CN109063840A - A kind of Interactive Dynamic inference method and device - Google Patents
A kind of Interactive Dynamic inference method and device Download PDFInfo
- Publication number
- CN109063840A CN109063840A CN201810752461.2A CN201810752461A CN109063840A CN 109063840 A CN109063840 A CN 109063840A CN 201810752461 A CN201810752461 A CN 201810752461A CN 109063840 A CN109063840 A CN 109063840A
- Authority
- CN
- China
- Prior art keywords
- decision
- factor
- model
- decision factor
- result
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Interactive Dynamic inference method and devices, this method comprises: step S1, identifies customer problem, according to customer problem determination and trigger corresponding scene, the Information Extraction Model and decision model of linkage triggering scene;Step S2 extracts the decision factor in customer problem, judges the decision factor for whether containing the decision model that has an impact in the decision factor, enters step S3 or directly storage according to judging result;Step S3, guidance user submits necessary information, the decision factor extracted in user's input carries out storage inside, the decision factor of storage inside is sent to decision model one by one, and judge whether to determine next decision factor by decision model, when the judgment result is yes, next decision factor is determined;Step S4 is triggered and is run the decision model;Step S5 obtains and exports the result of decision exported in the decision model and operation result, and the present invention can solve the challenge of user, promotes user experience.
Description
Technical field
The present invention relates to automatic question answering technical fields, more particularly to a kind of Interactive Dynamic inference method and device.
Background technique
Current intelligent Answer System mainly has question-response and more wheel two kinds of forms of session.Wherein, question-response is logical
It crosses and puts question to knowledge required for locating websites user, provide simple services for website user, when website user proposes problem, be
Problem answers not only are pushed out by system, but also also can all be pushed out knowledge relevant to this problem for user query;
And take turns session then by FAQs being organized into the knowledge of several process diagnosis types more, it is asked by guiding user to interact formula
It answers, finally provides service for user.
However, the form of session of current question-response, is limited by question and answer mode and round, it is general only to can solve simple knowledge
And problem, complicated individual subscriber can not be preferably solved the problems, such as using knowledge;The question and answer mode of more wheel sessions, is determined by being formed
Plan tree forms diagnosis type knowledge organization, common individual subscriber problem is finally able to solve, but application printenv learns, along tree
Root gets to downwards leaf node, that is, the result of decision, but it cannot provide user's service of customization, can not provide use
The result of decision accurately calculated is passed through at family.
Summary of the invention
In order to overcome the deficiencies of the above existing technologies, purpose of the present invention is to provide a kind of Interactive Dynamic reasoning sides
Method and device can identify decision factor in customer problem, save the trouble that deduplication is putd question to user, promote the body of user
It tests, and the challenge of user can be solved.
In view of the above and other objects, the present invention proposes a kind of Interactive Dynamic inference method, include the following steps:
Step S1 identifies customer problem, according to the determination of the customer problem of identification and triggers corresponding scene, linkage triggering field
The Information Extraction Model and decision model of scape;
Whether step S2 extracts the decision factor in customer problem, judge in the decision factor of customer problem containing having an impact
The decision factor of decision model enters step S3 according to judging result or directly stores the decision factor;
Step S3, guidance user submit necessary information, and the decision factor extracted in user's input carries out storage inside, will
The decision factor of storage inside is sent to decision model one by one, and by decision model judge whether to determine next decision because
Son determines next decision factor when the judgment result is yes;
Step S4 is triggered and is run the decision model;
Step S5 is obtained and is exported the result of decision exported in the decision model and operation result.
Preferably, step S1 further comprises:
It checks scene entry condition, receives the natural language data of user, and the natural language data are carried out semantic
Understand, with generate problem request, requested according to described problem, obtain scene entrance Candidate Set, and to scene entrance Candidate Set into
The sequence of row comprehensive score selects the highest scene of scoring as scene entrance.
Preferably, step S3 further comprises:
Issue the user with the inquiry for obtaining decision factor;
User is obtained to the feedback of inquiry, extracts the decision factor in user feedback;
The decision factor of extraction is subjected to storage inside;
The decision factor of storage inside is sent to decision model one by one;
Judge whether to determine next decision factor by decision model, when the judgment result is yes, according to determining for extraction
The plan factor determines next decision factor, when the judgment result is no, then enters step S5.
Preferably, step S2 further comprises:
The decision factor in customer problem is extracted using Information Extraction Model;
Judge the decision factor for whether containing the decision model that has an impact in the decision factor of the customer problem;
If the decision factor containing the decision model that has an impact in the decision factor of the customer problem, extracts the decision factor,
And the decision factor is subjected to storage inside;If not containing or lacking in the decision factor of the customer problem influences decision model
Decision factor then enters the step of issuing the user with the inquiry for obtaining decision factor.
