CN109815326B - Conversation control method and device - Google Patents
Conversation control method and device Download PDFInfo
- Publication number
- CN109815326B CN109815326B CN201910067893.4A CN201910067893A CN109815326B CN 109815326 B CN109815326 B CN 109815326B CN 201910067893 A CN201910067893 A CN 201910067893A CN 109815326 B CN109815326 B CN 109815326B
- Authority
- CN
- China
- Prior art keywords
- intention
- node
- tree
- attribute
- word
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a conversation control method and device. Wherein, the method comprises the following steps: acquiring an intention tree corresponding to the current conversation, wherein the intention tree indicates tasks required to be completed by the current conversation, and attribute nodes in the intention tree are associated with word slots; traversing the intention tree to obtain a traversal result; when the traversal result indicates that the intention tree has a vacant attribute node, inquiring the system action corresponding to the vacant attribute node, and sending the dialogue content corresponding to the system action; and after the traversal result indicates that all the attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree. The invention solves the technical problem that the effective filling of the associated word slot in the complex relation can not be realized in the task-type multi-turn dialogue in the correlation technique.
Description
Technical Field
The invention relates to the technical field of information processing, in particular to a conversation control method and device.
Background
Task-based multi-turn conversations generally refer to multiple turns of conversations needed by a user to complete a specific task (e.g., weather, flight) while chatting with a system, and these tasks are generally referred to as fields, each of which involves one or more field-related intents (e.g., the weather field may involve a query weather intention, a query air quality intention, etc.), and each of which is associated with zero or more slots (e.g., the query weather intention is associated with time, two slots in a city).
In the related art, when a task-based multi-turn conversation is processed, the conversation is completed by configuring basic information of an intention and a word slot associated with the intention in advance. FIG. 1 is a schematic diagram of an alternative dialog control according to the prior art, as shown in FIG. 1, the basic information of the intention, including the name of the intention and the alias of the intention, can be pre-configured, and the user can input the corresponding information through an input box; fig. 2 is a schematic diagram of another alternative conversation control according to the prior art, as shown in fig. 2, an intention associated word slot may be set, two word slots for inquiring the date and the city related to the weather are provided, each word slot must be filled completely to realize the weather inquiry, if a word slot is lacked, a feedback text corresponding to "clarification" may be sent, so that a user feeds back corresponding information, after the slot is filled, a corresponding action is triggered, a text may be replied, or a function (such as calling an external weather inquiry interface) is executed to give a corresponding conversation result to the user, such as feeding back the weather forecast of a certain city on a certain day.
When the task type multi-turn dialogue open platform configures the associated word slots of the intentions, the relation among the word slots is embodied by 'whether the word slots need to be filled', the method only relates to a small number of word slots for checking weather, and the task with simple relation among the word slots can be well processed. However, tasks including complex word slot relationships cannot be processed, for example, two word slots of "departure time" and "arrival time" in a flight reservation task only need to be satisfied, and a "departure city" and a "arrival city" must be completely filled, and complex relationships among multiple word slots in the task cannot be realized by using an existing associated word slot configuration method of an open platform, and other word slot management methods must be adopted.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the invention provides a conversation control method and device, which at least solve the technical problem that effective filling of associated word slots in complex relationships cannot be realized in task-type multi-turn conversations in the related technology.
According to an aspect of an embodiment of the present invention, there is provided a dialog control method including: acquiring an intention tree corresponding to a current conversation, wherein the intention tree indicates tasks required to be completed by the current conversation, and a word slot is associated with a property node in the intention tree; traversing the intention tree to obtain a traversal result; when the traversal result indicates that the intention tree has a vacant attribute node, inquiring a system action corresponding to the vacant attribute node, and sending conversation content corresponding to the system action; and after the traversal result indicates that all the attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree.
Further, obtaining an intent tree corresponding to the current dialog includes: constructing a plurality of domain forests corresponding to different domains, wherein each domain forest at least comprises an intention tree; and determining a domain forest corresponding to the current dialogue and an intention tree under the domain forest.
Further, the intent tree includes a plurality of nodes including at least one of: the intention node is used for indicating an intention target and corresponds to the task which needs to be completed by the current conversation; a relationship node for indicating a logical relationship between each intent node and attribute node in the intent tree; and the attribute nodes correspond to each word slot and are used for indicating the attributes of the word slots, wherein one word slot corresponds to one attribute node, and each attribute node corresponds to a system action.
