Movatterモバイル変換


[0]ホーム

URL:


CN108804643A - Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions - Google Patents

Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions
Download PDF

Info

Publication number
CN108804643A
CN108804643ACN201810570492.6ACN201810570492ACN108804643ACN 108804643 ACN108804643 ACN 108804643ACN 201810570492 ACN201810570492 ACN 201810570492ACN 108804643 ACN108804643 ACN 108804643A
Authority
CN
China
Prior art keywords
node
child node
child
session
branch
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
Application number
CN201810570492.6A
Other languages
Chinese (zh)
Inventor
杨凯程
张正
蒋宏飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Xuan Yi Science And Technology Co Ltd
Original Assignee
Beijing Xuan Yi Science And Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Xuan Yi Science And Technology Co LtdfiledCriticalBeijing Xuan Yi Science And Technology Co Ltd
Priority to CN201810570492.6ApriorityCriticalpatent/CN108804643A/en
Publication of CN108804643ApublicationCriticalpatent/CN108804643A/en
Priority to CN201811497347.6Aprioritypatent/CN109614474A/en
Pendinglegal-statusCriticalCurrent

Links

Landscapes

Abstract

A kind of methods that the embodiment of the present invention discloses the flow dispensing unit of more wheel session, carries out more wheel interactions using more wheel session flows that dispensing unit is configured, and the intelligent robot interactive systems applied to more wheel interactions.Wherein, flow dispensing unit includes start node, knowledge node, branch node and end node.Start node with the first child node for connecting, with from start node circulation to the first child node;First child node with branch node for connecting, with from the circulation of the first child node to branch node;Branch node with end node, the first child node or the second child node for connecting, with from branch node circulation to end node, the first child node or the second child node;Second child node with end node for connecting, with from the circulation of the second child node to end node.Above-mentioned flow dispensing unit simplifies the configuration of session flow, while intention is introduced in knowledge node, and interactive system is made to be enable to respond quickly user, provides more acurrate abundant answer.