Preferably, further include following steps before the decision factor by extraction carries out storage inside step:
Decision factor in the user feedback of extraction is compared with the decision factor of storage inside;
If skipping the problem containing the decision factor of the problem in the decision factor of storage inside, continuing the problem
Next problem;If not containing the decision factor of the problem in the decision factor of storage inside, the problem is enabled, by extraction
The decision factor contained in user feedback carries out storage inside.
Preferably, further include following steps before step S4:
Judge the decision model for exporting whether the default necessary information of a result of decision has extracted completely;
If the default necessary information for exporting any result of decision has extracted completely, enters step S4 and run the decision
Model, the S3 that otherwise gos to step guidance user submit necessary information.
Preferably, in step S4, the operation of the decision model includes:
Step S400, judges whether decision model contains mathematical computations or logical operation, has, and carries out mathematics or logic fortune
It calculates, enters step S401, nothing then carries out being directly entered step S401;
Step S401, by decision model the result of decision and operation result be sent to result node, by result node pair
The result of decision of decision model transmission is integrated with calculated result.
In order to achieve the above objectives, the present invention also provides a kind of Interactive Dynamic reasoning devices, comprising:
Ingress node, customer problem, according to the determination of the customer problem of identification and triggers corresponding scene, links for identification
Trigger the Information Extraction Model and decision model of scene;
Decision factor extracts and judging unit, for extracting the decision factor in customer problem, judges determining for customer problem
The decision factor for whether containing the decision model that has an impact in the plan factor, questions closely node according to judging result starting or is directly deposited
Storage;
Node is questioned closely, for guiding user to submit necessary information, the decision factor extracted in user's input is stored,
The decision factor of storage inside is sent to decision model one by one, and by decision model judge whether to determine next decision because
Son determines next decision factor when the judgment result is yes;
Decision model running unit, for triggering simultaneously operational decisions model;
Result node, for obtaining and exporting the result of decision exported in decision model and operation result.
Preferably, described device further includes condition node, for questioning closely the decision factor and inside that node extracts for described
The decision factor of storage is compared;If skipped when decision factor in the decision factor of storage inside containing the problem
The problem continues next problem of the problem, if not containing the decision factor of the problem in the decision factor of storage inside
The Shi Qiyong problem, and the decision factor contained in the user feedback of extraction is subjected to storage inside.
Preferably, described device further includes jumping node, for judging the decision model for exporting a result of decision
Whether default necessary information has extracted completely, and in judging result for for having exported the default necessary information of any result of decision
It extracts completely, then starts the decision model running unit, otherwise jump to the node of questioning closely and guide user's offer necessary
Information.
Compared with prior art, a kind of Interactive Dynamic inference method of the present invention and device are by extracting in customer problem
Decision factor, and the decision factor for influencing decision model is not contained or lacked in the decision factor for judge the customer problem
When, guidance user submits necessary information, and extracts the decision factor in user's input, and export decision by operational decisions model
As a result with operation result, the present invention can not only identify decision factor in customer problem, eliminate the fiber crops that repetition is putd question to user
It is tired, the experience of user is improved, and the challenge of user can be solved.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of Interactive Dynamic inference method of the present invention;
Fig. 2 is a kind of system architecture diagram of Interactive Dynamic reasoning device of the present invention;
Fig. 3 is the detail structure chart of Ingress node 202 in the specific embodiment of the invention;
Fig. 4 is the detail structure chart that node 203 is questioned closely in the specific embodiment of the invention;
Fig. 5 is a kind of system architecture diagram of Interactive Dynamic reasoning device specific embodiment of the present invention
Fig. 6 is the flow chart of the Interactive Dynamic inference method of the embodiment of the present invention.
Specific embodiment
Below by way of specific specific example and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand further advantage and effect of the invention easily by content disclosed in the present specification.The present invention can also pass through other differences
Specific example implemented or applied, details in this specification can also be based on different perspectives and applications, without departing substantially from
Various modifications and change are carried out under spirit of the invention.
Fig. 1 is a kind of step flow chart of Interactive Dynamic inference method of the present invention.As shown in Figure 1, a kind of friendship of the present invention
Mutual formula Dynamic Inference method, includes the following steps:
Step S1 identifies customer problem, according to the determination of the customer problem of identification and triggers corresponding scene, linkage triggering
The Information Extraction Model and decision model of scene.
