CN106599196B - Artificial intelligence dialogue method and system - Google Patents
Artificial intelligence dialogue method and system Download PDFInfo
- Publication number
- CN106599196B CN106599196B CN201611155286.6A CN201611155286A CN106599196B CN 106599196 B CN106599196 B CN 106599196B CN 201611155286 A CN201611155286 A CN 201611155286A CN 106599196 B CN106599196 B CN 106599196B
- Authority
- CN
- China
- Prior art keywords
- situation
- state
- dialogue
- user
- turn
- 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.)
- Active
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9032—Query formulation
- G06F16/90332—Natural language query formulation or dialogue systems
Abstract
The invention provides an artificial intelligence dialogue method and a system, wherein the artificial intelligence dialogue method comprises the following steps: acquiring multi-round dialogue information of an artificial intelligence dialogue system and a user; detecting situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model; the multi-turn dialogue situation tracking model determines the transition of situation states corresponding to the multi-turn dialogue according to the situation states corresponding to the multi-turn dialogue information; storing the transition of the context state in the multi-turn dialog context tracking model; deducing and estimating the current situation state faced by the user responded by the artificial intelligent dialogue system according to the stored situation state transition; and generating a proper response statement according to the current situation state, and replying the response statement to the user. The method fully considers the transfer of the situation state in the conversation process, can avoid the situations of incoherent semanteme, language obstruction and the like in the conversation process, and can improve the user experience.
Description
Technical Field
The invention relates to the technical field of intelligent conversation, in particular to an artificial intelligent conversation method and system.
Background
Today's society has entered an era of technology that is changing day by day. Among them, one of the most obvious signs is that high-tech products are widely introduced into the consumer market. They bring great convenience and enjoyment for work, life, communication, study and traffic of people. On the other hand, the popularization of high-tech products also brings great challenges to the promotion of the high-tech products, such as an artificial intelligence dialog system.
Most of the current artificial intelligence dialogue systems are dedicated to processing single-loop question-answer dialogue, and do not consider the transition of situation states between dialogue utterances, and cannot generate a strategy of robot dialogue according to the transition of situation states between dialogue utterances. The artificial intelligence dialogue communication is a series of processes of context-related and multi-topic situation state conversion, and the knowledge of the dialogue situation state transition is a very important basis for an answer mechanism. If the situation state transition between the conversation sentences is not considered, the situations of discontinuous conversation semanteme, language obstruction and the like can be generated in the conversation process, so that the times of carrying out multiple rounds of conversation between the user and the artificial intelligent conversation system can not be increased, and the use intention of the user is reduced.
Disclosure of Invention
In view of the above-mentioned drawbacks in the prior art, the present invention provides an artificial intelligence dialog method and system, which can detect the transition of the situation state during the dialog process, estimate the current situation state faced by the response user of the artificial intelligence dialog system according to the transition of the situation state, and select the response sentence conforming to the current situation state, so as to improve the response accuracy of the artificial intelligence dialog system and improve the user experience.
In a first aspect, the present invention provides an artificial intelligence dialog method, including:
acquiring multi-round dialogue information of an artificial intelligence dialogue system and a user;
detecting situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model;
the multi-turn dialogue situation tracking model determines the transition of situation states corresponding to the multi-turn dialogue information according to the situation states corresponding to the multi-turn dialogue information;
storing the transition of the context state in the multi-turn dialog context tracking model;
deducing and estimating the current situation state faced by the user responded by the artificial intelligent dialogue system according to the stored situation state transition;
and generating a proper response statement according to the current situation state, and replying the response statement to the user.
Optionally, the artificial intelligence dialog method further includes:
and (3) training a multi-round dialogue situation tracking model by adopting one or more of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule in advance.
Optionally, the estimating, according to the stored transition of the context state, a current context state faced by the user responded by the artificial intelligence dialog system includes:
predicting the possibility of the artificial intelligent dialog system responding to the situation state faced by the user according to the stored transfer of the situation state;
and selecting the situation state with the highest possible probability as the current situation state.
Optionally, the detecting the situation state corresponding to the multi-turn dialog information by using a pre-trained multi-turn dialog situation tracking model includes:
extracting the situation state characteristics corresponding to the multiple rounds of dialogue information by adopting one or more of the following modes:
extracting situation state characteristics of the text dialogue information corresponding to the multiple rounds of dialogue information by adopting a character analysis technology;
extracting the situation state characteristics of the voice conversation information corresponding to the multiple rounds of conversation information by adopting a voice recognition technology;
extracting the situation state characteristics of the image dialogue information corresponding to the multiple rounds of dialogue information by adopting an image recognition technology;
and detecting the situation state corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model according to the situation state characteristics.
Optionally, the context state includes: a combination of one or more of topic, intent, emotion, context.
Optionally, the generating a suitable response statement according to the current context state and replying to the user includes:
selecting at least one appropriate response sentence from an artificial intelligence response user list according to the current situation state;
predicting the possible probability of each suitable response sentence by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule;
and selecting the appropriate response sentence with the highest possible probability, and replying to the user.
In a second aspect, the present invention provides an artificial intelligence dialog system, including:
the dialogue information acquisition module is used for acquiring multi-round dialogue information between the artificial intelligence dialogue system and the user;
the detection module is used for detecting the situation state corresponding to the multi-round dialogue information by adopting a pre-trained multi-round dialogue situation tracking model;
the transfer determining module is used for determining the transfer of the situation states corresponding to the multi-turn dialogue information by the multi-turn dialogue situation tracking model according to the situation states corresponding to the multi-turn dialogue information;
a storage module to store the transition of the context state in the multi-turn dialog context tracking model;
an estimation module, configured to estimate, according to the stored transition of the context state, a current context state faced by the user responded by the artificial intelligence dialog system;
and the reply module is used for generating a proper response statement according to the current situation state and replying the response statement to the user.
Optionally, the system further includes:
and the training module is used for training the multi-round dialogue situation tracking model by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule in advance.
Optionally, the estimation module includes:
a situation state probability predicting unit, configured to predict a probability that the artificial intelligence dialog system will respond to the situation state faced by the user according to the stored transition of the situation state;
and the selecting situation state unit is used for selecting the situation state with the highest possible probability as the current situation state.
Optionally, the detecting module includes:
the feature extraction unit is used for extracting the situation state features corresponding to the multiple rounds of dialogue information by adopting one or more of the following modes:
extracting situation state characteristics of the text dialogue information corresponding to the multiple rounds of dialogue information by adopting a character analysis technology;
extracting the situation state characteristics of the voice conversation information corresponding to the multiple rounds of conversation information by adopting a voice recognition technology;
extracting the situation state characteristics of the image dialogue information corresponding to the multiple rounds of dialogue information by adopting an image recognition technology;
and the situation state detection unit is used for detecting the situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model according to the situation state characteristics.
Optionally, the context state includes: a combination of one or more of topic, intent, emotion, context.
Optionally, the reply module includes:
the response statement selection unit is used for selecting at least one appropriate response statement from an artificial intelligence response user list according to the current situation state;
a response sentence probability prediction unit for predicting a probability of each of the suitable response sentences by using one or a combination of a plurality of statistical learning algorithm, machine learning algorithm, deep learning algorithm, and artificial rule;
and the reply output unit is used for selecting the appropriate reply sentence with the highest possible probability and replying the reply sentence to the user.
According to the technical scheme, the invention provides an artificial intelligence conversation method, which comprises the following steps: acquiring multi-round dialogue information of an artificial intelligence dialogue system and a user; detecting situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model; the multi-turn dialogue situation tracking model determines the transition of situation states corresponding to the multi-turn dialogue according to the situation states corresponding to the multi-turn dialogue information; storing the transition of the context state in the multi-turn dialog context tracking model; deducing and estimating the current situation state faced by the user responded by the artificial intelligent dialogue system according to the stored situation state transition; and generating a proper response statement according to the current situation state, and replying the response statement to the user.
By the method, the situation states corresponding to the multiple rounds of dialogue information can be detected, the transition of the situation states corresponding to the multiple rounds of dialogue information is determined, the current situation state faced by the artificial intelligent dialogue system responding to the user can be estimated according to the transition of the situation states, and then a proper responding statement can be selected according to the current situation state, so that the situation states and the transition of the situation states in the dialogue process are fully considered, the situations of dialogue semantic incoherence, language obstruction and the like caused by the situation state transition are not considered, the times of multiple rounds of dialogue between the user and the artificial intelligent dialogue system can be increased, and the user experience and the use will of the user can be improved.
The artificial intelligence dialogue system provided by the invention has the same beneficial effects with the artificial intelligence dialogue method based on the same inventive concept.
Drawings
In order to more clearly illustrate the detailed description of the invention or the technical solutions in the prior art, the drawings that are needed in the detailed description of the invention or the prior art will be briefly described below. Throughout the drawings, like elements or portions are generally identified by like reference numerals. In the drawings, elements or portions are not necessarily drawn to scale.
FIG. 1 illustrates a flow chart of a first embodiment of the present invention providing an artificial intelligence dialog method;
FIG. 2 is a diagram illustrating a second embodiment of the present invention providing an artificial intelligence dialog system.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and therefore are only examples, and the protection scope of the present invention is not limited thereby.
It is to be noted that, unless otherwise specified, technical or scientific terms used herein shall have the ordinary meaning as understood by those skilled in the art to which the invention pertains.
For ease of understanding, the general inventive concept of the present invention is described herein:
the execution subject of the invention is an intelligent dialogue system which can be installed on carriers such as robots, intelligent terminals and the like. The invention provides an artificial intelligence dialogue method and system, which belong to the same invention concept on the whole, and the first step is to obtain the multi-turn dialogue information between the artificial intelligence dialogue system and the user; secondly, detecting situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model; thirdly, the multi-turn dialogue situation tracking model determines the transition of situation states corresponding to the multi-turn dialogue according to the situation states corresponding to the multi-turn dialogue information; a fourth step of storing the transition of the situation state in the multi-turn dialogue situation tracking model; fifthly, estimating the current situation state of the artificial intelligent dialogue system responding the user according to the stored situation state transition; and sixthly, generating a proper response statement according to the current situation state, and replying the response statement to the user. By utilizing the artificial intelligent dialogue method and the system, the transfer of the situation state in the dialogue process can be detected, the current situation state faced by the response user of the artificial intelligent dialogue system can be estimated according to the transfer of the situation state, and the transfer of the situation state in the dialogue process is fully considered, so that the current situation state can be more accurately determined, meanwhile, the response sentence which accords with the current situation state can be selected, the situations of incoherent semanteme, language obstruction and the like in the dialogue process can be avoided, the accuracy of the response user of the execution main body can be improved, the times of multi-round dialogue between the user and the execution main body can be further improved, the use desire of the user can be improved, and the user experience can be improved.
The invention provides an artificial intelligence dialogue method and system. Embodiments of the present invention will be described below with reference to the drawings.
Fig. 1 shows a flowchart of an artificial intelligence dialog method according to a first embodiment of the present invention. As shown in fig. 1, a first embodiment of the present invention provides an artificial intelligence dialog method, which includes the following steps:
step S101: and acquiring multi-round dialogue information of the artificial intelligence dialogue system and the user.
In this step, multi-turn dialog information of the artificial intelligence dialog system and the user is obtained, and the multi-turn dialog information comprises multi-turn dialog information with the user in a recent period, multi-turn dialog information in a current dialog turn, all dialog information with the user and the like. The dialog information includes: text dialog information, voice dialog information, image dialog information, or a combination of one or more thereof. For example, when a user has a conversation with a robot or an intelligent conversation system, the user inputs a picture and some text information and sends the picture and the text information to the robot or the intelligent conversation system; or the user sends a picture first and then inputs some characters, which are divided into two times. The user can input voice through a microphone on the robot or the intelligent dialogue system, so that the robot or the intelligent dialogue system obtains voice dialogue information with the user. For the acquisition of the image dialogue, the image dialogue information such as facial expressions, body postures and the like of the user can be acquired through the camera, and the image dialogue information can also be acquired through pictures input by the user.
Step S102: and detecting the situation state corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model.
Before this step, also include: and (3) training a multi-round dialogue situation tracking model by adopting one or more of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule in advance. The multi-turn dialog context tracking model comprises: a situation state detection model, a transition determination model and a current situation state estimation model. The contextual state comprising: a combination of one or more of a theme, an intent, an emotion, a context, etc. The context state detection model is trained according to one or more combinations of topics, intentions, emotions, contexts and the like of some conversations by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule. The context is representative words, such as "please help me", "i miss", etc.
In this step, when the context state detection model is used to detect the context state corresponding to the multiple rounds of session information, the context state characteristics of the multiple rounds of session information need to be used for detection. For example, if the situation status features of running, yoga, cycling, eating too much, weight rising, etc. appear in a session, the topic of the session can be detected as "losing weight" by using the situation status features. The situation state characteristics corresponding to the multiple rounds of dialogue information can be extracted by adopting one or more of the following modes: extracting situation state characteristics of the text dialogue information corresponding to the multiple rounds of dialogue information by adopting a character analysis technology; extracting the situation state characteristics of the voice conversation information corresponding to the multiple rounds of conversation information by adopting a voice recognition technology; and extracting the situation state characteristics of the image dialogue information corresponding to the multiple rounds of dialogue information by adopting an image recognition technology. And then, the situation state detection model analyzes the situation states corresponding to the multiple rounds of dialogue information according to the situation state characteristics. The multiple rounds of dialog information may correspond to one or more contextual states. The situation state detection model can obtain situation states corresponding to the multi-turn dialog information by corresponding the situation state characteristics with the situation states; one or more combinations of statistical learning algorithms, machine learning algorithms, deep learning algorithms, and artificial rules may also be employed to detect the contextual status. By the method for detecting the situation state, the information of one session can be comprehensively considered, and the accuracy of detecting the situation state can be improved.
In this step, the situation state corresponding to the multi-turn dialog information can also be directly detected according to the multi-turn dialog information. And converting the voice conversation information corresponding to the multiple rounds of conversation information into characters by adopting a voice recognition technology, and/or converting the image conversation information corresponding to the multiple rounds of conversation information into characters by adopting an image recognition technology. Then, in the situation state detection model, the artificial intelligence dialogue system detects the situation states corresponding to the multiple rounds of dialogue information by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule according to one or more combinations of characters corresponding to the voice, characters corresponding to the image and characters corresponding to the text. For example, in a session, "i am good fat, i am about to lose weight. "," you should do more. And if the characters are equal, the subject of the conversation can be directly judged to be weight loss according to the characters. The method for detecting the situation state reduces the extraction of situation state characteristics and can improve the detection speed, but the accuracy of detecting the situation state is not higher than that of the former method because the characters which can be analyzed are limited and too much information cannot be considered.
In the artificial intelligence dialog system, one or two of the above two methods for detecting the situation state can be adopted, which are all within the protection scope of the present invention.
In the process of detecting the situation state, the emotion detection model can adopt one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule to optimally train the emotion detection model according to the situation state, situation state characteristics, characters and the like, so that the artificial intelligent dialogue system can know users more and chat with the users more pertinently.
Step S103: and the multi-turn dialogue situation tracking model determines the transition of the situation states corresponding to the multi-turn dialogue information according to the situation states corresponding to the multi-turn dialogue information.
In this step, the determining of the transition of the context state corresponding to the multiple rounds of dialog information according to the context state corresponding to the multiple rounds of dialog information is performed in a transition determination model. The transfer determination model is also trained by adopting one or more combined methods of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule.
Wherein the context states corresponding to the multiple rounds of dialog information include at least one context state, and the transition of the context state includes: the situation state is not transferred, the situation state is transferred twice, the situation state is transferred three times and the like, and analyzing the transfer of the situation state further comprises analyzing the state from which the situation state is transferred to which state, so as to obtain the change rule of the situation state. And judging the transition of the situation state by analyzing the situation state corresponding to the multi-turn dialogue information. For example, in a period of multi-turn dialog, the situation state features of football, basketball, volleyball and the like appear, the theme of the multi-turn dialog is detected to be sports, and other themes are not involved in the period of dialog, so that the situation state can be judged not to be transferred. For example, in a session, the situation state features such as "football", "basketball", "volleyball" and the like appear, then the situation state features such as "rice", "sweet potato" and the like appear, two topics of the multiple sessions are detected, namely sports and eating respectively, and then the situation state can be judged to be transferred from sports to eating.
Step S104: storing the transition of the context state in the multi-turn dialog context tracking model.
In this step, the transfer of the situation state can be conveniently called later by storing the transfer of the situation state, the transfer of the situation state comprises some conversation characteristics of the user, the user can be known more by storing the conversation characteristics, and meanwhile, the transfer determination model can be optimally trained according to the transfer of the situation state, so that the performance of the transfer determination model is better, and the accuracy is higher.
In this step, the method also comprises the steps of storing the situation state, deleting the situation state and deleting the situation state. By storing the situation state, the situation state of the user in a period of time can be stored, and the situation state tracking model can be used for optimizing and training, so that the artificial intelligence system can know the user more. Deleting the situation state with the occurrence frequency lower than the preset frequency within a preset period of time. Deleting the transition of the situation state with the occurrence frequency lower than the preset frequency within a preset period of time. If one or more context states and/or transitions of context states occur less frequently than a preset frequency within a preset period of time, the context states and/or transitions of context states need to be deleted. By deleting the situation state and/or the transition of the situation state, the load of the artificial intelligence dialog system can be reduced, the memory occupied by the situation state and/or the transition of the situation state can be reduced, and the reaction speed of the artificial intelligence dialog system can be improved.
Step S105: and deducing and estimating the current situation state faced by the user responded by the artificial intelligent dialogue system according to the stored situation state transition.
In this step, first, the possible probability of the artificial intelligence dialog system responding to the situation state faced by the user needs to be predicted according to the stored transitions of the situation state, and then, the situation state with the highest possible probability is selected as the current situation state. This step is accomplished by the current context state estimation model. And according to the stored transfer of the situation state, predicting the possible probability of the artificial intelligent dialogue system responding to the situation state faced by the user by adopting a method of one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule, and then selecting the situation state with the highest possible probability as the current situation state. By the estimation method, the accuracy of predicting the current situation state can be improved.
For example, in a session, it is detected that the contextual state corresponding to the session is work or sports, and it can be predicted by the above method what the possible probability that the contextual state faced by the artificial intelligence session system is a contextual state such as eating, sleeping, etc., and if the possible probability of the contextual state of eating is higher than that of sleeping, eating is selected as the current contextual state.
When the situation state with the highest possible probability is selected, the situation states with the same probability may occur, and at this time, the situation states with the same probability may all be taken as the current situation state, and one of the situation states may also be randomly given as the current situation state. This is within the scope of the invention.
Wherein the situation state with the highest probability is the same as the latest situation state of the artificial intelligence dialog system, which indicates that the current situation state of the artificial intelligence dialog system is not transferred. For example, in a session, the user finally inputs "we go to play ball", detects that the current situation state is moving, and then predicts that the probability of the current situation state being moving is highest according to the above prediction method, so the system replies "do you play basketball? ", the context state has not changed in this session.
When estimating the current situation state, the current situation state can be directly estimated according to the transition of the situation states corresponding to the multiple rounds of dialog information, the transition of the situation states does not need to be stored, and the estimation process is similar to the step S105. And will not be described in detail herein.
When the current situation state is estimated, the current situation state can be directly estimated according to the situation states corresponding to the multi-turn dialogue information without using the transition of the situation states. According to the situation states corresponding to the multiple rounds of dialogue information, a method of one or a combination of multiple of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule can be directly adopted to predict the current situation state faced by the artificial intelligent dialogue system, and the current situation state can also be jointly judged according to the situation states corresponding to the multiple rounds of dialogue information and the transition of the situation states. The current context state may be the same as or different from the previous context state. This is within the scope of the invention.
By estimating the current situation state of the artificial intelligence dialog system responding to the user according to the stored situation state transition, the accuracy of predicting the current situation state can be improved, and the artificial intelligence dialog system is greatly helped when topics are automatically transferred. For example, in a session, the previous context states are working and tired, and then the context states when the artificial intelligence dialog system selects auto-shifting topics according to the above method are eating and sleeping. By this method, it is possible to prevent the user from feeling a sense of hesitation and thus losing the intention of conversation when the topic is automatically shifted.
When the user is in conversation with the artificial intelligence system, the situation state may not be transferred, and then the artificial intelligence system can reply in the situation state and can automatically transfer topics, which are within the protection scope of the invention.
Step S106: and generating a proper response statement according to the current situation state, and replying the response statement to the user.
In the step, firstly, according to the current situation state, at least one appropriate response statement is selected from an artificial intelligence response user list; then, predicting the possible probability of each appropriate response sentence by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule; and finally, selecting the appropriate response sentence with the highest possible probability, and replying to the user. The artificial intelligent answering user list is a responding sentence library established according to the contents of multiple chatting and the downloaded responding sentences. By generating an appropriate response sentence in this way, the accuracy of the response sentence can be increased.
In this step, an appropriate response sentence can be directly selected from the artificial intelligence response user list and returned to the user. The method is quick in recovery and small in occupied memory, but the accuracy is lower than that of the method.
In this step, one or more combination methods of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm, and an artificial rule may be further adopted to combine the characters to generate a suitable response sentence, and the response sentence is returned to the user. The method has large calculation amount and low recovery efficiency. The generation of the appropriate response statement according to the current situation state and the reply to the user can adopt one or more combinations of the three ways, and the invention is within the protection scope.
When the reply is sent to the user, the reply mode can be one or a combination of more of characters, images and voice, and the invention is within the protection scope of the invention. By using different reply modes for replying, the interest of the user in using the artificial intelligence dialog system can be improved, and the user experience can be improved.
In the first embodiment, an artificial intelligence dialog method is provided, and correspondingly, the application also provides an artificial intelligence dialog system. Please refer to fig. 2, which is a diagram illustrating an artificial intelligence dialog system according to a second embodiment of the present invention. Since the system embodiments are substantially similar to the method embodiments, they are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for relevant points. The system embodiments described below are merely illustrative.
A second embodiment of the present invention provides an artificial intelligence dialog system, including:
the dialogue information acquisition module 101 is used for acquiring multi-round dialogue information between the artificial intelligence dialogue system and the user;
the detecting module 102 is configured to detect a situation state corresponding to the multi-turn dialog information by using a pre-trained multi-turn dialog situation tracking model;
the transition determining module 103 is configured to determine, by the multi-turn dialog context tracking model, transition of context states corresponding to the multi-turn dialog information according to the context states corresponding to the multi-turn dialog information;
a storage module 104 for storing the transition of the context state in the multi-turn dialog context tracking model;
an estimation module 105, configured to estimate, according to the stored transition of the context state, a current context state faced by the user responded by the artificial intelligence dialog system;
and the reply module 106 is used for generating a suitable reply statement according to the current situation state and replying the reply statement to the user.
In an embodiment provided by the present invention, the system further includes:
and the training module is used for training the multi-round dialogue situation tracking model by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm, a deep learning algorithm and an artificial rule in advance.
In an embodiment provided by the present invention, the estimation module 105 includes:
a situation state probability predicting unit, configured to predict a probability that the artificial intelligence dialog system will respond to the situation state faced by the user according to the stored transition of the situation state;
and the selecting situation state unit is used for selecting the situation state with the highest possible probability as the current situation state.
In an embodiment of the present invention, the detecting module 102 includes:
the feature extraction unit is used for extracting the situation state features corresponding to the multiple rounds of dialogue information by adopting one or more of the following modes:
extracting situation state characteristics of the text dialogue information corresponding to the multiple rounds of dialogue information by adopting a character analysis technology;
extracting the situation state characteristics of the voice conversation information corresponding to the multiple rounds of conversation information by adopting a voice recognition technology;
extracting the situation state characteristics of the image dialogue information corresponding to the multiple rounds of dialogue information by adopting an image recognition technology;
and the situation state detection unit is used for detecting the situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model according to the situation state characteristics.
In one embodiment of the present invention, the context state includes: a combination of one or more of topic, intent, emotion, context.
In an embodiment provided by the present invention, the reply module 106 includes:
the response statement selection unit is used for selecting at least one appropriate response statement from an artificial intelligence response user list according to the current situation state;
a response sentence probability prediction unit for predicting a probability of each of the suitable response sentences by using one or a combination of a plurality of statistical learning algorithm, machine learning algorithm, deep learning algorithm, and artificial rule;
and the reply output unit is used for selecting the appropriate reply sentence with the highest possible probability and replying the reply sentence to the user.
The artificial intelligence dialogue system and the artificial intelligence dialogue method provided by the invention have the same inventive concept and the same beneficial effects, and are not repeated herein.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "for example," or "some examples," etc., mean 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 invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. 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. It is to be noted that the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer (which may be a personal computer, a server, or a network machine) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.
Claims (8)
1. An artificial intelligence dialog method, comprising:
acquiring multi-round dialogue information of an artificial intelligence dialogue system and a user;
detecting situation states corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model;
the multi-turn dialogue situation tracking model determines the transition of situation states corresponding to the multi-turn dialogue information according to the situation states corresponding to the multi-turn dialogue information;
storing the transition of the context state in the multi-turn dialog context tracking model;
deducing and estimating the current situation state faced by the user responded by the artificial intelligent dialogue system according to the stored situation state transition;
generating a proper response statement according to the current situation state, and replying the response statement to the user;
wherein the estimating the current situation state faced by the user responded by the artificial intelligence dialog system according to the stored transition of the situation state comprises:
predicting the possibility of the artificial intelligent dialog system responding to the situation state faced by the user according to the stored transfer of the situation state;
selecting the situation state with the highest possible probability as the current situation state;
deducing and estimating that the current situation state of the user is responded to by the artificial intelligent dialogue system according to the stored situation state transition and is finished by a current situation state deduction model; and predicting by adopting a method of one or more combinations of a statistical learning algorithm, a machine learning algorithm and an artificial rule according to the stored transfer of the situation state.
2. The method of claim 1, wherein before the step of detecting the situation status corresponding to the multi-turn dialog information by using the pre-trained multi-turn dialog situation tracking model, the method further comprises:
and (3) training a multi-round dialogue situation tracking model by adopting one or more of a statistical learning algorithm, a machine learning algorithm and an artificial rule in advance.
3. The method according to claim 1, wherein the detecting the context state corresponding to the multi-turn dialog information by using a pre-trained multi-turn dialog context tracking model comprises:
extracting the situation state characteristics corresponding to the multiple rounds of dialogue information by adopting one or more of the following modes:
extracting situation state characteristics of the text dialogue information corresponding to the multiple rounds of dialogue information by adopting a character analysis technology;
extracting the situation state characteristics of the voice conversation information corresponding to the multiple rounds of conversation information by adopting a voice recognition technology;
extracting the situation state characteristics of the image dialogue information corresponding to the multiple rounds of dialogue information by adopting an image recognition technology;
and detecting the situation state corresponding to the multi-turn dialogue information by adopting a pre-trained multi-turn dialogue situation tracking model according to the situation state characteristics.
4. The artificial intelligence dialog method of claim 1 wherein the contextual state comprises: a combination of one or more of topic, intent, emotion, context.
5. The method according to claim 1, wherein said generating an appropriate response sentence according to the current context state and replying to the user comprises:
selecting at least one appropriate response sentence from an artificial intelligence response user list according to the current situation state;
predicting the possible probability of each suitable response sentence by adopting one or more of a statistical learning algorithm, a machine learning algorithm and an artificial rule;
and selecting the appropriate response sentence with the highest possible probability, and replying to the user.
6. An artificial intelligence dialog system, comprising:
the dialogue information acquisition module is used for acquiring multi-round dialogue information between the artificial intelligence dialogue system and the user;
the detection module is used for detecting the situation state corresponding to the multi-round dialogue information by adopting a pre-trained multi-round dialogue situation tracking model;
the transfer determining module is used for determining the transfer of the situation states corresponding to the multi-turn dialogue information by the multi-turn dialogue situation tracking model according to the situation states corresponding to the multi-turn dialogue information;
a storage module to store the transition of the context state in the multi-turn dialog context tracking model;
an estimation module, configured to estimate, according to the stored transition of the context state, a current context state faced by the user responded by the artificial intelligence dialog system;
the reply module is used for generating a proper response statement according to the current situation state and replying the response statement to the user;
wherein the estimation module comprises:
a situation state probability predicting unit, configured to predict a probability that the artificial intelligence dialog system will respond to the situation state faced by the user according to the stored transition of the situation state;
and the selecting situation state unit is used for selecting the situation state with the highest possible probability as the current situation state.
7. The artificial intelligence dialog system of claim 6 wherein the system further comprises:
and the training module is used for training the multi-round dialogue situation tracking model by adopting one or more combinations of a statistical learning algorithm, a machine learning algorithm and an artificial rule in advance.
8. The artificial intelligence dialog system of claim 6 wherein the reply module comprises:
the response statement selection unit is used for selecting at least one appropriate response statement from an artificial intelligence response user list according to the current situation state;
a response sentence probability prediction unit for predicting a probability of each of the suitable response sentences by using one or a combination of a statistical learning algorithm, a machine learning algorithm, and an artificial rule;
and the reply output unit is used for selecting the appropriate reply sentence with the highest possible probability and replying the reply sentence to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611155286.6A CN106599196B (en) | 2016-12-14 | 2016-12-14 | Artificial intelligence dialogue method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611155286.6A CN106599196B (en) | 2016-12-14 | 2016-12-14 | Artificial intelligence dialogue method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106599196A CN106599196A (en) | 2017-04-26 |
| CN106599196B true CN106599196B (en) | 2021-07-16 |
Family
ID=58802451
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611155286.6A Active CN106599196B (en) | 2016-12-14 | 2016-12-14 | Artificial intelligence dialogue method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106599196B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133349B (en) * | 2017-05-24 | 2018-02-23 | 北京无忧创新科技有限公司 | One kind dialogue robot system |
| CN107193978A (en) * | 2017-05-26 | 2017-09-22 | 武汉泰迪智慧科技有限公司 | A kind of many wheel automatic chatting dialogue methods and system based on deep learning |
| CN109086282A (en) * | 2017-06-14 | 2018-12-25 | 杭州方得智能科技有限公司 | A kind of method and system for the more wheels dialogue having multitask driving capability |
| US20190122121A1 (en) * | 2017-10-23 | 2019-04-25 | AISA Innotech Inc. | Method and system for generating individual microdata |
| CN108268443B (en) * | 2017-12-21 | 2022-02-25 | 北京百度网讯科技有限公司 | Method and device for determining topic point transfer and acquiring reply text |
| CN108197274B (en) * | 2018-01-08 | 2020-10-09 | 合肥工业大学 | Abnormal personality detection method and device based on conversation |
| CN108197276B (en) * | 2018-01-08 | 2020-09-29 | 合肥工业大学 | Public emotion transfer distribution modeling method and device based on session |
| CN110659355A (en) * | 2018-06-29 | 2020-01-07 | 南京芝兰人工智能技术研究院有限公司 | Conversation control method and system |
| CN110874533B (en) * | 2018-08-14 | 2020-06-30 | 杭州智语网络科技有限公司 | Intention identification method based on keyword matching and response path after voice identification |
| CN109299237B (en) * | 2018-09-26 | 2020-06-16 | 苏州大学 | Cyclic network man-machine conversation method based on actor critic reinforcement learning algorithm |
| CN109522545B (en) * | 2018-10-11 | 2019-08-23 | 华东师范大学 | A kind of appraisal procedure that more wheels are talked with coherent property amount |
| CN110417637A (en) * | 2019-04-26 | 2019-11-05 | 成海林 | AI artificial intelligence augmentative communication technology |
| CN111108501B (en) * | 2019-12-25 | 2024-02-06 | 深圳市优必选科技股份有限公司 | Context-based multi-round dialogue method, device, equipment and storage medium |
| CN111475616B (en) * | 2020-03-13 | 2023-08-22 | 平安科技（深圳）有限公司 | Multi-round dialogue method and device based on dialogue state prediction and computer equipment |
| CN113495943B (en) * | 2020-04-02 | 2023-07-14 | 山东大学 | Man-machine dialogue method based on knowledge tracking and transferring |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101071418A (en) * | 2007-03-29 | 2007-11-14 | 腾讯科技（深圳）有限公司 | Chat method and system |
| CN101685454A (en) * | 2008-09-28 | 2010-03-31 | 华为技术有限公司 | Human-computer interactive method and system |
| CN104951428A (en) * | 2014-03-26 | 2015-09-30 | 阿里巴巴集团控股有限公司 | User intention recognition method and device |
| CN104965592A (en) * | 2015-07-08 | 2015-10-07 | 苏州思必驰信息科技有限公司 | Voice and gesture recognition based multimodal non-touch human-machine interaction method and system |
| CN105426436A (en) * | 2015-11-05 | 2016-03-23 | 百度在线网络技术（北京）有限公司 | Artificial intelligent robot based information provision method and apparatus |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6490698B1 (en) * | 1999-06-04 | 2002-12-03 | Microsoft Corporation | Multi-level decision-analytic approach to failure and repair in human-computer interactions |
| US20030158935A1 (en) * | 2002-02-15 | 2003-08-21 | Chaucer Chiu | Real-time situation report system and method |
| US8262474B2 (en) * | 2009-04-21 | 2012-09-11 | Mcmain Michael Parker | Method and device for controlling player character dialog in a video game located on a computer-readable storage medium |
| US20120296832A1 (en) * | 2011-05-16 | 2012-11-22 | Sap Ag | Defining agreements using collaborative communications |
| CN103617158A (en) * | 2013-12-17 | 2014-03-05 | 苏州大学张家港工业技术研究院 | Method for generating emotion abstract of dialogue text |
| EP2933071A1 (en) * | 2014-04-17 | 2015-10-21 | Aldebaran Robotics | Methods and systems for managing dialogs of a robot |
| CN104216954B (en) * | 2014-08-20 | 2017-07-14 | 北京邮电大学 | The prediction meanss and Forecasting Methodology of accident topic state |
| CN104899188A (en) * | 2015-03-11 | 2015-09-09 | 浙江大学 | Problem similarity calculation method based on subjects and focuses of problems |
| CN105094315B (en) * | 2015-06-25 | 2018-03-06 | 百度在线网络技术（北京）有限公司 | The method and apparatus of human-machine intelligence's chat based on artificial intelligence |
| CN106055662A (en) * | 2016-06-02 | 2016-10-26 | 竹间智能科技（上海）有限公司 | Emotion-based intelligent conversation method and system |
| CN106096717B (en) * | 2016-06-03 | 2018-08-14 | 北京光年无限科技有限公司 | Information processing method towards intelligent robot and system |
-
2016
- 2016-12-14 CN CN201611155286.6A patent/CN106599196B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101071418A (en) * | 2007-03-29 | 2007-11-14 | 腾讯科技（深圳）有限公司 | Chat method and system |
| CN101685454A (en) * | 2008-09-28 | 2010-03-31 | 华为技术有限公司 | Human-computer interactive method and system |
| CN104951428A (en) * | 2014-03-26 | 2015-09-30 | 阿里巴巴集团控股有限公司 | User intention recognition method and device |
| CN104965592A (en) * | 2015-07-08 | 2015-10-07 | 苏州思必驰信息科技有限公司 | Voice and gesture recognition based multimodal non-touch human-machine interaction method and system |
| CN105426436A (en) * | 2015-11-05 | 2016-03-23 | 百度在线网络技术（北京）有限公司 | Artificial intelligent robot based information provision method and apparatus |
Non-Patent Citations (1)
| Title |
|---|
| 信息检索系统中智能人机交互方法的研究;高海慧;《中国优秀硕士学位论文全文数据库 信息科技辑》;20141215(第 12 期);I138-473 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106599196A (en) | 2017-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106599196B (en) | Artificial intelligence dialogue method and system | |
| JP7210774B2 (en) | AVATOR BEHAVIOR CONTROL METHOD, DEVICE AND COMPUTER PROGRAM BASED ON TEXT | |
| CN107342078B (en) | Conversation strategy optimized cold start system and method | |
| CN106649704B (en) | Intelligent conversation control method and system | |
| JP7047087B2 (en) | Deep reinforcement learning framework for characterizing video content | |
| CN110364146B (en) | Speech recognition method, speech recognition device, speech recognition apparatus, and storage medium | |
| JP2021532515A (en) | Nomination method and equipment, computing equipment, and programs | |
| CN105704013A (en) | Context-based topic updating data processing method and apparatus | |
| CN111831798A (en) | Information processing method, information processing device, electronic equipment and computer readable storage medium | |
| CN110719525A (en) | Bullet screen expression package generation method, electronic equipment and readable storage medium | |
| CN109168003B (en) | Method for generating neural network model for video prediction | |
| CN110069612B (en) | Reply generation method and device | |
| CN111344717B (en) | Interactive behavior prediction method, intelligent device and computer readable storage medium | |
| CN112632242A (en) | Intelligent conversation method and device and electronic equipment | |
| Wang et al. | “Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens | |
| CN112183084A (en) | Audio and video data processing method, device and equipment | |
| CN110674398A (en) | Virtual character interaction method and device, terminal equipment and storage medium | |
| US20210081498A1 (en) | Artificial intelligence-based roleplaying experiences based on user-selected scenarios | |
| Cuayáhuitl et al. | Nonstrict hierarchical reinforcement learning for interactive systems and robots | |
| CN115809669B (en) | Dialogue management method and electronic equipment | |
| Kern et al. | Towards personalized mobile interruptibility estimation | |
| CN109670030B (en) | Question-answer interaction method and device | |
| JP5840186B2 (en) | Emotion search device, method and program | |
| CN114443821A (en) | Robot conversation method and device, electronic equipment and storage medium | |
| CN111506717B (en) | Question answering method, device, equipment and storage medium |
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 |