Description

Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions
Technical field
The present invention relates to artificial intelligence fields, and in particular to a kind of flow dispensing unit of more wheel sessions.In addition, the applicationIt further relates to one kind and taking turns exchange method and a kind of intelligent robot interactive system more.
Background technology
In intelligent answer field, other than the single-wheel session of common question-response form, common also has more wheel meetingsWords.This is because in practical applications, intelligent Answer System problem to be solved is likely to a complicated Diversification Type and knowsKnow, rather than simple question-response form.Developer needs pre-configured session flow, so that intelligent Answer SystemUser can be guided to complete dialogue step by step, to feed back to user most according to the session flow, in a manner of taking turns interaction moreWhole answer.
In the prior art, the flow of the more wheel sessions of configuration needs the developer's ability for understanding programming and algorithm flowIt can carry out, configuration process is cumbersome and inconvenient.In addition, in the flow of the more wheel sessions of configuration, the knowledge node in flow is typicallyThe preset signal language of developer.Intelligent Answer System will be prompted to language and be shown to user, then obtains user and is inputted according to signal languageInformation, determine next node that knowledge node will circulate further according to these information.By such method, intelligenceQuestion answering system can carry out more wheel sessions to guide user to input more information, recycle these information to generate answer, in this wayIntelligent Answer System responded in conversation procedure it is relatively slow.
Invention content
In order to solve the above-mentioned technical problem, the application provides a kind of new flow dispensing unit for being applied to more wheel sessions,To simplify the configuration process of session flow, while intention is introduced in knowledge node, thereby using family in the more wheel meetings of configurationThe type that can determine the node content specifically used when the flow of words according to actual application scenarios more flexiblely, makes profitWhen being applied in interactive system, interactive system can quickly be rung the more wheel session flows configured with the flow dispensing unitUsing family, more acurrate abundant answer is provided.
In a first aspect, a kind of flow dispensing unit of more wheel sessions is provided, including with lower component:
Start node, the trigger condition for obtaining more wheel sessions;
Knowledge node, including the first child node and the second child node, first child node are described for obtaining signal languageSignal language is the sentence that next round session information is inputted for boot sessions user, and second child node is for obtaining the first meaningFigure;
Branch node, for obtaining circulation condition, the next round session information got according to the circulation condition is trueThe information in the circulation direction of fixed first child node;And
End node, for being determined to the trigger condition during end node with circulation condition using start node circulationReply source, or determine reply source using first intention;
Wherein, start node with the first child node for connecting, to be saved from start node circulation to first sonPoint;First child node with branch node for connecting, with from first child node circulation to the branch node;Branch nodeFor being connect with end node, the first child node or the second child node, with from the branch node circulation to the end node,First child node or second child node;Second child node with end node for connecting, to be saved from second sonThe end node is arrived in point circulation.
With reference to first aspect, in first aspect in the first possible realization method, the first intention includes that slot is intended to,The slot intention includes at least one slot element;
When the first intention that the second child node is got is that slot is intended to, second child node is additionally operable to obtain slot intentionAll slot elements, and, a slot be intended to all slot elements all by session subscriber input session information it is filledIn the case of circulation to next node.
The first realization method with reference to first aspect, in second of possible realization method of first aspect, describedTwo child nodes are additionally operable to obtain leading question corresponding with each slot element, and the leading question is guiding user's input for fillingThe sentence of the session information of slot element.
With reference to first aspect and above-mentioned possible realization method, in first aspect in the third possible realization method, instituteThe first child node is stated to be additionally operable to limit the type of the next round session information inputted by session subscriber.
With reference to first aspect and above-mentioned possible realization method, in the 4th kind of possible realization method of first aspect, instituteIt includes second intention, entity or knowledge classification to state trigger condition, and the entity is word or phrase, and the knowledge classification is through excessiveThe professional knowledge of class.
Second aspect provides a kind of more wheel session streams that any flow dispensing unit using first aspect is configuredThe method that Cheng Jinhang takes turns interaction more, including:
Obtain the first round session information inputted by session subscriber;
Start to save if first round session information meets the trigger condition on start node, output in more wheel session flowsThe first signal language in the first child node that point flows directly to;
It obtains by session subscriber according to the first signal language input the second wheel session information;
If the second wheel session information meets the circulation item either one or two of being connect with the first child node on the first branch nodePart, it is determined that meet the next node that the first branch node of condition circulates;
If the next node that the first branch node circulates is the second child node, it is determined that second child nodeIt circulates to end node, answer is found out from the reply source that the end node determines.
In conjunction with second aspect, in second aspect in the first possible realization method, if the first branch node is circulatedThe next node arrived is the second child node, it is determined that second child node circulates includes to the step of end node:
If the next node that the first branch node circulates is the second child node, and in second child nodeFirst intention be that slot is intended to, then judge whether all slot elements that the slot is intended to have been filled;
If all slot elements have been filled, it is determined that the end node that second child node circulates.
Judge institute in second of possible realization method of second aspect in conjunction with the first realization method of second aspectAfter stating the step whether all slot elements that slot is intended to have been filled, further include:
If at least one slot element is not filled, it is sequentially output corresponding with the slot element not being filledLeading question;
The session information that user inputs according to the leading question is obtained successively;
Utilize session information filling slot element input by user.
In conjunction with second aspect and above-mentioned possible realization method, in second aspect in the third possible realization method, reallySurely after the step of meeting the next node that the first branch node of condition circulates, further include:
If the next node that the first branch node circulates is another first child node, export described anotherThe second signal language in a first child node;
Obtain the third round session information inputted according to the second signal language by session subscriber;
If third round session information satisfaction either one or two of connect the second branch node with another described first child nodeOn circulation condition, it is determined that meet the next node that the second branch node of condition circulates;
If the next node that the second branch node circulates is the second child node, it is determined that second child nodeIt circulates to end node, answer is found out from the reply source that the end node determines.
The third aspect, provides a kind of intelligent robot interactive system, and the interactive system includes:Utilize appointing for first aspectA kind of more wheel session flows, acquiring unit and processing unit that flow dispensing unit is configured;
The acquiring unit is for obtaining the first round session information inputted by session subscriber, and acquisition by session subscriberAccording to the first signal language input the second wheel session information;
The processing unit is used to meet the triggering item in more wheel session flows on start node in first round session informationIn the case of part, the first signal language in the first child node that start node flows directly to is exported;Session information is taken turns secondIn the case of meeting the circulation condition either one or two of being connect with the first child node on the first branch node, determination meets the of conditionThe next node that one branch node circulates;And the next node to circulate in the first branch node is secondIn the case of child node, determine that second child node circulates to end node, from the reply source that the end node determinesIn find out answer.
