Information processing method and device for intelligent session and electronic equipment

Abstract

The invention discloses an information processing method and device for an intelligent session and electronic equipment, and the method comprises the steps: determining a first session group which carries out a session with a user, the first session group comprises multiple rounds of sessions, and each round of session is a session between user input information and corresponding question and answer information; determining relevance between at least two rounds of sessions in the first session group and time information related to each round of session in the first session group in a session process; and determining the round of the second session group to be output based on the relevance and / or the time information.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to an information processing method, device and electronic device for intelligent conversation. Background Art

[0002] In the process of using artificial intelligence to have a conversation (dialogue) with a user, such as an AIGC (Artificial Intelligence Generated Content) conversation, a certain number of conversation information is usually generated first, such as generating a fixed 2-round conversation. Then, the subsequent conversation information is determined based on the existing conversation information, but the conversation rounds of the subsequent conversation information are usually fixed, such as 30 rounds of conversation, which has poor flexibility. This will cause the subsequent conversation information to be incompatible with the logic of the current conversation, and reduce intelligence. Summary of the invention

[0003] The present application provides an information processing method for intelligent conversation, including:

[0004] Determine a first conversation group for conversation with the user, wherein the first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information;

[0005] determining the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation;

[0006] Based on the correlation and / or the time information, a turn of the second conversation group to be output is determined.

[0007] Optionally, determining the correlation between at least two rounds of conversations in the first conversation group includes:

[0008] determining keywords in each round of conversation of the first conversation group;

[0009] Matching keywords in at least two rounds of conversations in the first conversation group to generate a matching result, wherein the matching result is used to characterize the similarity between the rounds of conversations;

[0010] The association is determined based on the matching results.

[0011] Optionally, determining the correlation between at least two rounds of conversations in the first conversation group includes:

[0012] Determining tag information for characterizing a conversation topic in each round of conversation of the first conversation group;

[0013] Determine a matching degree between tag information corresponding to at least two rounds of conversations of the first conversation group, and / or a contextual relationship between tag information corresponding to at least two rounds of conversations;

[0014] The association is determined based on the matching degree between the tag information and / or the contextual relationship between the tag information.

[0015] Optionally, the determining the correlation between at least two rounds of conversations in the first conversation group includes: determining, based on first user input information of a first round of conversations in the first conversation group, that the first conversation group is used to represent scene label information;

[0016] Based on the scene label information of the first conversation group and the label information corresponding to the round conversation in the first conversation group, determining a matching result between the round conversation in the first conversation group and the scene label information of the first conversation group;

[0017] The association is determined based on the matching results.

[0018] Optionally, determining the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation process includes:

[0019] Acquire first time interval information between adjacent rounds of conversations in the first conversation group, and second time interval information between question and answer information and user input information in each round of conversations in the first conversation group;

[0020] The time information is determined based on the first time interval information and / or the second time interval information.

[0021] Optionally, determining the round of the second conversation group to be output based on the time information includes:

[0022] determining whether the first time interval information exceeds a second threshold, and whether the second time interval information exceeds a third threshold;

[0023] When the first time interval information exceeds the second threshold, and / or the second time interval information exceeds the third threshold, the round of the second conversation group is reduced.

[0024] Optionally, determining the round of the second conversation group to be output based on the correlation and / or the time information includes:

[0025] determining a first data value for characterizing the association;

[0026] In a case where the first data value is greater than a first threshold, determining the content of the second conversation group based on the conversation topic content of the current conversation;

[0027] Based on the difference between the first data value and the first threshold, the round of the second conversation group is adjusted.

[0028] Optionally, the method further comprises:

[0029] After the session with the user ends, if the user input information is re-acquired, determining whether the re-acquired user input information is related to the previous session;

[0030] If it is determined that there is no dependency, then another session different from the existing one is restarted.

[0031] The present application also provides an information processing device for intelligent conversation, including:

[0032] A first determination module is configured to determine a first conversation group for conversation with the user, wherein the first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information;

[0033] a second determining module configured to determine the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation;

[0034] A processing module is configured to determine a turn of the second conversation group to be output based on the correlation and / or the time information.

[0035] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein the memory stores an executable program, and the processor executes the executable program to perform the steps of the method described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flowchart of the information processing method of the intelligent conversation according to the embodiment of the present application;

[0037] Figure 2 For the embodiments of this application Figure 1 Flow chart of a first embodiment of step S200;

[0038] Figure 3 For the embodiments of this application Figure 1 Flow chart of a second embodiment of step S200;

[0039] Figure 4 For the embodiments of this application Figure 1 Flow chart of a third embodiment of step S200;

[0040] Figure 5 For the embodiments of this application Figure 1 Flow chart of the fourth embodiment of step S200;

[0041] Figure 6 For the embodiments of this application Figure 1 Flow chart of a first embodiment of step S300;

[0042] Figure 7 For the embodiments of this application Figure 1 Flow chart of a second embodiment of step S300;

[0043] Figure 8 A flowchart of an embodiment of the information processing method of the intelligent conversation of the embodiment of the present application;

[0044] Fig. 9 This is a structural block diagram of the information processing device for intelligent conversation according to an embodiment of the present application. DETAILED DESCRIPTION

[0045] Various aspects and features of the present application are described herein with reference to the accompanying drawings.

