Conversation content summary generation method and device, electronic equipment and storage medium

Abstract

The invention provides a dialogue content summary generation method and device, electronic equipment and a storage medium. The method comprises the following steps: in response to a summary generation instruction for target dialogue content, obtaining the target dialogue content; wherein the target dialogue content comprises speaking content of a plurality of speakers; performing topic identification on the target dialogue content to obtain initial topic data of at least one initial topic; fusing the initial topic data of the at least one initial topic according to the association relationship of the speaking contents of the plurality of speakers to obtain target topic data of the at least one target topic; and performing natural language processing on the target topic data of each target topic to obtain a target summary statement of the target dialogue content. According to the method, the summary statement close to a natural language expression mode can be obtained, so that the summary of the dialogue content not only contains keywords, and the comprehensiveness and readability of the summary of the dialogue content are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer processing technology, and in particular to a method, device, electronic device and storage medium for summarizing and generating conversation content. Background Art

[0002] Online conversation is a common form of communication. In a group conversation environment, each speaker can speak freely. Therefore, a large amount of conversation content can often be formed in a short period of time. Summarizing these conversation contents can prevent group members from missing important information.

[0003] In the prior art, the summary of the conversation content can often only extract some repeatedly mentioned keywords, while the conversation relationship in the context and the emotional expression therein are often ignored, resulting in technical problems such as incomplete summary and poor readability. Summary of the invention

[0004] In view of this, an object of the present disclosure is to provide a method, device, electronic device and storage medium for generating a summary of conversation content, so as to improve the comprehensiveness and readability of the conversation content summary.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for generating a summary of conversation content, the method comprising: in response to a summary generation instruction for target conversation content, obtaining the target conversation content; wherein the target conversation content includes speech contents of multiple speakers; performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic; fusing the initial topic data of the at least one initial topic according to the correlation relationship between the speech contents of the multiple speakers to obtain target topic data of at least one target topic; performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0006] In a second aspect, an embodiment of the present disclosure provides a device for generating a summary of a conversation content, the device comprising: a response module, for obtaining the target conversation content in response to a summary generation instruction for the target conversation content; wherein the target conversation content includes speech contents of multiple speakers; an identification module, for performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic; a fusion module, for fusing the initial topic data of the at least one initial topic according to the correlation relationship between the speech contents of the multiple speakers to obtain target topic data of at least one target topic; and a processing module, for performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0007] In a third aspect, an embodiment of the present disclosure provides an electronic device, including a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the above-mentioned method for summarizing and generating conversation content.

[0008] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the above-mentioned method for summarizing and generating the conversation content.

[0009] The embodiments of the present disclosure bring the following beneficial effects:

[0010] The above-mentioned method, device, electronic device and storage medium for summarizing the conversation content extract initial topic data from the conversation content, then merge the topic data with related relationships in the speech content, and finally perform natural language processing to obtain a summary sentence that is close to natural language expression, so that the summary of the conversation content not only contains keywords, but also improves the comprehensiveness and readability of the summary of the conversation content.

[0011] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or understood by practicing the present disclosure. The purpose and other advantages of the present disclosure are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0012] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1 A flowchart of an embodiment of a method for generating a summary of conversation content in an embodiment of the present disclosure;

[0015] Figure 2 A schematic diagram of a device for generating a summary of conversation content provided by an embodiment of the present disclosure;

[0016] Figure 3 A schematic diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0018] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] For ease of understanding, the specific process of the embodiment of the present disclosure is described below. Figure 1 , an embodiment of the method for generating a summary of the conversation content in the embodiment of the present disclosure includes:

[0020] Step S10, in response to a summary generation instruction for a target conversation content, obtaining the target conversation content; wherein the target conversation content includes speech contents of multiple speakers;

[0021] It is understandable that in a group conversation environment, all group members can publish content as speakers, and the content published by the speakers is the speech content. For example, in online chat tools, group chat rooms can be formed, and members in the chat rooms can communicate in real time in the same conversation environment. For example, some game applications also have group chat functions, and game players can create or join groups to obtain a group conversation environment.

[0022] In this implementation, the target conversation content includes the speech content of multiple speakers, and each speaker can publish different content at different times. Therefore, the speech content of each speaker can include content published at at least one time, and the content published by each speaker at different times corresponds to a publishing time. It should be noted that the speaker can publish content in any form / format, such as voice, picture, video, text, emoticon, file, hyperlink, etc., which is not limited here.

[0023] In one embodiment, the summary generation instruction for the target conversation content can be automatically triggered or non-automatically triggered. When automatically triggered, it can be triggered when specific conditions are met. Specifically, the summary generation instruction can be triggered when the conversation content in the group meets specific conditions, for example, when the conversation content in the group reaches a certain number, when the time without new conversations in the group is greater than a certain time, when a certain time is reached, etc., and the specifics are not limited here.

[0024] In one embodiment, in response to a summary generation instruction for the target conversation content, when the target conversation content is obtained, in response to an arrival instruction of a preset time length, it is determined whether new speech content is generated in the target group within the preset time length; wherein all members of the target group can serve as speakers; if new speech content is generated, the new speech content generated within the preset time length is determined as the target conversation content.

[0025] In this embodiment, the summary generation instruction for the target conversation content is automatically triggered at a fixed time, for example, once every 2 hours. When the preset duration is reached, the preset duration reaching instruction is triggered to determine whether the target group has generated new speech content within the preset duration. If yes, the new speech content generated within the preset duration is determined as the target conversation content. If not, wait for the next preset duration reaching instruction. Among them, all members of the target group can publish content as speakers and obtain new speech content, which is not limited here.

[0026] In addition, in the method of non-automatically triggering summary generation instructions, the speech content can be specified as the target conversation content, the speech content within a specific time can be specified as the target conversation content, and the target conversation content can also be specified by other methods, which will not be repeated here.

[0027] Step S20: performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic;

[0028] It is understandable that the publication of personal speech is highly free, and in group conversations, topics / themes that group members are interested in are usually discussed. In this implementation, topics are extracted from the target conversation content to obtain initial topic data of at least one initial topic, and the initial topic data includes topic data of each initial topic. For example, assuming that part of the target conversation content is about a topic of "game props", then the "game props" can be used as initial topic data of an initial topic, and the specific details are not limited here.

[0029] In one embodiment, the initial topic data may include data such as topic objects, main opinions on the topic objects, positive opinions, negative opinions, and neutral opinions, among which the main opinions refer to the opinions expressed by most of the speech content, which are not specifically limited here. For example, assuming that the target conversation content includes: "The equipment updated today is so awesome!", "Yes, especially the new Flame Sword.", "I got it too! It's simply invincible."; then, the initial topic data of an initial topic identified from the target conversation content can be "

Flame Sword

Flame Sword

[0030] In one embodiment, the target conversation content may be subject to topic identification using a pre-trained artificial intelligence model or a topic identification algorithm to obtain initial topic data of at least one initial topic. Before subject identification is performed on the target conversation content, the target conversation content may also be subject to data cleaning to delete speech content in a specified format, such as emoticons, pictures, hyperlinks, etc. Duplicate speech content may also be deleted to improve the efficiency of summary generation.

[0031] In one embodiment, before performing topic identification on the target conversation content, the target conversation content may also be preprocessed to convert the speech content in a specified format into text to obtain the target conversation content in a plain text format for topic identification. For example, the speech content in a voice, video, or other format may be converted into text, and the specifics are not limited here.

[0032] In one embodiment, when performing topic identification on the target conversation content, topic identification in a specified field can be performed. For example, the target conversation content formed in a game application can specify topics in the game field, and can also be subdivided into topics in the game strategy field, game skill field, game props field, etc., which are not specifically limited here.

[0033] Furthermore, when determining the topic area to be identified, the topic area to be identified can be identified based on the name of the group and the topic area of ​​the group's historical conversation content, thereby performing topic identification for the specified topic area and obtaining initial topic data for at least one initial topic. For example, assuming that the name of a group is "XX Game Discussion Group", then when performing topic identification for the target conversation content of the group, it can be specified that only the topic of "XX Game" is identified. If the most discussed topic in the historical conversation content of a group is "XX Game Clearance Skills", then when performing topic identification for the target conversation content of the group, only the topic of "XX Game Clearance Skills" can be identified, so that the summary of the conversation content is more targeted.

[0034] Step S30: According to the correlation relationship between the speech contents of multiple speakers, the initial topic data of at least one initial topic are merged to obtain the target topic data of at least one target topic;

[0035] It is understandable that each identified initial topic may be for different speech contents, and there may be correlations between different speech contents. For speech contents with correlations, the identified initial topics may be merged to further condense and simplify the summary of the conversation content, making the summary more concise and easy to read.

[0036] In one embodiment, if the speech contents corresponding to any two or more initial topics are related, it is determined whether the topic objects of these initial topics are the same. If they are the same, the initial topic data of these initial topics are merged to obtain target topic data of at least one target topic, so that the topic data of the same topic can be merged into one, thereby improving the conciseness of the summary.

[0037] Step S40: Perform natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0038] In this embodiment, after the processing of steps S10-S30, the target topic data obtained may include vocabulary and / or sentences, which are key data of the target topic. This step performs natural language processing on the target topic data, which can improve the readability of the dialogue content summary and make it closer to the expression of natural language. In one embodiment, the target topic data of each target topic is processed by a pre-trained natural language model to obtain a target summary sentence of the target dialogue content.

[0039] The method for generating a summary of conversation content provided in the above-mentioned implementation method extracts initial topic data from the conversation content, then merges the topic data that are related to the speech content, and finally performs natural language processing to obtain a summary sentence that is close to natural language expression, so that the summary of the conversation content not only includes keywords, but also improves the comprehensiveness and readability of the summary of the conversation content.

[0040] Next, the method of generating topic data will be described in detail.

[0041] In one embodiment, based on the correlation relationship between the speech contents of multiple speakers, the initial topic data of at least one initial topic are fused to obtain the target topic data of at least one target topic, and then multiple first speech contents corresponding to each initial topic are obtained; the correlation relationship between the first speech contents of any initial topics is identified to determine at least one group of initial topics whose first speech contents have a correlation relationship; the initial topic data in each group of initial topics are fused to obtain the target topic data of at least one target topic.

[0042] In this implementation, each initial topic is identified based on multiple first speech contents. By identifying whether there is a correlation between the first speech contents of different initial topics, at least one group of initial topics whose first speech contents have a correlation is obtained, and then the initial topic data of the same group of initial topics are fused to obtain target topic data of at least one target topic. It can be seen that the number of target topics is less than or equal to the number of initial topics. When there is no correlation between the first speech contents of any initial topics, the initial topic is the target topic, and the initial topic data of the initial topic is the target topic data of the target topic. No specific limitation is made here.

[0043] In one embodiment, when identifying the association relationship between the first speech content of any initial topics, it can be determined whether there is an association relationship between the first speech content of any initial topics based on the entity information corresponding to each first speech content, and the initial topics corresponding to the first speech content with the same entity information are determined as the initial topics with the entity association relationship, thereby obtaining at least one group of initial topics with the first speech content with the association relationship.

[0044] It can be understood that entity information refers to specific things, objects, or concepts identified and classified from the speech content. For example, assuming the speech content is "I just got a new flaming sword!", then the entity information of the speech content is "flaming sword", which is not limited here.

[0045] In one embodiment, when the association relationship between the first speech content between any initial topics is identified and at least one group of initial topics with which the first speech content is associated is determined, the target conversation content is identified with a conversation chain based on the reference relationship and / or forwarding relationship between the speech contents in the target conversation content to obtain at least one conversation chain in the target conversation content; each conversation chain includes multiple speech contents; and the initial topics corresponding to the first speech contents belonging to the same conversation chain are determined as initial topics with which there is an associated relationship to obtain at least one group of initial topics.

[0046] In this implementation, the association relationship of the speech content includes a reference relationship and / or a forwarding relationship. When the association relationship of the first speech content between any initial topics is identified and at least one group of initial topics with which the first speech content has an association relationship is determined, the target conversation content is identified for a conversation chain based on the reference relationship and / or forwarding relationship between the speech contents in the target conversation content to obtain at least one conversation chain in the target conversation content; each conversation chain includes multiple speech contents; the initial topics corresponding to the first speech contents belonging to the same conversation chain are determined as initial topics with an association relationship to obtain at least one group of initial topics.

[0047] It is understandable that the speeches can reference and / or forward each other, thereby forming a reference relationship and / or a forwarding relationship. Through the reference relationship and / or the forwarding relationship, the conversation chain in the target conversation content can be identified, thereby obtaining at least one conversation chain, and then the initial topics corresponding to the first speech contents belonging to the same conversation chain are determined as the initial topics with an associated relationship, thereby obtaining at least one group of initial topics, so that the same discussion chain can form the same topic data, reducing the probability of repeated content in the summary content, and making the summary content more concise and easy to read.

[0048] Furthermore, when performing dialogue chain recognition on the target dialogue content, the context association algorithm can also be combined to perform context association recognition on the target dialogue content, so that the speech content that is contextually relevant and coherent is regarded as the same dialogue chain, making the topic extraction more accurate.

[0049] In one embodiment, before the step of identifying a dialogue chain for the target dialogue content according to the reference relationship and / or forwarding relationship between the speech contents in the target dialogue content to obtain at least one dialogue chain in the target dialogue content, the reference relationship and / or forwarding relationship between the speech contents in the target dialogue content may also be determined according to the reference mark and forwarding mark corresponding to the speech contents in the target dialogue content. The reference mark and forwarding mark may be identifiers of speech contents referenced or forwarded by the speech contents, which are not specifically limited here.

[0050] In this embodiment, the reference relationship and / or forwarding relationship between the speech contents can be determined by judging whether there are corresponding reference marks and / or forwarding marks for the speech contents. If the speech contents correspond to reference marks, it is determined that there is a reference relationship between the speech contents and the speech contents indicated by the reference marks. If the speech contents correspond to forwarding marks, it is determined that there is a forwarding relationship between the speech contents and the speech contents indicated by the forwarding marks.

[0051] In one embodiment, when the initial topic data in each group of initial topics are fused to obtain target topic data of at least one target topic, the initial topic data in each group of initial topics are semantically fused through a deep learning model to obtain target topic data of at least one target topic. It can be understood that semantic fusion refers to the semantic fusion of different text contents to obtain new text content, and the end-to-end deep learning model can recognize complex semantic structures and relationships, thereby achieving a higher level of semantic fusion and understanding capabilities and obtaining more accurate topic data.

[0052] In one embodiment, when topic identification is performed on the target conversation content to obtain initial topic data of at least one initial topic, the target conversation content is topic clustered to obtain initial topic data of at least one initial topic. Specifically, the target conversation content is topic clustered using a preset topic / theme clustering algorithm, such as latent dirichlet allocation (LDA) algorithm, linear discriminant analysis (LDA) algorithm, etc., which can be used for topic clustering, and the specifics are not limited here.

[0053] Furthermore, in one embodiment, when topic identification is performed on the target conversation content to obtain initial topic data of at least one initial topic, topic features of the target field are extracted for each speech content in the target conversation content through a preset topic feature extraction algorithm to obtain a topic feature vector corresponding to each speech content; all topic feature vectors are clustered through a preset clustering algorithm to obtain initial topic data of at least one initial topic.

[0054] In this implementation, the target domain topic features of each speech content are first extracted to obtain the topic features in vector form corresponding to each speech content, and then the topic feature vectors corresponding to all speech contents are clustered through a clustering algorithm, such as the clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise) and the mean clustering algorithm k-means, so as to obtain the initial topic data of at least one initial topic.

[0055] The following is a detailed description of the method of natural language processing.

[0056] In one embodiment, when natural language processing is performed on the target topic data of each target topic to obtain a target summary sentence of the target dialogue content, the language style information of the target reader is obtained; wherein the target summary sentence is generated for the target reader; natural language conversion is performed on the target topic data of each target topic through a preset natural language generation algorithm to obtain a first summary sentence of the target dialogue content; and the language style of the first summary sentence is adjusted according to the language style information of the target reader to obtain the target summary sentence of the target dialogue content.

[0057] In this implementation, in order to make the summary of the conversation content more in line with the language style of the target reader and more readable, the language style information of the target reader is first obtained. For example, the language style information can be a formal style or an informal style, etc. The target reader can be a member of the group or not (for a group with open conversation content), and the specific details are not limited here.

[0058] Next, the target conversation data is converted into natural language expressions to obtain the first summary language. Finally, the first summary sentence is adjusted to the language style of the target reader to obtain the target summary sentence. The target summary sentence is highly readable and extracts the essence of the topic in the conversation content, rather than just the presentation of keywords.

[0059] In one embodiment, when acquiring the language style information of the target reader, the target reader's historical speech content is acquired; the language style of the historical speech content is identified by a preset language style identification algorithm to obtain the language style information of the target reader. In this embodiment, the language style of the target reader can be identified based on the historical speech content, thereby obtaining the language style information, so that the generated dialogue content summary is more in line with the reader's reading habits and has higher readability.

[0060] Corresponding to the above method embodiment, see Figure 2 A schematic diagram of a device for generating a summary of a conversation content is shown, the device comprising: a response module 201, for obtaining the target conversation content in response to a summary generation instruction for the target conversation content; wherein the target conversation content includes the speech content of multiple speakers; an identification module 202, for performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic; a fusion module 203, for fusing the initial topic data of the at least one initial topic according to the correlation relationship between the speech contents of the multiple speakers to obtain target topic data of at least one target topic; and a processing module 204, for performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0061] The above-mentioned device for generating a summary of the conversation content extracts initial topic data from the conversation content, then merges the topic data that are related to the speech content, and finally performs natural language processing to obtain a summary sentence that is close to a natural language expression, so that the summary of the conversation content not only includes keywords, but also improves the comprehensiveness and readability of the summary of the conversation content.

[0062] Optionally, the above-mentioned fusion module 203 includes: an acquisition unit, used to acquire multiple first speech contents corresponding to each initial topic; an identification unit, used to identify the association relationship between the first speech contents between any initial topics, and determine at least one group of initial topics in which the first speech contents have an association relationship; a fusion unit, used to fuse the initial topic data in each group of initial topics to obtain target topic data of at least one target topic.

[0063] Optionally, the above-mentioned identification unit includes a function for: performing dialogue chain identification on the target dialogue content according to the reference relationship and / or forwarding relationship between the speech contents in the target dialogue content, to obtain at least one dialogue chain in the target dialogue content; each dialogue chain includes multiple speech contents; and determining the initial topics corresponding to the first speech contents belonging to the same dialogue chain as initial topics with an associated relationship, to obtain at least one group of initial topics.

[0064] Optionally, the above-mentioned device also includes: a determination module, which is used to determine the reference relationship and / or forwarding relationship between the speech contents in the target dialogue content according to the reference mark and forwarding mark corresponding to the speech content in the target dialogue content.

[0065] Optionally, the fusion unit is specifically used to: perform semantic fusion on the initial topic data in each group of initial topics through a deep learning model to obtain target topic data of at least one target topic.

[0066] Optionally, the above-mentioned identification module 202 is specifically used for: extracting topic features of the target field for each speech content in the target conversation content through a preset topic feature extraction algorithm to obtain a topic feature vector corresponding to each speech content; clustering all topic feature vectors through a preset clustering algorithm to obtain initial topic data of at least one initial topic.

[0067] Optionally, the processing module 204 includes: an acquisition unit, used to acquire language style information of the target reader; wherein the target summary sentence is generated for the target reader; a conversion unit, used to perform natural language conversion on the target topic data of each target topic through a preset natural language generation algorithm to obtain a first summary sentence of the target dialogue content; and an adjustment unit, used to adjust the language style of the first summary sentence according to the language style information of the target reader to obtain a target summary sentence of the target dialogue content.

[0068] Optionally, the acquisition unit is specifically used to: acquire historical speech content of the target reader; and perform language style recognition on the historical speech content by using a preset language style recognition algorithm to obtain language style information of the target reader.

[0069] Optionally, the above-mentioned response module 201 is specifically used to: respond to the arrival instruction of the preset time length, determine whether new speech content is generated in the target group within the preset time length; wherein, all members of the target group can serve as speakers; if new speech content is generated, then the new speech content generated within the preset time length is determined as the target conversation content.

[0070] This embodiment also provides an electronic device, including a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the above-mentioned method for summarizing and generating conversation content. The electronic device can be a server or a terminal device.

[0071] See also Figure 3 As shown, the electronic device includes a processor 100 and a memory 101, wherein the memory 101 stores machine executable instructions that can be executed by the processor 100, and the processor 100 executes the machine executable instructions to implement the above-mentioned method for summarizing and generating conversation content.

[0072] Further, Figure 3 The electronic device shown further includes a bus 102 and a communication interface 103 , and the processor 100 , the communication interface 103 and the memory 101 are connected via the bus 102 .

[0073] The memory 101 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 103 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used. The bus 102 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0074] The processor 100 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 100. The above processor 100 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present disclosure can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present disclosure can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 101, and the processor 100 reads the information in the memory 101 and completes the steps of the method of the above embodiment in combination with its hardware, for example:

[0075] In response to a summary generation instruction for a target conversation content, the target conversation content is obtained; wherein the target conversation content includes speech content of multiple speakers; topic recognition is performed on the target conversation content to obtain initial topic data of at least one initial topic; based on the correlation relationship between the speech content of multiple speakers, the initial topic data of at least one initial topic is merged to obtain target topic data of at least one target topic; natural language processing is performed on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0076] In this method, initial topic data is extracted from the conversation content, and then the topic data with correlation relationship with the speech content is merged, and finally natural language processing is performed to obtain a summary sentence close to natural language expression, so that the summary of the conversation content not only contains keywords, but also improves the comprehensiveness and readability of the summary of the conversation content.

[0077] Optionally, based on the correlation between the speech contents of multiple speakers, the initial topic data of at least one initial topic are merged to obtain the target topic data of at least one target topic, including: obtaining multiple first speech contents corresponding to each initial topic; identifying the correlation between the first speech contents of any initial topics, and determining at least one group of initial topics in which the first speech contents have a correlation; and fusing the initial topic data in each group of initial topics to obtain the target topic data of at least one target topic.

[0078] Optionally, the step of identifying the association relationship between the first speech contents between any initial topics and determining at least one group of initial topics with which the first speech contents are associated includes: identifying conversation chains on the target conversation contents based on the reference relationship and / or forwarding relationship between the speech contents in the target conversation contents to obtain at least one conversation chain in the target conversation contents; each conversation chain includes multiple speech contents; and determining the initial topics corresponding to the first speech contents belonging to the same conversation chain as initial topics with which they are associated to obtain at least one group of initial topics.

[0079] Optionally, before the step of performing conversation chain identification on the target conversation content based on the reference relationship and / or forwarding relationship between the speeches in the target conversation content to obtain at least one conversation chain in the target conversation content, the method further includes: determining the reference relationship and / or forwarding relationship between the speeches in the target conversation content based on the reference marks and forwarding marks corresponding to the speeches in the target conversation content.

[0080] Optionally, the step of fusing the initial topic data in each group of initial topics to obtain target topic data of at least one target topic includes: performing semantic fusion on the initial topic data in each group of initial topics through a deep learning model to obtain target topic data of at least one target topic.

[0081] Optionally, the step of performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic includes: extracting topic features of the target field for each speech content in the target conversation content by a preset topic feature extraction algorithm to obtain a topic feature vector corresponding to each speech content; and clustering all topic feature vectors by a preset clustering algorithm to obtain initial topic data of at least one initial topic.

[0082] Optionally, the step of performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target dialogue content includes: obtaining language style information of the target reader; wherein the target summary sentence is generated for the target reader; performing natural language conversion on the target topic data of each target topic through a preset natural language generation algorithm to obtain a first summary sentence of the target dialogue content; and adjusting the language style of the first summary sentence according to the language style information of the target reader to obtain the target summary sentence of the target dialogue content.

[0083] Optionally, the step of obtaining the target reader's language style information includes: obtaining the target reader's historical speech content; and performing language style recognition on the historical speech content using a preset language style recognition algorithm to obtain the target reader's language style information.

[0084] Optionally, in response to a summary generation instruction for the target conversation content, the step of obtaining the target conversation content includes: in response to an arrival instruction of a preset time length, determining whether new speech content is generated in the target group within the preset time length; wherein all members of the target group can serve as speakers; if new speech content is generated, then the new speech content generated within the preset time length is determined as the target conversation content.

[0085] This embodiment further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the above-mentioned method for summarizing and generating the conversation content, for example:

[0086] In response to a summary generation instruction for a target conversation content, the target conversation content is obtained; wherein the target conversation content includes speech content of multiple speakers; topic recognition is performed on the target conversation content to obtain initial topic data of at least one initial topic; based on the correlation relationship between the speech content of multiple speakers, the initial topic data of at least one initial topic is merged to obtain target topic data of at least one target topic; natural language processing is performed on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

[0087] In this method, initial topic data is extracted from the conversation content, and then the topic data with correlation relationship with the speech content is merged, and finally natural language processing is performed to obtain a summary sentence close to natural language expression, so that the summary of the conversation content not only contains keywords, but also improves the comprehensiveness and readability of the summary of the conversation content.

[0088] Optionally, based on the correlation between the speech contents of multiple speakers, the initial topic data of at least one initial topic are merged to obtain the target topic data of at least one target topic, including: obtaining multiple first speech contents corresponding to each initial topic; identifying the correlation between the first speech contents of any initial topics, and determining at least one group of initial topics in which the first speech contents have a correlation; and fusing the initial topic data in each group of initial topics to obtain the target topic data of at least one target topic.

[0089] Optionally, the step of identifying the association relationship between the first speech contents between any initial topics and determining at least one group of initial topics with which the first speech contents are associated includes: identifying conversation chains on the target conversation contents based on the reference relationship and / or forwarding relationship between the speech contents in the target conversation contents to obtain at least one conversation chain in the target conversation contents; each conversation chain includes multiple speech contents; and determining the initial topics corresponding to the first speech contents belonging to the same conversation chain as initial topics with which they are associated to obtain at least one group of initial topics.

[0090] Optionally, before the step of performing conversation chain identification on the target conversation content based on the reference relationship and / or forwarding relationship between the speeches in the target conversation content to obtain at least one conversation chain in the target conversation content, the method further includes: determining the reference relationship and / or forwarding relationship between the speeches in the target conversation content based on the reference marks and forwarding marks corresponding to the speeches in the target conversation content.

[0091] Optionally, the step of fusing the initial topic data in each group of initial topics to obtain target topic data of at least one target topic includes: performing semantic fusion on the initial topic data in each group of initial topics through a deep learning model to obtain target topic data of at least one target topic.

[0092] Optionally, the step of performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic includes: extracting topic features of the target field for each speech content in the target conversation content by a preset topic feature extraction algorithm to obtain a topic feature vector corresponding to each speech content; and clustering all topic feature vectors by a preset clustering algorithm to obtain initial topic data of at least one initial topic.

[0093] Optionally, the step of performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target dialogue content includes: obtaining language style information of the target reader; wherein the target summary sentence is generated for the target reader; performing natural language conversion on the target topic data of each target topic through a preset natural language generation algorithm to obtain a first summary sentence of the target dialogue content; and adjusting the language style of the first summary sentence according to the language style information of the target reader to obtain the target summary sentence of the target dialogue content.

[0094] Optionally, the step of obtaining the target reader's language style information includes: obtaining the target reader's historical speech content; and performing language style recognition on the historical speech content using a preset language style recognition algorithm to obtain the target reader's language style information.

[0095] Optionally, in response to a summary generation instruction for the target conversation content, the step of obtaining the target conversation content includes: in response to an arrival instruction of a preset time length, determining whether new speech content is generated in the target group within the preset time length; wherein all members of the target group can serve as speakers; if new speech content is generated, then the new speech content generated within the preset time length is determined as the target conversation content.

[0096] The computer program product of the conversation content summary generation method, device, electronic device and storage medium provided in the embodiments of the present disclosure includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be found in the method embodiments, which will not be repeated here.

[0097] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0098] In addition, in the description of the embodiments of the present disclosure, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present disclosure can be understood according to specific circumstances.

[0099] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0100] In the description of the present disclosure, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present disclosure and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present disclosure. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0101] Finally, it should be noted that the above embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above embodiments, those skilled in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for generating a summary of a conversation content, characterized in that: The method comprises: In response to a summary generation instruction for target conversation content, the target conversation content is acquired; wherein the target conversation content includes speech contents of multiple speakers; Performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic; According to the correlation relationship between the speech contents of the multiple speakers, the initial topic data of the at least one initial topic are merged to obtain target topic data of at least one target topic; Natural language processing is performed on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

2. The method according to claim 1, characterized in that The step of fusing the initial topic data of the at least one initial topic according to the correlation relationship between the speech contents of the multiple speakers to obtain the target topic data of at least one target topic includes: Obtain multiple first speech contents corresponding to each initial topic; Identify the association relationship between the first speech contents of any initial topics, and determine at least one group of initial topics in which the first speech contents have an association relationship; The initial topic data in each group of initial topics are merged to obtain target topic data of at least one target topic.

3. The method according to claim 2, characterized in that The step of identifying the association relationship between the first speech contents of any initial topics and determining at least one group of initial topics in which the first speech contents have an association relationship includes: According to the reference relationship and / or forwarding relationship between the speech contents in the target conversation content, the target conversation content is identified as a conversation chain to obtain at least one conversation chain in the target conversation content; each conversation chain includes a plurality of speech contents; Initial topics corresponding to the first speech contents belonging to the same dialogue chain are determined as initial topics with associated relationships, so as to obtain at least one group of initial topics.

4. The method according to claim 3, characterized in that: Before the step of performing dialog chain identification on the target dialog content according to the reference relationship and / or forwarding relationship between speech contents in the target dialog content to obtain at least one dialog chain in the target dialog content, the method further includes: According to the reference marks and / or forwarding marks corresponding to the speech contents in the target dialogue content, the reference relationship and / or forwarding relationship between the speech contents in the target dialogue content is determined.

5. The method according to claim 2, characterized in that: The step of fusing the initial topic data in each group of initial topics to obtain target topic data of at least one target topic includes: The initial topic data in each group of initial topics are semantically fused through a deep learning model to obtain target topic data of at least one target topic.

6. The method according to claim 1, characterized in that The step of performing topic identification on the target conversation content to obtain initial topic data of at least one initial topic includes: Extracting the topic features of the target domain from each speech content in the target conversation content by using a preset topic feature extraction algorithm to obtain a topic feature vector corresponding to each speech content; All topic feature vectors are clustered using a preset clustering algorithm to obtain initial topic data of at least one initial topic.

7. The method according to claim 1, characterized in that The step of performing natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content includes: Acquiring language style information of a target reader; wherein the target summary sentence is generated for the target reader; Performing natural language conversion on the target topic data of each target topic by using a preset natural language generation algorithm to obtain a first summary sentence of the target conversation content; The language style of the first summary sentence is adjusted according to the language style information of the target reader to obtain a target summary sentence of the target dialogue content.

8. The method according to claim 7, characterized in that Steps to obtain information about the target readers' language style include: Obtain historical speech content of target readers; The language style of the historical speech content is identified by a preset language style identification algorithm to obtain the language style information of the target readers.

9. The method according to claim 1, characterized in that: In response to a summary generation instruction for a target conversation content, the step of acquiring the target conversation content includes: In response to a preset time arrival instruction, determining whether new speech content is generated in the target group within the preset time; wherein all members of the target group can serve as speakers; If new speech content is generated, the new speech content generated within the preset time period is determined as the target dialogue content.

10. A device for generating a summary of conversation content, characterized in that: The device comprises: A response module, configured to obtain the target conversation content in response to a summary generation instruction for the target conversation content; wherein the target conversation content includes speech contents of multiple speakers; an identification module, configured to perform topic identification on the target conversation content to obtain initial topic data of at least one initial topic; A fusion module, configured to fuse the initial topic data of the at least one initial topic according to the correlation relationship between the speech contents of the multiple speakers, to obtain target topic data of at least one target topic; The processing module is used to perform natural language processing on the target topic data of each target topic to obtain a target summary sentence of the target conversation content.

11. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the method for summarizing and generating conversation content according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are called and executed by the processor, the computer-executable instructions prompt the processor to implement the method for summarizing and generating conversation content according to any one of claims 1 to 9.

Images