A conversation processing method and device

Abstract

The present invention discloses a dialogue processing method and device, the method comprising: obtaining first input information from a user, the first input information being used to represent first intention information; outputting first feedback information corresponding to the first input information, the first feedback information being used to answer the first intention information; obtaining second input information from the user, the second input information being used to represent second intention information, the second intention information being correlated with the first intention information; generating second feedback information when it is evaluated that the first feedback information can answer the first input information; outputting the second feedback information, the second feedback information being used to interpret the first feedback information; applying this method can optimize the dialogue process between intelligent customer service and users, so that the intelligent customer service can better solve problems for users.

Description

Technical Field

[0001] The present invention relates to the field of intelligent customer service technology, and in particular to a conversation processing method and device. Background Art

[0002] The service group of the intelligent customer service system is the users, especially the intelligent customer service that aims to solve certain tasks for users. When designing an intelligent system, it is often through a period of analysis, from the perspective of problem discovery and problem solving, to design a dialogue process system that is considered to be more reasonable to guide users to solve problems. However, different users have different experiences and ideas, which will affect users' perception and understanding of intelligent customer service, resulting in different choices and inputs made by different users in the dialogue process. In this case, it is easy for intelligent customer service and users to have deviations in their judgment of the degree of user understanding during the interaction process. Summary of the invention

[0003] The embodiments of the present invention provide a dialogue processing method and device, which can optimize the dialogue process between intelligent customer service and users, so that the intelligent customer service can better solve problems for users.

[0004] On the one hand, an embodiment of the present invention provides a dialogue processing method, which includes: obtaining first input information from a user, the first input information is used to represent first intention information; outputting first feedback information corresponding to the first input information, the first feedback information is used to answer the first intention information; obtaining second input information from the user, the second input information is used to represent second intention information, the second intention information is correlated with the first intention information; when it is evaluated that the first feedback information can answer the first input information, generating second feedback information; outputting the second feedback information, the second feedback information is used to interpret the first feedback information.

[0005] In one possible implementation, generating the second feedback information includes: obtaining second intention information corresponding to the second input information; obtaining answer dimension information corresponding to the first feedback information; evaluating the user's understanding of the first feedback information based on the second intention information and the answer dimension information; and generating the second feedback information based on the user's understanding of the first feedback information.

[0006] In one embodiment, the answer dimension information includes at least one of the following dimensions: a first dimension for characterizing the helpfulness of the answer, a second dimension for characterizing the degree of solution of the answer, a third dimension for characterizing the clarity of the answer, a fourth dimension for characterizing the degree of confusion of the conversation, and a fifth dimension for the user to characterize the degree of detail of the answer.

[0007] In one possible implementation, the evaluating the user's understanding of the first feedback information based on the second intention information and the answer dimension information includes: scoring the second intention information and the answer dimension information to obtain a first score; determining a preset understanding range corresponding to the first score, the preset understanding range being used to characterize the user's understanding of the first feedback information; wherein the understanding level includes at least a first understanding level characterizing a lack of understanding and a second understanding level characterizing a complete understanding.

[0008] In one possible implementation, second feedback information is generated based on the user's degree of understanding of the first feedback information, including: when the user does not understand the first feedback information, determining the user's third intention information based on the first input information, the first feedback information, and the second input information; and generating second feedback information based on the third intention information and the first intention information for interpreting the first feedback information.

[0009] On the other hand, an embodiment of the present invention provides a dialogue processing device, which includes: an acquisition module, used to obtain first input information from a user, the first input information is used to represent first intention information; an output module, used to output first feedback information corresponding to the first input information, the first feedback information is used to answer the first intention information; the acquisition module is also used to obtain second input information from the user, the second input information is used to represent second intention information, and the second intention information is correlated with the first intention information; a generation module, used to generate second feedback information when it is evaluated that the first feedback information can answer the first input information; the output module is used to output the second feedback information, and the second feedback information is used to interpret the first feedback information.

