Meeting minutes generation method and related device based on natural language processing

Abstract

The present application relates to the field of artificial intelligence technology, and provides a method for generating meeting minutes based on natural language processing and related devices, wherein the method includes: receiving a meeting minutes generation request sent by a user, wherein the meeting minutes generation request includes a meeting data set of a product briefing; obtaining a target text corresponding to the promotional product in the product briefing and text features of the target text, as well as a target image and image features of the target image based on the meeting data set; obtaining the minutes page style corresponding to the text features and the image features; and generating the meeting minutes of the product briefing based on the minutes page style, the target text, and the target image. The use of this application can improve the generation efficiency and accuracy of meeting minutes, which is conducive to the accumulation and precipitation of meeting minutes, as well as the promotional effect of promotional products.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and mainly to a method for generating meeting minutes based on natural language processing and related devices. Background Art

[0002] With the rapid development of the Internet, the Internet can be used to achieve goals such as brand and product marketing and customer service, and to hold meetings for publicity. For example, in the insurance field, product briefings for insurance products are broadcast through Internet videos. After the meeting, the minutes of the meeting often need to be sorted out, forwarded, and even followed up.

[0003] At present, the minutes of meetings are usually recorded manually by the recorders. Although a small number of them use voice recognition technology to recognize the speaker's voice and convert it into text records, the minutes of meetings are in a single form and it is difficult to improve the publicity effect. Summary of the invention

[0004] The embodiment of the present application provides a method and related device for generating meeting minutes based on natural language processing, which can improve the generation efficiency and accuracy of meeting minutes, facilitate the accumulation and sedimentation of meeting minutes, and improve the publicity effect of publicity products.

[0005] In a first aspect, an embodiment of the present application provides a method for generating meeting minutes based on natural language processing, wherein:

[0006] receiving a request for generating meeting minutes sent by a user, wherein the request for generating meeting minutes includes a meeting data set of a product briefing;

[0007] Acquire, based on the conference data set, a target text corresponding to the product promoted in the product briefing and text features of the target text, as well as a target image and image features of the target image;

[0008] Obtaining the minutes page style corresponding to the text feature and the image feature;

[0009] The minutes of the product briefing meeting are generated based on the minutes page style, the target text and the target image.

[0010] In a second aspect, an embodiment of the present application provides a device for generating meeting minutes based on natural language processing, wherein:

[0011] A receiving unit, configured to receive a request for generating a meeting minutes sent by a user, wherein the request for generating the meeting minutes includes a meeting data set for promoting a product;

[0012] An acquisition unit, configured to acquire, based on the conference data set, a target text corresponding to the product promoted in the product briefing and text features of the target text, as well as a target image and image features of the target image; and acquire a minutes page style corresponding to the text features and the image features;

[0013] A generating unit is used to generate the minutes of the product briefing based on the minutes page style, the target text and the target image.

[0014] In a third aspect, an embodiment of the present application provides a computer device, comprising a processor, a memory, a communication interface, and a computer program, wherein the computer program is stored in the memory, the computer program is configured to be executed by the processor, and the computer program includes instructions for some or all of the steps described in the first aspect.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute to implement some or all of the steps described in the first aspect.

[0016] Implementing the embodiments of the present application will have the following beneficial effects:

[0017] After adopting the above-mentioned method and related device for generating meeting minutes based on natural language processing, after receiving the meeting minutes generation request sent by the user, firstly, based on the meeting data set of the product briefing carried in the meeting minutes generation request, obtain the target text and text features of the target text corresponding to the product promoted in the product briefing, as well as the target image and image features of the target image. Then obtain the minutes page style corresponding to the text features and image features, and then generate the promotional page of the product briefing based on the minutes page style, target text and target image. In this way, the meeting minutes of the product briefing can be generated based on the meeting data set of the product briefing, which can avoid data omission or recording errors, improve the generation efficiency and accuracy of the meeting minutes, and facilitate the accumulation and precipitation of the meeting minutes. And the meeting minutes are generated based on the minutes page style obtained based on the text features of the target text corresponding to the promotional product and the image features of the target image in the meeting data set, which is conducive to improving the publicity effect of the promotional product. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] in:

[0020] Figure 1 A flowchart of a method for generating meeting minutes based on natural language processing provided in an embodiment of the present application;

[0021] Figure 2 A schematic diagram of the structure of a device for generating meeting minutes based on natural language processing provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. According to the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0025] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0026] The network architecture used in the embodiment of the present application includes a server and an electronic device. The embodiment of the present application does not limit the number of electronic devices and servers, and the server can provide services for multiple electronic devices at the same time. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The server may be implemented by a server cluster consisting of multiple servers.

[0027] The electronic device may be a personal computer (PC), a laptop or a smart phone, or an all-in-one machine, a handheld computer, a tablet computer (pad), a smart TV player terminal, a vehicle-mounted terminal or a portable device. The operating system of the electronic device on the PC side, such as an all-in-one machine, may include but is not limited to Linux system, Unix system, Windows series system (such as Windows XP, Windows 7, etc.), Mac OS X system (Apple computer operating system) and other operating systems. The operating system of the electronic device on the mobile side, such as a smart phone, may include but is not limited to Android system, IOS (Apple mobile phone operating system), Window system and other operating systems.

