CN113076407A - Information processing method and device - Google Patents
Information processing method and device Download PDFInfo
- Publication number
- CN113076407A CN113076407A CN202110301308.XA CN202110301308A CN113076407A CN 113076407 A CN113076407 A CN 113076407A CN 202110301308 A CN202110301308 A CN 202110301308A CN 113076407 A CN113076407 A CN 113076407A
- Authority
- CN
- China
- Prior art keywords
- information
- emotion
- query
- intention
- understanding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
- G06F40/35—Discourse or dialogue representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Accounting & Taxation (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Human Computer Interaction (AREA)
- Computing Systems (AREA)
- Machine Translation (AREA)
Abstract
The application discloses an information processing method and device, wherein the method comprises the following steps: obtaining first query information; performing intention understanding on the first query information to obtain first intention information and first emotion information corresponding to the first intention information; performing emotion analysis on the first query information to obtain second emotion information; when the first emotion information and the second emotion information are the same, the first intention information is determined as second intention information corresponding to the first query information.
Description
Technical Field
The present application relates to the field of natural language processing technologies, and in particular, to an information processing method and apparatus.
Background
In the process of understanding the problem proposed by the user by the intelligent customer service, the situation that the intelligent customer service understands the user problem incorrectly can occur due to the limitation of the intelligent customer service on answer coverage and classification algorithm. For example, when the user inputs "insert earphone with sound", the intelligent customer service will naturally classify "insert earphone with sound" into "earphone without sound" because the words "earphone" and "sound" are strong signal words in the category of "earphone without sound". Thereby outputting a solution corresponding to "no sound of headphone". The solution can not solve the user problem, and the user experience is influenced.
Content of application
The embodiment of the application provides an information processing method and device, and the effect of accurately understanding the intention of a user is achieved.
One aspect of the present application provides an information processing method, including: obtaining first query information; performing intention understanding on the first query information to obtain first intention information and first emotion information corresponding to the first intention information; performing emotion analysis on the first query information to obtain second emotion information; and when the first emotion information and the second emotion information are the same, determining the first intention information as second intention information corresponding to the first query information.
In another possible embodiment, the method further comprises: and discarding the first intention information when the first emotion information and the second emotion information are different.
In another embodiment, the performing emotion analysis on the first query information to obtain second emotion information includes: determining a query object corresponding to the first query information; and determining second emotion information corresponding to the query object, wherein the second emotion information is objective emotion information.
In another possible embodiment, the first emotion information and the second emotion information are at least one of the following emotions: objective negative emotion corresponding to the query object, objective positive emotion corresponding to the query object, and objective neutral emotion corresponding to the query object.
In another possible embodiment, the method further comprises: and determining first response information according to the second intention information, wherein the first response information comprises a query object and a solution corresponding to the query object.
In another embodiment, the performing intent understanding on the first query information and obtaining first intent information and first emotion information corresponding to the first intent information includes: understanding the first query information through an intention understanding model to obtain the first intention information and the first emotion information; wherein the intention understanding model is a multi-classification model obtained through machine learning training.
In another embodiment, the performing emotion analysis on the first query information to obtain second emotion information includes: analyzing the first query information through an emotion analysis model to obtain second emotion information; the emotion classification model is a three-classification model, and the three-classification model is obtained through machine learning training.
In another possible embodiment, the method further comprises: and when it is determined that the first query information does not have the corresponding second intention information, determining second response information according to the first query information, wherein the second response information is chatting information.
In another embodiment, the determining the second response information according to the first query information includes: carrying out dialogue generation on the second emotion information and the first query information through a dialogue generation model to obtain second response information; wherein the dialogue generating model is obtained through machine learning training.
Another aspect of the present application provides an information processing apparatus, including: an obtaining module, configured to obtain first query information; the understanding module is used for carrying out intention understanding on the first query information to obtain first intention information and first emotion information corresponding to the first intention information; the analysis module is used for carrying out emotion analysis on the first query information to obtain second emotion information; a determining module, configured to determine the first intention information as second intention information corresponding to the first query information when the first emotion information and the second emotion information are the same.
