A session analysis method and device, electronic equipment and storage medium

Abstract

The application relates to a conversation analysis method and device, an electronic device and a storage medium, wherein the method comprises the following steps: obtaining target conversation content; segmenting a target conversation set from the target conversation content; determining target correlation relationships between target conversation elements in the target conversation set; determining specified correlation relationships that satisfy preset correlation with the target correlation relationships in all candidate correlation relationships; and determining an association analysis result corresponding to the target conversation set according to the specified conversation set corresponding to each specified conversation element. The method provided in the application can expand conversation sentences on the basis of conversation intent correlation, so that the conversation sentences in the target conversation set are easier to understand, the efficiency and accuracy of communication are improved, and the technical problem that the user's appeal cannot be accurately reflected in the related art can be overcome.

Description

Technical Field

[0001] This application relates to the field of word processing technology, and more particularly to a conversation analysis method and apparatus, electronic device and storage medium. Background Technology

[0002] With the development of internet technology, more and more users are consulting, complaining or marketing online. However, some front-line service personnel, site managers, and analysts lack comprehensive knowledge. As a result, some staff cannot fully understand the needs of consumers based solely on their conversational statements.

[0003] In related technologies, professional customer service representatives manually tag the conversation after completing a customer's inquiry, making it easier for staff in subsequent processes, such as session management, to understand the customer's feedback. With the widespread adoption of WeChat Work, frontline employees will communicate with customers more frequently through WeChat. If manual tagging is also used, it will increase the workload of employees, and tags created by non-professional customer service representatives may not fully reflect the user's needs.

[0004] There is currently no effective solution to the technical problem that the relevant technologies fail to accurately reflect users' needs. Summary of the Invention

[0005] To address the aforementioned technical problem of failing to accurately reflect user demands, this application provides a session analysis method and apparatus, an electronic device, and a storage medium.

[0006] In a first aspect, embodiments of this application provide a session analysis method, including:

[0007] Obtain the target session content, wherein the target session content includes at least one session statement;

[0008] A target session set is segmented from the target session content, wherein all the session statements in the target session set satisfy a preset relationship;

[0009] Determine the target association relationships between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element;

[0010] Among all candidate associations, a specified association that satisfies a preset relevance with the target association is determined;

[0011] Based on the specified session set corresponding to each specified session element, the association analysis result corresponding to the target session set is determined, wherein the specified session element is the session element in the specified association relationship.

[0012] Optionally, as described above, segmenting the target session set from the target session content includes at least one of the following:

[0013] The target session content is segmented according to the keywords in the target session content to obtain at least one set of target sessions;

[0014] The target session content is segmented according to the issuance time of each session statement to obtain at least one target session set, wherein the time interval between any two temporally adjacent session statements in the same target session set is less than a preset interval upper limit.

[0015] Optionally, as described above, determining the target association relationship between each target session element in the target session set includes:

[0016] The target session set is input into a preset classification model to analyze the target session type corresponding to the target session set, wherein the target session type is used to indicate the type of event that the target session set is used to communicate;

[0017] Using a preset algorithm, a target core sentence is determined from all the conversation statements in the target conversation set, wherein the target core sentence is the conversation statement with the highest relevance to the conversation type among all the conversation statements in the target conversation set;

[0018] Based on the target scenario corresponding to the target session content, the target session element type is determined, and all target session elements in the target session set are identified according to the target session element type;

[0019] The target session elements are input into a preset relationship recognition model to obtain the target association relationship between each target session element.

[0020] Optionally, as described above, determining the specified association that satisfies a preset relevance with the target association from all candidate associations includes:

[0021] Determine all candidate session elements included in the knowledge graph that serve as the candidate associations, as well as the candidate associations between all the candidate session elements;

[0022] If all the target session elements are included among all the candidate session elements, and / or the candidate association includes the target association, the knowledge graph is determined as the specified association.

[0023] Optionally, as described above, after determining the specified association that satisfies a preset relevance with the target association from all candidate associations, the method further includes:

[0024] Determine the correspondence between the specified session element and the session element type;

[0025] The target session type, the target core sentence, and the corresponding relationship are used as feedback information for the target session set and output.

