CN113642294B - Method, apparatus, device, medium and product for generating meeting summary - Google Patents
Method, apparatus, device, medium and product for generating meeting summary Download PDFInfo
- Publication number
- CN113642294B CN113642294B CN202110930151.7A CN202110930151A CN113642294B CN 113642294 B CN113642294 B CN 113642294B CN 202110930151 A CN202110930151 A CN 202110930151A CN 113642294 B CN113642294 B CN 113642294B
- Authority
- CN
- China
- Prior art keywords
- information
- sample
- conference
- training
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000001914 filtration Methods 0.000 claims abstract description 38
- 238000013145 classification model Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims description 102
- 230000006870 function Effects 0.000 claims description 21
- 238000002372 labelling Methods 0.000 claims description 16
- 230000009471 action Effects 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 10
- 238000012952 Resampling Methods 0.000 claims description 9
- 230000004044 response Effects 0.000 claims description 7
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 12
- 238000004891 communication Methods 0.000 description 10
- 238000004590 computer program Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 239000002244 precipitate Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure provides a method, a device, equipment, a medium and a product for generating a meeting summary, and relates to the technical field of computers, in particular to the technical field of artificial intelligence. The specific implementation scheme is as follows: acquiring initial conference information; filtering the initial conference information to obtain target conference information; determining classification information of the target conference information based on the target conference information and a preset information classification model; the classification information at least comprises conference essential information and conference non-essential information; and generating a meeting summary based on the meeting point information and the meeting non-point information. The implementation mode can improve the generation efficiency of the meeting summary.
Description
Technical Field
The disclosure relates to the field of computer technology, and in particular to the field of artificial intelligence technology.
Background
Currently, in an office scenario, it is often necessary to organize meetings and generate corresponding meeting campaigns for the meeting content. The efficient meeting summary can trace back and precipitate the meeting, so that the working efficiency can be improved in an auxiliary mode.
The current generation mode of the meeting summary is generally manual summary of meeting contents, and the meeting summary is obtained. However, this way of manually summarizing meeting content to get a meeting summary has the problem that the meeting summary generation is inefficient.
Disclosure of Invention
The present disclosure provides a method, apparatus, device, medium, and article for generating a meeting summary.
According to an aspect of the present disclosure, there is provided a method for generating a meeting summary, comprising: acquiring initial conference information; filtering the initial conference information to obtain target conference information; determining classification information of the target conference information based on the target conference information and a preset information classification model; the classification information at least comprises conference essential information and conference non-essential information; and generating a meeting summary based on the meeting point information and the meeting non-point information.
According to another aspect of the present disclosure, there is provided an apparatus for generating a meeting summary, comprising: an information acquisition unit configured to acquire initial conference information; the information filtering unit is configured to filter the initial conference information to obtain target conference information; an information classification unit configured to determine classification information of the target conference information based on the target conference information and a preset information classification model; the classification information at least comprises conference essential information and conference non-essential information; and a meeting summary generation unit configured to generate a meeting summary based on the meeting point information and the meeting non-point information.
According to another aspect of the present disclosure, there is provided an electronic device including: one or more processors; a memory for storing one or more programs; the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of generating a meeting summary as any of the above.
According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a method as any one of the above for generating a meeting summary.
According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements a method as in any of the above for generating a meeting summary.
According to the technology disclosed by the invention, a method for generating the meeting summary is provided, and the meeting summary generation efficiency can be improved.
It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.
Drawings
The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:
FIG. 1 is an exemplary system architecture diagram in which an embodiment of the present disclosure may be applied;
FIG. 2 is a flow chart of one embodiment of a method for generating a meeting summary according to the present disclosure;
FIG. 3 is a schematic illustration of one application scenario of a method for generating a meeting summary according to the present disclosure;
FIG. 4 is a flow chart of another embodiment of a method for generating a meeting summary according to the present disclosure;
FIG. 5 is a schematic structural diagram of one embodiment of an apparatus for generating a meeting summary according to the present disclosure;
fig. 6 is a block diagram of an electronic device for implementing a method for generating a meeting summary in accordance with an embodiment of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
It should be noted that, without conflict, the embodiments of the present disclosure and features of the embodiments may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
As shown in fig. 1, a system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.
The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. The terminal devices 101, 102, 103 may be mobile phones, computers, tablet and other electronic devices, the terminal devices 101, 102, 103 may be installed with software for conducting online conferences, users may conduct multi-person conferences through the software, and the software may have an intelligent conference summary generation function. When the user selects to start the intelligent conference summary generation function, the terminal devices 101, 102, 103 can convert the voices of the users participating in the conference into initial conference information in text form, and send the initial conference information to the server 105 through the network 104, so that the server 105 returns to the corresponding conference summary.