In another possible embodiment, the apparatus further comprises: a discarding module, configured to discard the first intention information when the first emotion information and the second emotion information are different.
In another embodiment, the analysis module comprises: the determining submodule is used for determining a query object corresponding to the first query information; the determining submodule is further configured to determine second emotion information corresponding to the query object, where the second emotion information is objective emotion information.
In another possible embodiment, the first emotion information and the second emotion information are at least one of the following emotions: objective negative emotion corresponding to the query object, objective positive emotion corresponding to the query object, and objective neutral emotion corresponding to the query object.
In another embodiment, the determining module is further configured to determine first response information according to the second intention information, where the first response information includes a query object and a solution corresponding to the query object.
In another possible implementation, the understanding module includes: understanding the first query information through an intention understanding model to obtain the first intention information and the first emotion information; wherein the intention understanding model is a multi-classification model obtained through machine learning training.
In another embodiment, the analysis module comprises: analyzing the first query information through an emotion analysis model to obtain second emotion information; the emotion classification model is a three-classification model, and the three-classification model is obtained through machine learning training.
In another implementation manner, the determining module is further configured to determine, when it is determined that there is no corresponding second intention information for the first query information, second response information according to the first query information, where the second response information is chat information.
In another implementation, the determining module is further configured to perform dialog generation on the second emotion information and the first query information through a dialog generation model to obtain the second response information; wherein the dialogue generating model is obtained through machine learning training.
According to the information processing method and device, emotion analysis and intention understanding are carried out on the first query information, the first emotion information and the second emotion information are determined, and the first emotion information and the second emotion information are compared to determine whether the first intention information obtained through intention understanding can be used as second intention information corresponding to the first query information or not, so that the intention information can be accurately determined, the query requirements of a user are met, and the use experience of the user is improved.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present application will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the present application are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
in the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Fig. 1 is a schematic flow chart illustrating an implementation of an information processing method according to an embodiment of the present application;
fig. 2 is a schematic diagram of an implementation module of an information processing apparatus according to another embodiment of the present application.
Detailed Description
In order to make the objects, features and advantages of the present application more obvious and understandable, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Fig. 1 is a schematic diagram of an implementation flow of an information processing method according to an embodiment of the present application.
Referring to fig. 1, an aspect of the present application provides an information processing method, including: operation 101, obtaining first query information; operation 102, performing intention understanding on the first query information, and obtaining first intention information and first emotion information corresponding to the first intention information; operation 103, performing emotion analysis on the first query information to obtain second emotion information; in operation 104, in the case that the first emotion information and the second emotion information are the same, the first intention information is determined as second intention information corresponding to the first query information.
The information processing method provided by the method is applicable to the field of intelligent customer service, the intelligent customer service determines first emotion information corresponding to first intention information obtained through intention understanding and second emotion information corresponding to query information through emotion analysis and intention understanding of query information from a user, and determines whether the first intention information obtained through intention understanding can be used as the second intention information corresponding to the first query information through comparison of the first emotion information and the second emotion information. By the method, the intelligent customer service can understand the intention of the user more comprehensively, so that more accurate intention information corresponding to the first query information is determined, the query requirement of the user is met, and the use experience of the user is improved.
In operation 101, the first query information may be directly from the query information of the user, or may be information obtained by preprocessing the user. The first query message may be a voice message or a text message. The method does not limit the information content corresponding to the first query information, for example, according to different real-time scenes, the content of the first query information may be information related to the service field corresponding to the intelligent customer service, or information unrelated to the service field corresponding to the intelligent customer service. The first query information may or may not include information having substantial content.