[0026] Optionally, as described above, determining the association analysis result corresponding to the target session set based on the specified session set corresponding to each specified session element includes:

[0027] Among all candidate session sets, a first specified session set is determined that includes at least one of the specified session elements, wherein the candidate session set and the target session set are session sets corresponding to the same target scenario;

[0028] Obtain first semantic information corresponding to each of the first specified session sets, and obtain the association analysis result based on the first semantic information.

[0029] Optionally, as described above, after determining the specified association that satisfies a preset relevance with the target association from all candidate associations, the method further includes:

[0030] Based on the target session time of each session statement in the target session set, the target session time range corresponding to the target session set is determined;

[0031] Among all candidate session sets, a second specified session set is determined that the time range of the candidate sessions satisfies the preset time correlation requirement with the time range of the target session, wherein each candidate session set has a corresponding candidate session time range;

[0032] Obtain the second semantic information corresponding to each of the second specified session sets, and obtain the association analysis result based on the second semantic information.

[0033] Secondly, embodiments of this application provide a session analysis apparatus, including:

[0034] The acquisition module is used to acquire target session content, wherein the target session content includes at least one session statement;

[0035] A segmentation module is used to segment a target session set from the target session content, wherein all the session statements in the target session set satisfy a preset relationship;

[0036] The first determining module is used to determine the target association relationship between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element;

[0037] The second determining module is used to determine, from all candidate associations, a specified association that satisfies a preset correlation with the target association;

[0038] The third determining module is used to determine the association analysis result corresponding to the target session set based on the specified session set corresponding to each specified session element, wherein the specified session element is a session element in the specified association relationship.

[0039] Thirdly, embodiments of this application provide an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0040] The memory is used to store computer programs;

[0041] The processor, when executing the computer program, implements the method as described in any of the preceding descriptions.

[0042] Fourthly, embodiments of this application provide a computer-readable storage medium comprising a stored program, wherein the program, when executed, performs the method as described in any of the preceding claims.

[0043] This solution can be applied to prediction and optimization in the field of marketing intelligence technology. Compared with the prior art, the above-mentioned technical solution provided in this application embodiment has the following advantages: The method provided in this application embodiment can infer the target conversation elements in the target conversation set to obtain more specified conversation elements, which can make the specified conversation elements related to the target conversation elements, and thus make the conversation intent also related. Then, based on the specified conversation set corresponding to the specified conversation elements, the association analysis result corresponding to the target conversation set is determined. Furthermore, based on the correlation of conversation intent, the conversation statements can be expanded to make the conversation statements in the target conversation set easier to understand, improve the efficiency and accuracy of communication, and overcome the technical problem in related technologies that cannot accurately reflect the user's needs. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating a session analysis method provided in an embodiment of this application;

[0047] Figure 2 A flowchart illustrating a session analysis method provided in another embodiment of this application;

[0048] Figure 3 A flowchart illustrating a session analysis method provided as an application example of this application;

[0049] Figure 4 A block diagram of a session analysis device provided in an embodiment of this application;

[0050] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] According to one aspect of the embodiments of this application, a session analysis method is provided. Optionally, in this embodiment, the above-described session analysis method can be applied to a hardware environment consisting of a terminal and a server. The server is connected to the terminal via a network and can be used to provide services (such as analysis result analysis services, data storage services, etc.) to the terminal or clients installed on the terminal. A database can be set up on the server or independently of the server to provide data storage services to the server.

[0053] The aforementioned network may include, but is not limited to, at least one of the following: wired network, wireless network. The aforementioned wired network may include, but is not limited to, at least one of the following: wide area network, metropolitan area network, local area network. The aforementioned wireless network may include, but is not limited to, at least one of the following: Wi-Fi (Wireless Fidelity), Bluetooth. The terminal may not be limited to PC, mobile phone, tablet computer, etc.

[0054] The session analysis method of this application embodiment can be executed by a server, a terminal, or both. Alternatively, the session analysis method of this application embodiment can be executed by a client installed on the terminal.

[0055] Taking the session analysis method in this embodiment being executed by the server as an example, such as Figure 1 As shown in the figure, this application provides a session analysis method, including the following steps:

[0056] Step S101: Obtain the target session content, wherein the target session content includes at least one session statement.

[0057] The resource allocation method in this embodiment can be applied to scenarios where it is necessary to identify the analysis results expressed by the target conversation content in the target conversation content composed of one or more conversation statements. Examples include analyzing the content of online user inquiries, online user complaints, and online promotional content from other merchants or individuals. It can also be used for analyzing other types of content. This embodiment uses the analysis of online user complaints as an example to illustrate the above-described conversation analysis method. For other types of conversation content, the above-described conversation analysis method is equally applicable, provided there is no contradiction.

[0058] Taking the analysis of online user complaints as an example, the specific reasons for the user's complaint can be determined by identifying the content of the complaint.

[0059] When users communicate with customer service personnel through chat software, they will communicate with customer service personnel by sending one or more text messages or voice messages to express their needs.

[0060] In this case, the chat window between the user and customer service personnel will display all conversation statements, that is, the content of the target conversation will be displayed.

[0061] The target conversation content can be any chat content obtained by collecting all text or voice information from the chat window. Furthermore, each piece of text or voice information included in the target conversation content can be a single conversational statement.

[0062] Step S102: Segment the target session set from the target session content, wherein all session statements in the target session set satisfy a preset relationship.

[0063] After obtaining the target session content, if the target session content is too large or contains session statements used to consult different questions, the target session content can be segmented to form a target session set. Furthermore, the target session set can include at least one session statement, and all session statements in the target session set satisfy a preset relationship.

[0064] Optionally, the preset relationship can be: different conversation statements are used to inquire about the same question, or different conversation statements are located in the same time period.

[0065] For example, the target session set can be all the session statements used by a user to ask the same question; when the target session content includes a greeting statement 1 (e.g., hello) and a complaint statement 2 and a complaint statement 3, the target session set a (i.e., including session statement 2 and session statement 3) can be obtained by segmenting the content.

[0066] Step S103: Determine the target association relationship between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element.

[0067] After determining the target session set, the session elements can be extracted by using keyword recognition or feature extraction to obtain each session statement in the target session set, thereby obtaining all target session elements of the target session set. Then, based on the relationship between the target session elements, the target relationship can be obtained.

[0068] Optionally, when a user inquires about beauty-related questions, the target conversation element can be brand, category, or product-related information; when a user inquires about dining-related questions, the target conversation element can be restaurant name, coupons, etc.

[0069] For example, when the target session elements include "delivery", "cancellation", and "sundae", the preset model can determine that "cancellation" and "sundae" are associated. That is, in general, the cancellation is for sundae, not delivery. In the target association, "cancellation" and "sundae" are associated, while "delivery" is not associated with either "cancellation" or "sundae".

[0070] Step S104. Among all candidate relationships, identify the specified relationship that satisfies the preset relevance with the target relationship.

[0071] After obtaining the target relationship, since users may not be able to fully express what they want to say when asking questions, there may be some missing content, which may prevent customer service personnel from accurately understanding the true meaning of what the user wants to express. Therefore, from all candidate relationships, the specified relationship that meets the preset relevance with the target relationship is determined.

[0072] Preset relevance can be a criterion used to indicate the correlation between two relationships.

[0073] This ensures that the number of session elements in the candidate association is greater than or equal to the target session element, thereby achieving the purpose of reasoning and expanding the target association.

[0074] For example, if the target session content is "the sundae has melted", then the target session content contains the elements "sundae" and "melted". The associated elements of "sundae" are: associated category "cold drinks", associated brand "XXX", and associated region "XX city YY district"; the associated elements of "melted" are: associated product "ice cone" and "ice cube", and associated reasons for problems such as "too high temperature" and "too long delivery time".

[0075] As an optional implementation, the method described above, among all candidate associations, determines a specified association that satisfies a preset relevance to the target association, including the following steps:

[0076] Step S201: Determine all candidate session elements included in the knowledge graph as candidate associations, as well as the candidate associations between all candidate session elements.

[0077] In order to determine the specified associations that satisfy the preset relevance with the target association, the target session elements can be inferred through the candidate associations in the knowledge graph to obtain other session elements related to the target session set.

[0078] Candidate associations can be the associations between various candidate session elements in a knowledge graph.

[0079] In step S202, if all target session elements are included among all candidate session elements, and / or if the target association is included among the candidate associations, the knowledge graph is identified as the specified association.

[0080] In general, the number of candidate session elements in a knowledge graph is far greater than the number of target session elements.

[0081] Optionally, if the candidate session elements include the target session element, and / or the candidate associations include the target associations, the knowledge graph can be used as the established associations to infer the target associations and obtain more session elements related to the target session element.

[0082] For example, if a knowledge graph includes elements 1, 2, 3, 4, and 5, and element 1 is associated with elements 2 and 3 respectively, and element 3 is associated with elements 1, 4, and 5 respectively, and the target session element only includes element 1, and if all candidate session elements are assumed to include the target session element, then the knowledge graph can be identified as having a specified association relationship based on the fact that the knowledge graph includes the same element 1.

[0083] If the target session element includes related elements 2 and 4, and if the preset candidate association includes the target association, then the knowledge graph will be determined as the specified association. Since elements 2 and 3 in the knowledge graph are not related, the knowledge graph cannot be used as the specified association.

[0084] If a knowledge graph is defined as a specified association only if all candidate session elements include the target session element and the candidate associations include the target association, then when the session "Sundae melted" contains the mutually related target session elements "Sundae" and "melted", and the knowledge graph also includes the mutually related target session elements "Sundae" and "melted", and the related elements of "Sundae" are: related category "cold drinks", related brand "XXX", and related region "YYY", and the related elements of "melted" are: related products "ice cone" and "ice cube", and related reasons for problems such as "too high temperature" and "too long delivery time", then this knowledge graph includes both all target session elements and target associations, and can be used as a specified association.

[0085] The method in this embodiment can quickly determine the specified association relationship corresponding to the target session set, which in turn facilitates reasoning on the session elements included in the target session set to obtain more session elements corresponding to the target session set.

[0086] Step S105. Determine the association analysis results corresponding to the target session set based on the specified session set corresponding to each specified session element, wherein the specified session element is the session element in the specified association relationship.

[0087] After determining the specified association, all specified session elements in the specified association can be identified. Then, the specified session set including each specified session element can be identified. Finally, the semantics to be expressed by the target session set can be determined by combining the specified session set, and the association analysis results can be obtained.

[0088] The results of association analysis can be used to indicate the true intent that users corresponding to a target set of sessions want to express in their communications.

[0089] The method in this embodiment can infer the target session elements in the target session set to obtain more specified session elements. This can make the specified session elements related to the target session elements, and thus make the session intent also related. Then, based on the specified session set corresponding to the specified session elements, the association analysis results corresponding to the target session set can be determined. Furthermore, based on the correlation of session intent, the session statements can be expanded to make the session statements in the target session set easier to understand, thereby improving the efficiency and accuracy of communication.

[0090] As an optional implementation, as described above, step S102, which segments the target session set from the target session content, includes at least one of the following:

[0091] Step S301: Segment the target session content according to the keywords in the target session content to obtain at least one target session set.

[0092] After obtaining the target session content, all session statements in the target session content can be identified, and keywords in the target session content can be determined.

[0093] Keywords can be words used to indicate the start or end of a conversation. Examples include greetings, hello, are you there, goodbye, thank you, and so on. Furthermore, in this case, the target conversation set does not include the statements containing the keywords.

[0094] For example, when the target session content includes statements issued by the user as described below:

[0095] 1: Hello.

[0096] 2: My sanda has been turned into a sanda.

[0097] 3: How to compensate.

[0098] 4: Agreed, goodbye.

[0099] This allows us to determine that keywords are included in 1 and 4, and thus the target session set includes session statements 2 and 3.

[0100] Step S302: Segment the target session content according to the sending time of each session statement to obtain at least one target session set, wherein the time interval between any two temporally adjacent session statements in the same target session set is less than a preset interval upper limit.

[0101] After obtaining the content of the target session, the time of each session statement can be determined. Since, generally speaking, the closer the times of the session statements are, the higher the probability that they are used to query the same event, the target session set can be determined according to the requirement that the time interval between any two temporally adjacent session statements in the same target session set is less than the preset interval upper limit.