[0046] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but only as an example of an embodiment. Other modifications within the scope and spirit of the present application will occur to those skilled in the art.

[0047] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0048] These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0049] It should also be understood that although the present application has been described with reference to some specific examples, those skilled in the art will be able to readily implement many other equivalent forms of the present application.

[0050] The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.

[0051] Specific embodiments of the present application are described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments applied for are merely examples of the present application, which may be implemented in a variety of ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that obscure the present application. Therefore, the specific structural and functional details applied for herein are not intended to be limiting, but merely serve as a basis and representative basis for the claims to teach those skilled in the art to use the present application in a variety of ways with substantially any suitable detailed structure.

[0052] This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," all of which may refer to one or more of the same or different embodiments according to the present application.

[0053] The information processing method of an intelligent conversation in the embodiment of the present application can be applied to a variety of intelligent scenarios as needed, such as intelligent commodity sales scenarios, intelligent customer service scenarios, intelligent commodity after-sales scenarios, etc. Specifically, the intelligent conversation includes multiple conversation groups, which can predict the meaning of user input information based on the content and logical relationship of the existing conversation groups, thereby predicting the interaction behavior between the user and the artificial intelligence, and then adjusting the rounds of subsequent conversation groups that have not yet been output, including increasing or reducing the conversation rounds, so as to flexibly provide users with conversations that are adapted to the current conversation scenario to meet different user needs.

[0054] The information processing method of the intelligent session of the present application is described in detail below with reference to the accompanying drawings. Figure 1 Flow chart of the information processing method of the intelligent conversation according to the embodiment of the present application, as shown in Figure 1 As shown, the method comprises the following steps:

[0055] S100, determining a first conversation group for conversation with a user, wherein the first conversation group includes multiple rounds of conversations, each round of conversation being a conversation between user input information and corresponding question and answer information.

[0056] Exemplarily, the electronic device may be pre-set with a first conversation group according to a usage scenario of a conversation with a user, and the first conversation group is used to intelligently interact with the user. Of course, the content and rounds of the first conversation group can be adjusted according to actual needs, so as to be able to intelligently interact with user input information.

[0057] The first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information. The user input information can be information input by the user through the terminal when interacting with the electronic device (such as a server) or the corresponding intelligent system, and the question and answer information in the conversation is information asked by the electronic device or the corresponding intelligent system to the user, or information answered based on the user input information. For example, the question and answer information can be "How can I help you?" or "Today's activity information is..." The user input information and the corresponding question and answer information form a round of conversation.

[0058] The multiple rounds of conversations in the first conversation group are related to each other and can form relatively complete interactive information. For example, the contents of the multiple rounds of conversations are related to each other, which improves the consistency of the content in the process of interacting with the user input information, so that the user can feel the intelligence of the electronic device (or the corresponding intelligent system).

[0059] On the one hand, the conversation round content and conversation round in the first conversation group can be adjusted in the process of interacting with the user. For example, the conversation round of the first conversation group can be preset to 2 to 3 rounds. In the process of interacting with the user, as the conversation round increases, all the output conversations can be determined as the first conversation group. For example, if 5 rounds of conversations have been conducted with the user, all the output conversation content can be determined as the content of the first conversation group, and the rounds of the first conversation group can be adjusted to 5 rounds.

[0060] On the other hand, the conversation round content and conversation round in the first conversation group may not be adjusted. For example, the conversation round of the first conversation group may be preset to 2 to 3 rounds, and the round and the corresponding conversation content may not be adjusted.

[0061] S200: Determine the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation process.

[0062] Exemplarily, the first conversation group includes multiple conversation rounds, and at least two conversation rounds may be related. For example, if the meanings expressed in the first round and the second round are the same topic, it can be determined that the first round and the second round are related. Of course, the correlation may be strong or weak depending on the actual situation.

[0063] In one embodiment, when determining the correlation between at least two rounds of conversations, the semantic correlation relationship between each round of conversations in the first conversation group and the conversation basis information corresponding to each round of conversations can be obtained; based on the semantic correlation relationship and / or the conversation basis information, the correlation between at least two rounds of conversations is determined.

[0064] In another embodiment, when determining the correlation between at least two rounds of conversations, the correlation may be determined based on the degree of cohesion of semantic expressions between adjacent rounds of conversations in the first conversation group.

[0065] In yet another embodiment, when determining the correlation between at least two rounds of conversations, the first correlation may be determined based on the similarity between target objects targeted by each round of conversations in the first conversation group.

[0066] In this embodiment, each round of conversation in the first conversation group has time information associated therewith, including first time interval information between adjacent rounds of conversations, and second time interval information between question and answer information and user input information in each round of conversation in the first conversation group. If the numerical value corresponding to the first time interval information and / or the second time interval information is large, it indicates that the content coherence between the corresponding rounds of conversation is not strong, otherwise it indicates that the content of the corresponding rounds of conversation has strong logic.

[0067] S300: Determine a round of a second conversation group to be output based on the correlation and / or the time information.

[0068] Exemplarily, the second conversation group may be a conversation group to be output by the electronic device, and the rounds of the second conversation group are not set to be fixed, but the rounds of the second conversation group to be output may be determined based on the correlation between at least two rounds of conversations in the first conversation group and / or time information related to each round of conversations in the first conversation group.

[0069] In one embodiment, a preset round of a second conversation group may be predetermined, and then the preset round may be adjusted according to the correlation and / or the time information. In another embodiment, the preset round of a second conversation group may not be predetermined, but the preset round of the second conversation group may be determined based on a first comparison result of the correlation and the first relationship table, and / or a second comparison result of the time information and the second relationship table, wherein the first relationship table records a first correlation relationship between the correlation and the round of the second conversation group, and the second relationship table records a second correlation relationship between the time information and the round of the second conversation group.

[0070] For example, if the correlation between at least two rounds of conversations in the first conversation group is relatively strong, and the first time interval information between adjacent rounds of conversations in the first conversation group is relatively short, it means that the first conversation group has strong logic and intelligence in the process of interacting with the user, and relatively more rounds of the second conversation group can be determined based on the correlation and / or time information. If the correlation between at least two rounds of conversations in the first conversation group is relatively weak, and the first time interval information between adjacent rounds of conversations in the first conversation group is relatively long, it means that the first conversation group has poor logic and intelligence in the process of interacting with the user, and relatively fewer rounds of the second conversation group can be determined based on the correlation and / or time information.

[0071] After the round of the second conversation group is determined, the second conversation group can be output to further intelligently interact with the user. Since the round of the second conversation group is appropriately determined, the intelligence of the electronic device when interacting with the user can be effectively improved.

[0072] The information processing method of the intelligent conversation in the embodiment of the present application determines the conversation round of the output second conversation group that is adapted to the current user interaction scenario based on the correlation between at least two conversation rounds in the preset first conversation group and the time information related to each conversation round in the first conversation group, thereby making the subsequent conversation information adapted to the logic of the current conversation, effectively improving the intelligence.

[0073] In one embodiment of the present application, the determining of the correlation between at least two rounds of conversations in the first conversation group is as follows: Figure 2 As shown, the following steps are included:

[0074] S210: Determine keywords in each round of conversation of the first conversation group.

[0075] For example, the keywords may be key words, numbers, letters, etc. around which each word in a round of conversation is surrounded. The keywords in each round of conversation of the first conversation group may be determined according to specific application scenarios and user input information and / or user feedback on question and answer information output by the electronic device.

[0076] For example, in the scenario of artificial intelligence design and manufacturing, during the interaction between the electronic device (or intelligent system) and the user, the content of the first round of conversation in the first conversation group is: "Help me write an introduction related to artificial intelligence", the content of the second round of conversation in the first conversation group is: "Help me modify the design of the internal architecture...", and the content of the third round of conversation in the first conversation group is: "Help me add the latest translation of artificial intelligence papers". Artificial intelligence is mentioned in each round of conversation in the first conversation group, so the relevant words of artificial intelligence can be determined as keywords.

[0077] For another example, in a device technical support scenario, during the interaction between the electronic device (or intelligent system) and the user, each round of conversation in the first conversation group mentions specific information related to the device failure, so that the information related to the device failure can be determined as a keyword.

[0078] In another embodiment, the user's attention content actively input by the user may be obtained, such as obtaining the information actively input by the user such as "commodity price", "commodity price range", "price difference compensation", etc. Then, the user's attention content may be determined as keyword information.

[0079] S220, matching keywords in at least two rounds of conversations in the first conversation group to generate a matching result, wherein the matching result is used to characterize the similarity between the rounds of conversations.

[0080] For example, keywords in at least two rounds of conversations in the first conversation group may be matched for content similarity, such as matching keywords in the first round of conversations with keywords in the second round of conversations, or matching keywords in the second round of conversations with keywords in the third round of conversations. It can be determined whether the keywords involved in the matching have the same or similar content, so as to determine whether the subject content expressed in the corresponding rounds of conversations is similar. The degree of similarity between the corresponding rounds of conversations can then be determined.

[0081] In one embodiment, according to the keyword matching results, it is determined that the keywords of the two rounds of conversations are the same and both are content related to the commodity price, thereby determining that the two rounds of conversations have a high degree of similarity, and the similarity can be represented by a specific numerical value.

[0082] In another embodiment, according to the matching results of keywords, it is determined that the keywords of the two rounds of conversations are different, but have progressive meanings. For example, the keyword in the first round of conversations in the first conversation group is "new product release", and the keyword in the first round of conversations in the second conversation group is "new mobile phone". Therefore, based on the keyword, it can be determined that the similarity between the first round of conversations and the second round of conversations in the first conversation group is high, and the similarity can also be represented by a specific numerical value. The numerical value increases with the increase of the similarity.

[0083] S230: Determine the correlation based on the matching result.

[0084] Exemplarily, the matching result is used to characterize the similarity between rounds of conversations. Based on the matching result, a numerical value corresponding to the similarity between the keywords in at least two rounds of conversations in the first conversation group can be determined. If the numerical value exceeds a preset range, it can be determined that the two rounds of conversations in the first conversation group are associated. Otherwise, it can be determined that there is no association between the two rounds of conversations.

[0085] In one embodiment of the present application, the determining of the correlation between at least two rounds of conversations in the first conversation group is as follows: Figure 3 As shown, including:

[0086] S240: Determine tag information used to characterize a conversation topic in each round of conversation of the first conversation group.

[0087] Exemplarily, the conversation topic may be the main focus content in the corresponding conversation. Each round of conversation in the first conversation group has its own corresponding conversation topic, and the conversation topic can be determined by analyzing the semantics of each round of conversation in the first conversation group. For example, the conversation topic can be determined by semantically analyzing the user input information and the question and answer information output by the electronic device in each round of conversation.

[0088] Each conversation topic has its own corresponding label information, and different conversation topics have different label information. In one embodiment, based on the association between the conversation topic and the label information, a corresponding preset table is determined. After it is determined that each round of conversation in the first conversation group has its own corresponding conversation topic, the corresponding label information can be further determined based on the preset table.

[0089] S250: Determine a matching degree between tag information corresponding to at least two rounds of conversations in the first conversation group and / or a contextual relationship between tag information corresponding to at least two rounds of conversations.

[0090] Exemplarily, the tag information corresponding to at least two rounds of conversations in the first conversation group may be matched to determine the degree of association between the corresponding tag information. In one embodiment, the tag information may be represented by numerical values, and if the numerical values ​​are close, it indicates that the degree of association between the corresponding different tag information is relatively close, and if the numerical values ​​differ greatly, it indicates that the degree of association between the corresponding different tag information is relatively low.

[0091] In one embodiment, there may be a contextual relationship between the tag information corresponding to at least two rounds of conversations in the first conversation group. The contextual relationship represents the logical order and semantic cohesion, which can be determined by the intention understanding of the model in the electronic device (intelligent system). For example, the contextual relationship is determined by the logical order of the content between two adjacent rounds of conversations through a large model. This includes analyzing and understanding the words and sentences in the conversation to determine the logical order of the content, and then determining the contextual relationship.

[0092] In another embodiment, a template including the association between tag information and contextual relationship is preset, and the tag information can be matched with the template to determine the contextual relationship corresponding to the tag information. Of course, the template can be adjusted according to the specific usage scenario.

[0093] S260: Determine the correlation based on the matching degree between the tag information and / or the contextual relationship between the tag information.

[0094] Exemplarily, the matching degree between tag information and the contextual relationship between tag information can both characterize the correlation between at least two rounds of conversations in the first conversation group. Therefore, based on the matching degree between tag information and / or the contextual relationship between tag information, the correlation can be accurately determined.

[0095] In one embodiment of the present application, the determining of the correlation between at least two rounds of conversations in the first conversation group is as follows: Figure 4 As shown, including:

[0096] S270: Determine, based on first user input information of a first round of conversation in the first conversation group, that a first conversation group is used to represent scene label information.

[0097] Exemplarily, the first user input information of the first round of conversation in the first conversation group is information input by the user through the terminal during the interaction with the electronic device (or intelligent system), and is the user's intention expression during the interaction. For example, the first user input information may be asking about after-sales issues of a purchased mobile phone, and the current scene of the first conversation group may be determined as an after-sales scene, and the corresponding scene label information may be determined based on the after-sales scene, such as the scene label information may be "after-sales" or "mobile phone after-sales", or may be expressed in other non-textual ways.

[0098] In another embodiment, the user can actively input information to directly determine the label information used to characterize the scene, such as actively inputting the first user input information as "the intelligent customer service is required to determine this conversation as a related conversation for mobile phone after-sales service", and then the electronic device (or intelligent system) can directly determine the scene label information as "mobile phone after-sales service".

[0099] S280, based on the scene label information of the first conversation group and the label information corresponding to the round conversation in the first conversation group, determine the matching result between the round conversation in the first conversation group and the scene label information of the first conversation group.

[0100] Exemplarily, the scene label information is used to comprehensively represent the usage scenarios of each round of conversations in the first conversation group. Each round of conversations in the first conversation group also has its own label information, which is used to characterize the conversation topic of the corresponding round of conversations. The scene label information of the first conversation group can be matched with the label information corresponding to each round of conversations in the first conversation group. For example, if the scene label information is "mobile phone after-sales service", the label information corresponding to each round of conversations in the first conversation group may also be "mobile phone after-sales service" or other "pre-sales consultation". The intelligent system of the electronic device can compare the label information corresponding to each round of conversations of the scene label information respectively, so as to determine the corresponding matching results, such as the matching results can be "same", "similar", "different", etc. Of course, specific numerical values ​​can also be used to represent the degree of similarity or the degree of difference.

[0101] S290: Determine the correlation based on the matching result.

[0102] Exemplarily, the matching result characterizes the similarity or difference between the label information corresponding to the round conversation in the first conversation group and the scene label information of the first conversation group. The correlation between at least two round conversations in the first conversation group is determined based on the matching result. For example, if the matching result shows that the label information corresponding to the round conversation is the same as the scene label information of the first conversation group, it is determined that there is a strong correlation between at least two round conversations in the first conversation group. Otherwise, if the matching result shows that the label information corresponding to the round conversation is different from the scene label information of the first conversation group, it can be determined that there is a low correlation between at least two round conversations in the first conversation group.

[0103] In one embodiment of the present application, the determining of the correlation between at least two rounds of conversations in the first conversation group and the time information related to each round of conversations in the first conversation group during the conversation process is as follows: Figure 5 As shown, the following steps are included:

[0104] S2010: Obtain first time interval information between adjacent rounds of conversations in the first conversation group, and second time interval information between question and answer information and user input information in each round of conversation in the first conversation group.

[0105] Exemplarily, there is a certain time interval between adjacent rounds of conversations in the first conversation group, such as the time interval between the adjacent first round of conversations and the second round of conversations in the first conversation group. In this embodiment, the time interval can be determined as the first time interval information. If the time value corresponding to the first time interval information is shorter, it indicates that the possibility that the conversation content topics corresponding to the adjacent rounds of conversations are the same or similar is greater, such as the user quickly interacted with the intelligent system around the same conversation topic in a short period of time. If the time value corresponding to the first time interval information is longer, it indicates that the possibility that the conversation content topics corresponding to the adjacent rounds of conversations are the same or similar is smaller, such as the user conducted a pre-sales consultation on a mobile phone with the intelligent system at the first time point, and discussed the after-sales situation of the computer with the intelligent system at the second time point. The difference between the first time point and the second time point is relatively large.

[0106] In each round of conversation in the first conversation group, there is also a corresponding time interval between the user input information and the question and answer information output by the intelligent system. In this embodiment, the time interval can be determined as the second time interval information. When the user interacts normally with the intelligent system through the terminal, the time interval will not be too long. For example, when the user discusses the pre-sales service of the mobile phone with the intelligent system, the time interval between the user input information and the question and answer information output by the intelligent system is similar to the conversation interval between people. By obtaining the time of each round of conversation output in the first conversation group, and the time between the question and answer information and the user input information in each round of conversation, the first time interval information and the second time interval information can be obtained respectively.

[0107] S2020: Determine the time information based on the first time interval information and / or the second time interval information.

[0108] Exemplarily, based on the first time interval information and / or the second time interval information, various time factors can be comprehensively considered to accurately determine the time information related to each round of conversation in the first conversation group, and then determine the correlation between at least two rounds of conversation in the first conversation group based on the time information.

[0109] In one embodiment of the present application, the round of the second conversation group to be output is determined based on the time information, such as Figure 6 As shown, including:

[0110] S310, determining whether the first time interval information exceeds a second threshold, and whether the second time interval information exceeds a third threshold;

[0111] S320: When the first time interval information exceeds the second threshold and / or the second time interval information exceeds the third threshold, reduce the round of the second conversation group.

[0112] Exemplarily, the first time interval information is compared with the second threshold value. When it is determined that the first time interval information exceeds the second threshold value, it indicates that the time interval between adjacent rounds of conversations in the first conversation group is long. For example, if the time interval between the first round of conversations and the second round of conversations exceeds the normal conversation time between people (for example, exceeds one minute), it indicates that the conversation contents of adjacent rounds of conversations may be different. In this case, the rounds of the second conversation group can be reduced, thereby reducing the probability of conversation logic errors in the second conversation group and improving intelligence.

[0113] Similarly, the second time interval information is compared with the third threshold value. When it is determined that the second time interval information exceeds the third threshold value, it indicates that the time interval between the question and answer information and the user input information in each round of the first conversation group is long, which exceeds the normal conversation time between people. In this case, the number of rounds of the second conversation group can be reduced, thereby reducing the probability of conversation logic errors in the second conversation group and improving the intelligence of the intelligent system.

[0114] In one embodiment of the present application, the round of the second conversation group to be output is determined based on the correlation and / or the time information, such as Figure 7 As shown, the following steps are included:

[0115] S330, determining a first data value for characterizing the association;

[0116] Exemplarily, the correlation between at least two rounds of conversations in the first conversation group can be represented by a first data value, where the higher the first data value, the stronger the correlation, and the lower the first data value, the weaker the correlation. In this embodiment, a data value template can be pre-set, which includes the correlation relationship between correlations of different strengths and specific data values. For example, the corresponding data value of the correlation expressed as "extremely strong" is 100, the corresponding data value of the correlation expressed as "relatively strong" is 90, ..., and the corresponding data value of the correlation expressed as "extremely poor" is 10.

[0117] S340: When the first data value is greater than a first threshold, determine the content of the second conversation group based on the conversation topic content of the current conversation.

[0118] Exemplarily, the first threshold value can be preset according to actual usage scenarios and historical experience. When the first data value is greater than the first threshold value, it indicates that at least two rounds of conversations in the first conversation group have a strong correlation, and the conversation topic content of the current conversation is more likely to be involved by the user in the second round of conversation group to be output. Therefore, the content of the second conversation group can be determined based on the conversation topic content of the current conversation. For example, the content of the second conversation group can be formed based on the conversation topic content of the current conversation, as well as the logic of explanation between people and the progressiveness of content breadth and depth.

[0119] S350: Adjust the round of the second conversation group based on the difference between the first data value and the first threshold.

[0120] Exemplarily, the first data value indicates the correlation between at least two rounds of conversations in the first conversation group. On the one hand, when the first data value is greater than the first threshold, the higher the difference between the first data value and the first threshold, the higher the activeness of the user conversation, and the round of the second conversation group can be increased, such as increasing the original round by one round. On the other hand, when the first data value is less than the first threshold, the higher the difference between the first data value and the first threshold, the lower the activeness of the user conversation, and the round of the second conversation group can be reduced.

[0121] In one embodiment of the present application, Figure 8 As shown, the method further comprises the following steps:

[0122] S400, after the session with the user ends, if the user input information is re-acquired, determining whether the re-acquired user input information is related to the previous session;

[0123] S500: If it is determined that there is no correlation, restart another session different from the existing one.

[0124] Exemplarily, after the conversation with the user ends, the electronic device (or intelligent system) outputs evaluation information of the conversation in the hope that the user will evaluate the conversation. The user inputs the evaluation content, marking the end of the conversation.

[0125] If the user re-converses with the electronic device (or intelligent system) through the , the electronic device (or intelligent system) can re-acquire the user input information and compare the user input information with the subject content of the previous conversation to determine whether there is a correlation between the user input information and the previous conversation. For example, based on the specific content of the re-acquired user input information, it is determined that the subject content of the user's interaction this time is different from the subject content of the previous conversation, thereby determining that there is no correlation between the re-acquired user input information and the previous conversation, and the electronic device (or intelligent system) can restart another conversation that is different from the existing one. Avoid the previous conversation from interfering with the current conversation.

[0126] In another embodiment, after the session with the user ends, if the user input information is re-acquired, it is determined whether there is a correlation between the re-acquired user input information and the previous session, if it is determined that there is a correlation, then the interval between the time of the current user input information and the end of the previous session is obtained, and if the interval exceeds a fourth threshold, the round of the third session group to be output is reduced, or another session different from the existing one is restarted.

[0127] The present application also provides an information processing device for intelligent conversation, such as Fig. 9 As shown, including:

[0128] The first determining module is configured to determine a first conversation group for conversation with the user, wherein the first conversation group includes multiple rounds of conversations, and each round of conversation is a conversation between user input information and corresponding question and answer information.

[0129] Exemplarily, the electronic device may be pre-set with a first conversation group according to a usage scenario of a conversation with a user, and the first conversation group is used to intelligently interact with the user. Of course, the content and rounds of the first conversation group can be adjusted according to actual needs, so as to be able to intelligently interact with user input information.

[0130] The first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information. The user input information can be information input by the user through the terminal when interacting with the electronic device (such as a server) or the corresponding intelligent system, and the question and answer information in the conversation is information asked by the electronic device or the corresponding intelligent system to the user, or information answered based on the user input information. For example, the question and answer information can be "How can I help you?" or "Today's activity information is..." The user input information and the corresponding question and answer information form a round of conversation.

[0131] The multiple rounds of conversations in the first conversation group are related to each other and can form relatively complete interactive information. For example, the contents of the multiple rounds of conversations are related to each other, which improves the consistency of the content in the interactive process with the user input information, so that the user can feel the intelligence of the electronic device (or the corresponding system).

[0132] On the one hand, the first determination module can adjust the conversation round content and conversation round in the first conversation group during the interaction with the user. For example, the conversation round of the first conversation group can be preset to 2 to 3 rounds. In the process of interacting with the user, as the conversation round increases, the first determination module can determine all the output conversations as the first conversation group. For example, if 5 rounds of conversations have been conducted with the user, all the output conversation contents can be determined as the content of the first conversation group, and the rounds of the first conversation group are adjusted to 5 rounds.

[0133] On the other hand, the first determining module may not adjust the conversation round content and conversation round in the first conversation group. For example, the conversation round of the first conversation group may be preset to 2 to 3 rounds, and the round and the corresponding conversation content are not adjusted.

[0134] The second determining module is configured to determine the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation process.

[0135] Exemplarily, the first conversation group includes multiple conversation rounds, and at least two conversation rounds may be related. For example, the meanings expressed in the first round and the second round are the same topic, and the second determination module may determine that the first round and the second round are related. Of course, the correlation may be strong or weak depending on the actual situation.

[0136] In one embodiment, when determining the correlation between at least two rounds of conversations, the second determination module can obtain the semantic correlation relationship between each round of conversations in the first conversation group, as well as the conversation basis information corresponding to each round of conversations; and determine the correlation between at least two rounds of conversations based on the semantic correlation relationship and / or the conversation basis information.

[0137] In another embodiment, when determining the correlation between at least two rounds of conversations, the second determination module may determine the correlation based on the degree of cohesion of semantic expressions between adjacent rounds of conversations in the first conversation group.

[0138] In yet another embodiment, when determining the correlation between at least two rounds of conversations, the second determination module may determine the first correlation based on the similarity between target objects targeted by each round of conversations in the first conversation group.

[0139] In this embodiment, each round of conversation in the first conversation group has time information associated therewith, including first time interval information between adjacent rounds of conversations, and second time interval information between question and answer information and user input information in each round of conversation in the first conversation group. If the numerical value corresponding to the first time interval information and / or the second time interval information is large, it indicates that the content coherence between the corresponding rounds of conversation is not strong, otherwise it indicates that the content of the corresponding rounds of conversation has strong logic.

[0140] A processing module is configured to determine a turn of the second conversation group to be output based on the correlation and / or the time information.

[0141] Exemplarily, the second conversation group may be a conversation group to be output by the electronic device, and the rounds of the second conversation group may not be set to be fixed, but the processing module may determine the rounds of the second conversation group to be output based on the correlation between at least two rounds of conversations in the first conversation group and / or the time information related to each round of conversations in the first conversation group.

[0142] In one embodiment, the processing module may predetermine a preset round of a second conversation group, and then adjust the preset round according to the correlation and / or the time information. In another embodiment, the processing module may not predetermine a preset round of a second conversation group, but determine the preset round of the second conversation group based on a first comparison result between the correlation and the first relationship table, and / or a second comparison result between the time information and the second relationship table, wherein the first relationship table records a first correlation relationship between the correlation and the round of the second conversation group, and the second relationship table records a second correlation relationship between the time information and the round of the second conversation group.

[0143] For example, if the correlation between at least two rounds of conversations in the first conversation group is relatively strong, and the first time interval information between adjacent rounds of conversations in the first conversation group is relatively short, it means that the first conversation group has strong logic and intelligence in the process of interacting with the user, and then the processing module can determine relatively more rounds of the second conversation group based on the correlation and / or time information. If the correlation between at least two rounds of conversations in the first conversation group is relatively weak, and the first time interval information between adjacent rounds of conversations in the first conversation group is relatively long, it means that the first conversation group has poor logic and intelligence in the process of interacting with the user, and then the processing module can determine relatively fewer rounds of the second conversation group based on the correlation and / or time information.

[0144] After the round of the second conversation group is determined, the second conversation group can be output to further intelligently interact with the user. Since the round of the second conversation group is appropriately determined, the intelligence of the electronic device when interacting with the user can be effectively improved.

[0145] Optionally, the second determining module is further configured to:

[0146] determining keywords in each round of conversation of the first conversation group;

[0147] Matching keywords in at least two rounds of conversations in the first conversation group to generate a matching result, wherein the matching result is used to characterize the similarity between the rounds of conversations;

[0148] The association is determined based on the matching results.

[0149] Optionally, the second determining module is further configured to:

[0150] Determining tag information for characterizing a conversation topic in each round of conversation of the first conversation group;

[0151] Determine a matching degree between tag information corresponding to at least two rounds of conversations of the first conversation group, and / or a contextual relationship between tag information corresponding to at least two rounds of conversations;

[0152] The association is determined based on the matching degree between the tag information and / or the contextual relationship between the tag information.

[0153] Optionally, the second determining module is further configured to:

[0154] Determining, based on first user input information of a first round of conversation in the first conversation group, that the first conversation group is used to represent the scene label information;

[0155] Based on the scene label information of the first conversation group and the label information corresponding to the round conversation in the first conversation group, determining a matching result between the round conversation in the first conversation group and the scene label information of the first conversation group;

[0156] The association is determined based on the matching results.

[0157] Optionally, the second determining module is further configured to:

[0158] Acquire first time interval information between adjacent rounds of conversations in the first conversation group, and second time interval information between question and answer information and user input information in each round of conversations in the first conversation group;

[0159] The time information is determined based on the first time interval information and / or the second time interval information.

[0160] Optionally, the processing module is further configured to:

[0161] determining whether the first time interval information exceeds a second threshold, and whether the second time interval information exceeds a third threshold;

[0162] When the first time interval information exceeds the second threshold, and / or the second time interval information exceeds the third threshold, the round of the second conversation group is reduced.

[0163] Optionally, the processing module is further configured to:

[0164] determining a first data value for characterizing the association;

[0165] In a case where the first data value is greater than a first threshold, determining the content of the second conversation group based on the conversation topic content of the current conversation;

[0166] Based on the difference between the first data value and the first threshold, the round of the second conversation group is adjusted.

[0167] Optionally, the processing module is further configured to:

[0168] After the session with the user ends, if the user input information is re-acquired, determining whether the re-acquired user input information is related to the previous session;

[0169] If it is determined that there is no dependency, then another session different from the existing one is restarted.

[0170] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein the memory stores an executable program, and the processor executes the executable program to perform the steps of the method described above.

[0171] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. An information processing method for intelligent conversation, comprising: Determine a first conversation group for conversation with the user, wherein the first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information; determining the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation; Based on the correlation and / or the time information, a turn of the second conversation group to be output is determined.

2. The method according to claim 1, wherein determining the correlation between at least two rounds of conversations in the first conversation group comprises: determining keywords in each round of conversation of the first conversation group; Matching keywords in at least two rounds of conversations in the first conversation group to generate a matching result, wherein the matching result is used to characterize the similarity between the rounds of conversations; The association is determined based on the matching results.

3. The method according to claim 1, wherein determining the correlation between at least two rounds of conversations in the first conversation group comprises: Determining tag information for characterizing a conversation topic in each round of conversation of the first conversation group; Determine a matching degree between tag information corresponding to at least two rounds of conversations of the first conversation group, and / or a contextual relationship between tag information corresponding to at least two rounds of conversations; The association is determined based on the matching degree between the tag information and / or the contextual relationship between the tag information.

4. The method according to claim 3, wherein determining the correlation between at least two rounds of conversations in the first conversation group comprises: Determining, based on first user input information of a first round of conversation in the first conversation group, that the first conversation group is used to represent the scene label information; Based on the scene label information of the first conversation group and the label information corresponding to the round conversation in the first conversation group, determining a matching result between the round conversation in the first conversation group and the scene label information of the first conversation group; The association is determined based on the matching results.

5. The method according to claim 1, wherein determining the correlation between at least two rounds of conversations in the first conversation group and time information related to each round of conversations in the first conversation group during the conversation process comprises: Acquire first time interval information between adjacent rounds of conversations in the first conversation group, and second time interval information between question and answer information and user input information in each round of conversations in the first conversation group; The time information is determined based on the first time interval information and / or the second time interval information.

6. The method according to claim 5, wherein determining the round of the second conversation group to be output based on the time information comprises: determining whether the first time interval information exceeds a second threshold, and whether the second time interval information exceeds a third threshold; When the first time interval information exceeds the second threshold, and / or the second time interval information exceeds the third threshold, the round of the second conversation group is reduced.

7. The method according to claim 1, wherein determining the round of the second conversation group to be output based on the correlation and / or the time information comprises: determining a first data value for characterizing the association; In a case where the first data value is greater than a first threshold, determining the content of the second conversation group based on the conversation topic content of the current conversation; Based on the difference between the first data value and the first threshold, the round of the second conversation group is adjusted.

8. The method according to claim 1, further comprising: After the session with the user ends, if the user input information is re-acquired, determining whether the re-acquired user input information is related to the previous session; If it is determined that there is no dependency, then another session different from the existing one is restarted.

9. An information processing device for intelligent conversation, comprising: A first determination module is configured to determine a first conversation group for conversation with the user, wherein the first conversation group includes multiple rounds of conversations, each round of conversation is a conversation between user input information and corresponding question and answer information; a second determining module configured to determine the correlation between at least two rounds of conversations in the first conversation group, and time information related to each round of conversations in the first conversation group during the conversation; A processing module is configured to determine a turn of the second conversation group to be output based on the correlation and / or the time information. 10 . An electronic device comprising a memory and a processor, wherein an executable program is stored in the memory, and the processor executes the executable program to perform the steps of the method according to claim 1 .

Images