In operation 102, the first query information may be subjected to intent understanding through a model trained by machine learning, the intent understanding of the first query information may also be realized through keyword matching, and the intent understanding of the first query information may also be realized by adopting other methods to obtain the first intent information and the first emotion information. Wherein the first intention information refers to user intention in the first query information, including but not limited to question processing intention, chatting intention, or other intention. The first emotion information is determined according to the first intention information, and the corresponding user emotion information may be preset according to the intention content of the first intention information, and specifically, the emotion polarity tag corresponding to the first intention information may be preset to form a corresponding relationship. According to the preset answer corresponding to the first query information, the method can simultaneously determine the first intention information and the first emotion information through intention understanding, and also can determine the first intention information through intention understanding, and then determine the corresponding first emotion information according to the first intention information. It is to be added that, when there are multiple intentions in the first query information of the method, the method may obtain multiple first intention information and corresponding multiple first emotion information through intention understanding.
According to the preset requirement, the emotion polarity label of the method is a multi-classification emotion polarity label, specifically, the specific classification quantity of … … in one of a two-classification emotion polarity label, a three-classification emotion polarity label and a four-classification emotion polarity label is determined according to the actual situation polarity, and details are not repeated below.
In a specific implementation scenario, when the emotion polarity tag is a two-class emotion polarity tag, the two-class emotion polarity tag can be classified into a positive emotion polarity tag and a non-positive emotion polarity tag, and can also be classified into a negative emotion polarity tag and a non-negative emotion polarity tag.
In another implementation scenario, when the emotion polarity tag is a three-class emotion polarity tag, the three-class emotion polarity tag can be classified into a Positive (Positive) emotion polarity tag, a Neutral (Neutral) emotion polarity tag and a Negative (Negative) emotion polarity tag.
In operation 103, the method may perform emotion analysis on the first query information through the emotion analysis model to obtain second emotion information; the method can also carry out sentiment analysis on the first query information through keyword analysis to obtain second sentiment information; the method can also map and determine the second emotion information according to the keywords, and the method can also adopt other emotion analysis methods to determine the second emotion information. It is to be supplemented that, according to the preset required first emotion information type, the second emotion information of the method may be emotion information for user emotion, or emotion information for an object in the user evaluation query information, and it is required to satisfy that the first emotion information and the second emotion information belong to the same type of emotion information. For example, when the user says "find a good or troublesome effect," the mouse is moved too sensitively, "if the first emotion information is emotion information for the emotion of the user, the second emotion information corresponds to the negative emotion polarity tag, and if the first emotion information is emotion information of an object in the user evaluation query information, the second emotion information corresponds to the positive emotion polarity tag. In a specific implementation scenario of the method, the emotion information of the object in the query information evaluated by the user is used as the first emotion information and the second emotion information, namely, the emotion information referred by the method can be objectively evaluated emotion information. It can be understood that, when the first query information of the method has a plurality of query objects, the method can obtain the second emotion information corresponding to each query object through emotion analysis, that is, can obtain a plurality of second emotion information.
It should be noted that, there is no requirement for the order between operation 102 and operation 103 in the method, operation 102 may be executed first, operation 103 may be executed first, or operation 102 and operation 103 may be executed synchronously.
In operation 104, the smart customer service determines whether intention information corresponding to the query object corresponds to the user intention by determining whether the first emotion information and the second emotion information corresponding to the same query object are the same. Specifically, when it is determined that the first emotion information and the second emotion information are the same, it may be considered that the first intention information obtained by the method for intention understanding of the first query information meets the needs of the user, that is, the accuracy of the first intention information is verified, and the first intention information is determined as the second intention information corresponding to the first query information. The method can directly output the second intention information as the answer corresponding to the first query information, and can also output the second intention information after further processing so that the user can obtain the answer corresponding to the first query information. Further processing includes, but is not limited to, audio-video generation conversion, image generation conversion, solution link generation, and the like. To facilitate a further understanding of the above real-time approach, a specific implementation scenario is provided below for illustration.