The flow dispensing unit of above-mentioned more wheel sessions includes start node, knowledge node, branch node and terminates sectionThis four components of point.User when configuring the flow of a much more specific wheel sessions, can directly apply these components intoRow combination.It is starting with start node, the trigger conditions of mostly wheel sessions is set on start node, triggers entire more wheel sessionsFlow.Then Configuration knowledge node makes start node circulation to knowledge node.Knowledge node includes the first child node and secondChild node, the first child node are to input next round session letter for boot sessions user for obtaining signal language, the signal languageThe sentence of breath;Second child node is for obtaining first intention.Then be the first child node configure its will circulate toA few branch node, is arranged circulation condition on branch node, to be determined according to next round session information input by userWhich circulate from the first child node to branch node.Branch node can circulate to another the first child node, can also flowGo to the second child node.When branch node circulation is to another first child node, user can be guided to input more session lettersBreath, therefore can be that branch node configures its next first child node that will be circulated, the content of the first child node is set, i.e.,Signal language, to continue to guide user's input next round session information again.It is second when branch node circulates to the second child nodeChild node configures first intention.End node is finally configured, to terminate more wheel sessions.When a upper node for end node is theWhen one child node, trigger condition and circulation condition during the end node that circulated using start node determine reply source, make intelligenceEnergy question answering system can generate the final answer for feeding back to user from reply source.When a upper node for end node is secondWhen child node, determines reply source using the node content of second child node, that is, first intention, make intelligent Answer SystemThe final answer for feeding back to user can be generated from reply source.
User is directly configured and combines using these components according to the difference of application scenarios, you can configuration obtains oneThe flow of a complete more wheel sessions.More wheel session flows that configuration obtains are applied in intelligent Answer System, intelligent answerSystem can be based on this more wheel session flow, and by taking turns the form of interaction, guiding user inputs more information more, so that it is determined thatReply source.There are two types of the modes for determining reply source, and one is the information inputted in mostly wheel session using user to be replied to determineSource, another kind are that the information inputted in mostly wheel session using user determines first intention, recycle first intention to determine and replySource.After reply source determines, intelligent Answer System can generate final answer feedback to use from the specific reply sourceFamily.On the one hand, by introducing first intention so that the answer that intelligent Answer System ultimately generates is more abundant compared with originally;On the other hand, since it is determined the source of reply, intelligent Answer System are enable to respond quickly during conversating with user, togetherWhen provide more accurate answer, make dialogue more remarkable fluency.In addition, user can be flexible in the flow of the more wheel sessions of configurationGround determines the circulation direction between the knowledge node specifically used and node according to actual application scenarios, even if user has noAlgorithm basis, does not understand programming, being capable of flexibly and easily configuration session flow yet.
Description of the drawings
In order to illustrate more clearly of the technical solution of the application, letter will be made to attached drawing needed in the embodiment belowSingly introduce, it should be apparent that, for those of ordinary skills, without having to pay creative labor,Other drawings may also be obtained based on these drawings.
Fig. 1 is that the application takes turns in one embodiment of flow dispensing unit of session more, the configuration interface of dispensing unit;
Fig. 2 is by the enlarged drawing of the session flow configured in the configuration interface in Fig. 1, to be matched using the flow of the applicationSet the session flow instance that unit is configured;
Fig. 3 is that the application takes turns in one embodiment of flow dispensing unit of session more, the configuration interface of start node;
Fig. 4 is that the application takes turns in one embodiment of flow dispensing unit of session more, the first child node in knowledge nodeConfiguration interface;
Fig. 5 is that the application takes turns in one embodiment of flow dispensing unit of session more, the second child node in knowledge nodeConfiguration interface;
Fig. 6 is that the application takes turns in one embodiment of flow dispensing unit of session more, the configuration interface of branch node;
Fig. 7 is that the application takes turns in one embodiment of flow dispensing unit of session more, and the part of end node configures boundaryFace;
Fig. 8 is another session flow instance for being configured using the flow dispensing unit of the application;
Fig. 9 is the flow chart of one embodiment of the method for more wheel interactions in the application;
Figure 10 is the flow chart of the one of which realization method of S400 steps in more wheel exchange methods of the application;
Figure 11 is the structural schematic diagram of one embodiment of intelligent robot interactive system in the application.
Specific implementation mode
It elaborates below to embodiments herein.
Intention in the application refers to user view (User Intent), it is understood that in user's asked questionsThought is thought.If the central idea of different problems is the same, that is, belong to the same intention.For example, a problem is that " I buys paper urineTrousers post what express delivery Xiamen gives?", another problem is " what express delivery you post ", and both of these problems are although different, stillThe central idea of the two is the same, and therefore, being intended that for being identified from two problems is the same, is all that " what inquiry delivery usesExpress delivery ".In the prior art, intention, such as user can be identified from content input by user using intention assessment technologySeveral keywords are inputted in a search engine, these keywords and other useful information can be utilized to identify the intention of userAfterwards, then the more accurate search result of user is fed back to.For example, user inputs " water ", search engine can need to utilize thisRecord is clicked in a keyword and the previous inquiry of user, identifies that the intention of user refers to that " toner " still " is drunkWater ".
And the scheme of the application includes to configure for the first time by the flow dispensing unit for being intended to be introduced into more wheel sessionsThe more wheel session flows being intended to.Intelligent Answer System can determine problem using such more wheel session flows using being intended toReply source improve the response speed of answer to provide more accurate answer.
Referring to FIG. 1, in one embodiment of the present invention, a kind of flow dispensing unit of more wheel sessions, packet are providedIt includes with lower component:
Start node, the trigger condition for obtaining more wheel sessions;
Knowledge node, including the first child node and the second child node, first child node are described for obtaining signal languageSignal language is to input next round session sentence for boot sessions user, and second child node is for obtaining first intention;
Branch node, for obtaining circulation condition, the next round session information got according to the circulation condition is trueThe information in the circulation direction of fixed first child node;And
End node, for being determined to the trigger condition during end node with circulation condition using start node circulationReply source, or determine reply source using first intention;
Wherein, start node is with the first child node or the second child node for connect, with from the start node circulate toFirst child node or second child node;First child node with branch node for connecting, to be saved from first sonThe branch node is arrived in point circulation;Branch node with end node, the first child node or the second child node for connecting, with from instituteBranch node circulation is stated to the end node, first child node or second child node;Second child node be used forEnd node connects, with from second child node circulation to the end node.
In one embodiment, the interface of flow dispensing unit is please referred to Fig.1 to Fig. 2.Wherein, start node is for obtainingThe trigger conditions of more wheel sessions, trigger condition can be second intention, entity or knowledge classification, as shown in Figure 3.
Second intention can be identified using existing intention assessment technology from session information input by user.With silverFor row customer service field, second intention can be " how credit card is handled ", and user inputs " I wants to handle credit card " or " asksAsk credit card what is to be done reason " etc. session informations, can therefrom identify the same second intention, i.e., " how credit card is doneReason ".If in a session flow being configured, the trigger condition of start node is configured as second intention " credit cardHow to handle ", then when intelligent Answer System identifies the intention from the session of user, which is triggered, can be withIt circulates from the start node in the session flow to next node.Second intention herein, configuration personnel can be in configurationIt is chosen from a preset intention library, still by taking bank's customer service field as an example, preset intention library can include multiple defaultIntention, such as " how credit card is handled ", " how applying providing a loan ", " how to handle and report the loss " etc..