Further, there is an intention node corresponding to each intention tree, and the intention node is an intention node identification; the relationship node corresponds to a relationship node identifier or an attribute node identifier.
Further, the attribute nodes are shared among the target intention trees in the domain forests.
Further, obtaining an intent tree corresponding to the current dialog includes: analyzing the current conversation to obtain an analysis result, wherein the analysis result at least comprises: domain forest, intent target, slot value; and inquiring an intention tree corresponding to the field forest according to the analysis result, corresponding an intention target to the intention node, and filling the slot value into the corresponding word slot and the attribute node.
Further, obtaining the intention tree corresponding to the current dialog further comprises: acquiring historical slot filling information, wherein the historical slot filling information comprises slot filling information of each word slot associated with the intention tree in a historical time period; and updating the attribute nodes in the intention tree and the slot values of the word slots associated with the intention tree according to the historical slot filling information.
Further, traversing the intention tree to obtain a traversal result comprises: initializing a multi-layer empty list; starting from the intention node, accessing child nodes of the intention node, and adding the accessed child nodes into the empty list to obtain an element list, wherein the child nodes comprise: a relationship node and/or an attribute node; judging whether the element list only contains attribute nodes; under the condition that the element list not only contains attribute nodes, traversing each element in the element list to obtain a traversal result, wherein the element comprises at least one of the following elements: and element, and or element, in the case of the element being an and element, removing the child node from the element list and expanding the element list by the child node; and under the condition that the element is the OR element, counting the number n of child nodes corresponding to the OR element, copying the element list n times, removing the element list from an empty list, replacing the OR node with different child nodes for each copied element list, and adding the n element lists into the empty list, wherein n is a natural number.
Further, querying the system action corresponding to the vacant attribute node comprises: when the traversal result indicates that the intention tree has vacant attribute nodes, determining various word slot combinations according to the traversal result, wherein the word slot combinations indicate the combination modes of vacant word slots needing to be filled; acquiring a word slot combination with the least number of vacant word slots contained in the word slot combination; and selecting a vacant word slot from the word slot combination with the least number of vacant word slots, and determining the system action corresponding to the vacant word slot.
Further, each intention tree corresponds to an intention action, the intention action is used for indicating a dialog text corresponding to the task, or the intention action is used for indicating that query information corresponding to the task is acquired after a slot value corresponding to a word slot is transferred to a calling preset interface, wherein the dialog result comprises the dialog text and the query information.
According to another aspect of the embodiments of the present invention, there is also provided a dialog control device, including: the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring an intention tree corresponding to the current conversation, the intention tree indicates tasks required to be completed by the current conversation, and attribute nodes in the intention tree are associated with word slots; the traversal unit is used for traversing the intention tree to obtain a traversal result; the query unit is used for querying a system action corresponding to the vacant attribute node and sending the dialogue content corresponding to the system action when the traversal result indicates that the vacant attribute node exists in the intention tree; and the sending unit is used for sending the dialogue result corresponding to the intention tree after the traversal result indicates that all the attribute nodes in the intention tree are filled.
Further, the acquisition unit includes: the system comprises a construction unit, a tree management unit and a tree management unit, wherein the construction unit is used for constructing a plurality of field forests corresponding to different fields, and each field forest at least comprises an intention tree; and the determining unit is used for determining a field forest corresponding to the current dialogue and an intention tree under the field forest.
Further, the intent tree includes a plurality of nodes including at least one of: the intention node is used for indicating an intention target and corresponds to the task which needs to be completed by the current conversation; a relationship node for indicating a logical relationship between each intent node and attribute node in the intent tree; and the attribute nodes correspond to each word slot and are used for indicating the attributes of the word slots, wherein one word slot corresponds to one attribute node, and each attribute node corresponds to a system action.
Further, there is an intention node corresponding to each intention tree, and the intention node is an intention node identification; the relationship node corresponds to a relationship node identifier or an attribute node identifier.
Further, the attribute nodes are shared among the target intention trees in the domain forests.
Further, the acquisition unit includes: the analysis module is used for analyzing the current conversation to obtain an analysis result, wherein the analysis result at least comprises: domain forest, intent target, slot value; and the first query module is used for querying an intention tree corresponding to the field forest according to the analysis result, corresponding an intention target to the intention node and filling the slot value into the corresponding word slot and the attribute node.
Further, the acquiring unit further includes: the first acquisition module is used for acquiring historical slot filling information, wherein the historical slot filling information comprises slot filling information of each word slot associated with the intention tree in a historical time period; and the first updating module is used for updating the attribute nodes in the intention tree and the slot values of the word slots related to the intention tree according to the historical slot filling information.
Further, the traversal unit includes: the initialization module is used for initializing a multi-layer empty list; an access module, configured to access, starting from an intention node, a child node of the intention node, and add the accessed child node to the empty list to obtain an element list, where the child node includes: a relationship node and/or an attribute node; the judging module is used for judging whether the element list only contains attribute nodes; a traversing module, configured to traverse each element in the element list to obtain a traversal result when the element list does not include only an attribute node, where the element includes at least one of: and element, and or element, in the case of the element being an and element, removing the child node from the element list and expanding the element list by the child node; and under the condition that the element is the OR element, counting the number n of child nodes corresponding to the OR element, copying the element list n times, removing the element list from an empty list, replacing the OR node with different child nodes for each copied element list, and adding the n element lists into the empty list, wherein n is a natural number.
Further, the query unit includes: the determining module is used for determining various word slot combinations according to the traversal result when the traversal result indicates that the intention tree has vacant attribute nodes, wherein the word slot combinations indicate the combination modes of vacant word slots needing to be filled; the second acquisition module is used for acquiring the word slot combination with the least number of the vacant word slots in the word slot combination; and the selection module is used for selecting a vacant word slot from the word slot combination with the least number of vacant word slots and determining the system action corresponding to the vacant word slot.
Further, each intention tree corresponds to an intention action, the intention action is used for indicating a dialog text corresponding to the task, or the intention action is used for indicating that query information corresponding to the task is acquired after a slot value corresponding to a word slot is transferred to a calling preset interface, wherein the dialog result comprises the dialog text and the query information.
According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform any of the dialog control methods described above via execution of the executable instructions.
According to another aspect of the embodiments of the present invention, there is also provided a storage medium, the storage medium including a stored program, wherein when the program runs, a device on which the storage medium is located is controlled to execute any one of the above-mentioned dialog control methods.
In the embodiment of the invention, an intention tree corresponding to a current conversation is obtained, wherein the intention tree indicates a task which needs to be completed by the current conversation, a word slot is associated with an attribute node in the intention tree, the intention tree is traversed to obtain a traversal result, when the traversal result indicates that a vacant attribute node exists in the intention tree, a system action corresponding to the vacant attribute node is inquired, a conversation content corresponding to the system action is sent, and after the traversal result indicates that all attribute nodes in the intention tree are filled completely, a conversation result corresponding to the intention tree is sent. In the embodiment, multiple rounds of dialogs can be completed by using the intention tree, complex relationships can be dealt with, vacant attribute nodes can be filled, word slots are filled, associated word slots of complex relationships can be filled quickly and efficiently, and the technical problem that effective filling of associated word slots of complex relationships cannot be achieved in task-based multiple rounds of dialogs in the related technology is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a schematic illustration of an alternative dialog control according to the prior art;
FIG. 2 is a schematic diagram of another alternative dialog control according to the prior art;
FIG. 3 is a flow chart of a dialog control method according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an alternative intention tree, according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of another alternative intention tree in accordance with an embodiment of the present invention;
FIG. 6 is a schematic diagram of an alternative dialog control in accordance with an embodiment of the present invention;
FIG. 7 is a schematic diagram of a dialog control device according to an embodiment of the present invention;
FIG. 8 is a schematic diagram of an alternative electronic device according to an embodiment of the invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
To facilitate the understanding of the present invention, some terms or nouns related to the embodiments of the present invention are explained below:
NLU: neural Language Understanding of natural Language.
DST: dialog State Tracking.
DPL: dialog Policy Learning.
NLG: neural Language Generation natural Language Generation.
The method and the system can be applied to various task-based multi-turn conversation environments, and conversation between the user and the system is realized by completing a certain task (such as a weather-related task, a flight-related task and the like). Compared with the technical problem that word slot filling under a complex relation cannot be realized in the related technology, the word slot filling method and device can be used for solving the technical problem.
The embodiment can be applied to various software or hardware dialogues, such as a customer service system, an intelligent sound box and the like.
The present embodiment may relate to a task-based multi-turn dialog, which generally relates to technologies such as NLU, DST, DPL, NLG, etc., where NLU parses a user's speech into structured data, DST is used to track states in a user and system dialog process, such as intentions of the user at the current time and at a historical time, slot information provided by the user at the current time and at the historical time, and DPL executes a corresponding system action according to the states tracked by DST. The NLG further generates natural language feedback to the user based on this action, such as directly generating a piece of text or returning data of the invoked external interface to the user. These techniques are also referred to in the various implementation steps described below in this application.
In accordance with an embodiment of the present invention, there is provided a dialog control method embodiment, it being noted that the steps illustrated in the flowchart of the figure may be performed in a computer system such as a set of computer-executable instructions and that, although a logical ordering is illustrated in the flowchart, in some cases, the steps illustrated or described may be performed in an order different than here.
Fig. 3 is a flowchart of a dialog control method according to an embodiment of the present invention, as shown in fig. 3, the method including the steps of:
step S302, acquiring an intention tree corresponding to the current conversation, wherein the intention tree indicates tasks required to be completed by the current conversation, and a word slot is associated with a property node in the intention tree;
step S304, traversing the intention tree to obtain a traversal result;
step S306, when the traversal result indicates that the intention tree has a vacant attribute node, inquiring the system action corresponding to the vacant attribute node, and sending the dialogue content corresponding to the system action;
and step S308, after the traversal result indicates that all the attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree.
Through the steps, the intention tree corresponding to the current conversation can be obtained, wherein the intention tree indicates tasks required to be completed by the current conversation, the attribute nodes in the intention tree are associated with word slots, the intention tree is traversed to obtain a traversal result, when the traversal result indicates that the intention tree has the vacant attribute nodes, the system action corresponding to the vacant attribute nodes is inquired, the conversation content corresponding to the system action is sent, and after the traversal result indicates that all the attribute nodes in the intention tree are filled completely, the conversation result corresponding to the intention tree is sent. In the embodiment, multiple rounds of dialogs can be completed by using the intention tree, complex relationships can be dealt with, vacant attribute nodes can be filled, word slots are filled, associated word slots of complex relationships can be filled quickly and efficiently, and the technical problem that effective filling of associated word slots of complex relationships cannot be achieved in task-based multiple rounds of dialogs in the related technology is solved.
In the embodiment of the application, a field forest is constructed in advance for task-type multi-turn conversations, namely when an intention tree corresponding to a current conversation is obtained, a plurality of field forests corresponding to different fields are constructed, wherein each field forest at least comprises an intention tree; a domain forest corresponding to the current dialog and an intent tree under the domain forest are then determined. That is, the application may pre-construct a domain forest corresponding to each domain, and if a user initiates a current dialog, the domain forest corresponding to the current dialog may be determined first, and then an intention tree corresponding to the current dialog may be determined.
The term forest refers to a field forest established according to relationships between term slots and term slots that need to be filled to complete a conversation in different fields, for example, a forest having a "music field" corresponding to a music field, a forest having a "flight field" corresponding to a flight field, and the like. A domain forest includes at least one intention tree, such as "music domain" may include "music play", "music query", and the like, music-related intention trees.
The above steps will be explained below.
Step S302, an intention tree corresponding to the current conversation is obtained, wherein the intention tree indicates tasks required to be completed by the current conversation, and a word slot is associated with a property node in the intention tree.
The current dialog may be a dialog text obtained by a user through an input box or voice analysis, for example, the user asks "what is the weather forecast in tomorrow", and after the current dialog is analyzed, a corresponding field forest and an intention tree may be obtained.
Optionally, the intention tree includes a plurality of nodes, and in this application, the intention tree mainly includes three types of nodes, where each node includes a node identifier (e.g., a node number, which is used to uniquely identify a node and is generally described by a positive integer), a node name (a nickname of the node), a node major class (corresponding to three types of nodes, where the intention node major class may be Root, the relation node major class may be Logic, and the attribute node major class may be Slot), a node type, a node value, a node constraint condition, a child node number, and other attributes.
Wherein the plurality of nodes include at least one of:
1. the intention node is used for indicating an intention target and corresponds to a task which needs to be completed by the current conversation; the intention target refers to what the task is, such as flight inquiry, flight reservation, weather inquiry and the like, the nodes are only one and one in an intention tree, the node types, the node values and the node constraints are not required to be defined, the child node numbers are only one and the relationship node numbers are required. Optionally, the intent node is the intent node identification, as is the graph node numbering.
2. A relationship node for indicating a logical relationship between each of the intent nodes and the attribute nodes in the intent tree; the attribute nodes and the corresponding word slots are managed, that is, the relationship nodes can represent the logical relationship between the slots related to the target intention, such as the relationship of and, or, and the like, the node values and node constraint conditions of the relationship nodes do not need to be defined, the node type is one of the logical relationship of and, or, and the child node number is the relationship node number or the attribute node number and at least contains one child node. Optionally, the relationship node corresponds to a relationship node identifier or an attribute node identifier, such as a relationship node number and an attribute node number in a child node.
3. And the attribute nodes correspond to each word slot and are used for indicating the attributes of the word slots, wherein one word slot corresponds to one attribute node, and each attribute node corresponds to a system action. The attribute node represents a specific attribute of a word slot of the target intention, attribute nodes corresponding to two word slots, such as "time" and "city" related to the weather query intention are time and city, the node type of the attribute node is a customizable type, such as "character string", "integer", "floating point", "date", "time", and the like, and is used for describing the type of a node value, the node value represents a value actually stored by the node, such as "city", the attribute node can store a value of "Hangzhou state", the type of the value is a type of "character string", the node constraint condition is used for describing the range of a node value, such as the attribute node of the "date" type can limit the range of the date, the attribute node is necessarily a leaf node of the intention tree, and therefore, the child node number of the node is empty.
Fig. 4 is a schematic diagram of an optional intention tree according to an embodiment of the present invention, as shown in fig. 4, two intention trees are shown in a forest in the field, where a first intention tree corresponds to an intention node 1, a relationship node associated with the intention node 1 is a relationship node 1, a node associated with the relationship node 1 has a relationship node 2 and an attribute node 3, and a node associated with the relationship node 2 has an attribute node 1 and an attribute node 2, where the attribute node 1 corresponds to a system action 1 and the attribute node 2 corresponds to a system action 2; and the second intention tree corresponds to an intention node 2, the relationship node associated with the intention node 2 is a relationship node 3, and the nodes associated with the relationship node 3 are an attribute node 3 and an attribute node 4, wherein the attribute node 3 corresponds to a system action 3, and the attribute node corresponds to a system action 4.
The two intent trees shown in FIG. 4 correspond to intent action 1 and intent action 2, respectively.
In the embodiment of the application, the attribute nodes are shared among the target intention trees in the forest in the field, so that more convenient interactive experience can be provided for users in the same field. As shown in fig. 4, two intention trees, both of which may share the attribute node 3, i.e., the content indicated by the attribute node 3, may be used.
Each attribute node in the above-mentioned intent tree has a node value, a node type, and a node number, where the node value may be adapted to the value filled in the word slot, and may be the same, for example, a city name or time is filled in the word slot, and a corresponding city name or time is also filled in the node value of the corresponding attribute node.
Each attribute node in the intention tree corresponds to a system action, the system action is that the system acquires a node value of the attribute node, the system can be instructed to feed back information to the user terminal according to the filling condition of the attribute node according to the system action so as to acquire the attribute value (corresponding to the slot value of the word slot) corresponding to the attribute node, the feedback is carried out in the form of text, one or more texts corresponding to the system action can be set, and when the action is triggered, one text is randomly returned from the texts to serve as a reply. For example, in looking up a weather intention including two attributes of time and city, the text of the system action of the "time" attribute may be "ask you want to look up which day's weather? "," find what day? And one text corresponding to the preset system actions is randomly selected and returned to the user terminal to obtain the attribute value.
In another optional example of the present application, each intention tree corresponds to an intention action, and the intention action is used to indicate a dialog text corresponding to a task, or the intention action is used to indicate that query information corresponding to the task is obtained after a slot value corresponding to a word slot is passed to a preset invoking interface, where a dialog result includes the dialog text and the query information.
The above-mentioned intent action, which means that when the trigger condition of the intent target is satisfied, the action corresponding to the intent target is executed, a piece of text (similar to the system action) may be fed back or an external interface may be called to assist the execution of the action. The external interface can be any accessible service, for example, when the values of two attribute nodes of time and city are filled completely, the intention action of inquiring the weather interface is triggered, the values of two slots of time and city are used as parameters to be transmitted to the weather interface, weather information is returned, and finally the weather information is fed back to the user.