[0010] In one possible implementation, the generation module includes: an acquisition submodule, used to obtain second intention information corresponding to the second input information; the acquisition submodule, also used to obtain answer dimension information corresponding to the first feedback information; an evaluation submodule, used to evaluate the user's understanding of the first feedback information based on the second intention information and the answer dimension information; and a generation submodule, used to generate second feedback information based on the user's understanding of the first feedback information.

[0011] In one embodiment, the answer dimension information includes at least one of the following dimensions: a first dimension for characterizing the helpfulness of the answer, a second dimension for characterizing the degree of solution of the answer, a third dimension for characterizing the clarity of the answer, a fourth dimension for characterizing the degree of confusion of the conversation, and a fifth dimension for the user to characterize the degree of detail of the answer.

[0012] In one possible implementation, the evaluation submodule includes: a scoring unit, used to score the second intention information and the answer dimension information to obtain a first score; a determination unit, used to determine a preset understanding range corresponding to the first score, and the preset understanding range is used to characterize the user's understanding degree of the first feedback information; wherein the understanding degree includes at least a first understanding degree representing a lack of understanding and a second understanding degree representing a complete understanding.

[0013] In one possible implementation, the generation submodule includes: a determination unit, used to determine the user's third intention information based on the first input information, the first feedback information, and the second input information when the user does not understand the first feedback information; and a generation unit, used to generate second feedback information based on the third intention information and the first intention information, so as to interpret the first feedback information.

[0014] When a dialogue processing method and device provided by an embodiment of the present invention is applied, first input information for representing a user's problem is obtained, and then first feedback information is generated and output to provide feedback on the first input information. When it is determined that the first feedback information can answer the first input information and second input information is obtained from the user, second feedback information is generated and output based on the first input information and the second input information, and the second feedback information is used to interpret the first feedback information. The application of this method can optimize the dialogue process between the intelligent customer service and the user, so that the intelligent customer service can better solve problems for the user, and can indirectly guide the user to understand the intelligent customer service to learn how to use it, paving the way for better use of the intelligent customer service in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0016] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0017] Figure 1 A schematic diagram of an implementation flow of a conversation processing method according to an embodiment of the present invention;

[0018] Figure 2 A schematic diagram of an implementation flow of determining the degree of understanding in a method for processing a dialogue according to an embodiment of the present invention;

[0019] Figure 3 The present invention is a schematic diagram of an implementation module of a conversation processing device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] Figure 1 The present invention is a flowchart of a method for processing a conversation according to an embodiment of the present invention.

[0022] See also Figure 1 On the one hand, an embodiment of the present invention provides a dialogue processing method, the method comprising: operation 101, obtaining first input information from a user, the first input information is used to represent first intention information; operation 102, outputting first feedback information corresponding to the first input information, the first feedback information is used to answer the first intention information; operation 103, obtaining second input information from the user, the second input information is used to represent second intention information, and the second intention information is correlated with the first intention information; operation 104, when it is evaluated that the first feedback information can answer the first input information, generating second feedback information; operation 105, outputting second feedback information, the second feedback information is used to interpret the first feedback information.

[0023] The dialogue processing method provided by this method can be applied to intelligent customer service. In this method, the intelligent customer service obtains the first input information used to characterize the user's problem, and then generates and outputs the first feedback information to provide feedback on the first input information. When the intelligent customer service determines that the first feedback information can answer the first input information, when the intelligent customer service obtains the second input information from the user, the second feedback information is generated and output according to the first input information and the second input information. The second feedback information is used to interpret the first feedback information. The user can combine the first feedback information and the second feedback information to answer the first input information. The application of this method can optimize the dialogue process between the intelligent customer service and the user, so that the intelligent customer service can better solve the problem for the user, and can indirectly guide the user to understand the intelligent customer service to learn how to use it, paving the way for better use of the intelligent customer service in the future.

[0024] In operation 101 of the method, the intelligent customer service obtains first input information from the user. The user can input the first input information to the intelligent customer service by voice input or text input, and the intelligent customer service extracts information from the first input information to obtain first intention information in the first input information, where the first intention information is the intention corresponding to the first input information, including but not limited to the intention of chatting and the intention of solving problems.

