point text segment corresponding to the target content category determined most recently is spliced ​​to obtain new category content information under the target content category, the new category content information is used as the category content information, and the process returns to execute the aforementioned step S332 until the number of at least one key point text segment corresponding to the target content category determined most recently is not greater than the preset number.

[0069] Optionally, the preset number is 10-20.

[0070] In summary, by utilizing the powerful language understanding and generation capabilities of large language models, we ensure that the summary not only accurately captures the core points of the conference text information in the target conference information, but also is concise and informative. In addition, by iteratively extracting key point text segments, the quality of the key point text segments is further improved, making the information summary information based on the key point text segments more in line with user needs and application scenarios. At the same time, by setting a preset number, we ensure the comprehensiveness of the information and avoid information overload.

[0071] Continuing with the example in which the preset content categories include the financial performance category and the business progress category, and the speech text segment of the conference speaker includes segmented text information A and segmented text B, the segmented summary information A and the segmented summary information C can be directly spliced ​​to obtain the category content information A of the conference speaker under the financial performance category, and at the same time, the segmented summary information B and the segmented summary information D can be spliced ​​to obtain the category content information B of the conference speaker under the business progress category; then, a user template is generated according to the category content information A and its corresponding category label of the target content category, and the category content information B and its corresponding category label of the target content category, so as to generate a prompt (i.e., a first prompt) according to the user template and the system template, and the prompt is input into the pre-trained large language model, and the key point text segment corresponding to the financial performance category is extracted from the category content information A, and the key point text segment corresponding to the business progress category is extracted from the category content information B through the large language model.

[0072] S4, generating category summary information corresponding to each preset content category based on the preset content category and the key point information of each conference speech object;

[0073] In some optional embodiments, in the aforementioned S4, based on the preset content category and the key point information of each of the conference speech objects, category summary information corresponding to each of the preset content categories is generated, including: for the target content category under any preset content category, the key point information of each of the conference speech objects under the target content category is spliced ​​to obtain the category content information under the target content category, and the category content information under the target content category is determined as the category summary information under the target content category.

[0074] For example, the conference text information includes speech text information of conference speaker A and conference speaker B. After obtaining the key point text segment A corresponding to the speech text information of conference speaker A under the financial performance category and the key point text segment B corresponding to the speech text information of conference speaker A under the business progress category, and the key point text segment C corresponding to the speech text information of conference speaker B under the financial performance category and the key point text segment D corresponding to the speech text information of conference speaker B under the business progress category, the key point text segment A and the key point text segment C corresponding to the financial performance category can be directly spliced ​​to obtain the category summary information under the financial performance category. At the same time, the key point text segment B and the key point text segment D under the business progress category can be spliced ​​to obtain the category summary information under the business progress category.

[0075] In some other optional embodiments, in the aforementioned S4, generating category summary information corresponding to each preset content category based on the preset content category and the key point information of each conference speech object includes the following steps S41-S43:

[0076] S41, for any preset content category, splicing the key text segments of each conference speaker under the preset content category to obtain category content information under the preset content category;

[0077] S42, generating a second prompt according to the category content information under the preset content category;

[0078] Optionally, in the aforementioned S42, generating a second prompt based on the category content information under the preset content category includes the following steps: using a system template (system_template) used to indicate the relationship between the input content and output content of the large language model, and a user template (user_template) containing the category content information under the preset content category as the second prompt.

[0079] Optionally, the second prompt refers to instructions and context information for guiding the large language model to generate information summary information, wherein the second prompt includes a system template and a user template.

[0080] Optionally, for explanations of the system template and the user template, please refer to the aforementioned related explanations, which will not be elaborated here.

[0081] S43: extracting category summary information corresponding to the preset content category from the category content information under the preset content category by using the second prompt and the pre-trained large language model.

[0082] Continuing with the example of the speech text information of the conference text information including the conference speaker A and the conference speaker B, the key point text segment A and the key point text segment C corresponding to the financial performance category can be directly spliced ​​to obtain the category content information C under the financial performance category. At the same time, the key point text segment B and the key point text segment D under the business progress category can be spliced ​​to obtain the category content information D under the business progress category. Then, according to the category content information C and its corresponding category label of the target content category, and the category content information D and its corresponding category label of the target content category, a user template is generated, so as to generate a prompt (i.e., a second prompt) according to the user template and the system template, and by inputting the prompt into the pre-trained large language model, the category summary information corresponding to the financial performance category is extracted from the category content information C through the large language model, and the category summary information corresponding to the business progress category is extracted from the category content information D.

[0083] S5. Generate information summary information corresponding to the target conference information based on the category summary information corresponding to each of the preset content categories.

[0084] Optionally, the information summary information corresponding to the target conference information includes category summary information corresponding one-to-one to preset content categories.

[0085] Specifically, the information summary information corresponding to the target conference information includes: financial information summary information corresponding to the financial performance category in the preset content category, and business information summary information corresponding to the business progress category in the preset content category.

[0086] In summary, the news summary information is divided according to financial performance categories and business progress categories, allowing users to obtain the required information more quickly.

[0087] It should be understood that in order to facilitate the subsequent data processing and storage of the extracted category summary information, the category summary information is uniformly converted into a format. For specific operations, please refer to the following steps S001-S002:

[0088] S001. Convert the news summary information into JSON format using a large language model to obtain converted news summary information;

[0089] S002: storing the converted information summary information in an information repository, or displaying the converted information summary information.

[0090] In summary, by formatting the news summary information, for example, converting the news summary information into JSON format, it is convenient for subsequent data analysis and application integration.

[0091] In one embodiment, before obtaining the meeting text information of the target meeting information, different meeting messages can be obtained from different data sources in real time and written into the message queue so that the target meeting information can be obtained from the message queue later. It should be understood that the message queue stores different types of meeting messages obtained from different data sources. For example, taking the application scenario as a financial field scenario, the types of meeting messages include but are not limited to financial report meetings, shareholder meetings, investor relations meetings, etc. In order to improve the efficiency of extracting information summary information from the target type of meeting messages, this application will pre-process the meeting messages stored in the message queue to obtain the target meeting information. For the specific processing process, please refer to steps S01-S03:

[0092] S01. Acquire conference type information of a conference message, and if the conference type information includes a preset title and a preset conference identifier, determine that the conference message is a target conference message;

[0093] Optionally, the conference type information refers to relevant information of the conference message stored in the message queue, including but not limited to the conference title and conference type identifier.

[0094] Among them, the conference title refers to the title information of the conference message, which can be used to identify whether the subject of the conference message is related to the preset title corresponding to the subject type of conference message. Taking the application scenario of the financial field as an example, the preset title can be a financial report meeting or an earnings conference call, etc. After obtaining the conference titles corresponding to each conference message in the message queue, the conference titles of each conference message can be matched with the preset title to determine whether the subject of the conference message is related to the preset title corresponding to the subject type of conference message. For example, the preset title can be set to "Earnings Call Transcript" and "Earnings Conference Call Transcript", and the conference title of each conference message is matched with the preset title "Earnings Call Transcript" or "Earnings Conference Call Transcript" to identify the conference message containing the preset title.

[0095] The conference identifier is used to identify the conference category of the conference message. Specifically, the conference type includes but is not limited to a telephone conference type, a video conference type, etc. Taking the application scenario of the financial field as an example, the preset conference identifier can be a conference identifier of a telephone conference type. After obtaining the conference identifiers corresponding to each conference message in the message queue, the conference identifiers of each conference message can be matched with the preset conference identifier to determine whether the conference identifier of the conference message is a conference identifier of a telephone conference type.

[0096] S02, extracting language information used by the conference message, title information of the conference message, and text content of the conference message from the target conference message;

[0097] S03. Using the language information, the title information, and the text content as the target conference information.

[0098] Optionally, after the target conference information of the target conference message is acquired, the target conference information may be stored in a database, wherein the target conference information stored in the database may be extracted from the database at preset time periods for information summary processing.

[0099] To summarize, the present application obtains target meeting information by pre-processing the meeting messages stored in the message queue, realizes automatic identification and processing of messages from different data sources, and ensures through a screening mechanism that only relevant financial report meetings and conference call type meeting messages are used as target meeting information for information summary extraction and processing to obtain information summary information, thereby improving extraction efficiency.

[0100] Optionally, in order to improve the readability and usability of the information summary information and avoid wasting resources when storing the information summary information, the present application cleans the information summary information through a cleaning operation, and the specific operation can be referred to the following steps: the information summary information is cleaned to obtain more streamlined information summary information, wherein the cleaning process includes removing redundant information processing and removing secondary information processing.

[0101] Optionally, in order to avoid repeated conference summary extraction processing on the same target conference information, the method further includes: obtaining target conference information corresponding to the information summary information stored in the information library, and in the database, changing the identifier of the target conference information corresponding to the information summary information stored in the information library to a generated summary and entered into the information library.

[0102] In order to understand the relationship between database and information base, and clearly understand the relationship between the data stored in database and information base, please refer to Figure 2 , Figure 2A schematic diagram of the relationship between a database and an information library provided for an exemplary embodiment of the present application:

[0103] Filter out the target conference information from the message queue, store the target conference information in the database, obtain the target conference information from the database, extract and process the target conference information to obtain information summary information, store the information summary information in the information library, and at the same time mark the target conference information in the database as having generated a summary and having entered the information library.

[0104] In summary, by extracting target conference information data in a timed manner and avoiding information summary processing on target conference information data that has been extracted through status updates, manual intervention is reduced and repeated information summary extraction is ensured. Delays and errors are reduced, and the efficiency of information summary extraction is improved.

[0105] To understand more clearly how this application generates information summary information, please refer to Figure 3 , Figure 3 A schematic diagram of a flow chart of information summary information provided for an exemplary embodiment of the present application:

[0106] The speech text segments of each conference speaker are processed by length segmentation function and large language model to obtain key point information, and the key point information is processed again by length segmentation function and large language model to obtain information summary information, and the large language model is used to realize the Json format conversion of the information summary information. The converted information summary information containing financial information summary information corresponding to the financial performance category and business information summary information corresponding to the business progress category is cleaned to obtain more streamlined information summary information containing financial information summary information corresponding to the financial performance category and business information summary information corresponding to the business progress category. And output the streamlined information summary information.

[0107] In an alternative embodiment, in combination Figure 3 It can be seen from the content that this application provides another information summary information extraction method including the following steps S0001-S0009:

[0108] S0001. Acquire text content to be recognized. If the text content to be recognized contains a preset title and a conference logo, extract the language information used by the text content to be recognized, the title information of the text content to be recognized, and the text content;

[0109] S0002, using the language information, the title information and the text content as the target conference information;

[0110] S0003. Acquire conference text information of the target conference information, wherein the conference text information includes speech text information of at least one conference speaker;

[0111] S0004, segmenting the conference text information according to the conference speaker, and obtaining speech text segments of each conference speaker;

[0112] S0005. For each of the conference speakers, segment the speech text segments of the conference speaker to obtain a plurality of segmented text information, and based on a preset content category and each of the plurality of segmented text information, obtain segment summary information corresponding to each of the segmented text information under each of the preset content categories;

[0113] S0006. For a target content category under any preset content category, concatenate the segment summary information corresponding to each segment text information under the target content category to obtain the category content information under the target content category; generate a first prompt according to the category content information under the preset content category; extract the key point text segment corresponding to the target content category from the category content information under the target content category by using the first prompt and the pre-trained large language model, wherein the key point information of the conference speech object includes the key point text segment under the preset content category;

[0114] S0007. For any preset content category, concatenate the key text segments of each conference speaker under the preset content category to obtain the category content information under the preset content category; use the system template used to indicate the relationship between the input content and the output content of the large language model, and the user template containing the category content information under the preset content category as the second prompt; extract the category summary information corresponding to the preset content category from the category content information under the preset content category through the second prompt and the pre-trained large language model;

[0115] S0008. Generate information summary information corresponding to the target conference information based on the category summary information corresponding to each of the preset content categories;

[0116] S0009. Use a large language model to convert the information summary information into a json format to obtain converted information summary information; store the converted information summary information in an information library, or display the converted information summary information.

[0117] In summary, the solution provided by the present application is to obtain the conference text information of the target conference information, wherein the conference text information contains the speech text information of at least one conference speaker; the conference text information is segmented according to the conference speaker to obtain the speech text segments of each conference speaker; the key information of the speech text segments of each conference speaker is extracted respectively to obtain the key point information of each conference speaker; based on the preset content category and the key point information of each conference speaker, the category summary information corresponding to each preset content category is generated; based on the category summary information corresponding to each preset content category, the scheme of generating the information summary information corresponding to the target conference information is generated. The above scheme divides the target conference information into small-granularity speech text segments to obtain the key point information corresponding to the preset content category in the dimension of the small-granularity speech text segments, thereby avoiding the omission of key information, and then further extracts high-dimensional information from the obtained key point information to obtain the summary information of the target conference information, thereby ensuring that key information is not omitted and improving the accuracy of summary generation, and at the same time realizing the automatic extraction of information summary information in the conference text information, thereby improving the extraction efficiency of information summary information. It avoids the loss of information function caused by untimely extraction of information summary information, thereby saving resources. It can also be adapted to a variety of business scenarios. It is not only suitable for summaries of conference calls, but can also be applied to other fields that require content summaries, such as news reports, research reports, etc.

[0118] Specifically, the entire process from extracting target conference information to generating information summary information and then storing the information summary information is automated, reducing manual intervention and improving processing speed and accuracy. The use of a large language model improves the efficiency and quality of extracting information summary information, which can meet the needs of batch extraction of information summary information. The use of format conversion achieves consistency in the output of information summary information, and the use of a large language model also avoids the extraction of information summary information from being affected by human subjectivity, thereby improving the accuracy of the extracted information summary information.

[0119] Figure 4 A schematic diagram of a news summary information extraction device provided as an exemplary embodiment of the present application; wherein the device comprises:

[0120] An acquisition unit 21 is used to acquire conference text information of target conference information, wherein the conference text information includes speech text information of at least one conference speaker;

[0121] A segmentation unit 22, configured to segment the conference text information according to the conference speaker, to obtain speech text segments of each conference speaker;

[0122] An extraction unit 23 is used to extract key information from the speech text segments of each of the conference speakers to obtain key point information of each of the conference speakers;

[0123] A first generating unit 24 is used to generate category summary information corresponding to each preset content category based on the preset content category and key point information of each conference speech object;

[0124] The second generating unit 25 is configured to generate information summary information corresponding to the target conference information based on the category summary information corresponding to each of the preset content categories.

[0125] Optionally, in the device, the key point information of the conference speech object includes the key point text segments under the preset content category; the device is used to extract key information from the speech text segments of each of the conference speech objects respectively to obtain the key point information of each of the conference speech objects, specifically for: segmenting the speech text segments of the conference speech object for each of the conference speech objects to obtain multiple segmented text information; based on the preset content category and each of the segmented text information in the multiple segmented text information, obtaining the segmented summary information corresponding to each of the segmented text information under each of the preset content categories; for the target content category under any preset content category, generating the key point text segments of the conference speech object under the target content category according to the segmented summary information corresponding to all the segmented text information under the target content category.

[0126] Optionally, the device is used to generate a key point text segment of the conference speech object under the target content category based on the segmented summary information corresponding to all the segmented text information under the target content category, and is specifically used to: splice the segmented summary information corresponding to each of the segmented text information under the target content category to obtain the category content information under the target content category; generate a first prompt based on the category content information under the preset content category; and extract the key point text segment corresponding to the target content category from the category content information under the target content category through the first prompt and a pre-trained large language model.

[0127] Optionally, the device is used to generate category summary information corresponding to each preset content category based on the preset content category and the key point information of each conference speech object, and is specifically used to: for any preset content category, splice the key point text segments of each conference speech object under the preset content category to obtain the category content information under the preset content category; generate a second prompt based on the category content information under the preset content category; and extract the category summary information corresponding to the preset content category from the category content information under the preset content category through the second prompt and a pre-trained large language model.

[0128] Optionally, the device is used to generate a second prompt based on the category content information under the preset content category, specifically for: using a system template used to indicate the relationship between the input content and output content of the large language model, and a user template containing the category content information under the preset content category as the second prompt.

[0129] Optionally, the device is also used to: use a large language model to convert the information summary information into a json format to obtain converted information summary information; store the converted information summary information in an information library, or display the converted information summary information.

[0130] Optionally, the device is also used to: obtain conference type information of the conference message, and if the conference type information contains a preset title and conference identifier, determine that the conference message is a target conference message; extract the language information used by the conference message, the title information of the conference message, and the text content of the conference message from the target conference message; and use the language information, the title information, and the text content as the target conference information.

[0131] It should be understood that the device embodiment and the method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, no further description is given here. Specifically, the device may perform the above method embodiment, and the above and other operations and / or functions of each module in the device are the corresponding processes in each method in the above method embodiment, respectively, and no further description is given here for the sake of brevity.

[0132] The above describes the device of the embodiment of the present application from the perspective of the functional module in conjunction with the accompanying drawings. It should be understood that the functional module can be implemented in hardware form, can be implemented by instructions in software form, and can also be implemented by a combination of hardware and software modules. Specifically, the steps of the method embodiment in the embodiment of the present application can be completed by the hardware integrated logic circuit and / or software form instructions in the processor, and the steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or a combination of hardware and software modules in the decoding processor to perform. Optionally, the software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory, and completes the steps in the above method embodiment in conjunction with its hardware.

[0133] Figure 5 is a schematic block diagram of an electronic device provided in an embodiment of the present application, and the electronic device may include:

[0134] The memory 301 and the processor 302, the memory 301 is used to store the computer program and transmit the program code to the processor 302. In other words, the processor 302 can call and run the computer program from the memory 301 to implement the method in the embodiment of the present application.

[0135] For example, the processor 302 may be configured to execute the above method embodiments according to instructions in the computer program.

[0136] In some embodiments of the present application, the processor 302 may include but is not limited to:

[0137] General-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc.

[0138] In some embodiments of the present application, the memory 301 includes but is not limited to:

[0139] Volatile memory and / or non-volatile memory. Among them, the non-volatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0140] In some embodiments of the present application, the computer program may be divided into one or more modules, which are stored in the memory 301 and executed by the processor 302 to complete the method provided by the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0141] like Figure 5 As shown, the electronic device may also include:

[0142] The transceiver 303 may be connected to the processor 302 or the memory 301 .

[0143] The processor 302 may control the transceiver 303 to communicate with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices. The transceiver 303 may include a transmitter and a receiver. The transceiver 303 may further include an antenna, and the number of antennas may be one or more.

[0144] It should be understood that the various components in the electronic device are connected via a bus system, wherein the bus system includes not only a data bus but also a power bus, a control bus and a status signal bus.

[0145] The present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a computer, the computer can perform the method of the above method embodiment. In other words, the present application embodiment also provides a computer program product containing instructions, and when the instructions are executed by a computer, the computer can perform the method of the above method embodiment.

[0146] When software is used for implementation, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integration. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a digital video disc (digital video disc, DVD)), or a semiconductor medium (e.g., a solid state drive (solid state disk, SSD)), etc.

[0147] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0148] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the module is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0149] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. For example, each functional module in each embodiment of the present application may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0150] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A news summary information extraction method, characterized in that: include: Acquire conference text information of the target conference information, wherein the conference text information includes speech text information of at least one conference speaker; The conference text information is segmented according to the conference speaker to obtain speech text segments of each conference speaker; Extract key information from the speech text segments of each of the conference speakers to obtain key point information of each of the conference speakers; Based on the preset content categories and the key point information of each conference speech object, generating category summary information corresponding to each preset content category; Based on the category summary information corresponding to each of the preset content categories, information summary information corresponding to the target conference information is generated.

2. The method according to claim 1, characterized in that The key point information of the conference speech object includes the key point text segment under the preset content category; The extracting key information from the speech text segments of each of the conference speakers to obtain key point information of each of the conference speakers includes: For each of the conference speakers, segment the speech text segments of the conference speaker to obtain a plurality of segmented text information; Based on the preset content category and each of the segmented text information in the plurality of segmented text information, obtaining segmented summary information corresponding to each of the segmented text information under each of the preset content categories; For a target content category under any preset content category, a key point text segment of the conference speech object under the target content category is generated according to the segment summary information corresponding to all the segment text information under the target content category.

3. The method according to claim 2, characterized in that The step of generating a key point text segment of the conference speech subject under the target content category according to the segment summary information corresponding to all the segment text information under the target content category includes: splicing the segment summary information corresponding to each segment text information under the target content category to obtain category content information under the target content category; Generate a first prompt according to the category content information under the preset content category; By using the first prompt and the pre-trained large language model, a key point text segment corresponding to the target content category is extracted from the category content information under the target content category.

4. The method according to claim 2, characterized in that: The generating of category summary information corresponding to each preset content category based on the preset content category and the key point information of each conference speech object includes: For any preset content category, concatenate the key text segments of each conference speaker under the preset content category to obtain category content information under the preset content category; generating a second prompt according to the category content information under the preset content category; The category summary information corresponding to the preset content category is extracted from the category content information under the preset content category by using the second prompt and the pre-trained large language model.

5. The method according to claim 4, characterized in that The step of generating a second prompt according to the category content information under the preset content category includes: A system template for indicating the relationship between the input content and the output content of the large language model and a user template containing category content information under the preset content category are used as the second prompt.

6. The method according to claim 1, characterized in that The method further comprises: The information summary information is converted into JSON format using a large language model to obtain converted information summary information; The converted information summary information is stored in an information library, or the converted information summary information is displayed.

7. The method according to claim 1, characterized in that The method further comprises: Acquire conference type information of the conference message, and if the conference type information includes a preset title and a conference identifier, determine that the conference message is a target conference message; Extracting language information used by the conference message, title information of the conference message, and text content of the conference message from the target conference message; The language information, the title information and the text content are used as the target conference information.

8. A news summary information extraction device, characterized in that: include: An acquisition unit, configured to acquire conference text information of target conference information, wherein the conference text information includes speech text information of at least one conference speaker; A segmentation unit, configured to segment the conference text information according to the conference speaker, to obtain speech text segments of each conference speaker; An extraction unit, used to extract key information from the speech text segments of each of the conference speakers, to obtain key point information of each of the conference speakers; A first generating unit, configured to generate category summary information corresponding to each preset content category based on the preset content category and key point information of each conference speech object; The second generating unit is used to generate information summary information corresponding to the target conference information based on the category summary information corresponding to each of the preset content categories.

9. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images