The terminal devices 101, 102, 103 may be hardware or software. When the terminal devices 101, 102, 103 are hardware, they may be various electronic devices including, but not limited to, a vehicle-mounted computer, a vehicle-mounted tablet, a vehicle control device, and the like. When the terminal devices 101, 102, 103 are software, they can be installed in the above-listed electronic devices. Which may be implemented as multiple software or software modules (e.g., to provide distributed services), or as a single software or software module. The present invention is not particularly limited herein.
The server 105 may be a server providing various services, for example, the server 105 may acquire initial conference information transmitted by the terminal devices 101, 102, 103 through the network 104, filter the initial conference information to obtain target conference information, determine conference point information and conference non-point information based on the target conference information and the information classification model, generate a conference summary based on the conference point information and the conference non-point information, and return the conference summary to the terminal devices 101, 102, 103.
The server 105 may be hardware or software. When the server 105 is hardware, it may be implemented as a distributed server cluster formed by a plurality of servers, or as a single server. When server 105 is software, it may be implemented as a plurality of software or software modules (e.g., to provide distributed services), or as a single software or software module. The present invention is not particularly limited herein.
It should be noted that, the method for generating the meeting summary provided in the embodiments of the present disclosure may be performed by the terminal devices 101, 102, 103, or may be performed by the server 105, and the apparatus for generating the meeting summary may be provided in the terminal devices 101, 102, 103, or may be provided in the server 105.
It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
With continued reference to fig. 2, a flow 200 of one embodiment of a method for generating a meeting summary according to the present disclosure is shown. The method for generating the meeting summary of the embodiment comprises the following steps:
step 201, initial meeting information is obtained.
In the present embodiment, an executing body (such as the terminal devices 101, 102, 103 or the server 105 in fig. 1) can acquire conference information generated in a conference. The conference information can be various forms of information such as text, audio and video for describing conference contents. The conference may be an online conference based on online conference software, or may be an offline conference, which is not limited in this embodiment. Preferably, the executive body converts speaking voice of each participant in the conference process into corresponding text information based on a voice-to-text technology to obtain initial conference information. The initial conference information at this time is information for describing conference contents in text form. Or, the executing body can directly take the audio or video form in the conference process as the initial conference information, and then determine the corresponding classification information for the conference information to generate the conference summary.
Step 202, filtering the initial conference information to obtain target conference information.
In this embodiment, the initial conference information collected by the executing body is generated based on the original conference content in the conference process, which includes the discussion situation of each participant, and in the conference discussion process of each participant, a more spoken language is often used. Thus, the executing body may filter the initial conference information after acquiring the initial conference information, filtering out these spoken languages. These spoken languages may include, but are not limited to, the word of a natural language, the word of a repeated pronunciation, the word of a misspeaking, etc., which is not limited in this embodiment. The executive may use existing speech recognition techniques to process the words, such as filtering the words from the initial meeting information, to obtain target meeting information that complies with written terms.
In some optional implementations of this embodiment, filtering the initial meeting information to obtain the target meeting information may include: removing the spoken language from the initial conference information to obtain initial conference information after removal processing; and carrying out semantic analysis on the initial conference information after the removal processing, replacing the words which do not accord with the written expression, and adjusting the word sequence which does not accord with the written expression to obtain the target conference information. With this alternative implementation, the executing body may convert the spoken initial meeting information into the written meeting information by means of word filtering, word replacement, word order adjustment, etc. By classifying the information of the target meeting information which accords with the written expression, the meeting summary is generated, and the accuracy of the generated meeting summary can be further improved.
Step 203, determining classification information of the target conference information based on the target conference information and a preset information classification model; the classification information includes at least conference gist information and conference non-gist information.
In this embodiment, the preset information classification model is used to classify the target meeting information, so as to obtain classification information of the target meeting information, that is, obtain meeting essential information and meeting non-essential information in the target meeting information. The conference essential information refers to key information in a conference, and the conference non-essential information refers to common information in the conference. Meeting point information may include, but is not limited to, meeting to-do, meeting conclusion, etc., where meeting to-do refers to a to-do in a meeting that is decided to be completed later, and meeting conclusion is a conclusive description of the meeting. The conference non-essential information may be all other information except the conference essential information in the target conference information, or may be information obtained by further screening all other information, which is not limited in this embodiment.
Further, after the execution subject obtains the target conference information, the execution subject may input the target conference information into a preset information classification model, so that the preset information classification model outputs a category corresponding to each sentence in the target conference information. The executive main body can collect sentences belonging to the conference gist categories to obtain conference gist information, and collect sentences belonging to the conference non-gist to obtain conference non-gist information.
Step 204, generating a meeting summary based on the meeting point information and the meeting non-point information.
In this embodiment, after determining that the conference gist information and the conference non-gist information are obtained, the executing body may directly output the conference gist information as a conference summary, or may jointly output the conference gist information and the conference non-gist information as a conference summary.
In some optional implementations of the present embodiment, generating the meeting summary based on the meeting point information and the meeting non-point information may include: acquiring a target conference summary template, wherein the target conference summary template comprises a conference main point area and a conference non-main point area; and filling the conference gist information into conference gist areas in the conference summary template, and filling the conference non-gist information into conference non-gist areas in the conference summary template to obtain conference summary. By adopting the mode, the personalized conference summary template can be adopted, and conference essential information and conference non-essential information are filled into corresponding positions in the template, so that the display diversity of conference summary is improved.
With continued reference to fig. 3, a schematic diagram of one application scenario of a method for generating a meeting summary according to the present disclosure is shown. In the application scenario of fig. 3, each participant may perform an online conference, and in the application software for performing an online conference, a conference video screen 301 may be displayed, and the participant may control and display information such as a shared desktop or a face screen in the conference video screen 301. And during the conference discussion of the participants, the executing body may display the text contents of the discussion of the respective participants A, B, C in the conference information display area 302 in real time. The executing body may acquire the voice information of the participants, convert the voice information into text information, and output the text information in real time in the conference information display area 302. Optionally, the executing body may also filter the text information, so as to update the filtered text information synchronously in the output content of the conference information display area 302. The specific filtering method herein may refer to a method of filtering the initial conference information to obtain the target conference information, which is not described herein. And then, the execution main body can determine the text information obtained by filtering as input data of a preset information classification model to obtain classification information of the text information output by the preset information classification model, namely, the class corresponding to each text sentence in the text information, wherein the class comprises a conference conclusion, a conference to-be-handled and a conference non-essential point. The executing body may aggregate sentences of each category in the text information based on the classification information, and generate meeting summary 303. The meeting summary 303 includes a text statement set corresponding to the meeting conclusion, a text statement set corresponding to the meeting to-be-handled, and a text statement set corresponding to the meeting non-essential point.
According to the method for generating the conference summary, provided by the embodiment of the disclosure, the target conference information is obtained by filtering the initial conference information, the classification information of the target conference information is determined based on the target conference information and the preset information classification model, and the conference summary is generated based on the conference main point information and the conference non-main point information in the classification information, so that the automatic generation of the conference summary can be realized, and the generation efficiency of generating the conference summary is improved.
With continued reference to fig. 4, a flow 400 of another embodiment of a method for generating a meeting summary according to the present disclosure is shown. As shown in fig. 4, the method for generating a meeting summary of the present embodiment may include the steps of:
step 401, pre-training the pre-training model based on a preset pre-training task to obtain a target pre-training model after training.
In this embodiment, the executing body may preset a pre-training task that needs to train the pre-training model. The pre-training task is used for training the identification of key information in the specified information. The key information may include, but is not limited to, a key subject, a key object, a key action, etc. in the specified information, which is not limited in this embodiment. For example, the pre-training task may be a preset sentence set, each sentence in the sentence set includes a subject, an action, and an object, each sentence in the pre-training task is used as input data of a pre-training model, the training pre-training model outputs the subject, the action, and the object corresponding to the sentence, and at this time, the trained target pre-training model can accurately identify the subject, the action, and the object corresponding to each sentence.
In some optional implementations of the present embodiment, pre-training the pre-training model based on a preset pre-training task, obtaining a trained target pre-training model may include: determining task input data and task annotation data corresponding to a pre-training task; inputting the task input data into a pre-training model so that the pre-training model outputs a key information identification result corresponding to the task input data; based on the key information identification result and the task annotation data, model parameters of the pre-training model are continuously adjusted, and the target pre-training model after training is obtained.