[0025] In operation 102 of the present method, the intelligent customer service generates and outputs the first feedback information corresponding to the first input information according to the set dialogue processing method. Specifically, when the first intention information is a small talk intention, the first feedback information can be a small talk feedback content corresponding to the small talk topic; for example, when the first input information is "hello", the first feedback information can be "hello, how can I help you?". When the first intention information is a problem solving intention, the first feedback information can be the answer to the question corresponding to the user's question. It should be added that, according to the actual situation of the problem and the information storage of the intelligent customer service, the answer to the question may or may not solve the user's problem. For example, when the user's question is an answer that the intelligent customer service does not know, the first feedback information generated and output by the intelligent customer service can be "Sorry, I don't know for the time being." Further, the present method does not limit the output method of the first feedback information of the intelligent customer service, including but not limited to one or more of voice output, text output, picture output, video output, and hyperlink format output.

[0026] It should be added that the first input information and the first feedback information referred to in this method can be one or more. In one feasible implementation mode, when the intelligent customer service evaluates that the first feedback information can solve the first input information, the intelligent customer service will generate multiple feedback information to solve the first input information. At this time, there will be multiple rounds of dialogues between the intelligent customer service and the user, and these dialogues can all be regarded as the first input information and the first feedback information referred to in this method.

[0027] In operation 103 of the present method, after outputting the first feedback information, second input information from the user is obtained. The second input information is used to characterize the second intent information. Similarly, the second intent information may be a chat intention or a problem-solving intention. It is understandable that the second intent information may be correlated with the first intent information or may not be correlated with the first intent information. The correlation here can be determined by known conversation correlation analysis methods such as correlation templates, correlation analysis models, and intent understanding models. It should be noted that the present method may determine whether the second intent information is correlated with the first intent information after executing the evaluation that the first feedback information can answer the first input information, or it may determine whether the second intent information is correlated with the first intent information before executing the evaluation that the first feedback information can answer the first input information.

[0028] In operation 104 of the present method, the first input information and the first feedback information are evaluated to determine whether the first feedback information can answer the first input information. It should be noted that the answer to the question represented by the first feedback information may or may not be able to solve the first input information. The evaluation criteria of whether the first input information can be answered in this method can be set as whether the intelligent customer service has provided known information for the first input information. When the first feedback information is evaluated to be able to answer the first input information, the second feedback information is generated. It should be supplemented that when the second intent information is correlated with the first intent information, the second feedback information is generated based on the first input information and the second input information; when the second intent information is not correlated with the first intent information, the second feedback information can be generated based on the second intent information.

[0029] In operation 105 of the present method, second feedback information is output, and the second feedback information is used to interpret the first feedback information so that the user can more fully understand the answer to the question corresponding to the first input information in combination with the first feedback information and the second feedback information. It should be added that when there is no correlation between the second intention information and the first intention information, the second feedback information may have no association with the first feedback information, that is, the intelligent customer service can communicate with the user in a normal dialogue mode.

[0030] Figure 2 The present invention is a flowchart of a method for processing a conversation to determine the degree of understanding according to an embodiment of the present invention.

[0031] See also Figure 2 In one possible implementation, operation 105, generating second feedback information, includes: operation 1051, obtaining second intention information corresponding to the second input information; operation 1052, obtaining answer dimension information corresponding to the first feedback information; operation 1053, evaluating the user's understanding of the first feedback information based on the second intention information and the answer dimension information; operation 1054, generating second feedback information based on the user's understanding of the first feedback information.

[0032] When the first feedback information can be used to answer the first input information and the first input information and the second input information are correlated, the specific method of generating the second feedback information by the method includes: in operation 1051 and operation 1052, obtaining the second intention information corresponding to the second input information, and obtaining the answer dimension information corresponding to the first feedback information. Among them, the second intention information can be the intention information only for the second input information, that is, the second intention information is only used to evaluate the current intention of the second feedback information. The second intention information can also be information used to evaluate the current intention of the second input information and also to evaluate the correlation between the second input information and the first input information. The specific evaluation dimension of the second intention information can be adjusted according to actual needs. The answer dimension information is used to characterize all feedback information of the user in the dimension of question answering during the dialogue between the user and the intelligent customer service for the first input information and the first feedback information. Including but not limited to information such as user satisfaction, user understanding, user attitude, etc. That is, the information obtained based on the dimensional analysis of the first input information and the first feedback information is the answer dimension information; the information obtained by the intention analysis of the second input information and the first input information is the second intention information.

[0033] In operation 1053, by evaluating the user's understanding of the first feedback information, it can be understood whether the dialogue between the intelligent customer service and the user is misaligned, that is, whether the user can understand the information fed back by the intelligent customer service. It can be understood that when the evaluation shows that the user's understanding of the first feedback information is low, it can be considered that the user cannot understand or cannot fully understand the first feedback information, that is, the user cannot answer the question through the first feedback information. When the evaluation shows that the user's understanding of the first feedback information is high, it can be considered that the user can understand or roughly understand the first feedback information, that is, the user can answer the question through the first feedback information. The second intention information and the answer dimension information are combined to evaluate the user's understanding of the first feedback information based on this. The evaluation dimension basis and the evaluation information can evaluate whether the user understands the answer in many aspects, which can fully understand the user's understanding of the answer, which is conducive to improving the accuracy of the evaluation and avoiding evaluation errors.

[0034] In operation 1054, the second feedback information is generated according to the user's understanding of the first feedback information, which can supplement the content that the user does not understand in the first feedback information in a targeted manner, so that the user can have a more complete understanding of the first feedback information. Specifically, the understanding of the first feedback information can be determined as the user's understanding of each word in the first feedback information, and / or the first feedback information can also be determined as the user's understanding of the answer represented by the first feedback information. The supplement to the first feedback information can be an explanation of the words that the user does not understand in the first feedback information, or a further detailed description of the answer in the first feedback information. For example: in one implementation, when the user asks "What is the specific parameter information of mobile phone X?" for the first time, the intelligent customer service sends a hyperlink information named "parameter information" to the user as feedback information. The user does not understand and needs to click on the hyperlink information to see all the parameter information. After the intelligent customer service sends the hyperlink information named "parameter information" to the user, it can be evaluated as having answered the user's question. At this time, the user has a low understanding of the hyperlink information named "parameter information" sent by the intelligent customer service. In this case, the user may ask a second question "What is the specific parameter information of mobile phone X?" When the intelligent customer service determines that the user's first question has been answered, and the intelligent customer service knows from the second question and the first question and the corresponding feedback information that the user cannot understand the content of the first feedback information of the intelligent customer service, the intelligent customer service can send a text message or a picture message containing specific parameters to the user.

[0035] In one embodiment, the answer dimension information includes at least one of the following dimensions: a first dimension for characterizing the helpfulness of the answer, a second dimension for characterizing the degree of solution of the answer, a third dimension for characterizing the clarity of the answer, a fourth dimension for characterizing the degree of confusion of the conversation, and a fifth dimension for the user to characterize the degree of detail of the answer.

[0036] The specific dimensions of the answer dimension information of this method can be selected and adjusted as needed. Furthermore, in order to further improve the accuracy of the evaluation, this method can add an operation of collecting answer dimensions in the interaction between the first input information and the first feedback information. Specifically, the intelligent customer service can collect answer dimensions after replying to the answer corresponding to the first intention information. For example, the intelligent customer service can generate output information such as "Is the above answer helpful?", "Is the above communication clear?", "Please rate the above communication (1 / 100)" and other output information. It should be added that in order to facilitate classification and integration, each dimension information can be represented in a binary classification form through normalization processing, such as "first dimension → yes", "second dimension → no"; or represented in a scoring form, such as "first dimension → 20 points, second dimension → 0 points".

[0037] The answer dimension information is collected through the evaluation corresponding to the user feedback. To facilitate the understanding of the above answer dimension information, the dimension information in several cases is provided below for specific explanation.