[0102] For example, when the target session content includes statements issued by the user as described below, and the issuance time of each statement is as follows:

[0103] 1: Hello. November 28, 2021, 17:33:49

[0104] 2: Are there any discounts on sundaes? November 28, 2021, 17:33:54

[0105] 3: Hello. November 29, 2021, 17:34:52

[0106] 4: My sundae has turned into a sundae. November 29, 2021, 17:34:56

[0107] 5: How will compensation be provided? (November 29, 2021, 17:35:57)

[0108] 6: Agreed, goodbye. November 29, 2021, 17:37:00

[0109] It can be determined that both conversation statements 1 and 2 were sent on November 28, 2021, while conversation statements 3 to 6 were sent on November 29, 2021. Therefore, when the preset interval limit is 1 hour, the time interval between conversation statements 1 and 2 is less than 1 hour, the time interval between conversation statements 3 and 4 is greater than 1 hour, and the time interval between conversation statements 3 to 6 is less than 1 hour. Thus, we can obtain target conversation set I (including conversation statements 1 and 2) and target conversation set II (including conversation statements 3 to 6).

[0110] The method described in this embodiment provides a way to segment conversation statements. This method can increase the probability that all conversation statements in the same target conversation set are related to the same event, thereby improving the accuracy of subsequent analysis.

[0111] like Figure 2As shown, as an optional implementation, the method described above, step S103, which determines the target association relationships between each target session element in the target session set, includes the following steps:

[0112] Step S401: Input the target session set into the preset classification model and analyze the target session type corresponding to the target session set. The target session type is used to indicate the type of event that the target session set is used to communicate.

[0113] After obtaining the target session set, it can be classified according to a preset classification model. Optionally, the classification model can be pre-trained to classify the session content, and the classification type can be: complaint, sales pitch, inquiry, etc. By inputting the target session set into the classification model, the target session type corresponding to the target session set can be analyzed, thereby determining which type of session content corresponding to the target session set belongs to.

[0114] Step S402: Using a preset algorithm, determine the target core sentence among all the conversation statements in the target conversation set. The target core sentence is the conversation statement with the highest relevance to the conversation type among all the conversation statements in the target conversation set.

[0115] After obtaining the target conversation set, a preset algorithm for core sentence analysis can be used to determine the target core sentence from all conversation statements in the target conversation set. The target core sentence can be the sentence in a conversation that best reflects the target conversation type. The target algorithm can include relevant words related to each conversation type, and then use the number or proportion of relevant words in the conversation statements as the relevance. Therefore, the conversation statement with the most relevant words and / or the highest proportion of relevant words can be selected as the target core sentence.

[0116] For example, when the target session content includes statements made by the user as described below, and the target session type is determined to be "complaint" through the foregoing embodiments:

[0117] 1: Hello.

[0118] 2: My sundae has turned into a paste.

[0119] 3: Goodbye.

[0120] Since the target conversation type is determined in advance as "complaint", related words such as "how" can be identified. Then, the target core sentence is determined to include the related word "how" in "How did my sundae turn into this?", and thus the conversation statement 2 can be identified as the target core sentence.

[0121] Step S403: Determine the target session element type based on the target scenario corresponding to the target session content, and identify all target session elements in the target session set according to the target session element type.

[0122] After determining the target conversation content, it is also necessary to analyze it in conjunction with the specific scenario corresponding to the target conversation content. For example, when the target scenario is beauty, brands, categories, and products are used as target conversation element types, and all target conversation elements are identified in the target conversation set, such as brand-AA, category-personal care, and product-facial cleanser. When the target scenario is dining, restaurants and coupons can be used as target conversation element types, such as restaurant-BB and coupon-discount coupon.

[0123] Step S404: Input the target session elements into the preset relationship recognition model to obtain the target association relationship between each target session element.

[0124] After acquiring the target session elements, they can be input into a preset relationship recognition model to obtain the target association relationships between the various target session elements. Optionally, the relationship recognition model can be a neural network model that has been pre-trained with multiple training elements to meet a preset accuracy. After inputting the target session elements into the relationship recognition model, the model can determine whether there is an association relationship between any two target session elements, thereby obtaining the target association relationships between the various target session elements.

[0125] The method in this embodiment can quickly determine the target session type, target core sentence, target session elements, and target association relationships corresponding to the target session set.

[0126] As an optional implementation, the method described above, after determining the specified association that satisfies the preset relevance with the target association from all candidate associations in step S104, the method further includes the following steps:

[0127] Step S501: Determine the correspondence between the specified session element and the session element type;

[0128] Step S502: The target conversation type, target core sentence, and corresponding relationship are used as feedback information corresponding to the target conversation set and output.

[0129] After determining the target session type, target core sentence, target session element, target association relationship, and specified session element of the target session set through the methods in the foregoing embodiments, the correspondence between each specified session element and the session element type can be determined (e.g., sundae-product, takeout-service, etc.).

[0130] After determining the above information, the target conversation type, target core sentence, and corresponding relationship can be used as feedback information for the target conversation set and output so that the recipient of the target conversation set can obtain information beyond the conversation statements in the target conversation set. For example, for service personnel, this can help them better answer consumer questions, while for scene managers and analysts, it can help them to understand the main content of the conversation and the complete consumer demands in a timely and comprehensive manner.

[0131] As an optional implementation, as described above, step S105, which determines the association analysis result corresponding to the target session set based on the specified session set corresponding to each specified session element, includes the following steps:

[0132] Step S601: Among all candidate session sets, determine all first specified session sets that include at least one specified session element, wherein the candidate session set and the target session set are session sets corresponding to the same target scenario;

[0133] Step S602: Obtain the first semantic information corresponding to each first specified session set, and obtain the association analysis results based on the first semantic information.

[0134] Once the specified session element is identified, reasoning can be performed on the session statements in the target session set based on the specified session element.

[0135] Optionally, all candidate sessions corresponding to the same target scenario (e.g., restaurant, cosmetics, etc.) can be selected from the first specified session set. The first specified session set includes one or more specified session elements.

[0136] After obtaining all the first specified session sets, the first semantic information corresponding to the first specified session sets can be determined, and then all the first semantic information is used as part of the association analysis results.

[0137] The method in this embodiment can be used to determine the semantics expressed in the target session content by comprehensively judging the content mentioned in other sessions besides the target session content, based on the first semantic information corresponding to the first specified session set as the association analysis result.

[0138] As an optional implementation, the method described above, after determining the specified association that satisfies the preset relevance with the target association from all candidate associations in step S104, the method further includes the following steps:

[0139] Step S701: Determine the target session time range corresponding to the target session set based on the target session time of each session statement in the target session set.

[0140] After obtaining each target session set, the target session time for each session statement can be determined. Then, based on all the target session times, the target session time range corresponding to the target session set can be determined.

[0141] Optionally, the target session time range can be obtained by using the target session time of the earliest session statement and the target session time of the latest session statement in the target session set.

[0142] For example, when the target session content includes statements issued by the user as described below, and the issuance time of each statement is as follows:

[0143] 1: Hello. November 28, 2021, 17:33:49

[0144] 2: Are there any discounts on sundaes? November 28, 2021, 17:33:54

[0145] 3: Okay. November 28, 2021, 17:34:52

[0146] The target session time range is obtained (2021.11.28.17:33:49~2021.11.28.17:34:52).

[0147] Step S702: In all candidate session sets, determine all second specified session sets whose candidate session time ranges and target session time ranges meet the preset time correlation requirements. The candidate session set and the target session set are session sets corresponding to the same target scenario, and each candidate session set has a corresponding candidate session time range.

[0148] All second specified session sets can be selected from the candidate session sets corresponding to the same target scenario as the target session set. Furthermore, the second specified session set includes one or more specified session elements.

[0149] After obtaining all candidate session sets, the time range of each candidate session in each candidate session set can be determined. The method for obtaining the time range of each candidate session can refer to the method described in step S701 above, and will not be repeated here.

[0150] After obtaining the candidate session time range for each candidate session set, all second specified session sets whose candidate session time ranges and target session time ranges meet the preset time correlation requirements can be determined.

[0151] Meeting the preset time correlation requirement can be: the first difference between the earliest time of the candidate session time range and the earliest time of the target session time range is less than the first preset time difference (e.g., 1 hour, etc.), and / or the second difference between the latest time of the candidate session time range and the latest time of the target session time range is less than the second preset time difference (e.g., 1.5 hours, etc.). Furthermore, the first preset time difference and the second preset time difference can be the same or different.

[0152] For example, if the first preset time difference and the second preset time difference indicated in the preset time correlation requirement are both 1 hour, and the candidate session time range of candidate session set a is (2021.11.28.17:13:49~2021.11.28.17:44:52) and the target session time range is (2021.11.28.17:33:49~2021.11.28.17:34:52), then the first difference is 20 minutes and the second time difference is 10 minutes, both of which are less than 1 hour. Therefore, candidate session set a is used as the second designated session set.

[0153] Step S703: Obtain the second semantic information corresponding to each second specified session set, and obtain the association analysis results based on the second semantic information.

[0154] After obtaining all the second specified session sets, the second semantic information corresponding to the second specified session sets can be determined, and then all the second semantic information can be used as part of the association analysis results.

[0155] The method in this embodiment can determine a second specified session set corresponding to the target session set based on temporal correlation, and use the second semantic information corresponding to the second specified session set as the association analysis result, thereby achieving the purpose of comprehensively judging the semantics to be expressed in the target session content by using the content mentioned in other sessions besides the target session content.

[0156] like Figure 3 As shown, this application provides an example of applying any of the foregoing embodiments:

[0157] 1. Segmenting the conversation: First, the target conversation content between employees and customers in WeChat Work is segmented based on the conversation content (i.e., keywords), the time interval between each conversation statement, etc., to obtain at least one set of target conversations;

[0158] 2. Conversation Element Recognition: The segmented target conversation set is classified according to pre-trained conversation tags (i.e., conversation types), the core sentences of the conversation are extracted (the original words that can represent the content of the conversation are extracted from the conversation), entities (i.e., conversation elements), and entity relationships (i.e., association relationships) are identified according to the pre-trained relationship recognition model.

[0159] 3. Related Element Reasoning: Combining the candidate associations of each candidate session element in the knowledge graph, reasoning is performed on the elements contained in the target session set to deduce related entities (i.e., specified session elements) that have direct or indirect relationships.

[0160] 4. Related Session Analysis: After processing different sessions within the same time period using steps 1 to 2 to 3, sessions containing the same entity (i.e., the first specified session set) and sessions with the same relational characteristics (i.e., the second specified session set that meets the preset time correlation requirements) are connected.

[0161] 5. Feedback Element Analysis: The tags, core sentences of the conversation, conversation elements, and the types (i.e., correspondences) of the conversation elements analyzed in steps 2 and 3 are output as feedback elements. For frontline service personnel, this allows them to obtain elements outside the conversation and quickly grasp various elements related to the conversation, thereby helping them to better answer consumer questions. For session managers and analysts, it enables them to timely and comprehensively understand the main content of the conversation and the complete consumer demands.

[0162] 6. Feedback Reason Analysis: Based on the first and second specified session sets obtained in step 4, the correlation analysis results are obtained, and the output is the feedback reason analysis. Through feedback reason analysis, other sessions with the same entities as the current session can be found. By combining the content mentioned in other sessions, the reasons for the consumer's demands can be reconstructed from multiple dimensions.

[0163] like Figure 4 As shown, according to another embodiment of this application, a session analysis apparatus is also provided, comprising:

[0164] Module 1 is used to obtain the target session content, wherein the target session content includes at least one session statement;

[0165] Segmentation module 2 is used to segment the target session set from the target session content, wherein all session statements in the target session set satisfy a preset relationship;

[0166] The first determining module 3 is used to determine the target association relationship between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element;

[0167] The second determining module 4 is used to determine the specified association relationship, including the target association relationship, from all candidate association relationships;

[0168] The third determining module 5 is used to determine the association analysis result corresponding to the target session set based on the specified session set corresponding to each specified session element, wherein the specified session element is the session element in the specified association relationship.

[0169] Specifically, the detailed process by which each module in the device of this invention implements its function can be found in the relevant description in the method embodiment, and will not be repeated here.

[0170] According to another embodiment of this application, an electronic device is also provided, comprising: Figure 5 As shown, the electronic device may include: a processor 1501, a communication interface 1502, a memory 1503, and a communication bus 1504, wherein the processor 1501, the communication interface 1502, and the memory 1503 communicate with each other through the communication bus 1504.

[0171] Memory 1503 is used to store computer programs;

[0172] When the processor 1501 executes the program stored in the memory 1503, it implements the steps of the above method embodiment.

[0173] The bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0174] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0175] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0176] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0177] This application also provides a computer-readable storage medium, which includes a stored program, wherein the program executes the method steps of the above method embodiments when it runs.

[0178] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0179] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A session analysis method, characterized in that, include: Obtain the target session content, wherein the target session content includes at least one session statement; A target session set is segmented from the target session content, wherein all the session statements in the target session set satisfy a preset relationship; Determine the target association relationships between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element; Among all candidate associations, a specified association that satisfies a preset relevance with the target association is determined; Based on the specified session set corresponding to each specified session element, determine the association analysis result corresponding to the target session set, including: determining all first specified session sets including at least one specified session element in all candidate session sets, wherein the specified session element is a session element in the specified association relationship, and the candidate session set and the target session set are session sets corresponding to the same target scenario; obtaining first semantic information corresponding to each first specified session set, and obtaining the association analysis result based on the first semantic information, wherein the association analysis result is used to indicate the true intention of the user corresponding to the target session set in communication.

2. The method according to claim 1, characterized in that, The step of segmenting the target session set from the target session content includes at least one of the following: The target session content is segmented according to the keywords in the target session content to obtain at least one set of target sessions; The target session content is segmented according to the issuance time of each session statement to obtain at least one target session set, wherein the time interval between any two temporally adjacent session statements in the same target session set is less than a preset interval upper limit.

3. The method according to claim 1, characterized in that, Determining the target association relationship between each target session element in the target session set includes: The target session set is input into a preset classification model to analyze the target session type corresponding to the target session set, wherein the target session type is used to indicate the type of event that the target session set is used to communicate; Using a preset algorithm, a target core sentence is determined from all the conversation statements in the target conversation set, wherein the target core sentence is the conversation statement with the highest relevance to the conversation type among all the conversation statements in the target conversation set; Based on the target scenario corresponding to the target session content, the target session element type is determined, and all target session elements in the target session set are identified according to the target session element type; The target session elements are input into a preset relationship recognition model to obtain the target association relationship between each target session element.

4. The method according to claim 3, characterized in that, The step of determining a specified association that satisfies a preset relevance with the target association from all candidate associations includes: Determine all candidate session elements included in the knowledge graph that serve as the candidate associations, as well as the candidate associations between all the candidate session elements; If all the target session elements are included among all the candidate session elements, and / or the candidate association includes the target association, the knowledge graph is determined as the specified association.

5. The method according to claim 3, characterized in that, After determining, from all candidate associations, a specified association that satisfies a preset relevance to the target association, the method further includes: Determine the correspondence between the specified session element and the session element type; The target session type, the target core sentence, and the corresponding relationship are used as feedback information corresponding to the target session set and output.

6. The method according to claim 1, characterized in that, After determining, from all candidate associations, a specified association that satisfies a preset relevance to the target association, the method further includes: Based on the target session time of each session statement in the target session set, the target session time range corresponding to the target session set is determined; Among all candidate session sets, all second specified session sets whose candidate session time ranges satisfy the preset time correlation requirement with the target session time range are determined, wherein the candidate session set and the target session set are session sets corresponding to the same target scenario, and each candidate session set has a corresponding candidate session time range; Obtain the second semantic information corresponding to each of the second specified session sets, and obtain the association analysis result based on the second semantic information.

7. A conversation analysis device, characterized in that, include: The acquisition module is used to acquire target session content, wherein the target session content includes at least one session statement; A segmentation module is used to segment a target session set from the target session content, wherein all the session statements in the target session set satisfy a preset relationship; The first determining module is used to determine the target association relationship between each target session element in the target session set, wherein the target session element is used to indicate the characteristics of the target session set, and the target association relationship is used to indicate the logical relationship between each target session element; The second determining module is used to determine, from all candidate associations, a specified association that satisfies a preset correlation with the target association; The third determining module is used to determine the association analysis result corresponding to the target session set based on the specified session set corresponding to each specified session element, including: determining all first specified session sets including at least one specified session element in all candidate session sets, wherein the specified session element is a session element in the specified association relationship, and the candidate session set and the target session set are session sets corresponding to the same target scenario; obtaining first semantic information corresponding to each first specified session set, and obtaining the association analysis result based on the first semantic information, wherein the association analysis result is used to indicate the true intention of the user corresponding to the target session set in communication.

8. An electronic device, characterized in that, include: The processor, communication interface, memory, and communication bus are provided, wherein the processor, communication interface, and memory communicate with each other via the communication bus. The memory is used to store computer programs; The processor, when executing the computer program, implements the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6 when it is run.

Images