Conversation summary extraction method, device and storage medium

Abstract

The embodiment of the present invention provides a conversation summary extraction method, device and storage medium, which identifies the intent of each message in a conversation to be processed; obtains the first frequency and the second frequency of any intent, and obtains the evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is the frequency of the intent appearing in all intents of the conversation to be processed, and the second frequency is the frequency of the intent appearing in all intents of the global conversation within a predetermined period; and determines the summary of the conversation to be processed according to the evaluation score of each intent in the conversation to be processed. The embodiment of the present invention obtains the evaluation score for each intent in the conversation to be processed, and the evaluation score reflects the importance of each intent in the conversation to be processed and the uniqueness in the global conversation. The summary of the conversation to be processed is determined based on the evaluation score, which can filter out the intents with low importance in the conversation to be processed and low uniqueness in the global conversation, thereby improving the accuracy of extracting the conversation summary.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of Internet technology, and in particular to a conversation summary extraction method, device and storage medium. Background Art

[0002] In an online conversation, there are two parties, one asking question and one answering question. Both parties will continuously send one or more messages to each other, and a conversation is formed based on these messages. For multiple messages in a conversation, extracting intent classification from these messages and automatically and accurately summarizing the important information covered in this conversation based on various intents in the conversation is a very worthy issue in the industry.

[0003] In the prior art, after classifying the intent of each message in a conversation, a sequential sorting method, a frequency sorting method, or a comprehensive sorting method combining the above sorting methods is usually used to determine one or more intents as a summary of the conversation. The sequential sorting method includes a pre-order sorting method, a post-order sorting method, etc.

[0004] Existing conversation summary extraction methods only determine the conversation summary based on the position and / or frequency of the intent in the conversation. However, if a certain intent is high in frequency due to repeated message sending, or there may be common high-frequency intents on the conversation platform, it is not suitable as the conversation summary. Or some intents at specific positions are not suitable as the conversation summary. Existing conversation summary extraction methods may determine these intents as the conversation summary, resulting in inaccurate conversation summary, and thus unable to accurately determine the important information covered by the conversation based on the conversation summary. Summary of the invention

[0005] The embodiments of the present invention provide a conversation summary extraction method, device and storage medium, which are used to improve the accuracy of extracting conversation summaries, and further accurately determine important information covered by the conversation according to the conversation summaries.

[0006] In a first aspect, an embodiment of the present invention provides a method for extracting a conversation summary, comprising:

[0007] Identify the intent corresponding to each message in the pending session;

[0008] Obtaining a first frequency and a second frequency of any intent in the session to be processed, and obtaining an evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is a frequency at which the intent appears in all intents in the session to be processed, and the second frequency is a frequency at which the intent appears in all intents in the global session within a predetermined time period;

[0009] A summary of the session to be processed is determined according to the evaluation score of each intent in the session to be processed.

[0010] In a second aspect, an embodiment of the present invention provides a conversation summary extraction device, including:

[0011] An intent recognition module is used to identify the intent corresponding to each message in the session to be processed;

[0012] a processing module, configured to obtain a first frequency and a second frequency of any intent in the session to be processed, and obtain an evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is a frequency at which the intent appears in all intents in the session to be processed, and the second frequency is a frequency at which the intent appears in all intents in the global session within a predetermined period of time;

[0013] The summary determination module is used to determine the summary of the session to be processed according to the evaluation score of each intention in the session to be processed.

[0014] In a third aspect, an embodiment of the present invention provides an electronic device, including: at least one processor; and a memory;

[0015] The memory stores computer-executable instructions;

[0016] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method according to the first aspect.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the method described in the first aspect is implemented.

[0018] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program, which implements the method described in the first aspect when executed by a processor.

[0019] The conversation summary extraction method, device and storage medium provided by the embodiment of the present invention identify the intent corresponding to each message in the conversation to be processed; obtain the first frequency and the second frequency of any intent in the conversation to be processed, and obtain the evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is the frequency of the intent appearing in all the intents in the conversation to be processed, and the second frequency is the frequency of the intent appearing in all the intents in the global conversation within a predetermined period; and determine the summary of the conversation to be processed according to the evaluation score of each intent in the conversation to be processed. The embodiment of the present invention obtains an evaluation score for each intent in the conversation to be processed, and reflects the importance of each intent in the conversation to be processed and the uniqueness in the global conversation through the evaluation score, and then determines the summary of the conversation to be processed based on the evaluation score, which can screen out the intents with low importance in the conversation to be processed and low uniqueness in the global conversation, improve the accuracy of extracting the conversation summary, and then determine the important information covered by the conversation according to the conversation summary, and also facilitates distinguishing different conversations. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0021] Figure 1 A schematic diagram of an application scenario of a conversation summary extraction method provided by an embodiment of the present invention;