[0028] The electronic device can install and run the application, and the server can be a server corresponding to the application installed in the electronic device, providing application services for the application. The application can be a separately integrated application software, or a small program embedded in other applications, or a system on a web page, etc., which is not limited here.

[0029] The promotional products involved in this application may be products in the insurance field, or may be products in the real estate field, or may be products in the medical field, or may be products in the Internet field, etc., and are not limited here.

[0030] For example, after the product briefing for an insurance product is over, the relevant personnel upload the meeting data set of the product briefing and send a request to generate a meeting minutes carrying the meeting data set to the server through an electronic device. Then, the server can generate the meeting minutes of the promoted insurance product based on the meeting data set, and the insurance agent or staff can promote the insurance product by forwarding the meeting minutes, or the staff or team that organized the product briefing can store the meeting minutes for historical accumulation and business precipitation, which is conducive to improving the success rate of the next meeting.

[0031] The embodiment of the present application proposes a method for generating meeting minutes based on natural language processing, which can be executed by a device for generating meeting minutes based on natural language processing. The device can be implemented by software and / or hardware, and can generally be integrated in an electronic device or a server, which can improve the generation efficiency and display effect of meeting minutes, and is conducive to improving the promotion effect of products.

[0032] Please refer to Figure 1 , Figure 1 This is a flowchart of a method for generating meeting minutes based on natural language processing provided by the present application. Taking the method applied to a server as an example, the method includes the following steps S101 to S104, wherein:

[0033] S101: Receive a request for generating meeting minutes sent by a user, where the request for generating meeting minutes includes a meeting data set of a product briefing.

[0034] This application does not limit users and product briefings. Users can be planners or other staff members of the product briefing, or people participating in the product briefing, etc. A product briefing is a meeting attended by multiple people to explain the promotional products. The promotional products can refer to the aforementioned products in various fields. The product briefing can be in the form of online or offline. The people participating in the product briefing can at least include an explainer (or host), who refers to the person responsible for introducing the promotional products at the product briefing.

[0035] It should be noted that the number of explainers can be one or more. In addition to the explainers, the characters participating in the product briefing can also be interactive characters or other personnel, such as work characters, etc. Among them, interactive characters include people who interact with the explainer online, or people who participate in the product briefing discussion offline. For example, in the case where the product briefing is played in the form of a video, the interactive characters may include people who speak in the comment area or barrage area, or may include people who send comments such as likes and dislikes, or may include people who are interviewed online, etc.

[0036] In the embodiment of the present application, the meeting minutes generation request is used to generate the meeting minutes of the product briefing for promoting the product. The meeting minutes are used to record the general content of the product briefing, which can be forwarded and promoted by the relevant staff of the product briefing, or can be stored for reference and improvement by subsequent personnel.

[0037] The request for generating meeting minutes may include a meeting data set of a product briefing. The meeting data set may be in the form of text, image, audio, or video, etc., which is not limited here. It can be understood that when the meeting data set is the audio or video of the product briefing, the step of collating the meeting data set can be omitted, which can improve the efficiency of generating meeting minutes.

[0038] The meeting data set may include data for promoting the product. The data may include display data, such as presentation documents, promotional pages, etc. Among them, the display data may include the meeting theme, meeting location, meeting object, time information, etc. of the product briefing, which are not limited here. The data data may also be a promotional plan or discussion document for promoting the product. It can be understood that the data data is used to disclose the theme of the product briefing and other contents. Generating the minutes of the product briefing based on the data of the product briefing can improve the accuracy of generating the minutes, which is conducive to improving the accumulation and precipitation of data data.

[0039] The conference data set may also include audio or video of the speaker's explanation of the product. It is understandable that the speaker's audio or video can reflect the speaker's explanation style and product features. Generating the minutes of the product briefing based on the speaker's conference data set can improve the accuracy of the generated meeting minutes and help improve the explanation effect.

[0040] The meeting data set may also include interactive data of the promotional product. The interactive data is data of the interactive personnel, such as online participation video or audio, offline participation comment information, etc. It can be understood that the interactive data can reflect the response of the product briefing. Generating the minutes of the product briefing based on the interactive data can improve the accuracy of the generated minutes, which is conducive to improving the promotion effect.

[0041] S102: Acquire, based on the conference data set, a target text corresponding to the product promoted in the product briefing and text features of the target text, as well as a target image and image features of the target image.

[0042] In an embodiment of the present application, the target text refers to a text that can be used in the minutes of a meeting. The target text may include the meeting theme features of the product briefing, and may be determined based on the images and / or text corresponding to the data data, the audio and / or video of the lecturer, the discussion text of the interactive characters in the interactive data, and the like. The text features of the target text may also include attribute information of the promotional product. Exemplarily, the promotional product is a life insurance product, and the attribute information of the life insurance product may include the investment amount, investment method, payment period, insurance platform, participating objects, and the like. Furthermore, it may also include information such as the participation time and participation method of the promotional product. It can be understood that recording the minutes of a meeting in text can improve the readability of the meeting minutes display page.