In this application, entity refers to word or phrase, these words or phrase usually with the specific business in application scenariosIt is related.Such as in the application scenarios in bank's customer service field, word or phrase here can be " application ", " reporting the loss " etc..ConfigurationPersonnel can preset specific entity as trigger condition in start node, once include in session information input by userOne or more preset entities, then can trigger the session flow.
In this application, knowledge classification refers to having already passed through the professional knowledge of classification, for example, with bank's customer service fieldMay include credit card business, debit card business etc. for business system, so relevant professional knowledge can also be divided into creditCard class professional knowledge, debit card class professional knowledge etc..Configuration personnel can preset specific knowledge classification as trigger condition, comeTrigger the session flow.
Start node with the first child node for connecting, with from start node circulation to first child node.It needsIt is noted that start node can be also used for connecting with the second child node, to circulate from the start node to described secondChild node, the application are without limitation.But when start node is directly connect with the second child node, due to the second child nodeIt is connect with end node, so the session flow configured is the session flow of single-wheel interaction in general, it is unsuitableIt applies in the flow configuration of mostly wheel session.
Knowledge node includes the first child node and the second child node.First child node is for obtaining signal language, and firstChild node with branch node for connecting, with from first child node circulation to the branch node.When session subscriber inputsNext round session information when meeting the circulation condition on some branch node, just circulate from the first child node to this and meet phaseOn the branch node for the condition that should circulate.Configuration personnel can be specific to input according to the difference of the application scenarios of session flowSignal language, referring to FIG. 4, it is " it is much this year to may I ask you " that signal language, which can for example be arranged,.In addition, the first child node can also be usedIn the type for limiting next round session information input by user.The type of session information can be the date, the age or otherSpecific numerical value or character.For example, when the reply type for configuring the first child node of personnel depaly is age type, then user returnsMultiple information must be age type, such as " 20 years old ", such intelligent Answer System could the information based on the reply (i.e. by userThe next round session information of input) and the flow that has configured, to determine which branch node flowed to from the first child node.
Second child node is for obtaining first intention, and the second child node with end node for connecting, with from described secondChild node circulates to the end node.Referring to FIG. 5, configuration personnel can select in configuration from a preset intention libraryIt takes.By taking bank's customer service field as an example, preset intention library can include multiple preset intentions, for example " how credit card is doneReason ", " how applying providing a loan ", " how to handle and report the loss " etc..Herein, the second meaning is chosen with aforementioned in the intention library for choosing first intentionThe intention library of figure can be the same intention library, can also be different and be intended to library, and the application is not construed as limiting this.
For obtaining circulation condition, the next round session information got according to the circulation condition determines branch nodeThe information in the circulation direction of the first child node.Referring to FIG. 6, specific stream can be arranged in configuration personnel directly on branch nodeTurn condition.Circulation condition can in conjunction with >, < ,=, ≠ etc. oeprators describe, such as " > 18 " can also be with canonicalForm describes, such as " have wages ".Session next round session information input by user is compared with circulation condition, so that it mayTo determine which branch node circulation is arrived by the first child node.Branch node can be used for and end node, the first child nodeOr the second child node connection, with from branch node circulation to the end node, first child node or described secondChild node.
Branch node can be connect with the second child node with end node, and end node is arrived with circulation.Work as end nodeA upper node be branch node when, end node be used for using start node circulation end node during trigger conditionReply source is determined with circulation condition.When a upper node for end node is the second child node, end node is used to utilize theOne is intended to determine reply source.Here, reply source may include one by one knowledge, one by one professional knowledge classification and/Or knowledge base one by one.Knowledge, professional knowledge classification and knowledge base are stored in advance in intelligent Answer System.At itIn a kind of middle realization method, the configuration personnel of session flow can be according to the difference of application scenarios, to determine which is specifically selectedA reply source, to construct a correspondence between the first intention of each and reply source.It is saved when in session flowThe second child node that node content is first intention, when then circulation is to end node, the end node are arrived in point circulationReply source is determined using first intention.
For example, Fig. 8 is the example of the session flow gone out using the flow configuration of described dispensing unit in the present embodiment, Fig. 2For the example of another session flow configured.Assuming that trigger condition is second intention " how credit card is handled ".
In the example of fig. 8, the second child node is not included in the session flow, contains only start node, the first son sectionPoint " it is much this year to may I ask you?", " may I ask whether you have wages?" and " may I ask whether you have house?", branch node " > 18Year ", " < 18 years old ", " having wages ", " no wages ", " having house " and " no house " and end node.Wherein, most left in Fig. 8A upper node for the end node of side is branch node " having wages ", so, which is used to utilize start node streamTurn the trigger condition during end node and circulation condition, i.e. second intention " how credit card is handled ", circulation condition " > 18Year " and circulation condition " having wages ", reply source is determined according to this three.
In the example in figure 2, which includes 4 the second child nodes, and node content is respectively first intention 1" thering is wages credit card to handle rule more than 18 years old ", first intention 2 " handling rule without wages credit card more than 18 years old ", the first meaningFig. 3 " thering is house credit card to handle rule less than 18 years old " and first intention 4 " handling rule without house credit card less than 18 years old ".It is rightEach first intention, configuration personnel construct its correspondence between reply source in configuration flow.Such asFirst intention 1, corresponding reply source are the knowledge base A that a credit card handles field;It is right therewith for first intention 2The reply source answered is specific a few knowledge and knowledge base B.For the end node of the leftmost side, a node is thereonSecond child node.So using the node content of the second child node, that is, first intention 1 " had wages credit card more than 18 years oldHandle rule ", to determine reply source, i.e. knowledge base A.Then, intelligent Answer System finds out answer feedback to use from A againFamily, to solve the problems, such as similar " how credit card is handled " of user's proposition.
In addition, referring to FIG. 7, end node can be also used for configuring other subsequent actions, such as " turning artificial ", " waitingUser inputs " etc., to provide other solutions when intelligent Answer System can not provide suitable answer or can not provide answerCertainly scheme promotes user experience, preferably meets business demand.
The above-mentioned flow dispensing units for being applied to more wheel sessions, include start node, knowledge node, branch node andThis four components of end node.User can directly apply these when configuring the flow of a specific more wheel sessionsComponent is combined.It is starting with start node, the trigger condition of more wheel sessions is set on start node, it is entire more to triggerTake turns the flow of session.Then Configuration knowledge node makes start node circulation to knowledge node.Knowledge node includes the first child nodeWith the second child node, the first child node is to input next round for boot sessions user for obtaining signal language, the signal languageThe sentence of session information;Second child node is for obtaining first intention.Then configuring it for the first child node will circulateCirculation condition is arranged at least one branch node arrived on branch node, to be believed according to next round session input by userWhich branch node breath determination circulates from the first child node to.Branch node can circulate to another the first child node,It can circulate to the second child node.When branch node circulation is to another first child node, user's input can be guided moreSession information, therefore can be that branch node configures its next first child node that will be circulated, the first child node of settingContent, i.e. signal language, to continue to guide user's input next round session information again.When branch node circulation to the second child nodeWhen, configure first intention for the second child node.End node is finally configured, to terminate more wheel sessions.When upper the one of end nodeWhen a node is the first child node, trigger condition and circulation condition during the end node that circulated using start node are determined backMultiple source enables intelligent Answer System to generate the final answer for feeding back to user from reply source.When upper one of end nodeWhen node is the second child node, determines reply source using the node content of second child node, that is, first intention, make intelligenceEnergy question answering system can generate the final answer for feeding back to user from reply source.