[0038] The first dimension can be used to characterize the degree of help to users from the answers provided by the intelligent customer service. The first dimension can be evaluated by a model corresponding to the answer help evaluation. The answer help evaluation model can be obtained by training the classifier with the corresponding training data. The training data can mark the degree of help of the answer. It can also be done by asking the user whether the user's answer is helpful. It can also combine the model evaluation results and user feedback information to generate the first dimension information of the corresponding answer help. According to the needs, the dimension of evaluating the degree of help of the answer can have only two categories, such as helpful and unhelpful; it can also form multiple categories, for example, under helpful, it can be further classified into somewhat helpful, helpful, very helpful, etc.

[0039] The second dimension can be used to characterize the degree to which the answers provided by the intelligent customer service solve the user's problem. It should be clear that, unlike the first dimension, the first dimension focuses more on the user's subjective evaluation, and the second dimension can focus more on the objective evaluation of whether the answer can solve the user's problem. For example, when the first feedback information is "I can't answer, please ask xxx", although it may be helpful to the user, it cannot solve the user's problem, and the degree to which it solves the problem may be "unsolved". Similarly, the second dimension can be evaluated by a model corresponding to the answer solution evaluation. The answer solution evaluation model can be obtained by training the classifier with the corresponding training data, and the training data can be annotated with the answer solution degree; it can also be determined by asking the user to answer the answer to determine whether the problem can be solved, and the second dimension information corresponding to the answer solution degree can be generated by combining the model evaluation results and the user feedback information. Among them, according to needs, the dimension of evaluating the degree of answer solution can also have only two categories, such as solution and no solution; it can also form multiple categories, for example, under the solution dimension, it can be further classified into solving less than 20%, solving 21-50%, solving more than 50% but not reaching 100%, and completely solving.

[0040] The third dimension can be used to evaluate the clarity of the answers provided by the intelligent customer service. The clarity can be determined by asking the user whether it is clear and the user answering the answer to determine whether the question is clear. Further, the clarity can include the user's evaluation of whether the intelligent customer service is truly clear about the intention represented by the user's first input information, and can also include the user's evaluation of whether the user can clearly understand the intention represented by the first feedback information of the intelligent customer service. The third dimension information can be determined by combining the two angles of whether the intelligent customer service is truly clear about the intention represented by the user's first input information and whether the user can clearly understand the intention represented by the first feedback information of the intelligent customer service.

[0041] The fourth dimension can be used to evaluate the degree of confusion in the conversation between the user and the intelligent customer service. The evaluation method for the degree of confusion can be based on time information as the sorting, and each round of user input information and intelligent customer service output information is used as a set of data for evaluation to evaluate whether there is confusion between the user input information and the customer service output information of this round. Similarly, the intelligent customer service can also ask questions corresponding to the degree of confusion in the conversation, and the degree of confusion in the conversation can be determined by the user input.

[0042] The fifth dimension can be used to evaluate the level of detail of the answers provided by the intelligent customer service. The level of information can be evaluated according to the number of words in the intelligent customer service's answers, the amount of information in the intelligent customer service's answers, and other dimensions. The intelligent customer service can also ask questions corresponding to the level of answer information, and the user can input and determine the level of answer information.

[0043] It is clear that, according to needs, the answer dimension information referred to in the embodiment of the present method may also include a sixth dimension, a seventh dimension, an eighth dimension, etc. that represent the evaluation of answers in different dimensions, which will not be elaborated below.

[0044] In one possible implementation, operation 1053, evaluating the user's understanding of the first feedback information based on the second intent information and the answer dimension information, includes: first, scoring the second intent information and the answer dimension information to obtain a first score; then, determining a preset understanding range corresponding to the first score, the preset understanding range is used to characterize the user's understanding of the first feedback information; wherein the understanding level includes at least a first understanding level representing a lack of understanding and a second understanding level representing a complete understanding.

[0045] In operation 1053, the method first scores the user's understanding of the first feedback information by comprehensively analyzing the second intent information and the answer dimension information. The scoring process may first normalize the second intent information and the answer dimension information respectively, and then integrate the scores to obtain a first score. The first score may also be obtained by predicting a neural network model, which may be obtained by training with training data corresponding to the second intent information and the answer dimension information.

[0046] Then, the first score is matched with the preset understanding range to determine the user's understanding of the first feedback information. Among them, the preset understanding range can be two or more. When the preset understanding range is two, it can include a first understanding degree representing a lack of understanding and a second understanding degree representing complete understanding. For example, if the value range of the first score is 0 to 100, 0 to 50 can be used as the first understanding degree corresponding to lack of understanding, and 51 to 100 can be used as the second understanding degree corresponding to understanding. When the score obtained by scoring the second intent information and the answer dimension information is 40, within the range of 0 to 59, it can be considered that the user does not understand the first feedback information. When the score is 60, within the range of 60 to 100, it can be considered that the user understands the first feedback information. Furthermore, in order to make the evaluation of the user's understanding level more detailed, more than two preset understanding ranges can be set. For example, when the score is between 0 and 10, it can be determined that the user does not understand the first feedback information; when the score is between 11 and 30, it can be determined that the user can understand a small amount of the first feedback information; when the score is between 31 and 60, it can be determined that the user understands half of the first feedback information; when the score is between 61 and 90, it can be determined that the user understands most of the first feedback information; and when the score is between 91 and 100, it can be determined that the user basically or completely understands the first feedback information.

[0047] In one possible implementation, operation 1054 generates second feedback information based on the user's understanding of the first feedback information, including: first, when the user does not understand the first feedback information, determining the user's third intention information based on the first input information, the first feedback information, and the second input information; then, generating the second feedback information based on the third intention information and the first intention information for interpreting the first feedback information.

[0048] In operation 1054, first, when the user's understanding of the first feedback information is not understood according to the first score, that is, for the corresponding user, the first feedback information cannot answer the user's question, at this time, the intelligent customer service can re-determine the user's third intention information based on the first input information, the first feedback information, and the second input information. It can be understood that the third intention information is determined based on the first input information, the first feedback information, and the second input information, and the number of inputs is larger, which helps to more accurately determine the user's intention. Among them, the third intention information can be the same or partially the same as the first intention information and the second intention information, and the third intention information can also be a supplementary intention for the first feedback information. In the case where the degree of understanding between the intelligent customer service and the user is low, the third intention information can also be completely different from the first intention information and the second intention information. Then, the second feedback information is generated based on the third intention information and the first intention information, and the second feedback information is output for information interpretation of the first feedback information. For example, "What model of mobile phone am I using?", the first feedback information is "X", and the intelligent customer service evaluates that the first feedback information has answered the question of the first input information; after that, the user enters the second input information, such as "What X", it can be seen that the second input information is used for the user to express his or her misunderstanding of the first feedback information. The intelligent customer service can answer the question in other ways using the first input information, the first feedback information, and the second input information, such as "The model of mobile phone you are using is X".

[0049] It should be added that, when the user's understanding of the first feedback information is understood according to the first score, the second input information can be used as a new dialogue process, that is, the second feedback information can be feedback on the second input information, but it is irrelevant to the first input information. For example, the first input information is used to represent a question, such as "What is the model of the mobile phone?", and the first feedback information is "The model of the mobile phone is X", and the first feedback information has answered the question of the first input information; afterwards, the user enters the second input information, such as "OK", it can be seen that the second input information is used for small talk, and the intelligent customer service can generate a second feedback information only for the second input information, such as "Is there anything else I can help with?" to feedback the second input information.

[0050] To facilitate the understanding of the above implementation, a specific implementation scenario is provided below for illustration. In this scenario, the method is applied to the intelligent customer service of a mobile phone, i.e., bot, to answer questions of users.

[0051] For example, during a conversation, the following conversation took place:

[0052] Dialogue(1)[13:50:18.029 user(text)]:my phone is moto g6 plus I want to change my fingerprint sensor by moto service center.

[0053] Dialogue (2) [13:50:22.563 bot(richtext_question)]: [manage fingerprints].

[0054] Dialogue(3)[13:50:22.563 bot(richtext_related_questions)]:[fingerprintsensor to navigate][fingerprint sensor not working].

[0055] Dialogue (4) [13:50:22.563 bot(button)]: [helpful][unhelpful].

[0056] Dialogue (5) [13:50:52.490 user (quickresponse)]: helpful.

[0057] Dialogue (6)[13:50:52.583 bot(text)]:great to hear that worked.can i helpwith anything else? .

[0058] Dialogue (7) [13:50:52.583 bot (button)]: [yes] [no].

[0059] Dialogue (8) [13:51:06.921 user (text)]: no.

[0060] Dialogue (9) [13:51:07.079 bot (survey)]: ​​survey.

[0061] Dialogue (10) [13:51:37.099 user (survey)]: ​​[0] [yes] [0].

[0062] Dialogue(11)[13:51:37.138 bot(text)]:thanks for your feedback! .

[0063] Dialogue(12)[13:52:12.824 user(text)]:madam you can't respond my query.

[0064] As can be seen from the above, dialogue (1) is a question raised by the user, that is, the first input information, and dialogues (2)-(3) are the bot's answers to the user's questions and feedback on related questions, that is, the first feedback information. Third, it should be noted that in dialogue (2), the bot provides a hyperlink text, and the user clicks on the content of dialogue (2) to jump to the fingerprint management platform. Dialogue (4) and dialogue (6)-dialogue (11) are all bots asking users to obtain answer dimension information. Dialogue (12) is the second evaluation information input by the user. After the bot completes dialogue (11), it determines that the user's question has been answered. When the user enters dialogue (12), the bot obtains the current intention of dialogue (12) and integrates all feedback information on the answer dimension of dialogue (1)-dialogue (11) to obtain answer dimension information. Then, the answer dimension information and the current intention are used for scoring, and the scoring results are used to determine whether the dialogue between the bot and the user is misaligned, that is, whether the user can understand the bot's feedback information and whether the bot understands the user's intention. According to the above dialogue, the user cannot understand the bot's answer to the question in dialogue (2). At this time, the bot generates and outputs dialogue (13) [13:50:22.563 bot (richtext_question)]: Please click [manage fingerprints] and you will change the fingerprint sensor.

[0065] Figure 3 The present invention is a schematic diagram of an implementation module of a conversation processing device according to an embodiment of the present invention.

[0066] See also Figure 3 On the other hand, an embodiment of the present invention provides a dialogue processing device, which includes: an acquisition module 301, which is used to obtain first input information from a user, and the first input information is used to represent first intention information; an output module 302, which is used to output first feedback information corresponding to the first input information, and the first feedback information is used to answer the first intention information; the acquisition module 301 is also used to obtain second input information from the user, and the second input information is used to represent second intention information, and the second intention information is correlated with the first intention information; a generation module 303 is used to generate second feedback information when it is evaluated that the first feedback information can answer the first input information; the output module 302 is used to output the second feedback information, and the second feedback information is used to interpret the first feedback information.

[0067] In one embodiment, the generation module 303 includes: an acquisition submodule 3031, used to obtain second intention information corresponding to the second input information; the acquisition submodule 3031 is also used to obtain answer dimension information corresponding to the first feedback information; the evaluation submodule 3032 is used to evaluate the user's understanding of the first feedback information based on the second intention information and the answer dimension information; the generation submodule 3033 is used to generate the second feedback information based on the user's understanding of the first feedback information.

[0068] In one embodiment, the answer dimension information includes at least one of the following dimensions: a first dimension for characterizing the helpfulness of the answer, a second dimension for characterizing the degree of solution of the answer, a third dimension for characterizing the clarity of the answer, a fourth dimension for characterizing the degree of confusion of the conversation, and a fifth dimension for the user to characterize the degree of detail of the answer.

[0069] In one embodiment, the evaluation submodule 3032 includes: a scoring unit 30321, used to score the second intention information and the answer dimension information to obtain a first score; a first determination unit 30322, used to determine a preset understanding range corresponding to the first score, and the preset understanding range is used to characterize the user's understanding of the first feedback information; wherein the understanding level includes at least a first understanding level representing a lack of understanding and a second understanding level representing a complete understanding.

[0070] In one embodiment, the generation submodule 3033 includes: a second determination unit 30331, which is used to determine the user's third intention information based on the first input information, the first feedback information and the second input information when the user does not understand the first feedback information; a generation unit 30332, which is used to generate second feedback information based on the third intention information and the first intention information, so as to interpret the first feedback information.

[0071] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0072] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0073] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for processing a conversation, characterized in that: The method comprises: Obtaining first input information from a user, where the first input information is used to represent first intention information; Outputting first feedback information corresponding to the first input information, wherein the first feedback information is used to answer the first intention information; Obtaining second input information from a user, where the second input information is used to represent second intent information, and the second intent information is correlated with the first intent information; When it is evaluated that the first feedback information can answer the first input information, generating second feedback information; outputting the second feedback information, where the second feedback information is used to interpret the first feedback information; The generating the second feedback information comprises: Obtaining second intention information corresponding to the second input information; Obtaining answer dimension information corresponding to the first feedback information; evaluating the user's understanding of the first feedback information according to the second intention information and the answer dimension information; Second feedback information is generated according to the user's understanding of the first feedback information.

2. The method according to claim 1, characterized in that The answer dimension information includes at least one of the following dimensions: a first dimension used to characterize the helpfulness of the answer, a second dimension used to characterize the degree of solution of the answer, a third dimension used to characterize the clarity of the answer, a fourth dimension used to characterize the degree of confusion of the conversation, and a fifth dimension used by the user to characterize the degree of detail of the answer.

3. The method according to claim 1, characterized in that: The evaluating the user's understanding of the first feedback information according to the second intention information and the answer dimension information includes: Scoring the second intention information and the answer dimension information to obtain a first score; Determining a preset understanding range corresponding to the first score, where the preset understanding range is used to represent the user's understanding of the first feedback information; The degree of understanding includes at least a first degree of understanding representing a lack of understanding and a second degree of understanding representing a complete understanding.

4. The method according to claim 3, characterized in that Generating second feedback information according to the user's understanding of the first feedback information includes: When the user does not understand the first feedback information, determining the third intention information of the user according to the first input information, the first feedback information and the second input information; Second feedback information is generated according to the third intention information and the first intention information, so as to interpret the first feedback information.

5. A conversation processing device, characterized in that: The device comprises: An obtaining module, used to obtain first input information from a user, where the first input information is used to represent first intention information; an output module, configured to output first feedback information corresponding to the first input information, wherein the first feedback information is used to answer the first intention information; The obtaining module is further used to obtain second input information from the user, where the second input information is used to represent second intent information, and the second intent information is correlated with the first intent information; A generating module, configured to generate second feedback information when it is evaluated that the first feedback information can answer the first input information; an output module, configured to output the second feedback information, where the second feedback information is used to interpret the first feedback information; The generating module comprises: An obtaining submodule, used to obtain second intention information corresponding to the second input information; The obtaining submodule is further used to obtain answer dimension information corresponding to the first feedback information; an evaluation submodule, configured to evaluate the user's understanding of the first feedback information according to the second intention information and the answer dimension information; A generating submodule is used to generate second feedback information according to the user's understanding degree of the first feedback information.

6. The device according to claim 5, characterized in that The answer dimension information includes at least one of the following dimensions: a first dimension used to characterize the helpfulness of the answer, a second dimension used to characterize the degree of solution of the answer, a third dimension used to characterize the clarity of the answer, a fourth dimension used to characterize the degree of confusion of the conversation, and a fifth dimension used by the user to characterize the degree of detail of the answer.

7. The device according to claim 5, characterized in that The evaluation submodule comprises: a scoring unit, configured to score the second intention information and the answer dimension information to obtain a first score; A determination unit is used to determine a preset understanding range corresponding to the first score, wherein the preset understanding range is used to represent the user's understanding degree of the first feedback information; wherein the understanding degree at least includes a first understanding degree representing a lack of understanding and a second understanding degree representing a complete understanding.

8. The device according to claim 7, characterized in that The generating submodule comprises: a determining unit, configured to determine third intention information of the user according to the first input information, the first feedback information, and the second input information when the user does not understand the first feedback information; A generating unit is used to generate second feedback information according to the third intention information and the first intention information, so as to interpret the first feedback information.

Images