Specifically, step S1 further comprises:
Step S100 checks scene entry condition, receives the natural language data of user, and to the natural language data
Semantic understanding is carried out, to generate problem request;
Step S101, is requested according to described problem, obtains scene entrance Candidate Set, and is carried out to scene entrance Candidate Set comprehensive
Marking and queuing is closed, selects the highest scene of scoring as scene entrance.
Step S102, the highest scene of triggering scoring, linkage trigger the Information Extraction Model and decision model of the scene.
Since the targeted problem of different scenes is different, to be selected according to customer problem the strongest scene of specific aim into
Enter.For example, identifying that customer problem is " I will buy truck tire ", the scene entrance Candidate Set of acquisition includes scene 1 and field
Scape 2, scene 1 are that automobile tire buys scene, and scene 2 is that household electrical appliance buy scene, then can be according to customer problem " I of identification
Buy truck tire ", itself and scene 1 and 2 entrance of scene are preset into question synthesis scoring, select to score highest scene as
Scene entrance.Due to the comprehensive score highest of comprehensive score scene 1, so triggering scene 1, buys scene into automobile tire.
Whether step S2 extracts the decision factor in customer problem, judge in the decision factor of customer problem containing having an impact
The decision factor of decision model enters the step of guidance user submits necessary information according to judging result or directly storage should be certainly
The plan factor.
Specifically, step S2 further comprises:
Step S200 extracts the decision factor in customer problem using Information Extraction Model, and judges the customer problem
Decision factor in whether containing the decision factor of decision model of having an impact.In this step, the Information Extraction Model of scene is called
The decision factor in customer problem is extracted, and judges in customer problem " I will buy truck tire " whether containing decision model
Decision factor, the Information Extraction Model are then used to excavate, find, extracting the decision factor that user requests, contains in feedback;
Step S201, if the decision factor containing the decision model that has an impact in the decision factor of the customer problem, is extracted
The decision factor, and the decision factor is subjected to storage inside;Such as customer problem is that " I will buy truck tire " is assumed to take out
The decision factor of the customer problem taken is " truck ", if it influences also to contain " kilocalorie in the decision factor of decision model
The decision factor " truck " is then carried out storage inside by vehicle ";
Step S202, if do not contained in the decision factor of the customer problem or lack influence decision model decision because
Son, then the step of submitting necessary information into guidance user.
Preferably, further include following steps before step S200:
Judge that whether there is or not decision factors in the customer problem, if so, then entering step S200, is otherwise directly entered step S3.
Step S3, guidance user submit necessary information, and the decision factor extracted in user's input stores, will be internal
The decision factor of storage is sent to decision model one by one, and judges whether to determine next decision factor by decision model, when
Judging result is when being, to determine next decision factor.
Specifically, step S3 further comprises:
Step S300 issues the user with the inquiry for obtaining decision factor.
Step S301 obtains user to the feedback of inquiry, extracts the decision factor in user feedback;Specifically, in step
In 301, the decision factor in field feedback is extracted using Information Extraction Model.
The decision factor of extraction is carried out storage inside by step S302.That is when judge in step S2 extract use
There is no decision model that can be questioned closely that " you have into scene first problem with decision factor in decision factor in the problem of family
Without what brand requirements ", user can be answered by voice or text, and user answers " not having ", " will not have " to be stored to interior
Portion's storage unit A.
The decision factor of storage inside is sent to decision model by step S303 one by one;It, will be internal i.e. in step S303
The decision factor of storage unit A storage is sent to decision model one by one.
Decision model is also known as decision table or judgement table, is a kind of graphical tool in table shape, sentences suitable for description processing
The case where broken strip part is more, and each condition is combined with each other again, there are many decision schemes uses accurate and succinctly describes complex logic
Mode, will multiple conditions and these conditions meet after to execute that movement is corresponding, but different from the control in traditional program language
Sentence processed, decision table can clearly show multiple independent conditions and the direct connection of multiple movements.
Step S304 is judged whether to determine next decision factor by decision model, when the judgment result is yes, be determined
Next decision factor.That is, the decision factor of extraction is introduced into decision model, judge whether the result of decision out, no
It determines next decision factor when the result of decision again out, if the result of decision can be gone out, is directly entered next step step
S4。
Preferably, further include following steps before step 302:
The decision factor that step 301 extracts is compared with the decision factor of storage inside;
If skipping the problem containing the decision factor of the problem in the decision factor of storage inside, continue the problem
Next problem;
If not containing the decision factor of the problem in the decision factor of storage inside, the problem is enabled, is entered step
The decision factor contained in the user feedback of extraction is carried out internal preservation by S302.
Step S4, operational decisions model.