[0043] The target image refers to an image that can be used in the meeting minutes. The image features of the target image may include a product image for promoting a product, a scene image of a product briefing, or an image of an answer to a related question, etc. It can be understood that recording the meeting minutes with images can improve the readability of the meeting minutes display page.

[0044] The present application does not limit the method for obtaining the target text, the text features of the target text, the target image, and the image features of the target image. The conference data set can be classified to obtain a text data set and an image data set. Then, the target text is selected from the text data set, and the target image is selected from the image data set; then, the text features of the target text are obtained, and the image features of the target image are obtained. Alternatively, the text features of the text data set can be obtained first, and the image features of the image data set can be obtained; then, based on the text features of the text data set, the target text is selected from the text data set to obtain the text features of the target text, and based on the image features of the image data set, the target image is selected from the image data set to obtain the image features of the target image, etc.

[0045] The following is an explanation of the conference data set including the information data of the promotional product, the audio of the explanation of the promotional product, and the interactive data of the promotional product. It should be noted that this application does not limit the order of obtaining the target text and the text features of the target text, and the target image and the image features of the target image based on the information data, the audio of the explanation, and the interactive data. Optionally, steps A1 and A2 can be performed before steps B1 to B3 and steps C1 to C3.

[0046] In a possible example, the conference data set includes information data for promoting products, and step S102 may include the following steps A1 to A5, wherein:

[0047] A1. Classify the data to obtain a data text set and a data image set.

[0048] If the data includes images and texts, step A1 can be directly classified to obtain a data text set containing texts and a data image set containing images. If the data includes videos, the videos need to be split to obtain image data and audio data. Then, speech recognition is performed on the audio data to obtain text. Or further, the text corresponding to the image data is obtained. Then, the images and texts are classified to obtain a data text set containing texts and a data image set containing images. If the data includes audio, speech recognition is performed on the audio to obtain text. Or further, the images corresponding to the texts are obtained. Then, the images and texts are classified to obtain a data text set corresponding to the texts and a data image set corresponding to the images.

[0049] A2. Selecting a target text corresponding to the promotional product from the data text set.

[0050] In a possible example, step A2 may include the following steps A21 to A24, wherein:

[0051] A21. Perform named entity recognition on the data text in the data text set to obtain a text vector of the data text.

[0052] Among them, Named Entity Recognition (NER) is an important basic tool in application fields such as information extraction, question-answering systems, syntactic analysis, and machine translation. Generally speaking, the purpose of named entity recognition is to identify three major categories (entity class, time class, and number class) and seven minor categories (name, organization name, place name, time, date, currency, and percentage) of named entities in the text to be processed. For example, the sentence: Xiao Ming goes to school at 8 o'clock in the morning. Performing named entity recognition on this sentence, the information that should be extracted is {name: Xiao Ming; time: 8 o'clock in the morning; place: school}. In an embodiment of the present application, the text set obtained by performing named entity recognition on the data text is used as the text vector of the data text. Optionally, each entity in the text vector may include a weight corresponding to the part of speech of the entity.

[0053] In another possible example, step A21 may include: extracting a summary of the document text in the document text set to obtain a text vector of the document text.

[0054] Among them, summary extraction may include the above-mentioned named entity recognition method, or may include the following steps: determining the summary probability of each word in the document text; taking the word corresponding to the maximum value of the summary probability as the first word; taking the word after the first word in the document text as the second word; based on the combined summary probability of the first word and the second word, selecting at least one third word from the second word; obtaining a text sequence composed of the first word and the third word to obtain a text vector of the document text.

[0055] In the embodiments of the present application, the summary probability of a word refers to the probability that the word is part of the summary. The summary probability can be understood as the importance of the word, for example, the impact of the word on the change of the meaning of the sentence, whether the word belongs to a modifying word such as an adjective. In other words, the greater the summary probability, the higher the accuracy of the word in generating a summary. The present application does not limit the method for determining the summary probability of a word, which can be determined based on one or more of the word's part of speech, word meaning or word frequency, or can be obtained by evaluating the word based on a preset algorithm, for example, using a term frequency (TF)-inverse document frequency (IDF) algorithm.

[0056] The first character refers to the character corresponding to the maximum value of the summary probability, the second character is the character after the first character in the data text, and the third character is the character selected from the second character. The combined summary probability refers to the probability that the character and the character combination are part of the summary. In other words, the combined summary probability of the first character and the second character refers to the probability that the first character and the second character combination are part of the summary. It can be understood that the greater the combined summary probability of the first character and the second character, the greater the probability that the second character is combined with the first character to generate a summary of the data text. The present application does not limit the method for determining the combined summary probability, which can be determined based on one or more of the parts of speech, meanings or frequencies of the characters to be combined, or can be determined based on the matching values ​​of the combined characters and the characters in the preset dictionary, or obtained using the TF-IDF algorithm, etc.

[0057] It can be understood that in this example, the first word is selected based on the summary probability of the word. Then, the third word is selected from the second word after the first word in the data text in combination with the first word, so that the text sequence composed of the first word and the third word is used as the text vector of the data text. In this way, the accuracy of generating the summary can be improved, which is conducive to improving the accuracy of selecting the target text.

[0058] A22. Determine a first correlation value between the data text and the promotional product.

[0059] The first association value between the data text and the promotional product refers to the relevance of the data text for describing the promotional product. The larger the first association value, the more obvious the characteristics of the promotional product described are. The first association value can be determined based on the association value between the entity in the text vector of the data text and the promotional product, or the number of products corresponding to the data text can be obtained, and the first association value can be determined based on the number, etc., which is not limited here.

[0060] A23. Select a target data text from the data text based on the first association value.

[0061] The target data text may be the data text corresponding to the largest first association value, or may be the first N data texts, etc. The present application does not limit N, and may be a fixed value, or may be determined based on the number of data texts, etc. For example, if N is 3, the first association values ​​between the data texts and the promotional products are sorted in order of size, and the data texts corresponding to the first three first association values ​​are selected as the target data texts.

[0062] A24. Use the text vector corresponding to the target data text as the target text corresponding to the promotional product.

[0063] It can be understood that in steps A21 to A24, named entity recognition is first performed on each document text to obtain a text vector of each document text. Then, based on the association value between the document text and the promotional product, the target document text is selected from the document text. Then, the text vector corresponding to the target document text is used as the target text corresponding to the promotional product. In this way, using the text vector related to the promotional product as the target text can improve the accuracy of the target text selected as the meeting minutes.

[0064] A3. Acquire text features of the target text.

[0065] The text features of the target text may be determined based on the part of speech, meaning and frequency of each character or word in the target text. Alternatively, the text features may be determined based on the sentiment in the target text, etc., which is not limited here. In the process of selecting the target text, if some text features are obtained, the obtained text features may be fused.

[0066] A4. Acquire a target image corresponding to the promotional product from the data image set.

[0067] In a possible example, step A4 may include the following steps A41 to A44, wherein:

[0068] A41. Obtain a second association value between the document image in the document image set and the target text.

[0069] The second association value between the document image and the target text refers to the relevance of the document image for explaining the target text. The larger the second association value, the greater the probability that the document image is used to explain the target text. The second association value can be determined based on the similarity value between the text corresponding to the document image and the target text, or can be determined by obtaining the matching value between the target object in the document image and the entity in the target text, etc., which is not limited here.

[0070] A42. Obtain the important value of the data image.

[0071] The greater the importance value of the data image, the more beneficial effects can be brought by retaining the data image in the meeting minutes. The importance value of the data image can be determined based on the image intent of the data image, or can be determined based on the number of similar data images, or can be determined based on the clarity of the data image, etc., which is not limited here.

[0072] A43. Calculate the evaluation value of the material image based on the second relevance value and the importance value.

[0073] The evaluation value of the data image may include the product of the second correlation value and the importance value, the maximum value, the minimum value or the weighted average value, etc., which is not limited here.

[0074] A44. Based on the evaluation value of the data image, select a target image corresponding to the promotional product from the data image.

[0075] The target image may be the document image corresponding to the largest evaluation value, or may be the first M document images, etc. The present application does not limit M, and it may be a fixed value, or may be determined based on the number of document images, etc. For example, if M is 2, the evaluation values ​​of the document images are sorted in order of size, and the document images corresponding to the first two evaluation values ​​are selected as the target images.

[0076] It can be understood that in steps A41 to A44, the association value between the data image and the target text, as well as the importance value of the data image, are first obtained. Then, the evaluation value of the data image is calculated based on the association value and the importance value. Then, the target image corresponding to the promotional product is selected from the data images based on the evaluation value. In this way, the target image is selected based on the image related to the target text and the importance value of the image, which can improve the accuracy of the target image selected as the meeting minutes.

[0077] A5. Obtain image features of the target image.

[0078] The image features of the target text may be determined based on the image intent of the target image. Alternatively, the image features may be determined based on the composition of the object or the expression of the person in the target image, etc., which are not limited here. In the process of selecting the target image, if some image features are obtained, the obtained image features may be fused.

[0079] It should be noted that step A2 may be performed before step A4, or after step A4, or simultaneously with step A4.

[0080] It can be understood that in steps A1 to A5, the data data is first classified to obtain a data text set and a data image set. Then, the target text corresponding to the promotional product is selected from the data text set, and the target image corresponding to the promotional product is obtained from the data image set. Then, the text features of the target text are obtained, and the image features of the target image are obtained. In this way, obtaining from the data data of the promotional product is conducive to improving the accuracy of obtaining the target text and the text features of the target text, as well as the target image and the image features of the target image.

[0081] In a possible example, the conference data set also includes the audio of the speaker's explanation of the product promotion. Before step A3, the following steps B1 to B3 may also be included, wherein:

[0082] B1. Perform feature recognition on the audio of the explanation by the explainer to obtain the voice features of the explainer.

[0083] B2. Perform semantic recognition on the audio of the explanation by the explainer to obtain the speech text and language features of the explainer.

[0084] The voice characteristics of the guide include the characteristics of the guide's volume, pitch, loudness, etc. Language characteristics may include language emotions, the composition of language text, etc.

[0085] It should be noted that step B1 can be performed before step B2, step B1 can be performed after step B2, or step B1 can be performed simultaneously with step B2. After obtaining the voice features and language features of the tour guide, step A3 can also include: using the voice features and language features of the tour guide as text features of the target text. It can be understood that the voice features and language features of different characters can bring different atmospheres. Acquiring the text features of the tour guide's speech based on the tour guide's voice features and language features can improve the comprehensiveness and accuracy of acquiring text features.

[0086] B3. Selecting a target text corresponding to the promotional product from the voice text of the narrator.

[0087] The target text may include text or abstracts in the language text that are related to the promotional product.

[0088] It can be understood that in the above example, the voice features, speech text and language features of the tour guide are first obtained based on the tour guide's audio. Then, the target text corresponding to the promotional product is selected from the tour guide's speech text, and the voice features and language features of the tour guide are used as the text features of the target text. In this way, the accuracy of obtaining the target text and the text features of the target text can be improved.

[0089] It should be noted that if the conference data set includes a video of a speaker explaining a product, rather than an audio of the explanation, the explanation video can be split first to obtain an explanation image and an explanation audio, and then steps B1 to B3 are performed, as well as the step of obtaining a target text or a target image based on the explanation image.

[0090] In a possible example, the conference data set also includes the audio of the speaker's explanation of the product promotion. Before step A3, the following steps C1 to C3 may also be included, wherein:

[0091] C1. Classify the interactive data to obtain a first interactive text and an interactive image.

[0092] Among them, the first interactive text and interactive image can be obtained by classification based on the data type of the interactive data. For example, the interactive data includes images and texts, and step C1 can be directly classified to obtain the first interactive text and interactive image. If the interactive data includes video, it is necessary to split the video to obtain image data and audio data. Then perform speech recognition on the audio data to obtain text. Or further, obtain the text corresponding to the image data. Then classify the image and text to obtain the first interactive text and interactive image. If the data data includes audio, it is necessary to perform speech recognition on the audio to obtain text. Or further, obtain the image corresponding to the text. Then classify the image and text to obtain the first interactive text and interactive image.

[0093] C2. Identify a second interactive text corresponding to the interactive image.

[0094] The second interactive text may be obtained based on the composition of each object in the interactive image, or the facial expression features and body features of a character, etc., which are not limited here.

[0095] C3. Based on the first interactive text and the second interactive text, obtain a target text corresponding to the promotional product.

[0096] The target text may be similar information obtained by classifying the first interactive text and the second interactive text, or may be information selected from the content corresponding to each interactive theme, etc., which is not limited here.

[0097] In a possible example, step C3 may include the following steps C31 to C34, wherein:

[0098] C31. Classify the first interactive text and the second interactive text to obtain sub-text sets corresponding to at least two types of interactive topics.

[0099] The interactive theme may be various promotional products, or may be the names of different links, etc. It can be understood that selecting the target text based on the subtext set corresponding to the interactive theme can improve the comprehensiveness and accuracy of the selection.

[0100] C32. Analyze the sub-text set corresponding to the interactive theme to obtain the support rate and discussion volume of the interactive theme, as well as the target interactive text.

[0101] The support rate is used to describe the degree of support for the interactive topic by the interactive characters, and can be determined based on the lexical meaning of the text in the subtext set corresponding to the interactive topic, for example, the number of words indicating support, etc. The discussion volume is used to describe the degree of attention paid by the interactive characters to the interactive topic, and can be calculated based on the number of comment information, the number of interactive characters, the number of online characters, etc.

[0102] The target interactive text refers to the text corresponding to the sub-text set corresponding to the interactive theme, which can be understood as the comprehensive information corresponding to each text. The target text can be obtained by analyzing the support degree of the promotional product based on the semantics of the text in the sub-text set.

[0103] C33. Select a target interactive topic from the interactive topics based on the support rate and discussion volume of the interactive topic.

[0104] The target interactive topic may be the interactive topic corresponding to the maximum support rate, the interactive topic corresponding to the maximum discussion volume, or the first Q interactive topics. Alternatively, the evaluation value of the interactive topic may be calculated based on the support rate and discussion volume of the interactive topic, and then the interactive topic may be selected based on the evaluation value, etc., which is not limited here.

[0105] C34. Use the target interactive text of the target interactive theme as the target text corresponding to the promotional product.

[0106] It can be understood that in steps C31 to C34, the first interactive text and the second interactive text are first classified to obtain sub-text sets corresponding to various interactive topics. Then, based on the sub-text sets corresponding to various interactive topics, the support rate and discussion volume of the interactive topic, as well as the target interactive text, are obtained. Then, based on the support rate and discussion volume of the interactive topic, the target interactive topic is selected from the interactive topics, and the target interactive text of the target interactive topic is used as the target text corresponding to the promotional product. In this way, the promotional effect of the target text as a promotional product can be improved.

[0107] In steps C1 to C3, the interactive data is first classified to obtain the first interactive text and the interactive image. Then the second interactive text corresponding to the interactive image is identified. Then the target text corresponding to the promotional product is obtained based on the first interactive text and the second interactive text. In this way, the text corresponding to the interactive text and the interactive image is taken into consideration, which can improve the comprehensiveness of the target text.