In this implementation manner, the pre-training task may correspond to a plurality of preset sentences, and the execution subject may determine the plurality of preset sentences as task input data corresponding to the pre-training task. And each preset sentence can be marked with a subject, an action and an object of the sentence in advance. That is, the pre-labeled subjects, actions, and objects are task labeling data corresponding to the pre-training task. Then, the execution subject may input the task input data into the pre-training model, so that the pre-training model identifies key information in the task input data, that is, identifies the subject, the action, and the object corresponding to each sentence. And then, the execution main body can acquire a key information identification result which is output by the pre-training model and corresponds to the task input data, substitutes the key information identification result and the task annotation data into the loss function, and continuously adjusts model parameters of the pre-training model based on the difference between the key information identification result and the task annotation data so as to enable the key information identification result to continuously approximate to the task annotation data, so that the loss function is converged, and a trained target pre-training model is obtained.
In step 402, sample information is obtained.
In this embodiment, the sample information includes conference information in the history conference record, for example, the conference information may be a conference record in text form obtained by converting the voices of the participants in the history conference.
Step 403, determining sample non-essential information and sample essential information in the sample information based on preset information conditions.
In this embodiment, the preset information condition is used to perform rough screening on the sample information, perform preliminary screening on each category in the sample information, and preliminarily determine the information that is not essential to the sample in the sample information. Specifically, the preset information conditions may include, but are not limited to, the occurrence of a specified keyword, the length being less than a specified length threshold, the length being greater than a specified threshold, and the like, which is not limited in this embodiment. The specified keywords here are keywords in which preset non-essential information often appears. And since the shorter length is not normally a sample point, information of which the length is smaller than a specified length threshold is determined as sample non-point information. And since a longer length is also generally unsuitable as a sample point, information having a length greater than a specified length threshold is determined as sample non-point information. Further optionally, after the rough screening of the sample information based on the preset information condition, the execution subject may mark the rough screened sample non-essential information, output the rough screened sample essential information to the user, receive a further label of the rough screened sample essential information from the user, determine the information marked as the sample essential information by the user as final sample essential information, and determine the information marked as the sample non-essential information by the user as final sample non-essential information.
At step 404, responsive to determining that the difference in the amount of information between the sample non-point information and the sample point information is greater than a preset threshold, the sample information is resampled.
In this embodiment, since the number of pieces of information other than the gist in the conference information is generally larger than the number of pieces of information of the gist in practice, there is a problem of sample imbalance. After obtaining the sample non-point information and the sample point information, the execution subject may further determine an information amount difference between the sample non-point information and the sample point information, where the amount difference may be an amount difference of text sentences. Under the condition that the information quantity difference between the sample non-essential information and the sample essential information is larger than a preset threshold value, resampling is needed to be carried out on the sample information so as to realize sample equalization. The resampling process may be performed by various resampling techniques in the prior art, which is not limited in this embodiment. Specifically, the executing body may further determine an information amount difference between the sample non-essential information and the sample essential information after resampling, and if the information amount difference is greater than a preset threshold, the resampling process may be continued until the information amount difference is less than or equal to the preset threshold. Alternatively, the executing body may also calculate a ratio between the sample non-essential information and the sample essential information, preset a threshold corresponding to the ratio, and resample the sample information if the ratio is greater than the threshold corresponding to the ratio. The resampling process may include, but is not limited to, a process of reducing the number of non-essential information of the sample, a process of increasing the number of essential information of the sample, and the like, which is not limited in this embodiment.
In step 405, sample labeling data is determined based on the sample non-point information and the sample point information.
In this embodiment, the sample labeling data may be labeling data corresponding to each piece of information in the sample information, where the labeling data is a gist or a non-gist. The gist and non-gist herein may be further subdivided into finer classifications, such as gist including to-do and conclusion, which is not limited by the present embodiment. The execution subject can add corresponding category labels based on the sample non-essential information and the sample essential information to generate sample labeling data.
Step 406, training the target pre-training model based on the sample information and the sample labeling data, so that the target pre-training model outputs a sample classification corresponding to the sample information, and adjusting model parameters of the target pre-training model based on the sample classification, the sample labeling data and a preset loss function until the preset loss function converges, thereby obtaining a preset information classification model.
In this embodiment, the sample classification includes a sample point and a sample non-point; and the weight corresponding to the sample gist in the loss function is larger than the weight corresponding to the sample non-gist. The execution subject may input the sample information into the target pre-training model, to obtain information belonging to the sample gist in the sample information output by the target pre-training model and information belonging to the non-sample gist in the sample information, that is, sample classification corresponding to the sample information. And then, substituting the sample classification and sample labeling data corresponding to the sample information into a preset loss function by the execution subject, and continuously adjusting the model parameters of the target pre-training model until the preset loss function converges to obtain the preset information classification model.
In step 407, initial meeting information is obtained in response to determining that the information category of the initial meeting information is not the target meeting category.
In this embodiment, the target meeting class is a meeting that is not suitable for classification of points or non-points, for example, the target meeting class may be interviews. The executing body can determine the information category of the initial conference information based on the information analysis technology, and acquire the initial conference information when the information category is not the target conference category.
It should be noted that, in the model training stage, the history meeting record corresponding to the obtained sample information may also be a history meeting record that is not the target meeting category.
And step 408, filtering the language words in the initial conference information to obtain filtered initial conference information.
In this embodiment, the executing body may filter the word of the language in the initial conference information after acquiring the initial conference information. Optionally, the executing body may preset a to-be-filtered word list, and by matching the initial conference information with the to-be-filtered word list, the word which is matched with the to-be-filtered word list in the initial conference information may be filtered, so as to obtain the filtered initial conference information. The settings of the vocabulary of the speech and gas to be filtered may be determined based on both manual settings and historical meeting record mining. When the historical conference record is mined to determine the word to be filtered, the occurrence frequency of the word serving as an independent sentence in the historical conference record can be counted, the occurrence frequency of the word serving as a non-independent sentence in the historical conference record is counted, and then the word with high occurrence frequency is further mined.
Step 409, filtering the repeated words in the filtered initial conference information to obtain the target conference information.
In this embodiment, the executing body may determine the repeated words in the filtered initial conference information based on the existing text recognition technology, remove the repeated words, and implement filtering of the repeated words to obtain the target conference information.
In some optional implementations of this embodiment, filtering the repeated words in the filtered initial meeting information may include: determining meaningful repeated words in the filtered initial meeting information; and filtering the repeated words except the meaningful repeated words in the filtered initial conference information to obtain target conference information.
In this implementation manner, the executing body may preset a repeated word whitelist, where the repeated word whitelist is a set of meaningful repeated words, and the meaningful repeated words may be words such as "see", "consider", and the like. When the repeated words are filtered, the words matched with the repeated words in the repeated word white list in the filtered initial conference information can be determined, namely, the meaningful repeated words are determined. And then, the execution subject can remove the words of the non-meaningful repeated words without removing the meaningful repeated words, so as to obtain the target meeting information. The setting of the repeated word white list can be based on corpus mining of written text records, and the frequently-occurring meaningful repeated words meeting the specification are determined and added to the repeated word white list.
Step 410, determining classification information of the target conference information based on the target conference information and a preset information classification model; the classification information includes at least conference gist information and conference non-gist information.
In this embodiment, the information classification model obtained through the training in the model training step can output corresponding classification information based on the input target conference information. The classification information is a classification result of each information in the target conference information, namely, conference key point information belonging to the key point in the target conference information and conference non-key point information belonging to the non-key point in the target conference information are output.
Step 411, generating a meeting summary based on the meeting point information and the meeting non-point information.
In this embodiment, the detailed description of step 411 is referred to the detailed description of step 204, and will not be repeated here.
According to the method for generating the conference summary, which is provided by the embodiment of the disclosure, the initial conference information can be subjected to the word filtering and the repeated word filtering, so that more accurate target conference information can be obtained. And in the filtering of the repeated words, the removal of the meaningful repeated words is considered, so that the filtering accuracy of the repeated words is improved. In addition, in the training process of the information classification model, training of the model can be assisted based on a preset pre-training task, and the recognition capability of the model on key information is improved. And when the sample information category is determined, rough screening can be performed based on preset information conditions, sample classification can be obtained based on rough screening and manual labeling, and the accuracy of sample classification can be improved. In addition, aiming at the problem of sample imbalance, the samples can be resampled, and sample balancing is realized. And for meeting categories unsuitable for determining the key points and the non-key points, the method can directly filter the meeting categories without classifying the key points and the non-key points, and further improves the classification reliability.
With further reference to fig. 5, as an implementation of the method shown in the foregoing figures, the present disclosure provides an embodiment of an apparatus for generating a meeting summary, where the embodiment of the apparatus corresponds to the embodiment of the method shown in fig. 2, and the apparatus may be specifically applied to an electronic device such as a terminal device, a server, or the like.
As shown in fig. 5, the apparatus 500 for generating a meeting summary of the present embodiment includes: an information acquisition unit 501, an information filtering unit 502, an information classification unit 503, and a meeting summary generation unit 504.
The information acquisition unit 501 is configured to acquire initial conference information.
The information filtering unit 502 is configured to filter the initial conference information to obtain target conference information.
An information classification unit 503 configured to determine classification information of the target conference information based on the target conference information and a preset information classification model; the classification information includes at least conference gist information and conference non-gist information.
The meeting summary generating unit 504 is configured to generate a meeting summary based on the meeting point information and the meeting non-point information.
In some optional implementations of the present embodiment, the information filtering unit 502 is further configured to: filtering the language words in the initial conference information to obtain filtered initial conference information; and filtering repeated words in the filtered initial conference information to obtain target conference information.
In some optional implementations of this embodiment, the apparatus further includes: the model training unit is configured to pretrain the pretraining model based on a preset pretraining task to obtain a target pretraining model after training; training the target pre-training model based on the sample information and the sample labeling data to enable the target pre-training model to output sample classification corresponding to the sample information, and adjusting model parameters of the target pre-training model based on the sample classification, the sample labeling data and a preset loss function until the preset loss function converges to obtain a preset information classification model.
In some optional implementations of this embodiment, the apparatus further includes: a sample acquisition unit configured to acquire sample information; determining sample non-essential information and sample essential information in the sample information based on preset information conditions; sample annotation data is determined based on the sample non-point information and the sample point information.
In some optional implementations of the present embodiment, the sample acquisition unit is further configured to: and resampling the sample information in response to determining that the information amount difference between the sample non-point information and the sample point information is greater than a preset threshold.
In some optional implementations of the present embodiment, the sample classification includes a sample point and a sample non-point; and the weight corresponding to the sample gist in the loss function is larger than the weight corresponding to the sample non-gist.
In some optional implementations of the present embodiment, the model training unit is further configured to: determining task input data and task annotation data corresponding to a pre-training task; inputting the task input data into a pre-training model so that the pre-training model outputs a key information identification result corresponding to the task input data; based on the key information identification result and the task annotation data, model parameters of the pre-training model are continuously adjusted, and the target pre-training model after training is obtained.
In some optional implementations of the present embodiment, the information acquisition unit 501 is further configured to: in response to determining that the information category of the initial meeting information is not the target meeting category, initial meeting information is obtained.
It should be appreciated that the elements 501 to 504 recited in the means 500 for generating a meeting summary correspond to the respective steps in the method described with reference to fig. 2. Thus, the operations and features described above with respect to the method of in-vehicle calling are equally applicable to the apparatus 500 and the units contained therein, and are not described in detail herein.
According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.
Fig. 6 illustrates a schematic block diagram of an example electronic device 600 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.
As shown in fig. 6, the apparatus 600 includes a computing unit 601 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 602 or a computer program loaded from a storage unit 608 into a Random Access Memory (RAM) 603. In the RAM603, various programs and data required for the operation of the device 600 may also be stored. The computing unit 601, ROM 602, and RAM603 are connected to each other by a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
Various components in the device 600 are connected to the I/O interface 605, including: an input unit 606 such as a keyboard, mouse, etc.; an output unit 607 such as various types of displays, speakers, and the like; a storage unit 608, such as a magnetic disk, optical disk, or the like; and a communication unit 609 such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.
The computing unit 601 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 601 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as methods for generating meeting summaries. For example, in some embodiments, the method for generating a meeting summary may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and/or installed onto the device 600 via the ROM 602 and/or the communication unit 609. When the computer program is loaded into RAM 603 and executed by computing unit 601, one or more steps of the method for generating a meeting summary described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the method for generating meeting summary in any other suitable way (e.g., by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server incorporating a blockchain.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.
The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.
Claims (16)
1. A method for generating a meeting summary, comprising:
acquiring initial conference information;
filtering the initial conference information to obtain target conference information;
determining classification information of the target conference information based on the target conference information and a preset information classification model; the classification information at least comprises conference essential information and conference non-essential information;
generating a meeting summary based on the meeting point information and the meeting non-point information;
The preset information classification model is obtained through training based on the following steps:
pre-training the pre-training model based on a preset pre-training task to obtain a target pre-training model with completed training, wherein the pre-training task is used for training the identification of key information, and the key information comprises a key subject, a key object and a key action;
obtaining sample information, performing rough screening based on preset information conditions, determining sample non-essential information and sample essential information in the sample information, further marking the sample essential information which is subjected to rough screening in a user marking mode, and determining final sample non-essential information and sample essential information as sample marking data, wherein the preset information conditions comprise: the information of which the length is smaller than a first specified length threshold value or the length is larger than a second specified length threshold value is non-essential information;
training the target pre-training model based on sample information and sample labeling data to enable the target pre-training model to output sample classification corresponding to the sample information, and adjusting model parameters of the target pre-training model based on the sample classification, the sample labeling data and a preset loss function until the preset loss function converges to obtain the preset information classification model.
2. The method of claim 1, wherein the filtering the initial meeting information to obtain target meeting information comprises:
filtering the language words in the initial conference information to obtain filtered initial conference information;
and filtering the repeated words in the filtered initial conference information to obtain the target conference information.
3. The method of claim 2, wherein the filtering the repeated words in the filtered initial meeting information to obtain the target meeting information includes:
determining meaningful repeated words in the filtered initial meeting information;
and filtering the repeated words except the meaningful repeated words in the filtered initial conference information to obtain the target conference information.
4. The method of claim 1, wherein the method further comprises:
and resampling the sample information in response to determining that the information amount difference between the sample non-essential information and the sample essential information is greater than a preset threshold.
5. The method of claim 1, wherein the sample classification comprises a sample point and a sample non-point; and the weight corresponding to the sample point in the loss function is larger than the weight corresponding to the sample non-point.
6. The method of claim 1, wherein the pre-training model based on the preset pre-training task to obtain a trained target pre-training model, comprises:
determining task input data and task annotation data corresponding to the pre-training task;
inputting the task input data into the pre-training model so that the pre-training model outputs a key information identification result corresponding to the task input data;
and continuously adjusting model parameters of the pre-training model based on the key information identification result and the task annotation data to obtain the target pre-training model after training is completed.
7. The method of claim 1, wherein the acquiring initial meeting information comprises:
and acquiring the initial conference information in response to determining that the information category of the initial conference information is not the target conference category.
8. An apparatus for generating a meeting summary, comprising:
an information acquisition unit configured to acquire initial conference information;
the information filtering unit is configured to filter the initial conference information to obtain target conference information;
an information classification unit configured to determine classification information of the target conference information based on the target conference information and a preset information classification model; the classification information at least comprises conference essential information and conference non-essential information;
A meeting summary generation unit configured to generate a meeting summary based on the meeting point information and the meeting non-point information;
the preset information classification model is obtained through training based on the following steps:
pre-training the pre-training model based on a preset pre-training task to obtain a target pre-training model with completed training, wherein the pre-training task is used for training the identification of key information, and the key information comprises a key subject, a key object and a key action;
obtaining sample information, performing rough screening based on preset information conditions, determining sample non-essential information and sample essential information in the sample information, further marking the sample essential information which is subjected to rough screening in a user marking mode, and determining final sample non-essential information and sample essential information as sample marking data, wherein the preset information conditions comprise: the information of which the length is smaller than a first specified length threshold value or the length is larger than a second specified length threshold value is non-essential information;
training the target pre-training model based on sample information and sample labeling data to enable the target pre-training model to output sample classification corresponding to the sample information, and adjusting model parameters of the target pre-training model based on the sample classification, the sample labeling data and a preset loss function until the preset loss function converges to obtain the preset information classification model.
9. The apparatus of claim 8, wherein the information filtering unit is further configured to:
filtering the language words in the initial conference information to obtain filtered initial conference information;
and filtering the repeated words in the filtered initial conference information to obtain the target conference information.
10. The apparatus of claim 9, wherein the information filtering unit is further configured to:
determining meaningful repeated words in the filtered initial meeting information;
and filtering the repeated words except the meaningful repeated words in the filtered initial conference information to obtain the target conference information.
11. The apparatus of claim 8, wherein training the pre-set information classification model further comprises:
and resampling the sample information in response to determining that the information amount difference between the sample non-essential information and the sample essential information is greater than a preset threshold.
12. The apparatus of claim 8, wherein the sample classification comprises a sample point and a sample non-point; and the weight corresponding to the sample point in the loss function is larger than the weight corresponding to the sample non-point.
13. The apparatus of claim 8, wherein the pre-training model based on the preset pre-training task to obtain a trained target pre-training model, comprises:
determining task input data and task annotation data corresponding to the pre-training task;
inputting the task input data into the pre-training model so that the pre-training model outputs a key information identification result corresponding to the task input data;
and continuously adjusting model parameters of the pre-training model based on the key information identification result and the task annotation data to obtain the target pre-training model after training is completed.
14. The apparatus of claim 8, wherein the information acquisition unit is further configured to:
and acquiring the initial conference information in response to determining that the information category of the initial conference information is not the target conference category.
15. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.
16. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110930151.7A CN113642294B (en) | 2021-08-13 | 2021-08-13 | Method, apparatus, device, medium and product for generating meeting summary |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110930151.7A CN113642294B (en) | 2021-08-13 | 2021-08-13 | Method, apparatus, device, medium and product for generating meeting summary |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113642294A CN113642294A (en) | 2021-11-12 |
| CN113642294B true CN113642294B (en) | 2023-08-04 |
Family
ID=78421711
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110930151.7A Active CN113642294B (en) | 2021-08-13 | 2021-08-13 | Method, apparatus, device, medium and product for generating meeting summary |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113642294B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107679525A (en) * | 2017-11-01 | 2018-02-09 | 腾讯科技（深圳）有限公司 | Image classification method, device and computer-readable recording medium |
| CN109473103A (en) * | 2018-11-16 | 2019-03-15 | 上海玖悦数码科技有限公司 | A kind of meeting summary generation method |
| CN110866110A (en) * | 2019-09-20 | 2020-03-06 | 平安科技（深圳）有限公司 | Conference summary generation method, device, equipment and medium based on artificial intelligence |
| CN112364912A (en) * | 2020-11-09 | 2021-02-12 | 腾讯科技（深圳）有限公司 | Information classification method, device, equipment and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8266534B2 (en) * | 2007-06-28 | 2012-09-11 | Microsoft Corporation | Collaborative generation of meeting minutes and agenda confirmation |
| US9710940B2 (en) * | 2011-11-06 | 2017-07-18 | Sharp Laboratories Of America, Inc. | Methods, systems and apparatus for summarizing a meeting |
-
2021
- 2021-08-13 CN CN202110930151.7A patent/CN113642294B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107679525A (en) * | 2017-11-01 | 2018-02-09 | 腾讯科技（深圳）有限公司 | Image classification method, device and computer-readable recording medium |
| CN109473103A (en) * | 2018-11-16 | 2019-03-15 | 上海玖悦数码科技有限公司 | A kind of meeting summary generation method |
| CN110866110A (en) * | 2019-09-20 | 2020-03-06 | 平安科技（深圳）有限公司 | Conference summary generation method, device, equipment and medium based on artificial intelligence |
| CN112364912A (en) * | 2020-11-09 | 2021-02-12 | 腾讯科技（深圳）有限公司 | Information classification method, device, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 一种文本信息抽取技术的研究;王涛;贾媚;;计算机与网络(第09期);49-51 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113642294A (en) | 2021-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112926306B (en) | Text error correction method, device, equipment and storage medium | |
| CN110019742B (en) | Method and device for processing information | |
| CN114578969A (en) | Method, apparatus, device and medium for human-computer interaction | |
| CN111582360B (en) | Method, apparatus, device and medium for labeling data | |
| CN114841274B (en) | Language model training method and device, electronic equipment and storage medium | |
| CN111522944A (en) | Method, apparatus, device and storage medium for outputting information | |
| CN115099239B (en) | Resource identification method, device, equipment and storage medium | |
| CN114186681A (en) | Method, apparatus and computer program product for generating model clusters | |
| CN113642294B (en) | Method, apparatus, device, medium and product for generating meeting summary | |
| CN116150339A (en) | Dialogue method, dialogue device, dialogue equipment and dialogue storage medium | |
| CN114969195B (en) | Dialogue content mining method and dialogue content evaluation model generation method | |
| CN114490967B (en) | Training method of dialogue model, dialogue method and device of dialogue robot and electronic equipment | |
| CN114141236B (en) | Language model updating method and device, electronic equipment and storage medium | |
| CN116257690A (en) | Resource recommendation method and device, electronic equipment and storage medium | |
| CN116204624A (en) | Response method, response device, electronic equipment and storage medium | |
| CN113763968B (en) | Method, apparatus, device, medium, and product for recognizing speech | |
| CN115374063A (en) | File processing method, device, equipment and storage medium | |
| CN114218431A (en) | Video searching method and device, electronic equipment and storage medium | |
| CN114171063A (en) | Real-time telephone traffic customer emotion analysis assisting method and system | |
| CN113554062A (en) | Training method, device and storage medium of multi-classification model | |
| CN112632241A (en) | Method, device, equipment and computer readable medium for intelligent conversation | |
| CN113806541A (en) | Emotion classification method and emotion classification model training method and device | |
| CN113344405B (en) | Method, device, equipment, medium and product for generating information based on knowledge graph | |
| CN112836529B (en) | Method and device for generating target corpus sample | |
| CN114501112B (en) | Method, apparatus, device, medium, and article for generating video notes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |