CN115174285A - Conference record generation method and device and electronic equipment - Google Patents
Conference record generation method and device and electronic equipment Download PDFInfo
- Publication number
- CN115174285A CN115174285A CN202210882694.0A CN202210882694A CN115174285A CN 115174285 A CN115174285 A CN 115174285A CN 202210882694 A CN202210882694 A CN 202210882694A CN 115174285 A CN115174285 A CN 115174285A
- Authority
- CN
- China
- Prior art keywords
- sentence
- target
- conference
- text
- weight
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 230000015654 memory Effects 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 10
- 230000011218 segmentation Effects 0.000 claims description 7
- 230000001502 supplementing effect Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 abstract description 3
- 238000004364 calculation method Methods 0.000 description 16
- 239000000047 product Substances 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L12/00—Data switching networks
- H04L12/02—Details
- H04L12/16—Arrangements for providing special services to substations
- H04L12/18—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
- H04L12/1813—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast for computer conferences, e.g. chat rooms
- H04L12/1822—Conducting the conference, e.g. admission, detection, selection or grouping of participants, correlating users to one or more conference sessions, prioritising transmission
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L12/00—Data switching networks
- H04L12/02—Details
- H04L12/16—Arrangements for providing special services to substations
- H04L12/18—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
- H04L12/1813—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast for computer conferences, e.g. chat rooms
- H04L12/1831—Tracking arrangements for later retrieval, e.g. recording contents, participants activities or behavior, network status
Abstract
The specification discloses a conference record generation method, a conference record generation device and electronic equipment, wherein the method comprises the following steps: acquiring a target conference text; starting from the last sentence of the target conference text, looking up a first target sentence matched with a preset theme in the direction of the first sentence of the target conference text; starting from the target sentence and facing to the first sentence of the target conference text, the following operations are carried out on each sentence: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; taking the last sentence of the target sentence as the target sentence; and generating a conference record according to the screened sentences and the target sentences. The scheme adopts a backward-forward screening method, preferentially screens statements from a summary link in the later stage of a conference to generate a conference record, and can more efficiently generate the conference record which is closer to a theme and covers more conference key points.
Description
Technical Field
The present application relates to the field of intelligent conference technologies, and in particular, to a method and an apparatus for generating a conference record, and an electronic device.
Background
After each meeting, the meeting record is typically archived. The existing conference record generation method usually performs speech recognition on an audio file of a conference, removes spoken words, completes the supplement of context sentences, calculates the weight of each sentence one by one from a first sentence to a last sentence, and selects out the sentences used as conference records according to the weights.
However, the accuracy of the conference recordings generated by this method is low.
Disclosure of Invention
The embodiment of the application aims to provide a conference record generation method, a conference record generation device and electronic equipment, so as to solve the problem that the existing conference record generation method is low in accuracy.
To solve the above technical problem, a first aspect of the present specification provides a method for generating a conference record, including: acquiring a target conference text; starting from the last sentence of the target conference text, looking up a first target sentence matched with a preset theme in the direction of the first sentence of the target conference text; starting from the target sentence and facing to the first sentence of the target conference text, the following operations are carried out on each sentence: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; taking the last sentence of the target sentence as the target sentence; and generating a meeting record according to the screened sentences and the target sentences.
In some embodiments, calculating a sentence weight of a previous sentence of the target sentence according to a correlation of the previous sentence of the target sentence with the target sentence comprises: calculating the similarity between the last sentence of the target sentence and the target sentence; determining keywords in the last sentence of the target sentence and the weight of each keyword; and calculating the weight of the last sentence of the target sentence according to the similarity and the weight of each keyword.
In some embodiments, calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword includes: the weight of the last sentence of the target sentence is calculated according to the following formula:
In some embodiments, determining the keywords and the weights of the keywords in the previous sentence of the target sentence comprises: performing word segmentation on each sentence in the target conference text to obtain a plurality of words; calculating the weight of each word; determining the words with the weight reaching a first preset threshold value as the keywords.
In some embodiments, calculating the weight for each term includes: calculating the occurrence frequency of the target words in the target conference text; and calculating the weight of the target word according to the times.
In some embodiments, calculating the weight for each term includes: acquiring a global keyword set; the global keyword set comprises a plurality of global keywords, and each global keyword is obtained according to conference records of a plurality of conferences before the conference corresponding to the target conference text; and calculating the weight of the target word according to the number of the global keywords which are repeated with the target word in the global keyword set.
In some embodiments, after generating the meeting record according to the screened sentences, the method further includes: and adjusting a global keyword set according to the meeting record.
In some embodiments, calculating the weight for each term includes: acquiring the position of a sentence where a target word is in a target conference text; and calculating the weight of the target word according to the position.
In some embodiments, obtaining the position of the sentence where the target word is located in the target meeting text includes: starting from a first sentence of a target conference text, and numbering each sentence in the target conference text in sequence towards the direction of ending the target conference text; taking the number of the sentence where the target word is positioned as the position of the sentence where the target word is positioned in the target conference text; and/or determining a summarized sentence in the target conference text, and taking whether the sentence in which the target word is located is the summarized sentence as the position of the sentence in which the target word is located in the target conference text; and/or judging whether the role of a speaker corresponding to the statement of the target word in the conference is a key role, determining whether the target word is in the statement of the key role according to the judgment result, and taking whether the target word is in the statement of the key role as the position of the statement of the target word in the text of the target conference.
In some embodiments, in the step of performing an operation on each sentence starting from the target sentence and heading towards the first sentence of the target conference text, after performing the operation on each sentence, the method further includes: judging whether a preset stopping condition is reached or not; and after the preset stop condition is reached, canceling the operation of each sentence between the last sentence of the target sentence and the first sentence of the target conference text.
In some embodiments, determining whether the preset stop condition is reached includes: acquiring a keyword set of a target conference text; determining the number of keywords in the keyword set contained in the screened sentences; judging whether the number reaches a preset proportion of the total number of the keywords in the keyword set; in the case of this, it is determined that the predetermined stop condition is reached.
In some embodiments, determining whether the preset stop condition is reached includes: judging whether the sum of the weights of the screened sentences reaches a preset value; in the case where the predetermined value is reached, it is determined that the predetermined stop condition is reached.
In some embodiments, obtaining the target meeting text comprises: acquiring an audio file of a conference; carrying out voice recognition on the audio file to obtain a voice text of the conference; preprocessing the voice text of the conference; the preprocessing comprises removing spoken words and supplementing context sentences completely; segmenting the preprocessed voice text according to the conference theme; and taking the voice text corresponding to one of the divided conference subjects as a target conference text.
A second aspect of the present specification provides a conference record generating apparatus including: the acquisition unit is used for acquiring a target conference text; the searching unit is used for searching a first target sentence matched with a preset theme from the last sentence of the target conference text towards the direction of the first sentence of the target conference text; a processing unit, configured to, starting from the target sentence, perform the following operations on each sentence in a direction toward the first sentence of the target conference text: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; taking the last sentence of the target sentence as the target sentence; and the generating unit is used for generating the conference record according to the screened sentences and the target sentences.
In some embodiments, the processing unit comprises: the first calculating subunit is used for calculating the similarity between the last sentence of the target sentence and the target sentence; the first determining subunit is used for determining the keywords in the last sentence of the target sentence and the weight of each keyword; and the second calculating subunit is used for calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword.
In some embodiments, the second calculating subunit calculates the weight of the previous sentence of the target sentence according to the following formula:
In some embodiments, the first determining subunit comprises: the word segmentation subunit is used for segmenting words of each sentence in the target conference text to obtain a plurality of words; the third calculation subunit is used for calculating the weight of each word; and the second determining subunit is used for determining the words with the weight reaching the first preset threshold as the keywords.
In some embodiments, the third calculation subunit comprises: the fourth calculating subunit is used for calculating the frequency of the target words appearing in the target conference text; and the fifth calculating subunit is used for calculating the weight of the target word according to the times.
In some embodiments, the third calculation subunit comprises: the first acquisition subunit is used for acquiring a global keyword set; the global keyword set comprises a plurality of global keywords, and each global keyword is obtained according to conference records of a plurality of conferences before the conference corresponding to the target conference text; and the sixth calculating subunit is used for calculating the weight of the target word according to the number of the global keywords which are repeated with the target word in the global keyword set.
In some embodiments, the apparatus further comprises: and the adjusting unit is used for adjusting the global keyword set according to the meeting record.
In some embodiments, the third calculation subunit comprises: the second acquisition subunit is used for acquiring the position of the sentence where the target word is located in the target conference text; and the seventh calculation subunit is used for calculating the weight of the target word according to the position.
In some embodiments, the second obtaining subunit obtains the position of the sentence where the target word is located in the target conference text according to the following method: sequentially numbering each sentence in the target conference text from the first sentence of the target conference text towards the direction of the end of the target conference text; taking the number of the sentence where the target word is positioned as the position of the sentence where the target word is positioned in the target conference text; and/or determining a summarized sentence in the target conference text, and taking whether the sentence in which the target word is located is the summarized sentence as the position of the sentence in which the target word is located in the target conference text; and/or judging whether the role of the speaker corresponding to the sentence in which the target word is located in the conference is a key role, determining whether the target word is in the speech sentence of the key role according to the judgment result, and taking whether the target word is in the speech sentence of the key role as the position of the sentence in which the target word is located in the target conference text.
In some embodiments, the processing unit, starting from the target sentence and heading towards the first sentence of the target conference text, during the operation performed on each sentence, after performing the operation on each sentence, further includes: judging whether a preset stopping condition is reached or not; and after the preset stop condition is reached, canceling the operation of each sentence between the last sentence of the target sentence and the first sentence of the target conference text.
In some embodiments, determining whether the preset stop condition is reached includes: acquiring a keyword set of a target conference text; determining the number of keywords in the keyword set contained in the screened sentences; judging whether the number reaches a preset proportion of the total number of the keywords in the keyword set; in the case of this, it is determined that the predetermined stop condition is reached.
In some embodiments, determining whether the preset stop condition is reached includes: judging whether the sum of the weights of the screened sentences reaches a preset value; in the case where the predetermined value is reached, it is determined that the predetermined stop condition is reached.
In some embodiments, the obtaining unit comprises: the third acquisition subunit is used for acquiring an audio file of a conference; the identification subunit is used for carrying out voice identification on the audio file to obtain a voice text of the conference; the preprocessing subunit is used for preprocessing the voice text of the conference; the preprocessing comprises removing spoken words and supplementing context sentences completely; the segmentation subunit is used for segmenting the preprocessed voice text according to the conference subject; and the third determining subunit is used for taking the voice text corresponding to one of the divided conference subjects as the target conference text.
According to the conference record generation method, the conference record generation device and the electronic equipment, aiming at the characteristic that a conference summary link exists in the later stage of a conference, sentences matched with conference subjects are searched according to the sequence from back to front, the found sentence matched with the first sentence is used as a target sentence, the search can be continued after the target sentence is found, and the search workload is reduced; after the target sentence is found, the sentence weight of the previous sentence is calculated according to the correlation between the previous sentence and the next sentence in sequence from the target sentence towards the direction of the first sentence of the conference text, and whether the previous sentence is screened out for generating the conference record is determined according to the weight. Therefore, the conference record generation method is higher in efficiency and accuracy.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 shows a flowchart of a conference recording generation method provided in the present specification;
FIG. 2 is a diagram illustrating a plurality of target meeting texts for a meeting;
FIG. 3 illustrates a flow chart of a method of calculating a sentence weight of a previous sentence of a target sentence based on a correlation of the previous sentence of the target sentence with the target sentence;
fig. 4 shows a functional block diagram of a conference record generating apparatus provided in the present specification;
fig. 5 shows a functional block diagram of an electronic device provided by the present specification.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art without any inventive work based on the embodiments in the present application shall fall within the scope of protection of the present application.
The present specification provides a method for generating a conference record, which can be used in an online or offline intelligent conference system, and automatically generates a conference record according to an audio file obtained by recording speeches of speakers during a conference.
As shown in fig. 1, the method for generating a conference record includes the following steps:
s10: and acquiring a target conference text.
In some embodiments, the target meeting text may be obtained according to the following steps: firstly, recording a conference to obtain an audio file; then carrying out voice recognition on the audio file, converting the audio into characters and obtaining a voice text of the conference; the speech text of the conference is preprocessed, which may be, for example, to remove spoken words (e.g., kayings), to complete the context sentence (e.g., the BERT model using self0attention principle replaces words such as "they", "this", etc. in the previous sentence, the next sentence). The preprocessed text can be used as the target conference text.
In some embodiments, a conference includes a plurality of topics, and on the basis of the preprocessed text, the preprocessed text may be further divided into texts corresponding to a plurality of topics, where the text corresponding to each topic is used as the target conference text. That is, one conference may correspond to a plurality of target conference texts, one target conference text corresponding to one subject. For example, as shown in fig. 2, the conference text corresponding to the subject 3 may be used as a target conference text.
The preprocessed text is divided into texts corresponding to a plurality of subjects, and the following method can be adopted: identifying keywords or key sentences in the preprocessed text associated with topic switching, such as 'next topic', 'this topic is discussed here', and determining to segment the preprocessed text at the position of the keywords or key sentences associated with topic switching.
S20: and starting from the last sentence of the target conference text, searching a first target sentence matched with a preset theme in the direction of the first sentence of the target conference text.
Conference subjects are typically notified to participants prior to each conference.
In some embodiments, the intelligent conferencing system may be utilized to reserve and conference a meeting. When the conference is reserved, the conference theme can be input, the participants are added, and after the reservation is successful, the intelligent conference system can send the conference theme to the participants, so that the purpose of notifying the participants is achieved. The theme preset in S20 may be a conference theme input when the conference is reserved.
In some embodiments, the intelligent conference system is not used to reserve a conference and start the conference, but the participants are notified to start the conference in other manners (such as sending a short message), and when the conference is started, the recording device records the speech of the participants to obtain an audio file of the conference. In this case, the topic of the conference may be determined by identifying a statement related to the topic, for example, the conference moderator's utterance is: "the topic we have opened today is XXX" and the identified topic of the meeting is taken as the preset topic. Alternatively, the conference manager may manually input the subject of the conference, and use the input subject of the conference as a preset subject.
Fig. 2 shows a schematic diagram of a plurality of target conference texts corresponding to a conference, wherein a conference subject is shown below a horizontal line, a plurality of sentences in the conference text corresponding to the conference subject are shown above the horizontal line, sx represents the xth sentence in one target conference text, for example, S3 represents the third sentence, and so on. Of course, the target sentence under one subject may include tens, hundreds or even thousands of sentences, and the method provided in the present specification is illustrated here by taking only 5 sentences as an example.
Taking theme 3 as an example, when S20 is executed, it is first determined from the last sentence S5 whether S5 matches the preset theme, and if so, execution of S20 is stopped, and S5 is taken as the target sentence. If not, judging whether the S4 is matched with the preset theme or not, if so, stopping executing the S20, and taking the S4 as a target statement; if not, continuously judging whether the S3 is matched with the preset theme or not, and so on.
For a formal conference, a conference summary is usually generated at the end of the conference, and the conference summary basically corresponds to a conference subject, so that a summary language which can be quickly and efficiently found for the conference is found from the last sentence of the conference, and further, sentences used for generating conference records can be quickly and efficiently screened according to the correlation between other sentences and the summary language.
S30: starting from the target sentence and facing to the first sentence of the target conference text, the following operations are carried out on each sentence: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; and taking the last sentence of the target sentence as the target sentence.
Taking the conference text corresponding to the topic 3 in fig. 2 as an example, assuming that S5 is a target sentence, the sentence weight of S4 is calculated according to the correlation between S4 and S5, and it is determined whether to filter out S4 as the content of the conference record according to the sentence weight of S4. If yes, S4 is screened out, and the last sentence S3 of S4 is taken as a target sentence.
Then, according to the correlation between S3 and S4, the sentence weight of S3 is calculated, and whether S3 is taken as the content of the conference record is determined according to the sentence weight of S3. If not, directly taking S3 as a target statement.
Then, according to the correlation between S2 and S3, the sentence weight of S2 is calculated, and whether S2 is used as the content of the conference record is determined according to the sentence weight of S2. If yes, screening out S2 and taking S2 as a target statement.
S40: and generating a conference record according to the screened sentences.
In some embodiments, the screened sentences can be directly spliced into the conference record.
In other embodiments, the meeting record may be further generated from the filtered out statements.
According to the conference record generation method provided by the specification, for the characteristic that a conference summary link exists in the later stage of a conference, sentences matched with conference subjects are searched according to the sequence from back to front, the sentence matched with the first sentence is used as a target sentence, the search can be continued after the target sentence is found, and the search workload is reduced; after the target sentence is found, the sentence weight of the previous sentence is calculated according to the correlation between the previous sentence and the next sentence in sequence from the target sentence towards the direction of the first sentence of the conference text, and whether the previous sentence is screened out for generating the conference record is determined according to the weight. Therefore, the conference record generation method is higher in efficiency and accuracy.
In some embodiments, the sentence weight of the previous sentence of the target sentence is calculated according to the correlation between the previous sentence of the target sentence and the target sentence, and may be that only the similarity between the previous sentence of the target sentence and the target sentence is calculated and the similarity is taken as the weight of the previous sentence of the target sentence.
In other embodiments, when calculating the sentence weight of the previous sentence of the target sentence, the weight of each keyword included in the previous sentence of the target sentence may be further determined in addition to the similarity between the previous sentence of the target sentence and the target sentence. That is, as shown in fig. 3, calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence may include the following steps:
s310: and calculating the similarity between the last sentence of the target sentence and the target sentence.
Before calculating the similarity between two sentences, the two sentences may be mapped into vectors respectively, and then the vectors corresponding to the sentences are calculated.
For example, the similarity between two sentences is calculated according to the following formula:
wherein, similarity (S) i ,S j ) Representing the degree of similarity between two sentences, S i The last sentence representing the target sentence, S j Representing a target sentence, w k Denotes a word whereby S in the molecule i 、S j Representing the sentence itself, S in denominator i 、S j Representing the vector corresponding to the statement.
S320: and determining the keywords in the last sentence of the target sentence and the weight of each keyword.
S330: and calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword.
In some embodiments, the weight between the previous sentence of the target sentence and the target sentence may be added to the sum of the weights of the keywords to obtain the weight of the previous sentence of the target sentence. That is, the similarity and the weight of each keyword are directly summed to obtain the weight of the previous sentence of the target sentence.
In other embodimentsIn this way, the sum of the weights of the keywords and the similarity may be weighted and summed to obtain the weight of the previous sentence of the target sentence. That is to say that the first and second electrodes,
wherein ws (S) i ) Representing the weight of the sentence immediately preceding the target sentence, d is a weighting coefficient, similarity (S) i ,S j ) Weight (x) representing the similarity between the previous sentence of the target sentence and the target sentence n ) Representing the weight of the nth keyword in the previous sentence of the target sentence,
the sum of the weights of all the keywords in the previous sentence of the target sentence is represented, and the product operation is represented.
The present specification also provides a method of calculating a weight of a last sentence of a target sentence,
wherein ws (S) i ) Representing the weight of the sentence immediately preceding the target sentence, d is a weighting coefficient, similarity (S) i ,S j ) Weight (x) representing the similarity between the previous sentence of the target sentence and the target sentence n ) Representing the weight of the nth keyword in the previous sentence of the target sentence,
represents the sum of the weights of the keywords in the previous sentence of the target sentence, and represents the product operation, position (S) i ) The position of the last sentence of the target sentence in the target conference text is represented, N represents the length of the last sentence of the target sentence, for example, the length of the sentence can be measured by the number of words, and then N is the number of words contained in the last sentence of the target sentence.
In the weight calculation method, the position of the previous sentence of the target sentence in the target conference text is introduced, and the position can be expressed by a sentence number, that is, the sentences in the target conference text are numbered sequentially, and the number of the previous sentence of the target sentence is taken as the position of the previous sentence in the target conference text. According to the weight calculation method, the weight of the later sentences in the target conference text is further increased by introducing the sentence positions, and the probability of screening the later sentences is improved, so that the accuracy of the conference record can be improved for the conference with the summary link.
In addition, in the weight calculation method, the number of sentences of the target conference text is introduced as the denominator item, so that the situation that the weight is too high due to the fact that the sentences are long and contain more keywords can be avoided.
In some embodiments, the keywords in the sentences of the target conference text may be determined before S320 according to the following steps:
s321: and performing word segmentation on each sentence in the target conference text to obtain a plurality of words.
S322: the weight of each word is calculated.
In some embodiments, S322 may be: calculating the occurrence frequency of the target words in the target conference text; and calculating the weight of the target word according to the times.
For example, weight (x) = frequency (x) 1 )+duplicate(x 2 )+position(x 3 ) + k, where weight (x) is the weight of target word x, frequency (x) 1 ) Is based on the number x of occurrences of the target word in the conference text 1 The calculated value, duplicate (x) 2 ) Is based on the number x of the global keywords in the global keyword set that are repeated with the target word 2 The calculated value, position (x) 3 ) Is based on the position x of the sentence of the target word in the target meeting text 3 The resulting value, k, is a constant.
frequence(x 1 ) The calculation method of (2) may be:
wherein a is a coefficient and represents a multiplication operation.
In some embodiments, S322 may be: acquiring a global keyword set; the global keyword set comprises a plurality of global keywords, and each global keyword is obtained according to conference records of a plurality of conferences before the conference corresponding to the target conference text; and calculating the weight of the target word according to the number of the global keywords which are repeated with the target word in the global keyword set.
For example, duplicate (x) in the above formula for weight (x) 2 ) The calculation method of (2) may be:
where a is a coefficient, and denotes a multiplication operation.
Accordingly, after the conference record is generated, the global keyword set can be adjusted according to the conference record.
In the above embodiment, a global keyword set is introduced, and the keywords are screened by referring to the historical conference situation. For example, for a unit, the business field or research field of the unit is relatively determined, each meeting of the unit is in the business field or research field, and then keywords can be extracted according to the meeting records of the unit's past meetings, the keyword set of the past meetings is used as a global keyword set, and the global keyword set is used for guiding keyword extraction of later meetings. According to the scheme, the content of the field related to the unit can be accurately extracted, and the content is usually important content of the conference, so that the accuracy of conference recording can be improved.
In some embodiments, after obtaining the global keyword set according to the conference content of the past conference, the global keyword set may be further expanded, for example, by supplementing synonyms, expanding the global keyword set by using a term frequency-inverse file frequency algorithm (TF-IDF), or the like.
In other embodiments, S322 may be: acquiring the position of a sentence where a target word is in a target conference text; and calculating the weight of the target word according to the position.
The position of the sentence where the target word is located in the target conference text is obtained, which may be: sequentially numbering each sentence in the target conference text from the first sentence of the target conference text towards the direction of the end of the target conference text; taking the number of the sentence where the target word is positioned as the position of the sentence where the target word is positioned in the target conference text; and/or determining a summarized sentence in the target conference text, and taking whether the sentence in which the target word is located is the summarized sentence as the position of the sentence in which the target word is located in the target conference text; and/or judging whether the role of a speaker corresponding to the statement of the target word in the conference is a key role, determining whether the target word is in the statement of the key role according to the judgment result, and taking whether the target word is in the statement of the key role as the position of the statement of the target word in the text of the target conference.
For example, the number of sentences in the target conference text is counted as M, if the target word is in the sentence of the speech of the key role, the position of the target word may be set to M +1, if the target word is in the summarized sentence, the position of the target word may be set to M +2, and if the target word is in both the sentence of the speech of the key role and the summarized sentence, the position of the target word may be set to M +5. If the target word is not in the statement of the key role or in the summarized statement, the number of the statement in which the target word is located may be the position of the target word.
For example, in the above calculation formula of weight (x), position (x) 3 ) The calculation method of (2) may be:
wherein c is a coefficient, and represents the multiplication operation.
S323: determining the words with the weight reaching a first preset threshold value as the keywords.
The first preset threshold may be determined at the time of model training.
It should be noted that the weight of the word in S322 may be calculated by using the following formula, or may be calculated by deleting part of terms based on the formula and based on the modified formula: weight (x) = frequency (x) 1 )+duplicate(x 2 )+position(x 3 )+k。
In some embodiments, step S30 may start from the target sentence, perform the operation on each sentence in the direction of the first sentence of the target conference text, and terminate the operation after performing the operation on the first sentence of the target conference text.
In other embodiments, in the step S30, starting from the target sentence, and heading toward the direction of the first sentence of the target meeting text, in the process of performing an operation on each sentence, after performing an operation on each sentence, it may be determined whether a preset stop condition is reached; and after the preset stop condition is reached, canceling the operation of each sentence between the last sentence of the target sentence and the first sentence of the target conference text.
Taking the conference text corresponding to the topic 3 in fig. 2 as an example, assuming that S5 is a target sentence, the sentence weight of S4 is calculated according to the correlation between S4 and S5, and it is determined whether to filter out S4 as the content of the conference record according to the sentence weight of S4. If yes, S4 is screened out, and the last sentence S3 of S4 is taken as a target sentence. And judging whether a preset stop condition is reached.
If not, the sentence weight of S3 is calculated according to the correlation between S3 and S4, and whether S3 is taken as the content of the conference record is determined according to the sentence weight of S3. If not, directly taking S3 as a target statement. And judging whether a preset stop condition is reached.
If yes, the sentence weight of S2 is no longer calculated according to the correlation between S2 and S3, i.e., whether S2 is used as the content of the conference record or not is no longer determined, and whether S1 is used as the content of the conference record or not is no longer determined.
In some embodiments, the determining whether the preset stop condition is reached may be: acquiring a keyword set of a target conference text; determining the number of keywords in a keyword set contained in the screened sentences; judging whether the number reaches a preset proportion of the total number of the keywords in the keyword set; in the case of this, it is determined that the predetermined stop condition is reached.
That is, when the coverage rate of each screened sentence on the keywords in the target conference text keyword set reaches a predetermined ratio, a predetermined stop condition is reached.
In some embodiments, the determining whether the preset stop condition is reached may be: judging whether the sum of the weights of the screened sentences reaches a preset value; in the case where the predetermined value is reached, it is determined that the predetermined stop condition is reached.
In some embodiments, the determining whether the preset stop condition is reached may be: determining that the predetermined stop condition is reached when either of the following two conditions is satisfied: when the coverage rate of each screened sentence on the keywords in the target conference text keyword set reaches a preset proportion, and the sum of the weights of the screened sentences reaches a preset value.
The present specification provides a conference record generating apparatus that can be used to implement the conference record generating method shown in fig. 1. As shown in fig. 4, the apparatus includes an acquisition unit 10, a search unit 20, a processing unit 30, and a generation unit 40.
The acquisition unit 10 is used for acquiring a target conference text.
The searching unit 20 is configured to search for a first target sentence matching a preset topic from the last sentence of the target conference text towards the first sentence of the target conference text.
The processing unit 30 is configured to, starting from the target sentence, and heading towards the first sentence of the target conference text, perform the following operations for each sentence: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; and taking the last sentence of the target sentence as the target sentence.
The generating unit 40 is configured to generate a meeting record according to the selected sentence and the target sentence.
In some embodiments, the processing unit comprises: the first calculating subunit is used for calculating the similarity between the last sentence of the target sentence and the target sentence; the first determining subunit is used for determining the keywords in the last sentence of the target sentence and the weight of each keyword; and the second calculating subunit is used for calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword.
In some embodiments, the second calculating subunit calculates the weight of the previous sentence of the target sentence according to the following formula:
In some embodiments, the first determining subunit comprises: the word segmentation subunit is used for segmenting each sentence in the target conference text to obtain a plurality of words; the third calculation subunit is used for calculating the weight of each word; and the second determining subunit is used for determining the words with the weight reaching the first preset threshold as the keywords.
In some embodiments, the third calculation subunit comprises: the fourth calculating subunit is used for calculating the frequency of the target words appearing in the target conference text; and the fifth calculating subunit is used for calculating the weight of the target word according to the times.
In some embodiments, the third calculation subunit comprises: the first acquisition subunit is used for acquiring a global keyword set; the global keyword set comprises a plurality of global keywords, and each global keyword is obtained according to conference records of a plurality of conferences before the conference corresponding to the target conference text; and the sixth calculating subunit is used for calculating the weight of the target word according to the number of the global keywords which are repeated with the target word in the global keyword set.
In some embodiments, the apparatus further comprises: and the adjusting unit is used for adjusting the global keyword set according to the meeting record.
In some embodiments, the third calculation subunit comprises: the second acquisition subunit is used for acquiring the position of the sentence where the target word is located in the target conference text; and the seventh calculating subunit is used for calculating the weight of the target word according to the position.
In some embodiments, the second obtaining subunit obtains the position of the sentence where the target word is located in the target meeting text according to the following method: sequentially numbering each sentence in the target conference text from the first sentence of the target conference text towards the direction of the end of the target conference text; taking the number of the sentence where the target word is positioned as the position of the sentence where the target word is positioned in the target conference text; and/or determining a summarized sentence in the target conference text, and taking whether the sentence in which the target word is located is the summarized sentence as the position of the sentence in which the target word is located in the target conference text; and/or judging whether the role of the speaker corresponding to the sentence in which the target word is located in the conference is a key role, determining whether the target word is in the speech sentence of the key role according to the judgment result, and taking whether the target word is in the speech sentence of the key role as the position of the sentence in which the target word is located in the target conference text.
In some embodiments, the processing unit, starting from the target sentence and facing the direction of the first sentence of the target conference text, further includes, in the course of performing the operation on each sentence, after performing the operation on each sentence: judging whether a preset stopping condition is reached or not; and after the preset stop condition is reached, canceling the operation of each sentence between the last sentence of the target sentence and the first sentence of the target conference text.
In some embodiments, determining whether the preset stop condition is reached includes: acquiring a keyword set of a target conference text; determining the number of keywords in the keyword set contained in the screened sentences; judging whether the number reaches a preset proportion of the total number of the keywords in the keyword set; in the case of this, it is determined that the predetermined stop condition is reached.
In some embodiments, determining whether the preset stop condition is reached includes: judging whether the sum of the weights of the screened sentences reaches a preset value; in the case where the predetermined value is reached, it is determined that the predetermined stop condition is reached.
In some embodiments, the obtaining unit comprises: the third acquisition subunit is used for acquiring an audio file of a conference; the identification subunit is used for carrying out voice identification on the audio file to obtain a voice text of the conference; the preprocessing subunit is used for preprocessing the voice text of the conference; the preprocessing comprises removing spoken words and supplementing context sentences completely; the segmentation subunit is used for segmenting the preprocessed voice text according to the conference theme; and the third determining subunit is used for taking the voice text corresponding to one of the divided conference subjects as the target conference text.
The description and functions of the above units can be understood by referring to the contents of the conference record generation method, and are not described again.
An embodiment of the present invention further provides an electronic device, as shown in fig. 5, the electronic device may include a processor 51 and a memory 52, where the processor 51 and the memory 52 may be connected by a bus or in another manner, and fig. 5 takes the connection by the bus as an example.
The processor 51 may be a Central Processing Unit (CPU). The Processor 51 may also be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or combinations thereof.
The memory 52, which is a non-transitory computer-readable storage medium, may be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as program instructions/modules corresponding to the conference recording or identification method in the embodiment of the present invention (for example, the acquisition unit 10, the search unit 20, the processing unit 30, and the generation unit 40 shown in fig. 4). The processor 51 executes various functional applications and data processing of the processor by running non-transitory software programs, instructions and modules stored in the memory 52, that is, implements the conference record generation method in the above-described method embodiment.
The memory 52 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor 51, and the like. Further, the memory 52 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 52 may optionally include memory located remotely from the processor 51, and these remote memories may be connected to the processor 51 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more modules are stored in the memory 52 and when executed by the processor 51 perform a conference recording generation method as in the embodiment shown in fig. 1.
The details of the electronic device may be understood with reference to the corresponding descriptions and effects in the embodiments of fig. 1 or fig. 3, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk Drive (Hard Disk Drive, abbreviated as HDD), or a Solid State Drive (SSD); the storage medium may also comprise a combination of memories of the kind described above.
In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate a dedicated integrated circuit chip 2. Furthermore, nowadays, instead of manually manufacturing an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as ABEL (Advanced Boolean Expression Language), AHDL (alternate Hardware Description Language), traffic, CUPL (core universal Programming Language), HDCal, jhddl (Java Hardware Description Language), lava, lola, HDL, PALASM, rhyd (Hardware Description Language), and vhjh-Language (Hardware Description Language), which is currently used by Hardware compiler-Language-2. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments.
The systems, devices, modules or units described in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions.
For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing the present application.
From the above description of the embodiments, it is clear to those skilled in the art that the present application can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the present application may be essentially or partially implemented in the form of software products, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods of some parts of the embodiments of the present application.
The application is operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Although the present application has been described in terms of embodiments, those of ordinary skill in the art will recognize that there are numerous variations and permutations of the present application without departing from the spirit of the application, and it is intended that the appended claims encompass such variations and permutations without departing from the spirit of the application.
Claims (16)
1. A method for generating a conference record, comprising:
acquiring a target conference text;
starting from the last sentence of the target conference text, and looking for a first target sentence matched with a preset theme in the direction of the first sentence of the target conference text;
starting from the target sentence and facing to the first sentence of the target conference text, the following operations are carried out on each sentence: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; taking the last sentence of the target sentence as the target sentence;
and generating a conference record according to the screened sentences and the target sentences.
2. The method of claim 1, wherein calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence comprises:
calculating the similarity between the last sentence of the target sentence and the target sentence;
determining keywords in the last sentence of the target sentence and the weight of each keyword;
and calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword.
3. The method of claim 2, wherein calculating the weight of the previous sentence of the target sentence according to the similarity and the weight of each keyword comprises:
calculating the weight of the last sentence of the target sentence according to the following formula:
wherein ws (S) i ) Representing the weight of the sentence immediately preceding the target sentence, d is a weighting coefficient, similarity (S) i ,S j ) Weight (x) representing the similarity between the previous sentence of the target sentence and the target sentence n ) Representing the weight of the nth keyword in the previous sentence of the target sentence,
represents the sum of the weights of the keywords in the previous sentence of the target sentence, and represents the product operation, position (S) i ) And N represents the position of the last sentence of the target sentence in the target conference text, and the length of the last sentence of the target sentence.
4. The method of claim 2, wherein determining the keywords and the weights of the keywords in the previous sentence of the target sentence comprises:
performing word segmentation on each sentence in the target conference text to obtain a plurality of words;
calculating the weight of each word;
determining the words with the weight reaching a first preset threshold value as the keywords.
5. The method of claim 4, wherein calculating a weight for each term comprises:
calculating the occurrence frequency of the target words in the target conference text;
and calculating the weight of the target word according to the times.
6. The method of claim 4, wherein calculating a weight for each term comprises:
acquiring a global keyword set; the global keyword set comprises a plurality of global keywords, and each global keyword is obtained according to conference records of a plurality of conferences before the conference corresponding to the target conference text;
and calculating the weight of the target word according to the number of the global keywords which are repeated with the target word in the global keyword set.
7. The method of claim 6, wherein after generating the meeting record according to the selected sentences, further comprising:
and adjusting a global keyword set according to the meeting record.
8. The method of claim 4, wherein calculating a weight for each term comprises:
acquiring the position of a sentence where a target word is in a target conference text;
and calculating the weight of the target word according to the position.
9. The method of claim 8, wherein obtaining the position of the sentence in which the target word is located in the target meeting text comprises:
sequentially numbering each sentence in the target conference text from the first sentence of the target conference text towards the direction of the end of the target conference text; taking the number of the sentence where the target word is positioned as the position of the sentence where the target word is positioned in the target conference text;
and/or the presence of a gas in the gas,
determining a summarization statement in the target conference text, and taking whether the statement in which the target word is located is the summarization statement as the position of the statement in which the target word is located in the target conference text;
and/or the presence of a gas in the gas,
judging whether the role of the speaker in the conference corresponding to the sentence where the target word is located is a key role, determining whether the target word is in the speech sentence of the key role according to the judgment result, and taking whether the target word is in the speech sentence of the key role as the position of the sentence where the target word is located in the target conference text.
10. The method of claim 1, wherein, in the process of performing the operation on each sentence starting from the target sentence and heading towards the first sentence of the target conference text, after performing the operation on each sentence, the method further comprises:
judging whether a preset stopping condition is reached or not;
and after the preset stop condition is reached, canceling the operation of each sentence between the last sentence of the target sentence and the first sentence of the target conference text.
11. The method of claim 10, wherein determining whether a predetermined stop condition is reached comprises:
acquiring a keyword set of a target conference text;
determining the number of keywords in the keyword set contained in the screened sentences;
judging whether the number reaches a preset proportion of the total number of the keywords in the keyword set;
in the case of this, it is determined that the predetermined stop condition is reached.
12. The method of claim 10, wherein determining whether a predetermined stop condition is reached comprises:
judging whether the sum of the weights of the screened sentences reaches a preset value;
in the case where the predetermined value is reached, it is determined that the predetermined stop condition is reached.
13. The method of claim 1, wherein obtaining target meeting text comprises:
acquiring an audio file of a conference;
carrying out voice recognition on the audio file to obtain a voice text of the conference;
preprocessing the voice text of the conference; the preprocessing comprises removing spoken words and supplementing context sentences completely;
segmenting the preprocessed voice text according to the conference theme;
and taking the voice text corresponding to one of the divided conference subjects as a target conference text.
14. A conference record generating apparatus, comprising:
the acquisition unit is used for acquiring a target conference text;
the searching unit is used for searching a first target sentence matched with a preset theme from the last sentence of the target conference text towards the direction of the first sentence of the target conference text;
a processing unit, configured to, starting from the target sentence, perform the following operations on each sentence in a direction toward the first sentence of the target conference text: calculating the sentence weight of the previous sentence of the target sentence according to the correlation between the previous sentence of the target sentence and the target sentence, and determining whether the previous sentence of the target sentence is screened out for generating a conference record according to the sentence weight; taking the last sentence of the target sentence as the target sentence;
and the generating unit is used for generating the conference record according to the screened sentences and the target sentences.
15. An electronic device, comprising:
a memory and a processor, the processor and the memory being communicatively connected to each other, the memory having stored therein computer instructions, the processor implementing the steps of the method of any one of claims 1 to 13 by executing the computer instructions.
16. A computer storage medium storing computer program instructions which, when executed, implement the steps of the method of any one of claims 1 to 13.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210882694.0A CN115174285B (en) | 2022-07-26 | 2022-07-26 | Conference record generation method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210882694.0A CN115174285B (en) | 2022-07-26 | 2022-07-26 | Conference record generation method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115174285A true CN115174285A (en) | 2022-10-11 |
| CN115174285B CN115174285B (en) | 2024-02-27 |
Family
ID=83497573
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210882694.0A Active CN115174285B (en) | 2022-07-26 | 2022-07-26 | Conference record generation method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115174285B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117312612A (en) * | 2023-10-07 | 2023-12-29 | 广东鼎尧科技有限公司 | Multi-mode-based teleconference data recording method, system and medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052686A (en) * | 2018-01-26 | 2018-05-18 | 腾讯科技（深圳）有限公司 | A kind of abstract extraction method and relevant device |
| CN110335612A (en) * | 2019-07-11 | 2019-10-15 | 招商局金融科技有限公司 | Minutes generation method, device and storage medium based on speech recognition |
| CN110704608A (en) * | 2019-08-29 | 2020-01-17 | 中国平安人寿保险股份有限公司 | Text theme generation method and device and computer equipment |
| CN112836016A (en) * | 2021-02-05 | 2021-05-25 | 北京字跳网络技术有限公司 | Conference summary generation method, device, equipment and storage medium |
| US20210407499A1 (en) * | 2020-06-30 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Automatically generating conference minutes |
-
2022
- 2022-07-26 CN CN202210882694.0A patent/CN115174285B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052686A (en) * | 2018-01-26 | 2018-05-18 | 腾讯科技（深圳）有限公司 | A kind of abstract extraction method and relevant device |
| CN110335612A (en) * | 2019-07-11 | 2019-10-15 | 招商局金融科技有限公司 | Minutes generation method, device and storage medium based on speech recognition |
| CN110704608A (en) * | 2019-08-29 | 2020-01-17 | 中国平安人寿保险股份有限公司 | Text theme generation method and device and computer equipment |
| US20210407499A1 (en) * | 2020-06-30 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Automatically generating conference minutes |
| CN112836016A (en) * | 2021-02-05 | 2021-05-25 | 北京字跳网络技术有限公司 | Conference summary generation method, device, equipment and storage medium |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117312612A (en) * | 2023-10-07 | 2023-12-29 | 广东鼎尧科技有限公司 | Multi-mode-based teleconference data recording method, system and medium |
| CN117312612B (en) * | 2023-10-07 | 2024-04-02 | 广东鼎尧科技有限公司 | Multi-mode-based teleconference data recording method, system and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115174285B (en) | 2024-02-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11508366B2 (en) | Whispering voice recovery method, apparatus and device, and readable storage medium | |
| CN108735201B (en) | Continuous speech recognition method, device, equipment and storage medium | |
| CN110444198B (en) | Retrieval method, retrieval device, computer equipment and storage medium | |
| CN111241813B (en) | Corpus expansion method, apparatus, device and medium | |
| CN110765759B (en) | Intention recognition method and device | |
| CN108416032B (en) | Text classification method, device and storage medium | |
| WO2018045646A1 (en) | Artificial intelligence-based method and device for human-machine interaction | |
| WO2022121251A1 (en) | Method and apparatus for training text processing model, computer device and storage medium | |
| CN110070859B (en) | Voice recognition method and device | |
| CN111814770A (en) | Content keyword extraction method of news video, terminal device and medium | |
| US20230089308A1 (en) | Speaker-Turn-Based Online Speaker Diarization with Constrained Spectral Clustering | |
| CN110335608B (en) | Voiceprint verification method, voiceprint verification device, voiceprint verification equipment and storage medium | |
| WO2019148585A1 (en) | Conference abstract generating method and apparatus | |
| JP2020004382A (en) | Method and device for voice interaction | |
| CN112256822A (en) | Text search method and device, computer equipment and storage medium | |
| CN111126084B (en) | Data processing method, device, electronic equipment and storage medium | |
| CN115174285B (en) | Conference record generation method and device and electronic equipment | |
| CN112017643B (en) | Speech recognition model training method, speech recognition method and related device | |
| CN112397053B (en) | Voice recognition method and device, electronic equipment and readable storage medium | |
| CN115098729A (en) | Video processing method, sample generation method, model training method and device | |
| Chen et al. | End-to-end recognition of streaming Japanese speech using CTC and local attention | |
| WO2021047103A1 (en) | Voice recognition method and device | |
| CN114266247A (en) | Sensitive word filtering method and device, storage medium and electronic equipment | |
| CN111785259A (en) | Information processing method and device and electronic equipment | |
| CN112395414A (en) | Text classification method and training method, device, medium and equipment of classification model |
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 |