Specifically, the operation of the decision model includes the following steps:
Step S400, judges whether decision model contains mathematical computations or logical operation, has, and carries out mathematics or logic fortune
It calculates, enters step S401, nothing is then directly entered step S401;
Step S401, by decision model the result of decision and operation result be sent to result node, by result node pair
The result of decision of decision model transmission is integrated with calculated result.
Preferably, further include following steps before step S401:
Judge whether decision model has display format requirement, have, carries out Format adjusting, S401, nothing are entered step after adjustment
Then it is directly entered step S401.
Preferably, further include following steps before step S4:
Judge whether the decision model can export the result of decision, i.e., for exporting the default necessary letter an of result of decision
Whether breath has extracted complete;
If the default necessary information for exporting any result of decision has extracted completely, S4 operational decisions mould is entered step
Type and the operation result for exporting the decision model, otherwise jump procedure S3 guides user to submit necessary information again to user
Issue the inquiry for obtaining decision factor.
Such as when the decision factor in A is transmitted to decision model by step S3, the default necessary information of decision model is not yet
Completely, then will continue to that " you are how many at receptible price " questioned closely, user can be answered by voice or text, at this point, with
Family answers and will be stored to internal storage unit A " within 1000 yuan " " within 1000 yuan ", again by the decision in internal storage unit A
The factor is transmitted to decision model, if the default necessary information of decision model is not yet complete at this time, continues to question closely " tire adaptation ruler
It is very little to be how many ", condition 155mm, 165mm and 185mm are provided, user selects 185mm at this time, and being stored to internal storage unit A will
Decision factor is transmitted to decision model, until the default necessary information of " Michelin's BlahBlah tire " has been taken out in decision model
It takes whole.
At this time in step S4 operational decisions model, the result of decision is exported are as follows: " the information provided according to you, it is proposed that you buy rice
Its woods BlahBlah tire ".
Step S5 is obtained and is exported the result of decision exported in decision model and operation result.
Fig. 2 is a kind of system architecture diagram of Interactive Dynamic reasoning device of the present invention.As shown in Fig. 2, a kind of friendship of the present invention
Mutual formula Dynamic Inference device, comprising:
Ingress node 201, customer problem, according to the determination of the customer problem of identification and triggers corresponding scene for identification,
The Information Extraction Model and decision model of linkage triggering scene.
Specifically, as shown in figure 3, Ingress node 201 further comprises:
Customer problem recognition unit 2010 receives the natural language data of user, and right for checking scene entry condition
The natural language data carry out semantic understanding, to generate problem request.
Comprehensive score computing unit 2011 obtains scene entrance Candidate Set, user is asked for being requested according to described problem
Topic and the default problem of each scene entrance of scene entrance Candidate Set carry out comprehensive score calculating.
Since the targeted problem of different scenes is different, to be selected according to customer problem the strongest scene of specific aim into
Enter.For example, identifying that customer problem is " I will buy truck tire ", the scene entrance Candidate Set of acquisition includes scene 1 and field
Scape 2, scene 1 are that automobile tire buys scene, and scene 2 is that household electrical appliance buy scene, then can be according to customer problem " I of identification
Buy truck tire ", it is preset into question synthesis scoring with scene 1 and 2 entrance of scene.
Scenario triggered unit 2012, for the comprehensive score according to comprehensive score computing unit 2011 as a result, triggering is suitable
Scene, linkage triggering scene Information Extraction Model and decision model.Specifically, scenario triggered unit 2012 is to scene entrance
Candidate Set carries out comprehensive score sequence, selects the highest scene of scoring as scene entrance.For example, when according to comprehensive score result
When showing that the default question synthesis scoring of the entrance of customer problem and scene 1 is highest, scenario triggered unit 2012 then triggers scene
1, scene is bought into automobile tire, while linking and triggering the Information Extraction Model and decision model of scene 1.
As it can be seen that the present invention by scene entrance carry out Ingress node setting, judged by comprehensive score, can be more
Accurately identification user is intended to.
Decision factor extracts and judging unit 202 for extracting the decision factor in customer problem judges customer problem
The decision factor for whether containing the decision model that has an impact in decision factor, questions closely node 203 according to judging result starting or directly deposits
Store up the decision factor.Ingress node is first node of each scene, which is that operator creates in Scene Editor
After scene, first node established in the scene, the content of Ingress node is content related with scene, the work of Ingress node
Be match user statement it is whether related to the scene, if related to the scene, guide user into the scene friendship
Otherwise mutual process does not enter the scene interactivity process
Decision factor extracts and judging unit 202 specifically includes:
Decision factor extracting unit, for extracting the decision factor in customer problem using Information Extraction Model.In this hair
In bright, Information Extraction Model is then used to excavate, find, extracting the decision factor that user requests, contains in feedback;
Judging unit, in the decision factor for judging the customer problem whether the decision factor containing the decision model that has an impact
Judging result processing unit, the decision factor containing the decision model that has an impact in the decision factor of the customer problem
When, then the decision factor is extracted, and the decision factor is subjected to storage inside;Assuming that the decision factor of the customer problem extracted
For " weather ", influence that the decision factor " weather " is then carried out internal deposit also containing " weather " in the decision factor of decision model
Storage;It is not contained in the decision factor of the customer problem or lacks the decision factor for influencing decision model, then node is questioned closely in starting
203。
Preferably, decision factor extracts and judging unit 202 further includes that there are judging units for decision factor, for judging this
Whether there is or not decision factors in customer problem, if there is starting decision factor extracting unit, otherwise directly initiate and question closely node 203.
Node 203 is questioned closely, for guiding user to submit necessary information, the decision factor extracted in user's input is deposited
The decision factor of storage inside is sent to decision model one by one, and judges whether to determine next determine by decision model by storage
The plan factor determines next decision factor when the judgment result is yes.
It is specifically included as shown in figure 4, questioning closely node 203:
Inquiry unit 2031 issues the user with the inquiry for obtaining decision factor;
Feedback information extracting unit 2032 obtains user to the feedback of inquiry, extracts the decision factor in user feedback;Tool
Body, feedback information extracting unit 2032 extracts the decision factor in field feedback using Information Extraction Model;
The decision factor of extraction is carried out storage inside by storage unit 2033;
Decision factor transmission unit 2034, for the decision factor of storage inside to be sent to decision model one by one;
Decision factor determination unit 2035 determines next decision factor for judging whether by decision model, when sentencing
Disconnected result is when being, to determine next decision factor.That is, decision factor determination unit 2035 is by the decision factor of extraction
It is introduced into decision model, judges whether that the result of decision can be gone out, it is not possible to determine next decision again when going out the result of decision
The factor.
As it can be seen that the present invention is arranged by the node of questioning closely for the problem of questioning closely, bootable user's selection is provided in decision model
Decision factor.
Decision model running unit 204, for running the decision model.Specifically, the operation tool of the decision model
Body includes:
Judge whether decision model contains mathematical computations or logical operation, have, carry out mathematics or logical operation, nothing then will
The result of decision and operation result in decision model are sent to result node, the decision knot transmitted by result node to decision model
Fruit is integrated with calculated result.Preferably, by decision model the result of decision and operation result be sent to result node
Before, it also can determine whether decision model has display format requirement, have, carry out Format adjusting, without then by determining in decision model
Plan result and operation result are sent to result node.
Result node 205, for obtain decision model output the result of decision and operation result, and to the result of decision with
Operation result carries out integration output.The present invention is by the setting of result node, so that Interactive reasoning conclusion editable, makes to interact
Formula Dynamic Inference forms closed loop.
Preferably, as shown in figure 5, Dynamic Inference device when the interaction of the present invention, further includes: condition node 206, for will
The decision factor for questioning closely the extraction of node 203 is compared with the decision factor of storage inside;In the decision factor of storage inside
When decision factor containing the problem, then skip the problem, continue next problem of the problem, in storage inside decision because
The problem is enabled when decision factor in son not containing the problem, and the decision factor contained in the user feedback of extraction is carried out
Inside saves.
Preferably, Dynamic Inference device when the interaction of the present invention further include: node 207 is jumped, for judging decision model
The result of decision whether can be exported, i.e., whether the default necessary information that judgement is used to export a result of decision has extracted completely,
And then start decision model operation in judging result to have extracted completely for exporting the default necessary information of any result of decision
204 operational decisions model of unit simultaneously exports the result of decision and operation result, otherwise jumps to and questions closely the guidance user's offer of node 203
Necessary information obtains the inquiry of decision factor to issue the user with again.The present invention by jump node setting, it can be achieved that
Scene, question closely or condition between jump.
Fig. 6 is the flow chart of the Interactive Dynamic inference method of the embodiment of the present invention.Cooperation Fig. 6 further illustrates this below
The Interactive Dynamic reasoning process of invention:
1, it identifies customer problem, understands customer problem demand, trigger and enter respective fields scape.
Specifically, step 1 includes:
A, customer problem and each scene entrance are preset into problem and carries out comprehensive score;
B, comprehensive score calculated result is obtained;
C, triggering and comprehensive score result are highest scene.
Such as user puts question to: I will buy tire, and by comprehensive score, " I will buy tire " enters with automobile tire purchase scene
Mouth problem buys scene into automobile tire.
2, the decision factor contained in customer problem is extracted, judges whether the decision factor of customer problem contains default decision
The factor.
A, when into scene, the corresponding Information Extraction Model of scene synchronous averaging scene, and judge be in customer problem
The no decision factor containing the decision model that has an impact;
If b, containing the decision factor of decision model of having an impact in customer problem, extraction changes decision factor, and by decision
The factor is saved in storage inside A;
If the decision factor for influencing decision model c, is not contained or lacked in customer problem.Then enter guidance user to provide
Necessary information Step.
Call scene Information Extraction Model, judge in customer problem " I will buy tire " whether containing decision model certainly
Otherwise the plan factor enters next step if so, being then saved into internal storage unit A.
3, guidance user submits necessary information
A, the inquiry for obtaining decision factor is issued the user with;
B, the decision factor in user feedback is collected using Information Extraction Model;
C, the decision factor in user feedback that will acquire is compared with storage inside A
If 1., storage inside A contain the decision factor of the problem, skip the problem, continue the next of the problem
Problem.
If 2., storage inside A enable the problem not containing the decision factor of the problem, and to containing in feedback
Decision factor be saved in storage inside A.
When the decision factor in customer problem does not have decision model that can use decision factor, into the progress of scene first problem
" you are either with or without what brand requirements " are questioned closely, user can be answered by voice or text, and user answers " not having ", " will not be had
Have " it is stored to storage inside A, the decision factor in A is transmitted to decision model,
4, it sends decision model one by one by the decision factor of storage inside A, decision model saves decision factor;
5, judge whether decision model can export the result of decision, i.e., whether the necessary information of a result of decision has taken out out
Take whole, if the default necessary information of any result of decision is complete, otherwise output should as a result, return to above-mentioned steps
3.a.。
Because the default necessary information of decision model is not yet complete, continue to question closely " you are how many at receptible price ", use
Family can be answered by voice or text, at this point, user's answer will be stored to inside " within 1000 yuan " " within 1000 yuan " and deposit
Decision factor in internal storage unit A is transmitted to decision model by storage unit A, and the default necessary information of decision model is not yet
It is not yet complete, then continue that " tire fit size is how many " questioned closely, condition 155mm, 165mm and 185mm are provided, user selects at this time
Select 185mm, the decision factor in A be transmitted to decision model, at this time in decision model " Michelin's BlahBlah tire " it is pre-
If necessary information has extracted completely.
6, operational decisions model.
1., judge whether decision model contains mathematical computations or logical operation, have, carry out mathematics or logical operation, without then
It carries out in next step;
2., judge whether decision model has display format requirement, have, carry out Format adjusting, without then carrying out in next step.
When the default necessary information in decision model extracts completely, then operational decisions model, exports the result of decision are as follows: " root
The information provided according to you, it is proposed that you buy Michelin's BlahBlah tire "
7, the result of decision exported in decision model and operation result are obtained, and it is integrated.
8, answer is pushed.
In conclusion a kind of Interactive Dynamic inference method of the present invention and device by extract the decision in customer problem because
Son, and do not contained in the decision factor for judge the customer problem or lack influence decision model decision factor when, guidance
User submits necessary information, extract user input in decision factor, and by operational decisions model export the result of decision with
Operation result, the present invention can not only identify decision factor in customer problem, eliminate user and repeat the trouble asked, improve use
The experience at family, and the challenge of user can be solved.
Compared with prior art, the present invention has the advantage that
1, decision factor in the recognizable customer problem of the present invention, saves deduplication and asks, improve intelligence and user experience;
2, the present invention can solve the challenge of user using decision model using Interactive Dynamic reasoning;
3, the present invention reduces interaction round when crossing over different scenes by the setting of global variable;
4, the present invention passes through the setting of scene variable, provides necessary decision factor for decision model;
5, the setting of the invention by result node, so that Interactive reasoning conclusion editable;
6, the setting of the invention by jumping node, so that scene reusable, interaction is transportable.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.Any
Without departing from the spirit and scope of the present invention, modifications and changes are made to the above embodiments by field technical staff.Therefore,
The scope of the present invention, should be as listed in the claims.
Claims (10)
1. a kind of Interactive Dynamic inference method, includes the following steps:
Step S1 identifies customer problem, according to the determination of the customer problem of identification and triggers corresponding scene, linkage triggering scene
Information Extraction Model and decision model;
Whether step S2 extracts the decision factor in customer problem, judge in the decision factor of customer problem containing the decision that has an impact
The decision factor of model enters step S3 according to judging result or directly stores the decision factor;
Step S3, guidance user submit necessary information, and the decision factor extracted in user's input carries out storage inside, will be internal
The decision factor of storage is sent to decision model one by one, and judges whether to determine next decision factor by decision model, when
Judging result is when being, to determine next decision factor;
Step S4 is triggered and is run the decision model;
Step S5 is obtained and is exported the result of decision exported in the decision model and operation result.
2. a kind of Interactive Dynamic inference method as described in claim 1, which is characterized in that step S1 further comprises:
It checks scene entry condition, receives the natural language data of user, and semantic understanding is carried out to the natural language data,
To generate problem request, is requested according to described problem, obtain scene entrance Candidate Set, and integrate to scene entrance Candidate Set
Marking and queuing selects the highest scene of scoring as scene entrance.
3. a kind of Interactive Dynamic inference method as described in claim 1, which is characterized in that step S3 further comprises:
Issue the user with the inquiry for obtaining decision factor;
User is obtained to the feedback of inquiry, extracts the decision factor in user feedback;
The decision factor of extraction is subjected to storage inside;
The decision factor of storage inside is sent to decision model one by one;
Judge whether to determine next decision factor by decision model, when the judgment result is yes, according to the decision of extraction because
Son determines that next decision factor then enters step S5 when the judgment result is no.
4. a kind of Interactive Dynamic inference method as claimed in claim 3, which is characterized in that step S2 further comprises:
The decision factor in customer problem is extracted using Information Extraction Model;
Judge the decision factor for whether containing the decision model that has an impact in the decision factor of the customer problem;
If the decision factor containing the decision model that has an impact in the decision factor of the customer problem, the decision factor is extracted, and will
The decision factor carries out storage inside;If the decision for influencing decision model is not contained or lacked in the decision factor of the customer problem
The factor then enters the step of issuing the user with the inquiry for obtaining decision factor.
5. a kind of Interactive Dynamic inference method as claimed in claim 3, which is characterized in that in the decision by extraction because
Further include following steps before son carries out storage inside step:
Decision factor in the user feedback of extraction is compared with the decision factor of storage inside;
If skipping the problem containing the decision factor of the problem in the decision factor of storage inside, continue the next of the problem
A problem;If not containing the decision factor of the problem in the decision factor of storage inside, the problem is enabled, by the user of extraction
The decision factor contained in feedback carries out storage inside.
6. a kind of Interactive Dynamic inference method as described in claim 1, which is characterized in that before step S4, further include
Following steps:
Judge the decision model for exporting whether the default necessary information of a result of decision has extracted completely;
If the default necessary information for exporting any result of decision has extracted completely, enters step S4 and run the decision model
Type, the S3 that otherwise gos to step guidance user submit necessary information.
7. a kind of Interactive Dynamic inference method as described in claim 1, which is characterized in that in step S4, the decision
The operation of model includes:
Step S400, judges whether decision model contains mathematical computations or logical operation, has, and carries out mathematics or logical operation, into
Enter step S401, nothing then carries out being directly entered step S401;
Step S401, by decision model the result of decision and operation result be sent to result node, by result node to decision
The result of decision of model transmission is integrated with calculated result.
8. a kind of Interactive Dynamic reasoning device, comprising:
Ingress node, customer problem, according to the determination of the customer problem of identification and triggers corresponding scene for identification, linkage triggering
The Information Extraction Model and decision model of scene;
Decision factor extracts and judging unit, for extracting the decision factor in customer problem, judge the decision of customer problem because
The decision factor for whether containing the decision model that has an impact in son, questions closely node according to judging result starting or is directly stored;
Node is questioned closely, for guiding user to submit necessary information, the decision factor extracted in user's input is stored, will be interior
The decision factor of portion's storage is sent to decision model one by one, and judges whether to determine next decision factor by decision model,
When the judgment result is yes, next decision factor is determined;
Decision model running unit, for triggering simultaneously operational decisions model;
Result node, for obtaining and exporting the result of decision exported in decision model and operation result.
9. a kind of Interactive Dynamic reasoning device as claimed in claim 8, it is characterised in that: described device further includes conditional sections
Point, for the decision factor for questioning closely node extraction to be compared with the decision factor of storage inside;If storage inside
Decision factor in decision factor containing the problem when, then skip the problem, continue next problem of the problem, if interior
The problem is enabled when not containing the decision factor of the problem in the decision factor of portion's storage, and will be contained in the user feedback of extraction
Decision factor carry out storage inside.
10. a kind of Interactive Dynamic reasoning device as claimed in claim 8, which is characterized in that described device further includes jumping
Node, for judging whether default necessary information of the decision model for exporting a result of decision has extracted completely, and in sentencing
Disconnected result is to have extracted completely for exporting the default necessary information of any result of decision, then it is single to start the decision model operation
Otherwise member jumps to the node of questioning closely and user is guided to submit necessary information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810752461.2A CN109063840A (en) | 2018-07-10 | 2018-07-10 | A kind of Interactive Dynamic inference method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810752461.2A CN109063840A (en) | 2018-07-10 | 2018-07-10 | A kind of Interactive Dynamic inference method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109063840A true CN109063840A (en) | 2018-12-21 |
Family
ID=64819501
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810752461.2A Pending CN109063840A (en) | 2018-07-10 | 2018-07-10 | A kind of Interactive Dynamic inference method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109063840A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105701211A (en) * | 2016-01-13 | 2016-06-22 | 北京光年无限科技有限公司 | Question-answering system-oriented active interaction data processing method and system |
CN106056207A (en) * | 2016-05-09 | 2016-10-26 | 武汉科技大学 | Natural language-based robot deep interacting and reasoning method and device |
CN107944560A (en) * | 2017-12-08 | 2018-04-20 | 神思电子技术股份有限公司 | A kind of natural language semantic reasoning method |
CN108197191A (en) * | 2017-12-27 | 2018-06-22 | 神思电子技术股份有限公司 | A kind of scene of more wheel dialogues is intended to interrupt method |
CN108228764A (en) * | 2017-12-27 | 2018-06-29 | 神思电子技术股份有限公司 | A kind of single-wheel dialogue and the fusion method of more wheel dialogues |
-
2018
- 2018-07-10 CN CN201810752461.2A patent/CN109063840A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105701211A (en) * | 2016-01-13 | 2016-06-22 | 北京光年无限科技有限公司 | Question-answering system-oriented active interaction data processing method and system |
CN106056207A (en) * | 2016-05-09 | 2016-10-26 | 武汉科技大学 | Natural language-based robot deep interacting and reasoning method and device |
CN107944560A (en) * | 2017-12-08 | 2018-04-20 | 神思电子技术股份有限公司 | A kind of natural language semantic reasoning method |
CN108197191A (en) * | 2017-12-27 | 2018-06-22 | 神思电子技术股份有限公司 | A kind of scene of more wheel dialogues is intended to interrupt method |
CN108228764A (en) * | 2017-12-27 | 2018-06-29 | 神思电子技术股份有限公司 | A kind of single-wheel dialogue and the fusion method of more wheel dialogues |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106951468B (en) | Talk with generation method and device | |
Duijst | Can we improve the user experience of chatbots with personalisation | |
CN109446306A (en) | A kind of intelligent answer method of more wheels dialogue of task based access control driving | |
CN109062928A (en) | A kind of method and system that prompt recommendation is got on the bus a little | |
CN107329967A (en) | Question answering system and method based on deep learning | |
CN110909145B (en) | Training method and device for multi-task model | |
JP6851894B2 (en) | Dialogue system, dialogue method and dialogue program | |
CN110020426A (en) | User's consulting is assigned to the method and device of customer service group | |
CN110032623A (en) | The matching process and device of user's question sentence and knowledge dot leader | |
US20230394247A1 (en) | Human-machine collaborative conversation interaction system and method | |
CN107368548A (en) | Intelligent government affairs service interaction method and system | |
CN111191450A (en) | Corpus cleaning method, corpus entry device and computer-readable storage medium | |
CN112183098B (en) | Session processing method and device, storage medium and electronic device | |
CN108959531A (en) | Information search method, device, equipment and storage medium | |
CN110008308B (en) | Method and device for supplementing information for user question | |
CN110019742A (en) | Method and apparatus for handling information | |
CN113312468B (en) | Conversation mode-based conversation recommendation method, device, equipment and medium | |
CN111309914A (en) | Method and device for classifying multiple rounds of conversations based on multiple model results | |
CN116059646A (en) | Interactive expert guidance system | |
CN111259124A (en) | Dialogue management method, device, system and storage medium | |
CN110610627A (en) | Heuristic poetry learning method and device | |
CN106558309A (en) | A kind of spoken dialog strategy-generating method and spoken dialog method | |
CN115203393A (en) | Dialogue response method and system, electronic equipment and storage medium | |
CN112925894B (en) | Method, system and device for matching bid-asking questions in conversation | |
CN109063840A (en) | A kind of Interactive Dynamic inference method and device |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181221 |
|
RJ01 | Rejection of invention patent application after publication |