User is directly configured and combines using these components according to the difference of application scenarios, you can configuration obtains oneThe flow of a complete more wheel sessions.More wheel session flows that configuration obtains are applied in intelligent Answer System, intelligent answerSystem can be based on this more wheel session flow, and by taking turns the form of interaction, guiding user inputs more information more, so that it is determined thatReply source.There are two types of the modes for determining reply source, and one is the information inputted in mostly wheel session using user to be replied to determineSource, another kind are that the information inputted in mostly wheel session using user determines first intention, recycle first intention to determine and replySource.After reply source determines, intelligent Answer System can generate final answer feedback to use from the specific reply sourceFamily.On the one hand, by introducing first intention so that the answer that intelligent Answer System ultimately generates is more abundant compared with originally;On the other hand, since it is determined the source of reply, intelligent Answer System are enable to respond quickly during conversating with user, togetherWhen provide more accurate answer, make dialogue more remarkable fluency.In addition, user can be flexible in the flow of the more wheel sessions of configurationGround determines the circulation direction between the knowledge node specifically used and node according to actual application scenarios, even if user has noAlgorithm rudimentary knowledge does not understand programming, being capable of flexibly and easily configuration session flow yet.
In above-described embodiment, the second child node is for obtaining first intention, and optionally, first intention here can classifyBe intended to for slot intention, FAQ etc..
Wherein, slot is intended to include at least one slot element.When the first intention that the second child node is got is intended to for slotWhen, second child node be additionally operable to obtain slot be intended to all slot elements, and, a slot be intended to all slot elements allCirculation is to next node in the case of the session information that is inputted by session subscriber is filled.
For example, a slot is intended to " train ticket inquiry ", configuration personnel can be that slot intention is matched by the second child nodeSet at least one slot element.For example, slot element may include " city " and " age " etc., slot element " city " refers to that user is currentThe city at place, slot element " age " refer to the age of active user.When the information that user inputs in a wheel session containsWhen its city being currently located, intelligent Answer System therefrom identifies on the fills to " city " this slot element;When withWhen the information that family inputs in another wheel session contains the age of user, intelligent Answer System therefrom identifies the fillsTo " age " this slot element.If all slot elements that " train ticket inquiry " this slot is intended to are all by session subscriber inputSession information is populated, then can circulate from second child node to next node, that is, end node.
In order to preferably guide user's input to be capable of the session information of effective filling slot element, the second child node is additionally operable to obtainIt is the session information for guiding user's input for filling slot element to take leading question corresponding with each slot element, the leading questionSentence.In example as escribed above, it is that " could you tell me that you currently exist that leading question corresponding with slot element " city ", which can be arranged,Which city?", intelligent Answer System exports the leading question to user, includes to be currently located city to guide user's inputSession information, for filling the slot element.Similarly, leading question corresponding with slot element " age ", which can also be arranged, is" age that could you tell me you ".
In addition, the second child node can be also used for limiting the type of next round session information input by user, here underOne wheel refers to the wheel session after intelligent Answer System exports some corresponding leading question of slot element to user.With the first sonAs node limits the class types of next round session information input by user, the type of session information here can also be dayPhase, age or other specific numerical value or characters.
Other than slot is intended to, first intention can also be the other kinds of intentions such as FAQ intentions.FAQ refers to commonThe problem of and correspondence problem answer, common problem describes in the form of being intended to, i.e. FAQ is intended to.In general, FAQ is intended to toolThere is corresponding answer, and be stored in some position, as in intelligent Answer System or remote server is first-class, to makeThe form that source is replied for one of which is used for configuration session flow.
In second embodiment of the application, referring to FIG. 9, provide a kind of utilize takes turns the more wheel friendships of session flows progress moreMutual method, wherein the progress of any one of foregoing first embodiment flow dispensing unit may be used in more wheel session flowsConfiguration obtains.The executive agent of more wheel exchange methods can be intelligent Answer System, such as intelligent robot etc. can also beOther are able to carry out interactive device or system of this method, such as high in the clouds artificial intelligence service system etc., and the application does not do thisIt limits.It should be noted that when the main body for executing the above method is intelligent robot, it is not limited solely to the intelligence of entityRobot can also be virtual intelligent robot, such as network customer service robot etc..
By executing the exchange method, user can be rebooted after the problem of obtaining user and inputs more information, fromAnd using user in mostly wheel sessions the information that inputs determine user be intended to whether with second in the flow session that has configuredFirst intention in child node is identical.Then it recycles first intention to determine the reply source of problem, is searched from reply source or rawAt final answer feedback to user so that interactive device or system are enable to respond quickly during conversating with user,More acurrate abundant answer is provided, and makes dialogue more remarkable fluency.
Specifically, which includes the steps that following S100-S500.
S100:Obtain the first round session information inputted by session subscriber.
S200:If first round session information meets the trigger condition on start node in more wheel session flows, exportThe first signal language in the first child node that start node flows directly to.
S300:It obtains by session subscriber according to the first signal language input the second wheel session information.
S400:If the second wheel session information meets the stream either one or two of being connect with the first child node on the first branch nodeTurn condition, it is determined that meet the next node that the first branch node of condition circulates.
S500:If the next node that the first branch node circulates is the second child node, it is determined that described secondChild node circulates to end node, and answer is found out from the reply source that the end node determines.
The session information of above-mentioned session subscriber input can be word, can also be the other forms such as audio, picture, thisApplication does not limit the concrete form of session information input by user.In addition, during interaction, different rounds is usedFamily can also input session information with identical or different input form, and the application also do not limit this.
It, can be to the first round before judging whether first round session information meets trigger condition in the S200 the step ofSession information is pre-processed.For example, intention assessment is carried out to it, then the intention that will identify that is compared with trigger condition.Further for example, if first round session information input by user is speech form, and trigger condition is stated with written form,Then session information input by user can be converted to word, then carry out intention assessment, then the intention that will identify that and triggeringCondition is compared;Or without intention assessment, directly the word being converted to is compared with trigger condition.
With the step of S200 similarly, the S400 the step of in, judging whether the second wheel session information meets circulation itemWhen part, the second wheel session information can also be pre-processed, details are not described herein again.
In addition, in above-mentioned scheme, can also include the following steps, S700:What if the first branch node circulatedNext node is end node, then finds out answer from the reply source that end node determines.
In one implementation, if first intention is intended to for slot, the second child node will not flow directly under itOne node, i.e. end node, but need slot be intended in all slot elements all have been filled with after can circulate toNext node.If certain slots during slot is intended to are intended to fill not yet, that is to say, that indeed according to what is gotInformation, interactive system can't determine that the intention of active user is exactly the slot intention in the second child node, at this time cannot be directIt circulates and determines reply source to end node.Specifically, Figure 10 is referred to, the step of S400 may include the step of S410 and S420Suddenly.
S410:If the next node that the first branch node circulates is the second child node, and second sonFirst intention on node is intended to for slot, then judges whether all slot elements that the slot is intended to have been filled.
S420:If all slot elements have been filled, it is determined that the end node that second child node circulates.
When judging whether the slot element in slot intention has been filled, if it find that also at least one slot intention does not haveHave and do not have populated, then user can also be guided to input more session informations using leading question corresponding with slot element, so as to energyThe information enough further inputted using user fills the slot element not being filled.Specifically, after the S410 the step ofCan also include:
S430:If at least one slot element is not filled, it is sequentially output and the slot element that is not filledCorresponding leading question;
S440:The session information that user inputs according to the leading question is obtained successively;
S450:Utilize session information filling slot element input by user.
In the step of above-mentioned S450, existing slot fill method may be used using session information filling slot element, thisPlace does not repeat.It is intended to for slot it should be noted that the method in this realization method can essentially be applied to all first intentionsThe second child node on, the second child node for circulating rather than just the first branch node.
Optionally, refer to Fig. 9, in above-mentioned any exchange method, the S400 the step of after, can also include:
S610:If the next node that the first branch node circulates is another first child node, institute is exportedState the second signal language in another first child node;
S620:Obtain the third round session information inputted according to the second signal language by session subscriber;
S630:If third round session information satisfaction either one or two of connect the second branch with another described first child nodeCirculation condition on node, it is determined that meet the next node that the second branch node of condition circulates;
S640:If the next node that the second branch node circulates is the second child node, it is determined that described secondThe end node that child node circulates finds out answer from the reply source that the end node determines.
In addition, in above-mentioned scheme, if the next node that the second branch node circulates is end node, fromAnswer is found out in the reply source that end node determines.In fact, before circulating to the second child node, start node and secondMay include the interaction of arbitrary round between child node, that is, the signal language in the first child node of output, to guide user to input meetingInformation is talked about, the next node of branch node and branch node is then determined using session information input by user.Until circulationWhen to some branch node, the next node of the branch node is the second child node or is directly end node.
Exchange method above-mentioned is further illustrated by taking more wheel session flows shown in Fig. 2 as an example, it is assumed that triggering item belowPart is second intention " how credit card is handled ", and the executive agent of exchange method is intelligent Answer System.
Intelligent Answer System gets the first round session information " I wants to handle credit card " inputted by session subscriber, passes throughIntention assessment identifies intention " how credit card is handled " from first round session information.Due in the intention and session flowTrigger condition (i.e. second intention " how credit card is handled ") on start node is identical, so triggering the session flow, connects downCarry out intelligent Answer System based on the session flow to be interacted with user.First, output start node flow directly to theSignal language on one child node a1, i.e., " it is much this year to may I ask you?".Then, it obtains and is saved according to the first son by session subscriberSignal language input the second wheel session information of point a1, it is assumed that user inputs " 20 years old ".First two first point of child node a1 connectionsZhi Jiedian b1 and b2, circulation condition are " > 18 years old " and " < 18 years old " respectively.Intelligent Answer System is by input by user " 20Year " is compared with two circulation conditions respectively, and as a result the second wheel session information meets the circulation condition of " > 18 years old ", so fromThe first branch node b1 of " > 18 years old " is arrived in first child node a1 circulations, and determines that first branch node b1 circulates downOne node, i.e. another first child node a2 " may I ask whether you have wages ".It is circulated by the first branch node b1Next node is another first child node, rather than the second child node, so export on another first child node a2Second signal language " may I ask whether you have wages?".It obtains and is believed according to the third round session that the second signal language inputs by session subscriberBreath, it is assumed that user inputs " no wages ".Another first child node a2 connections two second branch nodes b3 and b4, circulate itemPart is " having wages " and " no wages " respectively.Intelligent Answer System by " no wages " input by user respectively with two circulation conditionsIt is compared, as a result the second wheel session information meets the circulation condition of " no wages ", so flowed from another first child node a2It goes to the second branch node b4 of " no wages ", and determines the next node that second branch node b4 circulates, i.e., theTwo child node c1 " handled rule " more than 18 years old without wages credit card.The next section to be circulated by the second branch node b3Point is the second child node, so intelligent Answer System can determine the end node d1 that the second child node c1 circulates, finallyAnswer is found out from the reply source that the end node d1 is determined.
Accordingly with the exchange method in second embodiment, in the third embodiment of the application, Figure 11 is referred to,A kind of intelligent robot interactive system is provided, which can be applied to user take turns to interact more, the interaction systemSystem includes:More wheel session flows, the acquiring unit configured using any one of one embodiment flow dispensing unitAnd processing unit;
The acquiring unit 2 is used to obtain the first round session information inputted by session subscriber, and obtains and used by sessionFamily is according to the first signal language input the second wheel session information;
The processing unit 3 is used to meet the triggering item in more wheel session flows on start node in first round session informationIn the case of part, the first signal language in the first child node that start node flows directly to is exported;Session information is taken turns secondIn the case of meeting the circulation condition either one or two of being connect with the first child node on the first branch node, determination meets the of conditionThe next node that one branch node circulates;And the next node to circulate in the first branch node is secondIn the case of child node, determine that second child node circulates to end node, from the reply source that the end node determinesIn find out answer.
Optionally, the next node that the processing unit 3 is additionally operable to circulate in the first branch node is the second sonNode, and the first intention in second child node is to judge that all slots that the slot is intended to are wanted in the case that slot is intended toWhether element has been filled;In the case where all slot elements have been filled, determine what second child node circulatedEnd node.
Optionally, the processing unit 3 is additionally operable in the case where at least one slot element is not filled, successivelyOutput leading question corresponding with the slot element not being filled;And utilize session information filling slot element input by user.
The acquiring unit 2 is additionally operable to obtain the session information that user inputs according to the leading question successively.
Optionally, the next node that the processing unit 3 is additionally operable to circulate in the first branch node is anotherIn the case of first child node, the second signal language in another described first child node is exported;It is full in third round session informationIn the case of the sufficient circulation condition either one or two of being connect with another described first child node on the second branch node, determines and meetThe next node that second branch node of condition circulates;And the next section to circulate in the second branch nodeIt in the case that point is the second child node, determines that second child node circulates to end node, is determined from the end nodeReply source in find out answer.
The acquiring unit 2 is additionally operable to obtain the third round session information inputted according to the second signal language by session subscriber.
The same or similar parts between the embodiments can be referred to each other in this specification.Invention described above is realThe mode of applying is not intended to limit the scope of the present invention..

Claims (10)

CN201810570492.6A2018-06-052018-06-05Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessionsPendingCN108804643A (en)

Priority Applications (2)

Application NumberPriority DateFiling DateTitle
CN201810570492.6ACN108804643A (en)2018-06-052018-06-05Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions
CN201811497347.6ACN109614474A (en)2018-06-052018-12-07Multi-round conversation process configuration unit and method and intelligent robot interaction system

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN201810570492.6ACN108804643A (en)2018-06-052018-06-05Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions

Publications (1)

Publication NumberPublication Date
CN108804643Atrue CN108804643A (en)2018-11-13

Family

ID=64088683

Family Applications (2)

Application NumberTitlePriority DateFiling Date
CN201810570492.6APendingCN108804643A (en)2018-06-052018-06-05Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions
CN201811497347.6APendingCN109614474A (en)2018-06-052018-12-07Multi-round conversation process configuration unit and method and intelligent robot interaction system

Family Applications After (1)

Application NumberTitlePriority DateFiling Date
CN201811497347.6APendingCN109614474A (en)2018-06-052018-12-07Multi-round conversation process configuration unit and method and intelligent robot interaction system

Country Status (1)

CountryLink
CN (2)CN108804643A (en)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN109739605A (en)*2018-12-292019-05-10北京百度网讯科技有限公司 Method and apparatus for generating information
CN109829729A (en)*2019-01-182019-05-31青牛智胜(深圳)科技有限公司A kind of intelligence outgoing call system and method
CN109857910A (en)*2019-01-072019-06-07平安科技(深圳)有限公司Generation method, device, computer equipment and the storage medium of XML file
CN109992655A (en)*2019-03-292019-07-09深圳追一科技有限公司Intelligent customer service method, apparatus, equipment and storage medium
CN110377720A (en)*2019-07-262019-10-25中国工商银行股份有限公司The more wheel exchange methods of intelligence and system
CN110704594A (en)*2019-09-272020-01-17北京百度网讯科技有限公司Task type dialogue interaction processing method and device based on artificial intelligence
CN110955769A (en)*2019-12-172020-04-03联想(北京)有限公司Processing flow construction method and electronic equipment
CN111090733A (en)*2019-12-242020-05-01科大讯飞股份有限公司Human-computer interaction method, device, equipment and readable storage medium
WO2020103345A1 (en)*2018-11-192020-05-28众安信息技术服务有限公司Session flow processing method and apparatus, and computer-readable storage medium
CN111831799A (en)*2019-04-232020-10-27北京京东尚科信息技术有限公司 Information processing method, information processing device, and information processing system
CN112035623A (en)*2020-09-112020-12-04杭州海康威视数字技术股份有限公司Intelligent question and answer method and device, electronic equipment and storage medium
CN112084315A (en)*2020-09-072020-12-15腾讯科技(深圳)有限公司Question-answer interaction method, device, storage medium and equipment
CN112612462A (en)*2020-12-292021-04-06平安科技(深圳)有限公司Method and device for adjusting phone configuration, electronic equipment and storage medium
CN114328849A (en)*2020-09-302022-04-12北京彩智科技有限公司Method and device for multi-turn conversation and electronic equipment

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN110444292B (en)*2019-07-292022-04-08北京爱医生智慧医疗科技有限公司Information question-answering method and system
CN110737762A (en)*2019-10-092020-01-31尹曦old people personal information assistant system based on voice interaction
CN111984355A (en)*2020-08-202020-11-24第四范式(北京)技术有限公司Method and device for realizing man-machine multi-turn conversation
CN113268973B (en)*2021-06-012021-11-16北京太极华保科技股份有限公司Man-machine multi-turn conversation method and device

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
US20040044515A1 (en)*2002-08-302004-03-04Michael MetcalfAutomated natural language inference system
CN104408639A (en)*2014-10-222015-03-11百度在线网络技术(北京)有限公司Multi-round conversation interaction method and system
CN106776649B (en)*2015-11-242020-02-14中科国力(镇江)智能技术有限公司Intelligent question-answer multi-round interaction method and system based on visual flow chart
CN105589848A (en)*2015-12-282016-05-18百度在线网络技术(北京)有限公司Dialog management method and device
CN107665708B (en)*2016-07-292021-06-08科大讯飞股份有限公司Intelligent voice interaction method and system
CN107180080B (en)*2017-04-282018-10-16北京神州泰岳软件股份有限公司A kind of intelligent answer method and device of more interactive modes
CN108090177B (en)*2017-12-152020-05-05上海智臻智能网络科技股份有限公司Multi-round question-answering system generation method, equipment, medium and multi-round question-answering system

Cited By (19)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
WO2020103345A1 (en)*2018-11-192020-05-28众安信息技术服务有限公司Session flow processing method and apparatus, and computer-readable storage medium
CN109739605A (en)*2018-12-292019-05-10北京百度网讯科技有限公司 Method and apparatus for generating information
CN109857910A (en)*2019-01-072019-06-07平安科技(深圳)有限公司Generation method, device, computer equipment and the storage medium of XML file
CN109857910B (en)*2019-01-072024-03-26平安科技(深圳)有限公司XML file generation method and device, computer equipment and storage medium
CN109829729A (en)*2019-01-182019-05-31青牛智胜(深圳)科技有限公司A kind of intelligence outgoing call system and method
CN109992655A (en)*2019-03-292019-07-09深圳追一科技有限公司Intelligent customer service method, apparatus, equipment and storage medium
CN109992655B (en)*2019-03-292021-04-13深圳追一科技有限公司Intelligent customer service method, device, equipment and storage medium
CN111831799A (en)*2019-04-232020-10-27北京京东尚科信息技术有限公司 Information processing method, information processing device, and information processing system
WO2020215914A1 (en)*2019-04-232020-10-29北京京东尚科信息技术有限公司Information processing method, information processing device, and information processing system
CN110377720A (en)*2019-07-262019-10-25中国工商银行股份有限公司The more wheel exchange methods of intelligence and system
CN110704594A (en)*2019-09-272020-01-17北京百度网讯科技有限公司Task type dialogue interaction processing method and device based on artificial intelligence
CN110955769A (en)*2019-12-172020-04-03联想(北京)有限公司Processing flow construction method and electronic equipment
CN111090733A (en)*2019-12-242020-05-01科大讯飞股份有限公司Human-computer interaction method, device, equipment and readable storage medium
CN111090733B (en)*2019-12-242022-12-06科大讯飞股份有限公司Human-computer interaction method, device, equipment and readable storage medium
CN112084315A (en)*2020-09-072020-12-15腾讯科技(深圳)有限公司Question-answer interaction method, device, storage medium and equipment
CN112084315B (en)*2020-09-072024-06-11腾讯科技(深圳)有限公司Question-answer interaction method, device, storage medium and equipment
CN112035623A (en)*2020-09-112020-12-04杭州海康威视数字技术股份有限公司Intelligent question and answer method and device, electronic equipment and storage medium
CN114328849A (en)*2020-09-302022-04-12北京彩智科技有限公司Method and device for multi-turn conversation and electronic equipment
CN112612462A (en)*2020-12-292021-04-06平安科技(深圳)有限公司Method and device for adjusting phone configuration, electronic equipment and storage medium

Also Published As

Publication numberPublication date
CN109614474A (en)2019-04-12

Similar Documents

PublicationPublication DateTitle
CN108804643A (en)Flow dispensing unit, method and the intelligent robot interactive system of more wheel sessions
US10891438B2 (en)Deep learning techniques based multi-purpose conversational agents for processing natural language queries
US7454469B2 (en)Method and system for instant messaging Bots specification using state transition methodology and XML
CN110413758B (en)Session framework construction method and device based on machine learning
CN112084315B (en)Question-answer interaction method, device, storage medium and equipment
CN110708358B (en) Session message processing method, electronic device, and computer-readable storage medium
CN107977236B (en)Question-answering system generation method, terminal device, storage medium and question-answering system
CN109101545A (en)Natural language processing method, apparatus, equipment and medium based on human-computer interaction
CN110020426A (en)User&#39;s consulting is assigned to the method and device of customer service group
CN113132214B (en)Dialogue method, dialogue device, dialogue server and dialogue storage medium
CN109308357A (en) A method, apparatus and device for obtaining answer information
DE112012001696T5 (en) A method of providing self-support services using information from a viral source
DE112012001675T5 (en) Method for providing vendor-independent support services
DE112012001669T5 (en) A method for providing dynamic and proactive support services
CN110046230A (en)Generate the method for recommending words art set, the method and apparatus for recommending words art
DE112012001683T5 (en) One-touch platform for product registration and support
DE112012001672T5 (en) A method of providing support services over an available communication channel based on customer settings and client settings
CN110019731A (en)Intelligent interactive method, device, terminal device and storage medium
CN109492152A (en)Push method, apparatus, computer equipment and the storage medium of customized content
CN111090733A (en)Human-computer interaction method, device, equipment and readable storage medium
CN118535688A (en)Information processing method based on multi-agent system and multi-agent system
CN117593069A (en) An information generation method, device, computer equipment and medium
CN120162404A (en) A question-answering method, device, equipment and medium based on intelligent agent
CN118413505A (en)Session interaction method and device, electronic equipment and storage medium
CN112231027A (en)Task type session configuration method and system

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
WD01Invention patent application deemed withdrawn after publication
WD01Invention patent application deemed withdrawn after publication

Application publication date:20181113


[8]ページ先頭

©2009-2025 Movatter.jp