[0022] Figure 2 A flowchart of a method for extracting a conversation summary provided by an embodiment of the present invention;

[0023] Figure 3 A flowchart of a conversation summary extraction method provided by another embodiment of the present invention;

[0024] Figure 4 A flowchart of a conversation summary extraction method provided by another embodiment of the present invention;

[0025] Figure 5 A flowchart of a conversation summary extraction method provided by another embodiment of the present invention;

[0026] Figure 6 A structural diagram of a conversation summary extraction device provided by an embodiment of the present invention;

[0027] Figure 7 A structural diagram of an electronic device for executing a method for extracting a conversation summary provided by an embodiment of the present invention.

[0028] The above drawings show clear embodiments of the present disclosure, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present disclosure in any way, but to illustrate the concepts of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0029] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0030] In the prior art, after classifying the intent of each message in a conversation, a sequential sorting method, a frequency sorting method, or a comprehensive sorting method combining the above sorting methods is usually used to determine one or more intents as a summary of the conversation. The sequential sorting method includes a pre-order sorting method, a post-order sorting method, etc.

[0031] Specifically, the basic assumption of the pre-order sorting method is that for a conversation, the first intent is the most important, so the first N intents are selected as the summary of the conversation.

[0032] The basic assumption of the post-order sorting method is that the last intent in a conversation is the most important, so the last N intents are selected as the summary of the conversation.

[0033] The basic assumption of the frequency ranking method is that for a conversation, the more frequently occurring intent is more important. Therefore, the top N intents ranked from high to low in frequency are selected as the summary of the conversation.

[0034] The comprehensive ranking method combines the order of intent appearance and the corresponding frequency, multiplies it by the relevant weight to obtain a comprehensive score, and then further selects the top N intents as the summary.

[0035] Among the existing conversation summary extraction methods, the sequential sorting method only determines the conversation summary based on the position of the intent in the current conversation, the frequency sorting method only determines the conversation summary based on the frequency of the intent in the current conversation, and the comprehensive sorting method only determines the conversation summary based on the position and frequency of the intent in the current conversation. There are the following defects:

[0036] For the pre-order sorting method and the post-order sorting method, if the most important information during the conversation may exist in the middle of the conversation or other positions, in this case, it is difficult to obtain the most important intention as a summary.

[0037] For the frequency sorting method, high-frequency intents can be obtained. If there is a situation of repeated message sending, the high-frequency intent may be mistakenly extracted as the important content of the conversation.

[0038] In addition, both the order sorting method and the frequency sorting method are only compared within the conversation, without comparing the overall situation of the conversation platform. When there are common high-frequency intents on the conversation platform or intents concentrated in a specific position of the conversation (such as greetings and requesting manual services at the beginning of the conversation, and thank you and requesting evaluations at the end of the conversation), such intents can usually be ignored and cannot be used as summary features of the conversation. However, the order sorting and frequency sorting methods often cannot filter out these intents, and extract the common features as summaries, making it difficult to distinguish the differences between conversations through summaries.

[0039] Therefore, the conversation summary extraction method in the prior art has the problem of extracting inaccurate conversation summaries, and thus cannot accurately determine the important information covered by the conversation based on the conversation summary.

[0040] In view of the above technical problems of the prior art, in the embodiment of the present disclosure, after identifying the intent corresponding to each message in the to-be-processed session, the frequency of occurrence of any intent in the to-be-processed session in all intents in the to-be-processed session (recorded as the first frequency) and the frequency of occurrence in all intents in the global session within a predetermined period of time (recorded as the second frequency) can be obtained, and then the evaluation score of the intent is obtained based on the first frequency and the second frequency of the intent, wherein the first frequency can reflect the importance of the intent in the to-be-processed session, and the second frequency can reflect the uniqueness of the intent in the global session, so the evaluation score can comprehensively reflect the intent. The importance of the intent in the session to be processed and the uniqueness of the intent in the global session. The higher the evaluation score, the higher the importance of the intent in the session to be processed and the uniqueness of the intent in the global session. The lower the evaluation score, the lower the importance of the intent in the session to be processed and the uniqueness of the intent in the global session. Then, the intents in the session to be processed can be sorted according to the evaluation score of each intent, and the summary of the session to be processed can be determined according to the sorting, which improves the accuracy of extracting the session summary, and then the important information covered by the session can be determined according to the session summary, which is also convenient for distinguishing different sessions.

[0041] The specific application scenarios of the conversation summary extraction method provided by the embodiment of the present invention are as follows: Figure 1As shown, it includes terminal device 101a, terminal device 101b, server 102, and database 103. When terminal device 101a and terminal device 101b have a conversation, the conversation message of the conversation can be obtained through server 102, and server 102 can also store the conversation message of the conversation in database 103. When server 102 needs to extract a conversation summary for a conversation to be processed, it can identify the intent corresponding to each message in the conversation to be processed, obtain the first frequency and the second frequency of any intent in the conversation to be processed, and obtain the evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is the frequency of the intent appearing in all intents of the conversation to be processed, and the second frequency is the frequency of the intent appearing in all intents of the global conversation within a predetermined period; and determine the summary of the conversation to be processed according to the evaluation score of each intent in the conversation to be processed.

[0042] The following specific embodiments are used to describe in detail the technical solutions of the present invention and how the technical solutions of the present application solve the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0043] Figure 2 This is a flow chart of a conversation summary extraction method provided by an embodiment of the present invention. This embodiment provides a conversation summary extraction method, the execution subject of which can be any electronic device, and the specific steps of the conversation summary extraction method are as follows:

[0044] S201. Identify the intent corresponding to each message in the conversation to be processed.

[0045] In this embodiment, the conversation to be processed can be any conversation, such as a conversation between a user and customer service, a conversation between a consumer and a merchant, etc. The conversation to be processed includes multiple messages, and any existing intent recognition method can be used to identify the intent corresponding to each message, such as a natural language recognition algorithm, which is not limited here.

[0046] S202: Obtain a first frequency and a second frequency of any intention in the conversation to be processed, and obtain an evaluation score of the intention according to the first frequency and the second frequency;

[0047] The first frequency is the frequency at which the intent appears in all intents of the to-be-processed session, and the second frequency is the frequency at which the intent appears in all intents of the global session within a predetermined period of time.

[0048] In this embodiment, for any intent in the session to be processed, its frequency of appearance in all intents of the session to be processed (recorded as the first frequency) and its frequency of appearance in all intents of the global session within a predetermined period of time (recorded as the second frequency) can be obtained, and then the evaluation score of the intent is obtained based on the first frequency and the second frequency of the intent.

[0049] The first frequency may reflect the importance of the intent in the session to be processed, and the second frequency may reflect the uniqueness of the intent in the global session, where the global session is all sessions within a predetermined period, and the predetermined period may be the period in which the session to be processed is located, or any other period. Therefore, the evaluation score is related to the first frequency and the second frequency, and thus the evaluation score may comprehensively reflect the importance of the intent in the session to be processed and the uniqueness of the intent in the global session, and the higher the evaluation score, the higher the importance of the intent in the session to be processed and the uniqueness of the intent in the global session, and the lower the evaluation score, the lower the importance of the intent in the session to be processed and the uniqueness of the intent in the global session.

[0050] S203: Determine a summary of the session to be processed according to the evaluation score of each intention in the session to be processed.

[0051] In this embodiment, after obtaining the evaluation score of each intention in the session to be processed, the summary of the session to be processed may be determined according to the evaluation score of each intention.

[0052] Optionally, the intents in the pending session may be sorted in descending order according to the evaluation score of each intent in the pending session; from all the intents in the pending session, the first preset number of intents are selected according to the sorting, and determined as the summary of the pending session. The first preset number of intents are the preset number of intents with the highest evaluation scores. These intents are highly important in the pending session and unique in the global session, and are representative to a certain extent, including important information covered by the pending session. The preset number can be set according to actual needs.

[0053] Of course, the intents in the to-be-processed session may also be sorted in ascending order according to the evaluation score of each intent in the to-be-processed session; from all the intents in the to-be-processed session, the intents that are arranged at the rear by a preset number according to the sorting are selected to be determined as the summary of the to-be-processed session. Alternatively, the summary of the to-be-processed session may also be determined in other ways based on the evaluation score of each intent in the to-be-processed session, which will not be described in detail here.

[0054] The conversation summary extraction method provided in the present embodiment identifies the intent corresponding to each message in the conversation to be processed; obtains the first frequency and the second frequency of any intent in the conversation to be processed, and obtains the evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is the frequency of the intent appearing in all the intents in the conversation to be processed, and the second frequency is the frequency of the intent appearing in all the intents in the global conversation within a predetermined period; and determines the summary of the conversation to be processed according to the evaluation score of each intent in the conversation to be processed. The present embodiment obtains an evaluation score for each intent in the conversation to be processed, and the evaluation score reflects the importance of each intent in the conversation to be processed and the uniqueness in the global conversation, and then determines the summary of the conversation to be processed based on the evaluation score, thereby screening out intents with low importance in the conversation to be processed and low uniqueness in the global conversation, improving the accuracy of extracting the conversation summary, and then determining the important information covered by the conversation according to the conversation summary, and facilitating the distinction between different conversations.

[0055] In an optional embodiment, based on the above embodiment, the step of obtaining the evaluation score of the intention according to the first frequency and the second frequency in S202 may specifically include:

[0056] A first ratio of the first frequency to the second frequency is obtained, and the first ratio is determined as an evaluation score of the intention.

[0057] In this embodiment, the evaluation score of the intent is equal to the ratio of the first frequency to the second frequency (recorded as the first ratio). The larger the first frequency and the smaller the second frequency, the larger the evaluation score, that is, the larger the frequency of occurrence in all intents of the to-be-processed session (the greater the importance of the intent in the to-be-processed session), the smaller the frequency of occurrence in all intents of the global session within a predetermined period (the smaller the uniqueness of the intent in the global session), the larger the evaluation score; on the contrary, the smaller the first frequency and the larger the second frequency, the smaller the evaluation score, that is, the smaller the frequency of occurrence in all intents of the to-be-processed session (the smaller the importance of the intent in the to-be-processed session), the greater the frequency of occurrence in all intents of the global session within a predetermined period (the greater the uniqueness of the intent in the global session), the smaller the evaluation score. In this way, the intents with low importance in the to-be-processed session and low uniqueness in the global session can be screened out based on the evaluation score.

[0058] In another alternative embodiment, if Figure 3 As shown, based on the above embodiment, the step of obtaining the evaluation score of the intention according to the first frequency and the second frequency in S202 may specifically include:

[0059] S301, obtaining location information of the intention in the session to be processed, and obtaining a corresponding preset location weight according to the location information;

[0060] S302, obtaining a first ratio of the first frequency to the second frequency;

[0061] S303: Obtain a product of the first ratio and the preset position weight, and determine the product as an evaluation score of the intention.

[0062] In this embodiment, considering that some specific intentions may be concentrated in certain specific positions of the conversation, while the intentions at some specific positions are more important, and the intentions at some specific positions are less important, different preset position weights can be pre-configured for the intentions at different positions of the conversation, and the preset position weights can be set according to actual conditions.

[0063] For any intent in the session to be processed, when obtaining the evaluation score of the intent, the location information of the intent in the session to be processed can be obtained, and then the corresponding preset location weight can be obtained according to the location information; the evaluation score of the intent is determined based on the preset location weight, the first frequency and the second frequency: that is, the evaluation score of the intent = (first frequency / second frequency) * preset location weight. By considering the location weight in the evaluation score, the evaluation score of the intent at an important position can be higher, while the evaluation score of the intent at an unimportant position can be lower, so as to avoid that some intents at unimportant positions are determined as session summaries.

[0064] It should be noted that, in this embodiment, the order of S301 and S302 is not limited, and they can be executed successively or simultaneously.

[0065] Based on any of the above embodiments, Figure 4 As shown, the first frequency of any intent in the to-be-processed session can be obtained through the following process, specifically including:

[0066] S401, obtaining the frequency of occurrence of the intent in all intents of the session to be processed;

[0067] S402, obtaining the total frequency of all the intentions of the session to be processed;

[0068] S403: Obtain a second ratio of the frequency to the total frequency, and determine the second ratio as the first frequency of the intention.

[0069] In this embodiment, after identifying the intent corresponding to each message in the session to be processed, the frequency of each intent appearing in all the intents of the session to be processed can be counted, and the total frequency of all the intents of the session to be processed can also be counted. For any intent, the frequency of the intent appearing in all the intents of the session to be processed can be calculated based on the frequency of the intent appearing in all the intents of the session to be processed and the total frequency of all the intents of the session to be processed, that is, the first frequency, and the formula is: first frequency = frequency of the intent appearing in all the intents of the session to be processed / total frequency of all the intents of the session to be processed, that is, the second ratio. It should be noted that the order of S401 and S402 is not limited in this embodiment, and they can be executed one after the other or simultaneously.

[0070] Based on any of the above embodiments, Figure 5 As shown, the second frequency of any intent in the to-be-processed session can be obtained through the following process, specifically including:

[0071] S501, obtaining the number of sessions in which the intent appears in the global session within a predetermined period of time;

[0072] S502, obtaining the total number of global sessions within a predetermined period;

[0073] S503: Obtain a third ratio of the sum of the number of sessions in which the intent occurs and a preset value to the total number of sessions, and determine the third ratio as a second frequency of the intent, wherein the preset value is a positive number.

[0074] In this embodiment, for any intent, the number of sessions in which the intent appears in the global session within the predetermined period can be counted, and the total number of sessions in the global session within the predetermined period can also be counted. When calculating the second frequency, that is, the frequency of appearance in all intents of the global session within the predetermined period, it is usually sufficient to calculate the ratio of the number of sessions in which the intent appears in the global session within the predetermined period to the total number of sessions in the global session within the predetermined period. However, considering that the first ratio of the first frequency and the second frequency needs to be obtained in the above embodiment, and the number of sessions in which the intent appears in the global session within the predetermined period may be 0, in order to avoid the denominator of the first ratio of the first frequency and the second frequency being 0, a preset value is added to the numerator of the second frequency, and the preset value is a positive number, that is, the second frequency of the intent = (the number of sessions in which the intent appears in the global session within the predetermined period + the preset value) / the total number of sessions in the global session within the predetermined period, that is, the third ratio. Optionally, the preset value is 1, then the second frequency of the intent = (the number of sessions in which the intent appears in the global session within the predetermined period + 1) / the total number of sessions in the global session within the predetermined period.

[0075] Based on any of the above embodiments, after determining the summary of the session to be processed, the following may also be included:

[0076] According to the summary of the session to be processed, tag information is set for relevant users in the session to be processed.

[0077] In this embodiment, after obtaining the summary of the session to be processed, the content that the relevant users are concerned about can be understood. Therefore, tag information can be set for the relevant users in the session to be processed based on the summary of the session to be processed, so as to facilitate the subsequent classification of users based on the tag information, or push information based on the tag information, etc. Of course, user big data analysis, business information insights, etc. can also be achieved based on the session summary.

[0078] Figure 6 1 is a structural diagram of a conversation summary extraction device according to an embodiment of the present invention. The conversation summary extraction device provided in this embodiment can execute the processing flow provided in the method embodiment, such as Figure 6 As shown, the conversation summary extraction device 600 includes an intention recognition module 601 , a processing module 602 , and a summary determination module 603 .

[0079] Wherein, the intention identification module 601 is used to identify the intention corresponding to each message in the conversation to be processed;

[0080] The processing module 602 is configured to obtain a first frequency and a second frequency of any intent in the session to be processed, and obtain an evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is a frequency at which the intent appears in all intents in the session to be processed, and the second frequency is a frequency at which the intent appears in all intents in the global session within a predetermined period of time;

[0081] The summary determination module 603 is used to determine the summary of the session to be processed according to the evaluation score of each intent in the session to be processed.

[0082] In an optional embodiment, when acquiring the evaluation score of the intention according to the first frequency and the second frequency, the processing module 602 is configured to:

[0083] A first ratio of the first frequency to the second frequency is obtained, and the first ratio is determined as an evaluation score of the intention.

[0084] In another optional embodiment, when acquiring the evaluation score of the intention according to the first frequency and the second frequency, the processing module 602 is configured to:

[0085] Obtaining location information of the intention in the session to be processed, and obtaining a corresponding preset location weight according to the location information;

[0086] Acquire a first ratio of the first frequency to the second frequency;

[0087] The product of the first ratio and the preset position weight is obtained, and the product is determined as the evaluation score of the intention.

[0088] Based on any of the above embodiments, when acquiring the first frequency of any intention in the conversation to be processed, the processing module 602 is used to:

[0089] Obtaining the frequency of occurrence of the intent in all intents of the session to be processed;

[0090] Obtain the total frequency of all intents of the session to be processed;

[0091] A second ratio of the frequency to the total frequency is obtained, and the second ratio is determined as the first frequency of the intention.

[0092] Based on any of the above embodiments, when acquiring the second frequency of any intention in the conversation to be processed, the processing module 602 is used to:

[0093] Obtain the number of sessions in which the intent appears in the global session within a predetermined period;

[0094] Get the total number of global sessions within the scheduled time period;

[0095] Obtain a third ratio of the sum of the number of sessions in which the intent occurs and a preset value to the total number of sessions, and determine the third ratio as a second frequency of the intent, wherein the preset value is a positive number.

[0096] Based on any of the above embodiments, when determining the summary of the session to be processed according to the evaluation score of each intent in the session to be processed, the summary determination module 603 is used to:

[0097] Sort the intents in descending order according to the evaluation score of each intent in the session to be processed;

[0098] From all the intents of the session to be processed, the intents arranged in front of a preset number are selected according to the sorting, and determined as the summary of the session to be processed.

[0099] Based on any of the above embodiments, after determining the summary of the session to be processed, the summary determination module 603 is further configured to:

[0100] According to the summary of the session to be processed, tag information is set for relevant users in the session to be processed.

[0101] The conversation summary extraction device provided in the embodiment of the present invention can be specifically used to perform the above Figure 2-5The specific functions of the provided method embodiment will not be repeated here.

[0102] The conversation summary extraction device provided by the embodiment of the present invention identifies the intent corresponding to each message in the conversation to be processed; obtains the first frequency and the second frequency of any intent in the conversation to be processed, and obtains the evaluation score of the intent according to the first frequency and the second frequency; wherein the first frequency is the frequency of the intent appearing in all the intents in the conversation to be processed, and the second frequency is the frequency of the intent appearing in all the intents in the global conversation within a predetermined period; and determines the summary of the conversation to be processed according to the evaluation score of each intent in the conversation to be processed. This embodiment obtains an evaluation score for each intent in the conversation to be processed, and reflects the importance of each intent in the conversation to be processed and the uniqueness in the global conversation through the evaluation score, and then determines the summary of the conversation to be processed based on the evaluation score, thereby screening out intents with low importance in the conversation to be processed and low uniqueness in the global conversation, improving the accuracy of extracting the conversation summary, and then determining the important information covered by the conversation according to the conversation summary, and facilitating the distinction between different conversations.

[0103] Figure 7 The electronic device provided by the embodiment of the present invention can execute the processing flow provided by the embodiment of the method for extracting a session summary, such as Figure 7 As shown, the electronic device 70 includes a memory 71, a processor 72, and a computer program; wherein the computer program is stored in the memory 71 and is configured to execute the conversation summary extraction method described in the above embodiment by the processor 72. In addition, the electronic device 70 may also have a communication interface 73 for receiving control instructions and / or data.

[0104] Figure 7 The electronic device of the illustrated embodiment can be used to execute the technical solution of the above-mentioned conversation summary extraction method embodiment, and its implementation principle and technical effect are similar and will not be repeated here.

[0105] In addition, this embodiment further provides a non-transitory computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to implement the conversation summary extraction method described in the above embodiment.

[0106] In addition, this embodiment further provides a computer program product, including a computer program, which implements the conversation summary extraction method described in the above embodiment when executed by a processor.

[0107] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0109] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0110] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), online storage, disk or optical disk and other media that can store program codes.

[0111] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0112] The above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, rather than to limit them. Although the embodiments of the present invention are described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

[0113] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be regarded as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0114] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A conversation summary extraction method, characterized in that: include: Identify the intent corresponding to each message in the pending session; Acquire a first frequency and a second frequency of any intention in the to-be-processed session, and acquire an evaluation score of the intention according to the first frequency and the second frequency; The first frequency is the frequency of the intent appearing in all intents of the to-be-processed session, the second frequency is the frequency of the intent appearing in all intents of the global session within a predetermined period, and the global session is all sessions within the predetermined period; A summary of the session to be processed is determined according to the evaluation score of each intent in the session to be processed.

2. The method according to claim 1, characterized in that The acquiring the evaluation score of the intention according to the first frequency and the second frequency includes: A first ratio of the first frequency to the second frequency is obtained, and the first ratio is determined as an evaluation score of the intention.

3. The method according to claim 1, characterized in that The acquiring the evaluation score of the intention according to the first frequency and the second frequency includes: Obtaining location information of the intention in the session to be processed, and obtaining a corresponding preset location weight according to the location information; Acquire a first ratio of the first frequency to the second frequency; The product of the first ratio and the preset position weight is obtained, and the product is determined as the evaluation score of the intention.

4. The method according to claim 2 or 3, characterized in that: Obtaining a first frequency of any intent in the session to be processed includes: Obtaining the frequency of occurrence of the intent in all intents of the session to be processed; Obtain the total frequency of all intents of the session to be processed; A second ratio of the frequency to the total frequency is obtained, and the second ratio is determined as the first frequency of the intention.

5. The method according to claim 2 or 3, characterized in that: Obtaining a second frequency of any intent in the session to be processed includes: Obtain the number of sessions in which the intent appears in the global session within a predetermined period; Get the total number of global sessions within the scheduled time period; Obtain a third ratio of the sum of the number of sessions in which the intent occurs and a preset value to the total number of sessions, and determine the third ratio as a second frequency of the intent, wherein the preset value is a positive number.

6. The method according to any one of claims 1 to 3, characterized in that: The step of determining a summary of the session to be processed according to the evaluation score of each intent in the session to be processed includes: Sort the intents in descending order according to the evaluation score of each intent in the session to be processed; From all the intents of the session to be processed, the intents arranged in front of a preset number are selected according to the sorting, and determined as the summary of the session to be processed.

7. The method according to any one of claims 1 to 3, characterized in that: After determining the summary of the session to be processed, the method further includes: According to the summary of the session to be processed, tag information is set for relevant users in the session to be processed.

8. A conversation summary extraction device, characterized in that: include: An intent recognition module is used to identify the intent corresponding to each message in the session to be processed; A processing module, configured to obtain a first frequency and a second frequency of any intention in the conversation to be processed, and obtain an evaluation score of the intention according to the first frequency and the second frequency; The first frequency is the frequency of the intent appearing in all intents of the to-be-processed session, the second frequency is the frequency of the intent appearing in all intents of the global session within a predetermined period, and the global session is all sessions within the predetermined period; The summary determination module is used to determine the summary of the session to be processed according to the evaluation score of each intention in the session to be processed.

9. The device according to claim 8, characterized in that When acquiring the evaluation score of the intention according to the first frequency and the second frequency, the processing module is used to: A first ratio of the first frequency to the second frequency is obtained, and the first ratio is determined as an evaluation score of the intention.

10. The device according to claim 8, characterized in that When acquiring the evaluation score of the intention according to the first frequency and the second frequency, the processing module is used to: Obtaining location information of the intention in the session to be processed, and obtaining a corresponding preset location weight according to the location information; Acquire a first ratio of the first frequency to the second frequency; The product of the first ratio and the preset position weight is obtained, and the product is determined as the evaluation score of the intention.

11. The device according to claim 9 or 10, characterized in that When acquiring the first frequency of any intention in the to-be-processed session, the processing module is used to: Obtaining the frequency of occurrence of the intent in all intents of the session to be processed; Obtain the total frequency of all intents of the session to be processed; A second ratio of the frequency to the total frequency is obtained, and the second ratio is determined as the first frequency of the intention.

12. The device according to claim 9 or 10, characterized in that When acquiring the second frequency of any intention in the to-be-processed session, the processing module is used to: Obtain the number of sessions in which the intent appears in the global session within a predetermined period; Get the total number of global sessions within the scheduled time period; Obtain a third ratio of the sum of the number of sessions in which the intent occurs and a preset value to the total number of sessions, and determine the third ratio as a second frequency of the intent, wherein the preset value is a positive number.

13. The device according to any one of claims 8 to 10, characterized in that When determining the summary of the session to be processed according to the evaluation score of each intention in the session to be processed, the summary determination module is used to: Sort the intents in descending order according to the evaluation score of each intent in the session to be processed; From all the intents of the session to be processed, the intents arranged in front of a preset number are selected according to the sorting, and determined as the summary of the session to be processed.

14. The device according to any one of claims 8 to 10, characterized in that After determining the summary of the session to be processed, the summary determination module is further used to: According to the summary of the session to be processed, tag information is set for relevant users in the session to be processed.

15. An electronic device, characterized in that: include: at least one processor; and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method according to any one of claims 1 to 7.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executable instructions, the method according to any one of claims 1 to 7 is implemented.

17. A computer program product, characterized in that The program product comprises a computer program, which implements the method according to any one of claims 1 to 7 when executed by a processor.

Images