In the specific implementation scenario, when the intelligent customer service acquires the query information from the user that the computer has no sound but the earphone is inserted with sound, the intention information i, namely the loudspeaker has no sound, and the intention information i, namely the earphone has no sound, can be acquired by performing intention understanding on the query information. The first passive emotion polarity tag is corresponding to 'loudspeaker silence' according to the first intention information, and the second passive emotion polarity tag is corresponding to 'earphone silence' according to the second intention information. And performing emotion analysis on the query information to obtain a negative emotion polarity label three corresponding to 'no sound in computer' and a positive emotion polarity label four corresponding to 'sound in earphone'. By comparison, it can be determined that the tag one and the tag three corresponding to the intention information one "speaker is silent" are identical, and the intention is determined as the user intention. And the second label and the fourth label corresponding to the intention information one, namely 'no sound of earphone', are different, the intention is not the intention of the user. The method may further process the user intent to determine a solution corresponding to the user intent "horn silent" and output the solution to solve the user problem.
In another possible embodiment, the method further comprises: in operation 105, in the case that the first emotion information and the second emotion information are different, the first intention information is discarded.
When it is determined that the first emotion information and the second emotion information are different, that is, the first intention information obtained by intention understanding of the query object corresponding to the first emotion information and the second emotion information is not the user intention, in this case, the method may directly discard the intention information without further intention understanding of the query object. Under the condition that the first intention information is eliminated, intention understanding is conducted on the query information corresponding to the query object again, the third intention information and the corresponding third emotion information are determined again, and whether the third intention information is the user intention or not is determined by comparing the third emotion information with the second emotion information.
It should be added that, in some cases, when the user intends not to solve the purpose of the problem or there is no solution, the method may preset an answer corresponding to the purpose of not solving the problem and output the answer for feedback to the user. Namely, the answer output by the method aiming at the user intention can be written according to the actual situation.
In another possible implementation, the performing sentiment analysis on the first query information to obtain second sentiment information in operation 103 includes: firstly, determining a query object corresponding to first query information; then, second emotion information corresponding to the query object is determined, wherein the second emotion information is objective emotion information.
According to the method, the first emotion information and the second emotion information both refer to objective emotion information corresponding to a query object. The method specifically analyzes objective emotion information corresponding to a query object in the first query information under the condition of emotion analysis on the first query information, wherein the type of emotion information is targeted, and the polarity is well determined.
Specifically, the method determines a query object corresponding to the first query information. Specifically, the query object can be predicted through a machine learning training model, and can also be determined through keyword extraction. In one specific implementation, the query object is selected as a query object corresponding to a service domain of the intelligent customer service bureau. For example, for all sub-accessories or intelligent customer service related to computer accessories, when the query information is that "i can not control, the mouse movement is too sensitive", and the query object determined in the method is that "mouse related".
After the query object is determined, the method can determine second emotion information corresponding to the query object, specifically second objective emotion information, namely an emotion polarity tag used for evaluating the objective condition of the query object, if the query information is 'internet surfing speed is slow', the query object is 'network speed', and the emotion polarity tag of the query object is 'negative'. If the query information is 'earphone silence', the query object is 'earphone related', and the emotion polarity tag of the query object is 'negative'. If the query information is 'mouse movement sensitive', the query object is 'mouse relevant', and the emotion polarity label of the query object is 'positive'. If the query information is 'mouse use condition', the query object is 'mouse relevant', and the emotion polarity label of the query object is 'neutral'. If the query information is 'mouse unresponsive', the query object is 'mouse related', and the emotion polarity label of the query object is 'negative'. According to the method, the first emotion information is determined according to the first intention information, and the second emotion information is determined according to the first query information. And comparing the second emotion information with the first emotion information to determine whether the emotion corresponding to the first intention information accords with the emotion corresponding to the first query information, and if not, determining that the intention of the first query information is wrongly understood. The intention information may be discarded or redetermined. If so, the intention of the first query information is considered to be correctly understood, and subsequent operations including, but not limited to, determining an answer corresponding to the first query information may be performed based on the intention information. It can be understood that according to the classification condition of the emotion information and the number of the emotion polarity labels, the method can also have the condition of multi-classification emotion polarity labels such as two-classification emotion polarity labels, four-classification emotion polarity labels … and the like.
In another possible embodiment, the first emotion information and the second emotion information are at least one of the following emotions: objective negative feelings corresponding to the query object, objective positive feelings corresponding to the query object, and objective neutral feelings corresponding to the query object.
In the case where the first emotion information and the second emotion information are classified into emotions using the three classification polarity tags, there may be a case where:
when the polarity tag comprises an objective negative emotion corresponding to the query object, namely an objective negative emotion polarity tag, for example, when the first intention information corresponding to the user is 'internet surfing speed slow', the first emotion information corresponding to the first intention information is a negative emotion polarity tag.
When the polarity tag comprises objective positive emotion corresponding to the query object, namely an objective positive emotion polarity tag, and if the first intention information corresponding to the user is 'mouse movement sensitivity', the first emotion information corresponding to the first intention information is a positive emotion polarity tag.
When the polar tag includes an objective neutral emotion corresponding to the query object, if the first intention information corresponding to the user is the "mouse use method", the first emotion information corresponding to the first intention information is the neutral emotion polar tag.
In another possible implementation, the performing sentiment analysis on the first query information to obtain second sentiment information in operation 103 includes: analyzing the first query information through an emotion analysis model to obtain second emotion information; the emotion classification model is a three-classification model, and the three-classification model is obtained through machine learning training.
According to the method, three classification models are adopted to carry out sentiment analysis on the first query information, so that the first query information is identified based on the three classification polarity labels. The three-classification model is used for classifying the first query information into one of objective negative feelings corresponding to the query object, objective positive feelings corresponding to the query object and objective neutral feelings corresponding to the query object. The three-classification model can be obtained through training of training samples corresponding to the first query information and the emotion information.
In another possible embodiment, the method further comprises: and determining first response information according to the second intention information, wherein the first response information comprises a query object and a solution corresponding to the query object.
After determining the second intention information satisfying the user intention, the method may determine the first answer information according to the second intention information. The first answer information refers to answer information corresponding to the user query information. The first response information may be determined in advance by a technician, and when the second intention information is determined, the first response information corresponding to the second intention information is directly determined. The first response information may also be generated by a dialog generation model obtained by machine learning training, and corresponding first response information is generated by inputting the second intention information. The first response information may also be determined in advance by a technician, and when the second intention information is determined, the first response information corresponding to the second intention information is determined through keyword mapping.
According to different scenarios, the first response information includes the following conditions:
when the user query information includes a query object and a query question corresponding to the query object, the first response information includes the query object and a solution corresponding to the query question.
When the user query information does not include the query object but includes the specific query question corresponding to the query object, the query object may be determined according to the specific query question, the second intention information corresponding to the query information may be determined, and the first response information may be determined to include the query object and the solution corresponding to the query question according to the second intention information.
When the query information of the user includes a query object and a query question corresponding to the query object, but the query question cannot be solved, the first response information includes the query object and an answer corresponding to the query question, where the answer may be an answer with a comfort or chatty nature.
That is, the solution referred to by the method may be a solution that can solve the user query problem, or a solution that cannot solve the user problem.
In another implementation, the performing an intention understanding on the first query information and obtaining first intention information and first emotion information corresponding to the first intention information in operation 102 includes: understanding the first query information through an intention understanding model to obtain first intention information and first emotion information; wherein, the intention understanding model is a multi-classification model, and the multi-classification model is obtained through machine learning training.
The method can train the intention understanding model corresponding to the field of the intelligent customer service station service through machine learning, and the intention understanding model can comprise a plurality of models corresponding to the category of the query object and can also be a single model.
When the intention understanding model may include a plurality of models corresponding to categories of query objects, the method may first determine a query object corresponding to the first query information, and determine the intention understanding model corresponding to the query object according to the query object. For example, when the first query information is "slow internet speed", the query object is "network-related", and the first query information is input into an intention understanding model corresponding to the network-related query object for intention understanding, so as to determine corresponding intention information and emotion information. It is to be understood that, in the case where the intention understanding model may include a plurality of models corresponding to the categories of the query object, the training samples used by the intention understanding model of each category are selected as the training samples corresponding to the category. The intention understanding model corresponding to the network correlation is trained by adopting the corpus sample corresponding to the network correlation.
When the intention understanding model is a single model, the model is trained by the language material samples related to the field served by the intelligent customer service department, for example, the language material samples related to the after-sale of the computer are trained by the language material samples including the relevance categories of network speed, sound box, mouse, screen, keyboard and the like, so as to obtain the intention understanding model related to the after-sale of the computer. After the intelligent customer service obtains the query information from the user, the query information can be directly subjected to intention understanding through an intention understanding model so as to determine corresponding intention information and emotion information.
In another possible embodiment, the method further comprises: and when the first query information is determined to have no corresponding second intention information, determining second response information according to the first query information, wherein the second response information is chatting information.
In an implementation manner, the method may perform object extraction through the first query information to determine whether a preset query object exists in the first query information, and when the preset query object does not exist in the first query information, it may be determined that the first query information does not have corresponding second intention information, that is, the current user intention is unrelated to the service field corresponding to the intelligent customer service, and may perform intention prediction and emotion analysis on the first query information to determine the current subjective emotion of the user, and determine corresponding second response information, that is, the chat information, to feed back the user. The first query information can also be directly input into a user to perform dialog generation by a dialog generation model which is trained by machine learning and used for generating a chat dialog, so as to determine corresponding second response information, namely the chat information, so as to perform feedback on the user.
In another implementation scenario, the first query information has a query object, but when the first emotion information obtained after intention understanding is different from the second emotion information, the second response information may be determined according to the first query information, specifically, in order to further improve feedback accuracy, preset times of intention understanding may be set, and when the first emotion information obtained after preset times of intention understanding is different from the second emotion information all the time, it may be determined that there is no corresponding second intention information in the first query information, and at this time, the second response information is determined according to the first query information, and the second response information is chatting information. The generated second response message includes, but is not limited to, "do i not hear clearly", "please say again", "the question i do not understand well, help you to transfer to customer service? "and the like.
In another embodiment, determining the second response information according to the first query information includes: carrying out dialogue generation on the second emotion information and the first query information through a dialogue generation model to obtain second response information; wherein the dialogue generating model is obtained through machine learning training.
The dialogue generating model refers to a dialogue generating model used for generating chatting information and is obtained by training corpus samples corresponding to the second emotion information and the first query information. Based on the second emotion information, the chat conversation generated by the method can be closer to the query information of the user, and the use experience of the user is improved.
Fig. 2 is a schematic diagram of an implementation module of an information processing apparatus according to another embodiment of the present application.
Referring to fig. 2, another aspect of the present application provides an information processing apparatus, comprising: an obtaining module 201, configured to obtain first query information; an understanding module 202, configured to perform intention understanding on the first query information, and obtain first intention information and first emotion information corresponding to the first intention information; the analysis module 203 is configured to perform emotion analysis on the first query information to obtain second emotion information; a determining module 204, configured to determine the first intention information as second intention information corresponding to the first query information when the first emotion information and the second emotion information are the same.
In another possible embodiment, the apparatus further comprises: a discarding module 205, configured to discard the first intention information when the first emotion information and the second emotion information are different.
In another embodiment, the analysis module 203 comprises: the query object corresponding to the first query information is determined; and the system is also used for determining second emotion information corresponding to the query object, wherein the second emotion information is objective emotion information.
In another possible embodiment, the first emotion information and the second emotion information are at least one of the following emotions: objective negative feelings corresponding to the query object, objective positive feelings corresponding to the query object, and objective neutral feelings corresponding to the query object.
In another possible implementation, the determining module 204 is further configured to determine first response information according to the second intention information, where the first response information includes the query object and the solution corresponding to the query object.
In another possible implementation, the understanding module 202 includes: understanding the first query information through an intention understanding model to obtain first intention information and first emotion information; wherein, the intention understanding model is a multi-classification model, and the multi-classification model is obtained through machine learning training.
In another embodiment, the analysis module 203 comprises: analyzing the first query information through an emotion analysis model to obtain second emotion information; the emotion classification model is a three-classification model, and the three-classification model is obtained through machine learning training.
In another implementation, the determining module 204 is further configured to determine, when it is determined that there is no corresponding second intention information for the first query information, second response information according to the first query information, where the second response information is chat information.
In another implementation, the determining module 204 is further configured to perform dialog generation on the second emotion information and the first query information through a dialog generation model to obtain second response information; wherein the dialogue generating model is obtained through machine learning training.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. An information processing method, the method comprising:
obtaining first query information;
performing intention understanding on the first query information to obtain first intention information and first emotion information corresponding to the first intention information;
performing emotion analysis on the first query information to obtain second emotion information;
and when the first emotion information and the second emotion information are the same, determining the first intention information as second intention information corresponding to the first query information.
2. The method of claim 1, further comprising:
and discarding the first intention information when the first emotion information and the second emotion information are different.
3. The method of claim 1, wherein performing sentiment analysis on the first query information to obtain second sentiment information comprises:
determining a query object corresponding to the first query information;
and determining second emotion information corresponding to the query object, wherein the second emotion information is objective emotion information.
4. The method of claim 1, the first sentiment information and the second sentiment information being at least one of:
objective negative emotion corresponding to the query object, objective positive emotion corresponding to the query object, and objective neutral emotion corresponding to the query object.
5. The method of claim 1, further comprising:
and determining first response information according to the second intention information, wherein the first response information comprises a query object and a solution corresponding to the query object.
6. The method of claim 1, wherein the understanding the intent of the first query information, obtaining first intent information and first emotion information corresponding to the first intent information, comprises:
understanding the first query information through an intention understanding model to obtain the first intention information and the first emotion information;
wherein the intention understanding model is a multi-classification model obtained through machine learning training.
7. The method of claim 1, wherein performing sentiment analysis on the first query information to obtain second sentiment information comprises:
analyzing the first query information through an emotion analysis model to obtain second emotion information;
the emotion classification model is a three-classification model, and the three-classification model is obtained through machine learning training.
8. The method of claim 1, further comprising:
and when it is determined that the first query information does not have the corresponding second intention information, determining second response information according to the first query information, wherein the second response information is chatting information.
9. The method of claim 8, the determining second response information from the first query information comprising:
carrying out dialogue generation on the second emotion information and the first query information through a dialogue generation model to obtain second response information;
wherein the dialogue generating model is obtained through machine learning training.
10. An information processing apparatus, the apparatus comprising:
an obtaining module, configured to obtain first query information;
the understanding module is used for carrying out intention understanding on the first query information to obtain first intention information and first emotion information corresponding to the first intention information;
the analysis module is used for carrying out emotion analysis on the first query information to obtain second emotion information;
a determining module, configured to determine the first intention information as second intention information corresponding to the first query information when the first emotion information and the second emotion information are the same.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110301308.XA CN113076407B (en) | 2021-03-22 | 2021-03-22 | Information processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110301308.XA CN113076407B (en) | 2021-03-22 | 2021-03-22 | Information processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113076407A true CN113076407A (en) | 2021-07-06 |
| CN113076407B CN113076407B (en) | 2023-07-21 |
Family
ID=76613035
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110301308.XA Active CN113076407B (en) | 2021-03-22 | 2021-03-22 | Information processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113076407B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105893344A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | User semantic sentiment analysis-based response method and device |
| CN106599998A (en) * | 2016-12-01 | 2017-04-26 | 竹间智能科技（上海）有限公司 | Method and system for adjusting response of robot based on emotion feature |
| CN107944008A (en) * | 2017-12-08 | 2018-04-20 | 神思电子技术股份有限公司 | A kind of method that Emotion identification is carried out for natural language |
| CN110910898A (en) * | 2018-09-15 | 2020-03-24 | 华为技术有限公司 | Voice information processing method and device |
| CN110931006A (en) * | 2019-11-26 | 2020-03-27 | 深圳壹账通智能科技有限公司 | Intelligent question-answering method based on emotion analysis and related equipment |
| CN111199732A (en) * | 2018-11-16 | 2020-05-26 | 深圳Tcl新技术有限公司 | Emotion-based voice interaction method, storage medium and terminal equipment |
-
2021
- 2021-03-22 CN CN202110301308.XA patent/CN113076407B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105893344A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | User semantic sentiment analysis-based response method and device |
| CN106599998A (en) * | 2016-12-01 | 2017-04-26 | 竹间智能科技（上海）有限公司 | Method and system for adjusting response of robot based on emotion feature |
| CN107944008A (en) * | 2017-12-08 | 2018-04-20 | 神思电子技术股份有限公司 | A kind of method that Emotion identification is carried out for natural language |
| CN110910898A (en) * | 2018-09-15 | 2020-03-24 | 华为技术有限公司 | Voice information processing method and device |
| CN111199732A (en) * | 2018-11-16 | 2020-05-26 | 深圳Tcl新技术有限公司 | Emotion-based voice interaction method, storage medium and terminal equipment |
| CN110931006A (en) * | 2019-11-26 | 2020-03-27 | 深圳壹账通智能科技有限公司 | Intelligent question-answering method based on emotion analysis and related equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113076407B (en) | 2023-07-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108509619B (en) | Voice interaction method and device | |
| CN106446045B (en) | User portrait construction method and system based on dialogue interaction | |
| CN107767869B (en) | Method and apparatus for providing voice service | |
| CN104598644B (en) | Favorite label mining method and device | |
| WO2019084810A1 (en) | Information processing method and terminal, and computer storage medium | |
| CN108388553B (en) | Method for eliminating ambiguity in conversation, electronic equipment and kitchen-oriented conversation system | |
| CN109241332B (en) | Method and system for determining semantics through voice | |
| CN106847279A (en) | Man-machine interaction method based on robot operating system ROS | |
| CN111901627B (en) | Video processing method and device, storage medium and electronic equipment | |
| Kaushik et al. | Automatic sentiment detection in naturalistic audio | |
| CN112417158A (en) | Training method, classification method, device and equipment of text data classification model | |
| CN111881283A (en) | Business keyword library creating method, intelligent chat guiding method and device | |
| JP2008537396A (en) | Apparatus and method for user communication in a communication system | |
| CN114722822B (en) | Named entity recognition method, named entity recognition device, named entity recognition equipment and named entity recognition computer readable storage medium | |
| CN117079299A (en) | Data processing method, device, electronic equipment and storage medium | |
| CN113806501B (en) | Training method of intention recognition model, intention recognition method and equipment | |
| CN112786028A (en) | Acoustic model processing method, device, equipment and readable storage medium | |
| CN112069786A (en) | Text information processing method and device, electronic equipment and medium | |
| CN113076407A (en) | Information processing method and device | |
| CN113255368A (en) | Method and device for emotion analysis of text data and related equipment | |
| CN114547264A (en) | News diagram data identification method based on Mahalanobis distance and comparison learning | |
| CN114242045A (en) | Deep learning method for natural language dialogue system intention | |
| CN114297380A (en) | Data processing method, device, equipment and storage medium | |
| CN114676705A (en) | Dialogue relation processing method, computer and readable storage medium | |
| CN111985238A (en) | Answer generation method and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |