CN107273406A - Dialog process method and device in task dialogue system - Google Patents
Dialog process method and device in task dialogue system Download PDFInfo
- Publication number
- CN107273406A CN107273406A CN201710289869.6A CN201710289869A CN107273406A CN 107273406 A CN107273406 A CN 107273406A CN 201710289869 A CN201710289869 A CN 201710289869A CN 107273406 A CN107273406 A CN 107273406A
- Authority
- CN
- China
- Prior art keywords
- dialogue
- task
- epicycle
- action
- last round
- 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
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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Information Transfer Between Computers (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The invention discloses the dialog process method and device in a kind of task dialogue system, wherein, this method includes：Obtain epicycle user message；Obtain the dialogue action of task dialogue system feedback in last round of dialogue；According to the dialogue action of task dialogue system feedback in epicycle user message and last round of dialogue, the dialogue action of task dialogue system feedback in generation epicycle dialogue；Show the dialogue action of task dialogue system feedback in epicycle dialogue.Dialog process method in task dialogue system provided in an embodiment of the present invention, the dialogue action talked with reference to the dialogue action generation epicycle of upper wheel dialogue, dialogue state need not be pre-defined for the task dialogue system of different field so that this method has very strong cross-cutting transfer ability.
Description
Technical field
The present invention relates to a kind of dialog process method in human-computer interaction technique field, more particularly to task dialogue system and
Device.
Background technology
Conversational system (Dialogue System) is a kind of man-machine interactive system based on natural language.By talking with system
System, people can use natural language and computer to carry out many wheel interactions to complete specific task, such as information inquiry, service acquisition
Deng.Conversational system provides a kind of more natural, easily man-machine interaction mode, is widely used in the fields such as vehicle-mounted, household, customer service
Scape.
Wherein, conversational system can be divided into three classes chat conversations system, question answer dialog system and task according to usage scenario
Conversational system.In chat conversations system, machine can what is said or talked about according to user, carries out related reply, allows and chats under progress
Go, but chat has no specific purpose in itself.Chat conversations system is generally used in the scenes such as chat robots, is entertained for people
With kill time.Question answer dialog system then to answer user automatically the problem of as target, answer comes from specific knowledge base, lead to
Often in the form of question-response, it is adaptable to the scene such as search engine.Task dialogue system is more more multiple than both the above conversational system
It is miscellaneous, the purpose is to allow machine to carry out many wheel interactions with people, understand the intention of people, and help people to complete certain task.Task
Conversational system is generally used for the scenes such as intelligent assistant, helps user to complete task, such as inquiry day by way of taking turns dialogue more
Gas, managing schedule, make a reservation.
Task dialogue system is relatively complicated in conversational system, mainly there is following challenge.First, Task
Dialogue generally many wheels, this require system can when each round interact, according to user it is current what is said or talked about with context come
Comprehensive descision simultaneously performs rational action.Therefore, it is very crucial to the modeling of context in Task conversational system.Secondly, chat
Its conversational system needs only assure that the correlation of reply, interest, dialogue can be allowed to continue, and Task dialogue is most
Whole purpose is to aid in user and goes to perform specific task, and this requirement to accuracy is very high.Finally, different types of task it
Between have very big difference, how to allow Task conversational system that there is cross-cutting versatility, the problem of be one very challenging.
Common Task conversational system generally uses the scheme based on state machine.Specifically, according to the characteristics of task,
Conversational system defines a variety of dialogue states.In each dialogue state, system has corresponding executable action.Institute
Meaning is acted, i.e., specific message, calls application interface (Application are such as replied in the behavior that system can be taken
Programming Interface, API) etc. belong to action.Based on this, the working method of Task conversational system is as follows：
Under current dialogue states, system according to system after current what is said or talked about the rational action of selection one, the action executing of user according to
The state transition rule defined in state machine jumps to next dialogue state, waits user's input next time.
However, inventor has found that the conversational system based on state machine at least has following ask during the present invention is realized
Topic：First, the definition of state depends on domain knowledge, it is necessary to which professional person puts into substantial amounts of energy to design.Secondly, for multiple
Miscellaneous task, state is often very more, and the state machine designed can be extremely complex, it is difficult to safeguards.Finally, times of different field
The corresponding state of business conversational system differs widely, and will redesign state machine for different fields, not possess cross-cutting
Versatility.Therefore, many wheel conversational systems based on state machine are often only applicable to relatively simple task scene.
The content of the invention
It is contemplated that at least solving one of technical problem in correlation technique to a certain extent.
Therefore, it is an object of the present invention to propose a kind of dialog process method in task dialogue system, this method
The dialogue action talked with reference to the dialogue action generation epicycle of upper wheel dialogue, it is not necessary to for the task dialogue system of different field
Pre-defined dialogue state so that this method has very strong cross-cutting transfer ability.
Second object of the present invention is to propose the dialog process device in a kind of task dialogue system.
Third object of the present invention is to propose a kind of server.
The present invention the 4th aims at a kind of computer-readable recording medium of proposition.
For up to above-mentioned purpose, first aspect present invention embodiment proposes the dialog process side in a kind of task dialogue system
Method, including：Obtain epicycle user message；Obtain the dialogue action of task dialogue system feedback described in last round of dialogue；According to
The dialogue action of task dialogue system feedback described in the epicycle user message and last round of dialogue, institute in generation epicycle dialogue
State the dialogue action of task dialogue system feedback；Show the dialogue action of task dialogue system feedback described in epicycle dialogue.
Dialog process method in task dialogue system according to embodiments of the present invention, obtains epicycle user message, obtains
The dialogue action of task dialogue system feedback in last round of dialogue, according to task dialogue in epicycle user message and last round of dialogue
The dialogue action of system feedback, the dialogue action of task dialogue system feedback in generation epicycle dialogue, and display epicycle dialogue
The dialogue action of middle task dialogue system feedback, thus, is moved with reference to the dialogue that the dialogue action generation epicycle that upper wheel is talked with is talked with
Make, it is not necessary to which the task dialogue system for different field pre-defines dialogue state, and this method can apply to any field
Task dialogue, that is to say, that this method has very strong cross-cutting transfer ability.
It is preferred that, pair of the task dialogue system feedback according to the epicycle user message and last round of dialogue
Words are acted, the dialogue action of task dialogue system feedback described in generation epicycle dialogue, including：According to last round of dialogue
The dialogue action of task dialogue system feedback and the epicycle user message, extract reality related to task in the epicycle dialogue
Body information, and the corresponding entity vector of result generation is extracted according to entity；The epicycle user message is carried out at text vector
Reason, to generate the text vector of the epicycle user message；Obtain the hidden vector of last round of dialogue；Depth based on training in advance
Neural network model, the institute according to the text vector, the entity be vectorial, in last round of dialogue hidden vectorial, last round of dialogue
Task dialogue system feedback described in the dialogue action vector generation epicycle dialogue for the dialogue action for stating task dialogue system feedback
Dialogue action.
It is preferred that, the deep neural network model is that the Recognition with Recurrent Neural Network based on long short-term memory (LSTM) is built in advance
Vertical.
It is preferred that, the dialogue action of the task dialogue system feedback according to last round of dialogue and the epicycle are used
Family message, extracts the entity information of the epicycle dialogue, including：Task dialogue system is anti-according to the last round of dialogue
The dialogue action of feedback and the epicycle user message determine the task of user；Judge in the epicycle user message whether there is with
The entity key of the related entity elements of the task；If in the presence of extracting the epicycle from the epicycle user message
The entity information related to task in dialogue.
For up to above-mentioned purpose, second aspect of the present invention embodiment proposes the dialog process dress in a kind of task dialogue system
Put, including：First acquisition module, for obtaining epicycle user message；Second acquisition module, for obtaining institute in last round of dialogue
State the dialogue action of task dialogue system feedback；Generation module, for according in the epicycle user message and last round of dialogue
The dialogue action of the task dialogue system feedback, the dialogue action of task dialogue system feedback described in generation epicycle dialogue；
Display module, the dialogue action for showing task dialogue system feedback described in epicycle dialogue.
Dialog process device in task dialogue system according to embodiments of the present invention, obtains epicycle user message, obtains
The dialogue action of task dialogue system feedback in last round of dialogue, according to task dialogue in epicycle user message and last round of dialogue
The dialogue action of system feedback, the dialogue action of task dialogue system feedback in generation epicycle dialogue, and display epicycle dialogue
The dialogue action of middle task dialogue system feedback, thus, is moved with reference to the dialogue that the dialogue action generation epicycle that upper wheel is talked with is talked with
Make, it is not necessary to which the task dialogue system for different field pre-defines dialogue state, and this method can apply to any field
Task dialogue, that is to say, that this method has very strong cross-cutting transfer ability.
It is preferred that, the generation module, including：Extracting unit, for the task dialogue system according to last round of dialogue
The dialogue action of system feedback and the epicycle user message, extract entity information related to task in the epicycle dialogue, and
Result is extracted according to entity and generates corresponding entity vector；Text vector unit, for being carried out to the epicycle user message
Text vector processing, to generate the text vector of the epicycle user message；Acquiring unit, for obtaining the hidden of last round of dialogue
Vector；Generation unit, for the deep neural network model based on training in advance, according to the text vector, the entity to
Amount, the dialogue of the dialogue action of task dialogue system feedback described in hidden vectorial, the last round of dialogue of last round of dialogue act to
The dialogue action of task dialogue system feedback described in amount generation epicycle dialogue.
It is preferred that, the deep neural network model is that the Recognition with Recurrent Neural Network based on long short-term memory (LSTM) is built in advance
Vertical.
It is preferred that, the extracting unit, specifically for：The task dialogue system feedback according to the last round of dialogue
Dialogue action and the epicycle user message determine the task of user；Judge to whether there is and institute in the epicycle user message
State the related entity key of task；If in the presence of, extracted from the epicycle user message in epicycle dialogue with task
Related entity information.
For up to above-mentioned purpose, third aspect present invention embodiment proposes a kind of server, including：One or more processing
Device；Storage device, for storing one or more programs, when one or more of programs are by one or more of processors
Perform so that one or more of processors are realized at the dialogue in the task dialogue system of first aspect present invention embodiment
Reason method.
Fourth aspect present invention embodiment provides a kind of computer-readable recording medium, is stored thereon with computer journey
Sequence, it is characterised in that the program is realized when being executed by processor in the task dialogue system of first aspect present invention embodiment
Dialog process method.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Brief description of the drawings
Fig. 1 is the flow chart of the dialog process method in the task dialogue system of one embodiment of the invention；
Fig. 2 is the step S13 of one embodiment of the invention refined flow chart；
Fig. 3 is the schematic diagram of the dialog process process in the task dialogue system of one embodiment of the invention；
Fig. 4 is the structural representation of the dialog process device in the task dialogue system of one embodiment of the invention；
Fig. 5 is the structural representation of the dialog process device in the task dialogue system of another embodiment of the present invention.
Embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and be not considered as limiting the invention.
Below with reference to the accompanying drawings dialog process method and device in the task dialogue system of the embodiment of the present invention is described.
Fig. 1 is the flow chart of the dialog process method in the task dialogue system of one embodiment of the invention.
As shown in figure 1, the dialog process method in the task dialogue system can include following steps：
S11, obtains epicycle user message.
In the scene of task dialogue, the epicycle user message that user inputs in dialog interface is obtained.
S12, obtains the dialogue action of task dialogue system feedback in last round of dialogue.
, wherein it is desired to understand, the dialogue action that last round of centering task dialogue system is returned can include specific disappear
Breath, calls application interface etc..
S13, is acted according to the dialogue of task dialogue system feedback in epicycle user message and last round of dialogue, generates epicycle
The dialogue action of task dialogue system feedback in dialogue.
In one embodiment of the invention, according to task dialogue system feedback in epicycle user message and last round of dialogue
Dialogue action, generation epicycle dialogue in task dialogue system feedback dialogue action process, as shown in Fig. 2 can include：
S21, according to the dialogue action of task dialogue system feedback in last round of dialogue and epicycle user message, extracts epicycle
The entity information related to task in dialogue, and the corresponding entity vector of result generation is extracted according to entity.
In one embodiment of the invention, according to the dialogue action of task dialogue system feedback in last round of dialogue and this
Wheel user message determines the task of user, can determine whether to whether there is the reality of the entity elements related to task in epicycle user message
Body keyword, if in the presence of the entity information related to task in the dialogue of extraction epicycle from epicycle user message.
Wherein, the dimension of entity vector is that the total quantitys of the entity elements in the entity elements set corresponding to task is determined
, for example, task is predetermined air ticket, it is assumed that the entity elements set corresponding to the task includes 5 entity elements, then entity
The dimension of vector is 5, i.e. entity vector is 5 dimensional vectors.
Wherein, an entity elements (occur in entity vector per the entity elements set corresponding to one-dimensional representation task
For 1, occur without as 0).
, wherein it is desired to explanation, in task conversational system, user wishes that the task that task dialogue system is performed can be with
Determined by epicycle dialog information and/or other dialog history information.
For example, epicycle user message is " helping me to buy Zhangbei County capital to the air ticket in Shanghai ", by being carried out to epicycle user message
It is intended to analysis, you can determine that user's is intended that predetermined air ticket, i.e. task dialogue system can determine that task is predetermined air ticket, root
The entity elements set related to predetermined air ticket is can determine that according to predetermined air ticket, it is assumed that entity elements set include departure city,
Land city, departure date, target date and Flight Information.By analysis can determine in epicycle user message " Beijing " and
" Shanghai " is the entity information related to predetermined air ticket this task, and the corresponding entity elements of keyword " Beijing " is " play
Fly city ", " Shanghai " corresponding entity elements are " landing city ".
In another example, epicycle user message is " Zhang San, ID card No. 123456199001010001 ", it is assumed that obtain upper one
The dialogue action of task dialogue system feedback is particular message in wheel dialogue, and particular message please provide to receive and seize the opportunity people's information,
Then extracted according to the dialogue action of task dialogue system feedback in last round of dialogue and epicycle user message in epicycle dialogue with appointing
Entity elements " seizing the opportunity people's name " corresponding keyword of " predetermined air ticket " correlation of being engaged in is " Zhang San ", entity elements " identification card number
The corresponding keyword of code " is " 123456199001010001 ".
, wherein it is desired to extract the entity information influence task pair related to task in explanation, epicycle user message
The selection and execution of telephone system dialogue action.
After entity information related to task in extracting epicycle dialogue, result can be extracted according to entity and generate corresponding reality
Body vector.
, wherein it is desired to explanation, can be by way of based on entity decimation rule or statistics according to last round of dialogue
The dialogue action of middle task dialogue system feedback and epicycle user message, extract entity letter related to task in epicycle dialogue
Breath.
S22, carries out text vector processing, to generate the text vector of epicycle user message to epicycle user message.
Wherein, text vector is the corresponding vector of epicycle user message.
Generally, even if epicycle user message does not refer to entity, epicycle user message also expresses certain implication in itself,
The selection that this is acted to dialogue is helpful.Appoint for example, not included when epicycle user message is " good, to order this class ", in text
What entity, but express and to order the intention of some flight.
In an embodiment of the present invention, can in order that more accurate dialogue action can be fed back by obtaining task dialogue system
Text vector processing is carried out to epicycle user message by prior art, to generate the text vector of epicycle user message.
S23, obtains the hidden vector of last round of dialogue.
Wherein, the vector corresponding to the state of the last round of dialogue of hidden vector representation of last round of dialogue.
, wherein it is desired to explanation, hidden vectorial dimension is set in advance, for example, in training deep neural network mould
Hidden vectorial dimension is manually set during type.
S24, the deep neural network model based on training in advance is vectorial, last round of dialogue according to text vector, entity
Appoint in hidden vectorial, last round of dialogue in the raw dialogue action vector generation epicycle dialogue of the dialogue action of task dialogue system feedback
The dialogue action of conversational system of being engaged in feedback.
Wherein, deep neural network model is that the Recognition with Recurrent Neural Network based on long short-term memory (LSTM) is pre-established.
, wherein it is desired to explanation, in the deep neural network model based on training in advance, according to text vector, entity
In hidden vectorial, the last round of dialogue of vectorial, last round of dialogue the raw dialogue of the dialogue action of task dialogue system feedback act to
Amount generation epicycle dialogue in task dialogue system feedback dialogue action while, can also generate epicycle dialogue it is corresponding it is hidden to
Amount.
That is, in vectorial, last round of dialogue hidden vectorial, the last round of dialogue by text vector, entity task pair
The dialogue that the dialogue action of telephone system feedback is raw acts vector input deep neural network model, passes through deep neural network model
Epicycle can be generated and talk with dialogue action of the corresponding hidden vector sum task dialogue system for epicycle conversational feedback.
S14, the dialogue action of task dialogue system feedback in display epicycle dialogue.
In order to which more clearly description is of the invention, with reference to Fig. 3 to the dialogue in the task dialogue system of the embodiment
Processing method is described.
Wherein, deep neural network model includes LSTM units, the deep neural network model based on training in advance, root
The dialogue action of task dialogue system feedback according to text vector, entity be vectorial, in last round of dialogue hidden vectorial, last round of dialogue
The dialogue action detailed process for generating task dialogue system feedback in epicycle dialogue is as follows：
First, initial hidden vectorial h is obtained0With initial dialog action vector a0。
, wherein it is desired to explanation, initial hidden vectorial h0With initial dialog action vector a0In each element be 0,
That is, initial hidden vectorial h0With initial dialog action vector a0Be complete zero vector.
Then, from the n-th=1 wheel beginning of conversation, step S1 is performed, step S1 is：By epicycle user message qnWith it is last round of
Dialogue action vector ai-1Entity abstraction module is inputted, output entity extracts vector en；Front-wheel user message q will be worked asnInput text
Vectorization module, obtains text vector vn；By last round of hidden vectorial hn-1, last round of dialogue action vector an-1, epicycle entity takes out
Amount of orientation en, epicycle text vector vnSpliced, input LSTM units, with by LSTM units obtain epicycle talk with it is hidden to
Measure hnThe dialogue action vector a talked with epicyclen, and the dialogue action vector talked with by action executing module according to epicycle
anCorresponding dialogue action is performed, and shows the dialogue action of task dialogue system feedback in epicycle dialogue.That is, LSTM
The output of unit is the hidden vectorial h of epicycle dialoguenThe dialogue action vector a talked with epicyclen。
If user continues have input, n is increased by 1, and perform step S1.If user does not input, terminate.
It in summary it can be seen, the task in generation epicycle dialogue of the dialog process method in the task dialogue system of the implementation
During the dialogue action of conversational system feedback, pass through task dialogue system feedback in epicycle user message and last round of dialogue
Dialogue action, generation epicycle dialogue in task dialogue system feedback dialogue action, thus, with reference to it is upper wheel dialogue dialogue move
Make the dialogue action of generation epicycle dialogue, it is not necessary to which the task dialogue system for different field pre-defines dialogue state, makes
Obtaining this method has very strong cross-cutting transfer ability.
Dialog process method in task dialogue system according to embodiments of the present invention, obtains epicycle user message, obtains
The dialogue action of task dialogue system feedback in last round of dialogue, according to task dialogue in epicycle user message and last round of dialogue
The dialogue action of system feedback, the dialogue action of task dialogue system feedback in generation epicycle dialogue, and display epicycle dialogue
The dialogue action of middle task dialogue system feedback, thus, is moved with reference to the dialogue that the dialogue action generation epicycle that upper wheel is talked with is talked with
Make, it is not necessary to which the task dialogue system for different field pre-defines dialogue state so that this method has very strong across neck
Domain migration ability.
In order to realize above-described embodiment, the invention also provides the dialog process device in a kind of task dialogue system.
Fig. 4 is the structural representation of the dialog process device in the task dialogue system of one embodiment of the invention.
As shown in figure 4, the dialog process device in the task dialogue system is obtained including the first acquisition module 110, second
Module 120, generation module 130 and display module 140, wherein：
First acquisition module 110 is used to obtain epicycle user message.
Second acquisition module 120 is used for the dialogue action for obtaining task dialogue system feedback in last round of dialogue.
, wherein it is desired to understand, the dialogue action that last round of centering task dialogue system is returned can include specific disappear
Breath, calls application interface etc..
Generation module 130 is used for dynamic according to the dialogue of task dialogue system feedback in epicycle user message and last round of dialogue
Make, the dialogue action of task dialogue system feedback in generation epicycle dialogue.
Display module 140 is used for the dialogue action for showing task dialogue system feedback in epicycle dialogue.
In one embodiment of the invention, on the basis of shown in Fig. 4, as shown in figure 5, generation module 130 can be wrapped
Extracting unit 131, text vector unit 132, acquiring unit 133 and generation unit 134 are included, wherein：
Extracting unit 131 is used to be disappeared according to the dialogue action and epicycle user of task dialogue system feedback in last round of dialogue
Breath, extracts entity information related to task in epicycle dialogue, and extract the corresponding entity vector of result generation according to entity.
Wherein, the dimension of entity vector total quantity of entity elements in the entity elements set corresponding to task is determined
, for example, task is predetermined air ticket, it is assumed that the entity elements set corresponding to the task includes 5 entity elements, then entity
The dimension of vector is 5, i.e. entity vector is 5 dimensional vectors.
Wherein, an entity elements (occur in entity vector per the entity elements set corresponding to one-dimensional representation task
For 1, occur without as 0).
Text vector unit 132 is used to carry out text vector processing to epicycle user message, is disappeared with generating epicycle user
The text vector of breath.
Acquiring unit 133 is used for the hidden vector for obtaining last round of dialogue.
Wherein, the vector corresponding to the state of the last round of dialogue of hidden vector representation of last round of dialogue.
, wherein it is desired to explanation, hidden vectorial dimension is set in advance, for example, in training deep neural network mould
Hidden vectorial dimension is manually set during type.
Generation unit 134 be used for the deep neural network model based on training in advance, according to text vector, entity vector,
The dialogue action vector generation of the dialogue action of task dialogue system feedback in hidden vectorial, the last round of dialogue of last round of dialogue is originally
The dialogue action of task dialogue system feedback in wheel dialogue.
In one embodiment of the invention, deep neural network model is the circulation god based on long short-term memory (LSTM)
Pre-established through network.
In one embodiment of the invention, it is determined that user task when, extracting unit 131 specifically for：According to upper
The dialogue action of task dialogue system feedback and epicycle user message determine the task of user in one wheel dialogue, judge epicycle user
It whether there is the entity key related to task in message, if in the presence of from epicycle user message in the dialogue of extraction epicycle
The entity information related to task.
, wherein it is desired to the explanation of the dialog process embodiment of the method in explanation, the foregoing system to task dialogue
The dialog process device in the task dialogue system of the embodiment is also applied for, its realization principle is similar, and here is omitted.
Dialog process device in task dialogue system according to embodiments of the present invention, obtains epicycle user message, obtains
The dialogue action of task dialogue system feedback in last round of dialogue, according to task dialogue in epicycle user message and last round of dialogue
The dialogue action of system feedback, the dialogue action of task dialogue system feedback in generation epicycle dialogue, and display epicycle dialogue
The dialogue action of middle task dialogue system feedback, thus, is moved with reference to the dialogue that the dialogue action generation epicycle that upper wheel is talked with is talked with
Make, it is not necessary to which the task dialogue system for different field pre-defines dialogue state, and this method can apply to any field
Task dialogue, that is to say, that this method has very strong cross-cutting transfer ability.
On the device in above-described embodiment, wherein modules perform the concrete mode of operation in relevant this method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
For device embodiment, because it corresponds essentially to embodiment of the method, so related part is real referring to method
Apply the part explanation of example.Device embodiment described above is only schematical, wherein illustrating as separating component
Unit can be or may not be physically separate, the part shown as unit can be or may not be
Physical location, you can with positioned at a place, or can also be distributed on multiple NEs.Can be according to the actual needs
Some or all of module therein is selected to realize the purpose of disclosure scheme.Those of ordinary skill in the art are not paying wound
In the case that the property made is worked, you can to understand and implement.
To realize above-described embodiment, the invention also provides a kind of server.
The server includes：One or more processors；Storage device, for storing one or more programs, when one
Or multiple programs are executed by one or more processors so that one or more processors are realized in above-mentioned task dialogue system
Dialog process method.
A kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is by processor
The dialog process method in above-mentioned task dialogue system is realized during execution.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means to combine specific features, structure, material or the spy that the embodiment or example are described
Point is contained at least one embodiment of the present invention or example.In this manual, to the schematic representation of above-mentioned term not
Identical embodiment or example must be directed to.Moreover, specific features, structure, material or the feature of description can be with office
Combined in an appropriate manner in one or more embodiments or example.In addition, in the case of not conflicting, the skill of this area
Art personnel can be tied the not be the same as Example or the feature of example and non-be the same as Example or example described in this specification
Close and combine.
In addition, term " first ", " second " are only used for describing purpose, and it is not intended that indicating or implying relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can express or
Implicitly include at least one this feature.In the description of the invention, " multiple " are meant that at least two, such as two, three
It is individual etc., unless otherwise specifically defined.
Any process described otherwise above or method description are construed as in flow chart or herein, represent to include
Module, fragment or the portion of the code of one or more executable instructions for the step of realizing specific logical function or process
Point, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not be by shown or discussion suitable
Sequence, including according to involved function by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Represent in flow charts or logic and/or step described otherwise above herein, for example, being considered use
In the order list for the executable instruction for realizing logic function, it may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system including the system of processor or other can be held from instruction
The system of row system, device or equipment instruction fetch and execute instruction) use, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicate, propagate or pass
Defeated program is for instruction execution system, device or equipment or the dress for combining these instruction execution systems, device or equipment and using
Put.The more specifically example (non-exhaustive list) of computer-readable medium includes following：Electricity with one or more wirings
Connecting portion (electronic installation), portable computer diskette box (magnetic device), random access memory (RAM), read-only storage
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device, and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, can even is that can be in the paper of printing described program thereon or other are suitable for computer-readable medium
Medium, because can then enter edlin, interpretation or if necessary with it for example by carrying out optical scanner to paper or other media
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each several part of the present invention can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In embodiment, the software that multiple steps or method can in memory and by suitable instruction execution system be performed with storage
Or firmware is realized.If, and in another embodiment, can be with well known in the art for example, realized with hardware
Any one of row technology or their combination are realized：With the logic gates for realizing logic function to data-signal
Discrete logic, the application specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method is carried
Rapid to can be by program to instruct the hardware of correlation to complete, described program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
In addition, each functional unit in each embodiment of the invention can be integrated in a processing module, can also
That unit is individually physically present, can also two or more units be integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, it would however also be possible to employ the form of software function module is realized.The integrated module is such as
Fruit is realized using in the form of software function module and as independent production marketing or in use, can also be stored in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..Although having been shown and retouching above
Embodiments of the invention are stated, it is to be understood that above-described embodiment is exemplary, it is impossible to be interpreted as the limit to the present invention
System, one of ordinary skill in the art can be changed to above-described embodiment, change, replace and become within the scope of the invention
Type.
Claims (10)
1. a kind of dialog process method in task dialogue system, it is characterised in that comprise the following steps：
Obtain epicycle user message；
Obtain the dialogue action of task dialogue system feedback described in last round of dialogue；
The dialogue action of the task dialogue system feedback according to the epicycle user message and last round of dialogue, generates epicycle
The dialogue action of task dialogue system feedback described in dialogue；
Show the dialogue action of task dialogue system feedback described in epicycle dialogue.
2. the method as described in claim 1, it is characterised in that described according in the epicycle user message and last round of dialogue
The dialogue action of the task dialogue system feedback, the dialogue action of task dialogue system feedback described in generation epicycle dialogue,
Including：
The dialogue action of the task dialogue system feedback according to last round of dialogue and the epicycle user message, are extracted described
The entity information related to task in epicycle dialogue, and the corresponding entity vector of result generation is extracted according to entity；
Text vector processing is carried out to the epicycle user message, to generate the text vector of the epicycle user message；
Obtain the hidden vector of last round of dialogue；
Deep neural network model based on training in advance, it is vectorial, last round of dialogue according to the text vector, the entity
In the dialogue action vector generation epicycle dialogue of the dialogue action of task dialogue system feedback described in hidden vectorial, last round of dialogue
The dialogue action of the task dialogue system feedback.
3. method as claimed in claim 2, it is characterised in that the deep neural network model is to be based on long short-term memory
(LSTM) what Recognition with Recurrent Neural Network was pre-established.
4. method as claimed in claim 2, it is characterised in that the task dialogue system according to last round of dialogue is anti-
The dialogue action of feedback and the epicycle user message, extract the entity information of the epicycle dialogue, including：
The dialogue action of the task dialogue system feedback according to the last round of dialogue and the epicycle user message are determined
The task of user；
Judge the entity key with the presence or absence of the entity elements related to the task in the epicycle user message；
If in the presence of extracting entity information related to task in epicycle dialogue from the epicycle user message.
5. the dialog process device in a kind of task dialogue system, it is characterised in that including：
First acquisition module, for obtaining epicycle user message；
Second acquisition module, the dialogue for obtaining task dialogue system feedback described in last round of dialogue is acted；
Generation module, the dialogue for the task dialogue system feedback according to the epicycle user message and last round of dialogue
Action, the dialogue action of task dialogue system feedback described in generation epicycle dialogue；
Display module, the dialogue action for showing task dialogue system feedback described in epicycle dialogue.
6. device as claimed in claim 5, it is characterised in that the generation module, including：
Extracting unit, dialogue action and the epicycle user for the task dialogue system feedback according to last round of dialogue
Message, extracts entity information related to task in the epicycle dialogue, and extract the corresponding entity of result generation according to entity
Vector；
Text vector unit, for carrying out text vector processing to the epicycle user message, to generate the epicycle user
The text vector of message；
Acquiring unit, the hidden vector for obtaining last round of dialogue；
Generation unit, for the deep neural network model based on training in advance, according to the text vector, the entity to
Amount, the dialogue of the dialogue action of task dialogue system feedback described in hidden vectorial, the last round of dialogue of last round of dialogue act to
The dialogue action of task dialogue system feedback described in amount generation epicycle dialogue.
7. device as claimed in claim 6, it is characterised in that the deep neural network model is to be based on long short-term memory
(LSTM) what Recognition with Recurrent Neural Network was pre-established.
8. device as claimed in claim 6, it is characterised in that the extracting unit, specifically for：
The dialogue action of the task dialogue system feedback according to the last round of dialogue and the epicycle user message are determined
The task of user；
Judge to whether there is the entity key related to the task in the epicycle user message；
If in the presence of extracting entity information related to task in epicycle dialogue from the epicycle user message.
9. a kind of server, including：
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are by one or more of computing devices so that one or more of processors are real
The existing dialog process method in the task dialogue system as described in any in claim 1-4.
10. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is by processor
The dialog process method in the task dialogue system as described in any in claim 1-4 is realized during execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710289869.6A CN107273406B (en) | 2017-04-27 | 2017-04-27 | Dialogue processing method and device in task dialogue system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710289869.6A CN107273406B (en) | 2017-04-27 | 2017-04-27 | Dialogue processing method and device in task dialogue system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107273406A true CN107273406A (en) | 2017-10-20 |
| CN107273406B CN107273406B (en) | 2020-06-05 |
Family
ID=60073603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710289869.6A Active CN107273406B (en) | 2017-04-27 | 2017-04-27 | Dialogue processing method and device in task dialogue system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107273406B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241265A (en) * | 2018-09-17 | 2019-01-18 | 四川长虹电器股份有限公司 | A kind of field recognition methods and system towards more wheel inquiries |
| CN109582767A (en) * | 2018-11-21 | 2019-04-05 | 北京京东尚科信息技术有限公司 | Conversational system processing method, device, equipment and readable storage medium storing program for executing |
| CN109684451A (en) * | 2018-12-24 | 2019-04-26 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN110162608A (en) * | 2019-02-26 | 2019-08-23 | 腾讯科技（深圳）有限公司 | The method of computer implemented dialogue management, equipment and computer readable storage medium for dialogue management |
| CN110211572A (en) * | 2019-05-14 | 2019-09-06 | 北京来也网络科技有限公司 | Dialog control method and device based on intensified learning |
| CN111427993A (en) * | 2019-11-15 | 2020-07-17 | 西安海的电子科技有限公司 | Multi-conversation implementation method based on android system |
| CN112084300A (en) * | 2020-08-07 | 2020-12-15 | 北京三快在线科技有限公司 | Response information output method and device, electronic equipment and readable storage medium |
| CN113821620A (en) * | 2021-09-18 | 2021-12-21 | 湖北亿咖通科技有限公司 | Multi-turn conversation task processing method and device and electronic equipment |
| CN114265920A (en) * | 2021-12-27 | 2022-04-01 | 北京易聊科技有限公司 | Intelligent robot dialogue method and system based on signals and scenes |
| CN115617972A (en) * | 2022-12-14 | 2023-01-17 | 成都明途科技有限公司 | Robot dialogue method, device, electronic equipment and storage medium |
| CN115659994B (en) * | 2022-12-09 | 2023-03-03 | 深圳市人马互动科技有限公司 | Data processing method and related device in human-computer interaction system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102165518A (en) * | 2008-08-27 | 2011-08-24 | 罗伯特·博世有限公司 | System and method for generating natural language phrases from user utterances in dialog systems |
| CN104871150A (en) * | 2012-07-20 | 2015-08-26 | 韦韦欧股份有限公司 | Method of and system for inferring user intent in search input in a conversational interfaction system |
| CN105138710A (en) * | 2015-10-12 | 2015-12-09 | 金耀星 | Chat agent system and method |
| CN105704013A (en) * | 2016-03-18 | 2016-06-22 | 北京光年无限科技有限公司 | Context-based topic updating data processing method and apparatus |
| CN106528522A (en) * | 2016-08-26 | 2017-03-22 | 南京威卡尔软件有限公司 | Scenarized semantic comprehension and dialogue generation method and system |
-
2017
- 2017-04-27 CN CN201710289869.6A patent/CN107273406B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102165518A (en) * | 2008-08-27 | 2011-08-24 | 罗伯特·博世有限公司 | System and method for generating natural language phrases from user utterances in dialog systems |
| CN104871150A (en) * | 2012-07-20 | 2015-08-26 | 韦韦欧股份有限公司 | Method of and system for inferring user intent in search input in a conversational interfaction system |
| CN105138710A (en) * | 2015-10-12 | 2015-12-09 | 金耀星 | Chat agent system and method |
| CN105704013A (en) * | 2016-03-18 | 2016-06-22 | 北京光年无限科技有限公司 | Context-based topic updating data processing method and apparatus |
| CN106528522A (en) * | 2016-08-26 | 2017-03-22 | 南京威卡尔软件有限公司 | Scenarized semantic comprehension and dialogue generation method and system |
Non-Patent Citations (1)
| Title |
|---|
| JASON D. WILLIAMS 等: ""Hybrid Code Networks: practical and efficient end-to-end dialog control with supervised and reinforcement learning"", 《HTTPS://ARXIV.ORG/ABS/1702.03274》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241265B (en) * | 2018-09-17 | 2022-06-03 | 四川长虹电器股份有限公司 | Multi-round query-oriented field identification method and system |
| CN109241265A (en) * | 2018-09-17 | 2019-01-18 | 四川长虹电器股份有限公司 | A kind of field recognition methods and system towards more wheel inquiries |
| CN109582767A (en) * | 2018-11-21 | 2019-04-05 | 北京京东尚科信息技术有限公司 | Conversational system processing method, device, equipment and readable storage medium storing program for executing |
| CN109582767B (en) * | 2018-11-21 | 2024-05-17 | 北京京东尚科信息技术有限公司 | Dialogue system processing method, device, equipment and readable storage medium |
| CN109684451A (en) * | 2018-12-24 | 2019-04-26 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN109684451B (en) * | 2018-12-24 | 2022-01-14 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN110162608A (en) * | 2019-02-26 | 2019-08-23 | 腾讯科技（深圳）有限公司 | The method of computer implemented dialogue management, equipment and computer readable storage medium for dialogue management |
| CN110211572A (en) * | 2019-05-14 | 2019-09-06 | 北京来也网络科技有限公司 | Dialog control method and device based on intensified learning |
| CN111427993A (en) * | 2019-11-15 | 2020-07-17 | 西安海的电子科技有限公司 | Multi-conversation implementation method based on android system |
| CN112084300A (en) * | 2020-08-07 | 2020-12-15 | 北京三快在线科技有限公司 | Response information output method and device, electronic equipment and readable storage medium |
| CN113821620B (en) * | 2021-09-18 | 2023-12-12 | 亿咖通(湖北)技术有限公司 | Multi-round dialogue task processing method and device and electronic equipment |
| CN113821620A (en) * | 2021-09-18 | 2021-12-21 | 湖北亿咖通科技有限公司 | Multi-turn conversation task processing method and device and electronic equipment |
| CN114265920A (en) * | 2021-12-27 | 2022-04-01 | 北京易聊科技有限公司 | Intelligent robot dialogue method and system based on signals and scenes |
| CN115659994B (en) * | 2022-12-09 | 2023-03-03 | 深圳市人马互动科技有限公司 | Data processing method and related device in human-computer interaction system |
| CN115617972A (en) * | 2022-12-14 | 2023-01-17 | 成都明途科技有限公司 | Robot dialogue method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107273406B (en) | 2020-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107273406A (en) | Dialog process method and device in task dialogue system | |
| CN107193865A (en) | Natural language is intended to understanding method and device in man-machine interaction | |
| CN106095834A (en) | Intelligent dialogue method and system based on topic | |
| CN107329967A (en) | Question answering system and method based on deep learning | |
| CN105912600B (en) | Question and answer knowledge base and its method for building up, intelligent answer method and system | |
| CN111712834A (en) | Artificial intelligence system for inferring realistic intent | |
| CN107704482A (en) | Method, apparatus and program | |
| CN105096953B (en) | Realize the multilingual audio recognition method being used in mixed way | |
| CN109543010A (en) | The interactive method and system of fused data library inquiry | |
| CN109165386A (en) | A kind of Chinese empty anaphora resolution method and system | |
| CN109408800A (en) | Talk with robot system and associative skills configuration method | |
| Braun et al. | Towards a Framework for Classifying Chatbots. | |
| Windiatmoko et al. | Developing facebook chatbot based on deep learning using rasa framework for university enquiries | |
| CN113268610B (en) | Intent jump method, device, equipment and storage medium based on knowledge graph | |
| CN117573834B (en) | Multi-robot dialogue method and system for software-oriented instant service platform | |
| CN110046230A (en) | Generate the method for recommending words art set, the method and apparatus for recommending words art | |
| Angelov et al. | E-commerce distributed chatbot system | |
| CN109408799A (en) | Semantic decision-making technique and system | |
| CN110427454A (en) | Text mood analysis method and device, electronic equipment and non-transient storage media | |
| TW201820172A (en) | System, method and non-transitory computer readable storage medium for conversation analysis | |
| Wohlgenannt et al. | Extracting social networks from literary text with word embedding tools | |
| CN108628908A (en) | The method, apparatus and electronic equipment of sorted users challenge-response boundary | |
| CN112199486A (en) | Task type multi-turn conversation method and system for office scene | |
| CN113918698A (en) | Customer service processing system and device | |
| Mohammed et al. | Chatbot system architecture |
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 | ||
| CP03 | Change of name, title or address |
Address after: Room 138, building 3, 33 Leshan Road, Xuhui District, Shanghai 200030 Patentee after: Shanghai laiyibert Network Technology Co.,Ltd. Address before: 100081 701, block a, Rongke information, Xueyuan South Road, Haidian District, Beijing Patentee before: SHANGHAI BENYING NETWORK TECHNOLOGY CO.,LTD. |
|
| CP03 | Change of name, title or address |