[0108] It should be noted that if the conference data set includes interactive data for promoting a product, and there is no interactive text in the interactive data but there is an interactive image, then step C2 can be executed, and then the target text corresponding to the promotion product can be obtained based on the second interactive text obtained in step C2, and the target image can be obtained based on the interactive image. The present application does not limit the method for obtaining the target image, and the second interactive text can be classified and selected based on the interactive theme in the second interactive text corresponding to the interactive image, and then the target image can be formed based on the selected text, etc., which is not limited here.

[0109] If the conference data set includes interactive data for promoting products, and the interactive data contains interactive texts but no interactive images, the target text corresponding to the promotional product can be obtained based on the interactive texts. In addition, the interactive texts can be classified and selected based on the interactive topics, and then the target images can be formed based on the selected texts, etc., which are not limited here.

[0110] S103: Obtain the minutes page style corresponding to the text feature and the image feature.

[0111] In the embodiment of the present application, the minutes page style refers to the style of the display page of the minutes of the product briefing, which may include the image style and text style in the display page, or further may include the image style of each image and the text style of each text.

[0112] The style of the minutes page can be generated using a pre-trained page generation model. This application does not limit the page generation model. In one possible example, the page generation model is a generative adversarial network (GAN). The training method of the GAN may include the following steps: obtaining a reference meeting minutes page from the network; grouping the reference meeting minutes page to obtain a training data set and a test data set; performing unsupervised learning based on the training data set to obtain an initialization model of the GAN; performing supervised learning on the initialization model of the GAN based on the test data set to obtain an initial model of the GAN.

[0113] The reference meeting minutes page may include various styles, and each style carries label information of image features and text features. The label information can be obtained by classification based on the pre-trained CNN convolutional neural network. In this way, obtaining the minutes page style through the trained GAN is conducive to improving the display effect of the generated meeting minutes.

[0114] S104: Generate meeting minutes of the product briefing based on the minutes page style, the target text, and the target image.

[0115] After obtaining the minutes page style, the target text and target image can be processed based on the minutes page style so that the target text and target image meet the corresponding minutes page style. This application does not limit the order of the target text and target image in the meeting minutes. It can be determined based on the historical meeting minutes page, or can be obtained by sorting the corresponding summaries of the target text and target image, or can be determined based on the occurrence time of the target text and target image, etc.

[0116] exist Figure 1In the method shown, after receiving the request for generating meeting minutes sent by the user, the target text and text features of the target text, as well as the target image and image features of the target image corresponding to the product promoted in the product briefing are first obtained based on the meeting data set of the product briefing carried in the request for generating meeting minutes. Then the minutes page style corresponding to the text features and image features is obtained, and then the promotional page of the product briefing is generated based on the minutes page style, the target text and the target image. In this way, the meeting minutes of the product briefing can be generated based on the meeting data set of the product briefing, which can avoid data omissions or recording errors, improve the generation efficiency and accuracy of the meeting minutes, and facilitate the accumulation and precipitation of the meeting minutes. Moreover, the meeting minutes are generated based on the minutes page style obtained based on the text features of the target text corresponding to the promotional product and the image features of the target image in the meeting data set, which is conducive to improving the publicity effect of the promotional product.

[0117] The method of the embodiment of the present application is described in detail above, and the device of the embodiment of the present application is provided below.

[0118] and Figure 1 The embodiment shown is consistent with that shown in FIG. Figure 2 , Figure 2 This is a schematic diagram of the structure of a meeting minutes generation device based on natural language processing proposed in this application. Figure 2 As shown, the above-mentioned conference minutes generation device 200 based on natural language processing includes:

[0119] The receiving unit 201 is used to receive a meeting minutes generation request sent by a user, wherein the meeting minutes generation request includes a meeting data set for promoting a product;

[0120] The acquisition unit 202 is used to acquire the target text and text features of the target text corresponding to the product promoted in the product briefing based on the conference data set, as well as the target image and image features of the target image; and acquire the minutes page style corresponding to the text features and the image features;

[0121] The generating unit 203 is used to generate the minutes of the product briefing based on the minutes page style, the target text and the target image.

[0122] In a possible example, the conference data set includes data on the promotional product, and the acquisition unit 202 is specifically used to classify the data data to obtain a data text set and a data image set; select a target text corresponding to the promotional product from the data text set; obtain text features of the target text; obtain a target image corresponding to the promotional product from the data image set; and obtain image features of the target image.

[0123] In a possible example, the acquisition unit 202 is specifically used to perform named entity recognition on the data text in the data text set to obtain a text vector of the data text; determine a first association value between the data text and the promotional product; based on the first association value, select a target data text from the data text; and use the text vector corresponding to the target data text as the target text corresponding to the promotional product.

[0124] In a possible example, the acquisition unit 202 is specifically used to obtain a second association value between the data image in the data image set and the target text; obtain an importance value of the data image; calculate an evaluation value of the data image based on the second association value and the importance value; and select a target image corresponding to the promotional product from the data images based on the evaluation value of the data image.

[0125] In a possible example, the conference data set also includes the audio of the commentator's explanation of the promotional product, and the acquisition unit 202 is specifically used to perform feature recognition on the commentator's audio to obtain the commentator's voice features; perform semantic recognition on the commentator's audio to obtain the commentator's voice text and language features; select the target text corresponding to the promotional product from the commentator's voice text; and use the commentator's voice features and language features as text features of the target text.

[0126] In a possible example, the conference data set also includes interactive data of the promotional product, and the acquisition unit 202 is specifically used to classify the interactive data to obtain a first interactive text and an interactive image; identify a second interactive text corresponding to the interactive image; and acquire a target text corresponding to the promotional product based on the first interactive text and the second interactive text.

[0127] In a possible example, the acquisition unit 202 is specifically used to classify the first interactive text and the second interactive text to obtain sub-text sets corresponding to at least two types of interactive topics; analyze the sub-text sets corresponding to the interactive topics to obtain the support rate and discussion volume of the interactive topics, as well as the target interactive text; select a target interactive topic from the interactive topics based on the support rate and discussion volume of the interactive topic; and use the target interactive text of the target interactive topic as the target text corresponding to the promotional product.

[0128] The detailed execution process of each unit in the natural language processing-based conference minutes generation device 200 can be found in the execution steps in the aforementioned method embodiment, which will not be repeated here.

[0129] and Figure 1 For details, please refer to Figure 3 , Figure 3 Schematic diagram of the structure of a computer device provided in an embodiment of the present application. Figure 3 As shown, the computer device 300 includes a processor 310, a memory 320 and a communication interface 330. The processor 310, the memory 320 and the communication interface 330 are connected to each other via a bus 350. Figure 2 The related functions implemented by the receiving unit 201 shown can be implemented through the communication interface 330. Figure 2 The related functions implemented by the acquisition unit 202 and the generation unit 203 shown can be implemented by the processor 310.

[0130] The memory 320 stores a computer program 340, which is configured to be executed by the processor 310. The computer program 340 includes instructions for executing the following steps:

[0131] receiving a request for generating meeting minutes sent by a user, wherein the request for generating meeting minutes includes a meeting data set of a product briefing;

[0132] Acquire, based on the conference data set, a target text corresponding to the product promoted in the product briefing and text features of the target text, as well as a target image and image features of the target image;

[0133] Obtaining the minutes page style corresponding to the text feature and the image feature;

[0134] The minutes of the product briefing meeting are generated based on the minutes page style, the target text and the target image.

[0135] In a possible example, the conference data set includes information data of the promotional product. In acquiring the target text and text features of the target text corresponding to the promotional product in the product briefing based on the conference data set, as well as the target image and image features of the target image, the computer program 340 specifically includes instructions for executing the following steps:

[0136] Classifying the data to obtain a data text set and a data image set;

[0137] Selecting a target text corresponding to the promotional product from the data text set;

[0138] Acquire text features of the target text;

[0139] Acquire a target image corresponding to the promotional product from the data image set;

[0140] The image features of the target image are obtained.

[0141] In a possible example, in the aspect of selecting the target text corresponding to the promotional product from the data text set, the computer program 340 specifically includes instructions for executing the following steps:

[0142] Performing named entity recognition on the document text in the document text set to obtain a text vector of the document text;

[0143] Determining a first correlation value between the material text and the promotional product;

[0144] Based on the first association value, selecting a target data text from the data text;

[0145] The text vector corresponding to the target document text is used as the target text corresponding to the promotional product.

[0146] In a possible example, in the aspect of acquiring the target image corresponding to the promotional product from the data image set, the computer program 340 specifically includes instructions for executing the following steps:

[0147] Acquire a second association value between the document image in the document image set and the target text;

[0148] Obtaining the importance value of the data image;

[0149] calculating an evaluation value of the material image based on the second relevance value and the importance value;

[0150] Based on the evaluation value of the data image, a target image corresponding to the promotional product is selected from the data image.

[0151] In a possible example, the conference data set also includes the audio of the speaker's explanation of the promotional product. Before acquiring the text features of the target text, the computer program 340 also includes instructions for performing the following steps:

[0152] Performing feature recognition on the audio of the explanation by the explainer to obtain the voice features of the explainer;

[0153] Performing semantic recognition on the audio of the explanation by the explainer to obtain the speech text and language features of the explainer;

[0154] Selecting a target text corresponding to the promotional product from the voice text of the narrator;

[0155] In terms of obtaining the text features of the target text, the computer program 340 specifically includes instructions for executing the following steps:

[0156] The voice features and language features of the narrator are used as text features of the target text.

[0157] In a possible example, the conference data set further includes interactive data of the promotional product. Before acquiring the text features of the target text, the computer program 340 further includes instructions for executing the following steps:

[0158] Classifying the interactive data to obtain a first interactive text and an interactive image;

[0159] Identifying a second interactive text corresponding to the interactive image;

[0160] Based on the first interactive text and the second interactive text, a target text corresponding to the promotional product is obtained.

[0161] In a possible example, in the aspect of acquiring the target text corresponding to the promotional product based on the first interactive text and the second interactive text, the computer program 340 specifically includes instructions for executing the following steps:

[0162] Classifying the first interactive text and the second interactive text to obtain sub-text sets corresponding to at least two types of interactive topics;

[0163] Analyze the subtext set corresponding to the interactive topic to obtain the support rate and discussion volume of the interactive topic, as well as the target interactive text;

[0164] Selecting a target interactive topic from the interactive topics based on the support rate and discussion volume of the interactive topics;

[0165] The target interactive text of the target interactive theme is used as the target text corresponding to the promotional product.

[0166] An embodiment of the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, which enables a computer to execute part or all of the steps of any method recorded in the method embodiment, and the computer includes an electronic device or a server.

[0167] The present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute to implement some or all steps of any method described in the method embodiment. The computer program product can be a software installation package, and the computer includes an electronic device or a server.

[0168] In the above embodiment, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of blockchain nodes, etc. For example, the blockchain may store meeting data sets and meeting minutes of a product briefing, page generation models, etc., which are not limited here.

[0169] The blockchain referred to in the embodiments of the present application is a new application model of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods, each of which contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0170] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modes involved are not necessarily required by the present application.

[0171] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0172] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as at least one unit or component 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, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0173] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on at least one network unit. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0174] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software program mode.

[0175] If the integrated unit is implemented in the form of a software program model and sold or used as an independent product, it can be stored in a computer-readable memory. According to this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk or optical disk, etc., and other media that can store program codes.

[0176] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for generating meeting minutes based on natural language processing, characterized in that: include: receiving a request for generating a meeting minutes sent by a user, wherein the request for generating the meeting minutes includes a meeting data set of a product briefing, and the meeting data set includes information data for promoting the product; Based on the conference data set, the target text corresponding to the promotional product in the product briefing and the text features of the target text, as well as the target image and the image features of the target image are obtained, including: classifying the data data to obtain a data text set and a data image set; selecting the target text corresponding to the promotional product from the data text set; obtaining the text features of the target text, the text features of the target text including the attribute information of the promotional product, the attribute information of the promotional product including the investment amount, investment method, payment period, insurance platform, and participating object; obtaining the target image corresponding to the promotional product from the data image set; obtaining the image features of the target image, the image features of the target image including the product image of the promotional product; Obtaining the minutes page style corresponding to the text feature and the image feature; Generate the minutes of the product briefing based on the minutes page style, the target text and the target image; The step of selecting the target text corresponding to the promotional product from the data text set includes: Performing named entity recognition on the document texts in the document text set to obtain text vectors of the document texts; determining a first association value between the document texts and the promotional product; selecting a target document text from the document texts based on the first association value; and using the text vector corresponding to the target document text as the target text corresponding to the promotional product; The step of acquiring the target image corresponding to the promotional product from the data image set includes: Obtain a second association value between a data image in the data image set and the target text; obtain an importance value of the data image; calculate an evaluation value of the data image based on the second association value and the importance value; and select a target image corresponding to the promotional product from the data images based on the evaluation value of the data image.

2. The method according to claim 1, characterized in that The conference data set also includes the audio of the speaker's explanation of the promotional product. Before obtaining the text features of the target text, the method also includes: Performing feature recognition on the audio of the explanation by the explainer to obtain the voice features of the explainer; Performing semantic recognition on the audio of the explanation by the explainer to obtain the speech text and language features of the explainer; Selecting a target text corresponding to the promotional product from the voice text of the narrator; The acquiring of the text features of the target text comprises: The voice features and language features of the narrator are used as text features of the target text.

3. The method according to claim 1, characterized in that The conference data set also includes interactive data of the promotional product. Before acquiring the text features of the target text, the method further includes: Classifying the interactive data to obtain a first interactive text and an interactive image; Identifying a second interactive text corresponding to the interactive image; Based on the first interactive text and the second interactive text, a target text corresponding to the promotional product is obtained.

4. The method according to claim 3, characterized in that The acquiring, based on the first interactive text and the second interactive text, a target text corresponding to the promotional product includes: Classifying the first interactive text and the second interactive text to obtain sub-text sets corresponding to at least two types of interactive topics; Analyze the subtext set corresponding to the interactive topic to obtain the support rate and discussion volume of the interactive topic, as well as the target interactive text; Selecting a target interactive topic from the interactive topics based on the support rate and discussion volume of the interactive topics; The target interactive text of the target interactive theme is used as the target text corresponding to the promotional product.

5. A device for generating conference minutes based on natural language processing, the device being used to execute the method according to any one of claims 1 to 4, characterized in that: include: A receiving unit, configured to receive a request for generating a meeting minutes sent by a user, wherein the request for generating the meeting minutes includes a meeting data set for promoting a product; An acquisition unit, configured to acquire, based on the conference data set, a target text corresponding to the product promoted in the product briefing and text features of the target text, as well as a target image and image features of the target image; and acquire a minutes page style corresponding to the text features and the image features; A generating unit is used to generate the minutes of the product briefing based on the minutes page style, the target text and the target image.

6. A computer device, characterized in that: The method comprises a processor, a memory and a communication interface, wherein the memory stores a computer program, the computer program is configured to be executed by the processor, and the computer program comprises instructions for executing the steps in any one of the methods of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which enables a computer to execute to implement the method according to any one of claims 1 to 4.

Images