The intended actions in this application are divided into two categories: first, a text is directly fed back, which is the same as the system action; secondly, an external interface is called, and the interface can be built by a user or can be ready made on a network; when the intended slot values are all filled, such as two slots looking up weather intent { "time": "today", "city": the 'Hangzhou' is filled with values, so that the parameters of the slot can be directly transmitted to the interface to return to the weather of the Hangzhou today and then return to the user terminal.
In the method, one field forest is configured, the field forests are used for managing a plurality of intention trees in the same field, word slot information sharing among different intentions in the same field is realized, the word slots are managed in the form of the intention trees, the expression of complex word slot relation is realized, the defect of the conventional on-line associated word slot configuration is overcome, and the possibility of complex task configuration is improved.
For the above step S302, acquiring the intention tree corresponding to the current dialog includes: analyzing the current conversation to obtain an analysis result, wherein the analysis result at least comprises: domain forest, intent target, slot value; and inquiring an intention tree corresponding to the field forest according to the analysis result, corresponding an intention target to an intention node, and filling the slot value into the corresponding word slot and the attribute node.
After the current dialogue is analyzed, what the user wants to inquire, the domain forest and the intention target to which the user belongs can be determined, the determined groove value is determined, then the intention tree corresponding to the current domain can be determined, the corresponding intention tree in the domain forest is called, the attribute node needing to be filled in the intention tree and the associated word groove are determined, and the determined groove value is filled in the corresponding word groove and the attribute value of the attribute node.
As another optional example of the present application, obtaining the intention tree corresponding to the current dialog further comprises: acquiring historical slot filling information, wherein the historical slot filling information comprises slot filling information of each word slot associated with the intention tree in a historical time period; and updating the attribute nodes in the intention tree and the slot values of the word slots associated with the intention tree according to the historical slot filling information.
Namely, the attribute values of the attribute nodes of the current intention tree and the slot values of the word slots associated with the intention tree can be updated according to the slot filling information acquired in the history process. For example, a mode of "intention inheritance" or "slot value inheritance" can be adopted to realize more complete context management, so as to solve the problem that the confidence of the intention in the NLU parsing result is too low, at this time, the intention of the previous round of the user can be inherited to the current round of the dialog through the intention inheritance, and the intention of "looking for the weather" of the previous round of the dialog can be inherited through the "intention inheritance" or "slot value inheritance", so that the intention tree of "looking for the weather" can be continuously accessed, and the completion of the task corresponding to the current dialog is promoted.
And step S304, traversing the intention tree to obtain a traversal result.
When traversing the intention tree, the method may traverse the slot values placed in the word slots, place the slot values in the attribute of the "node value" of the corresponding attribute node, and then decide according to the attribute value filling state of the attribute node of the intention tree.
Optionally, traversing the intention tree to obtain a traversal result includes: initializing a multi-layer empty list; starting from the intention node, accessing child nodes of the intention node, and adding the accessed child nodes into an empty list to obtain an element list, wherein the child nodes comprise: a relationship node and/or an attribute node; judging whether the element list only contains attribute nodes; under the condition that the element list not only contains attribute nodes, traversing each element in the element list to obtain a traversal result, wherein the element comprises at least one of the following elements: and element, and or element, in the case of element and element, remove the child node from the list of elements, and expand the list of elements through the child node; under the condition that an element is an OR element, counting the number n of child nodes corresponding to the OR element, copying the element list n times, removing the element list from an empty list, replacing the OR node with different child nodes for each copied element list, and adding the n element lists into the empty list, wherein n is a natural number.
The multi-layer empty list in the present application may refer to an empty list of two layers. During the process, the method comprises the following four steps, wherein in the first step, a two-layer empty list L is initialized: [ ] ], can then start from the intent node, access its child node, and add the child node to the empty list. As can be seen from FIG. 4, the child nodes obtained after traversal from the intent node 1 of the first intent tree include: the relationship node 1, whose child nodes can be obtained after traversal from the intention node 2 of the second intention tree, includes: a relationship node 3; thus, the steps are sequentially performed. After traversal, the relationships between the traversed nodes and the corresponding nodes may be stored in a list, resulting in an element list.
And secondly, checking whether each element list in the L only contains attribute nodes, if so, jumping to the fourth step, and if not, entering the third step.
Thirdly, traversing each element list in the L, firstly traversing each element (which is a relationship node or an attribute node) of the element list, 1, if the element is an AND element, firstly removing child nodes from the list, and then expanding the list by using the child nodes; 2, if the element is an "OR" element, counting the number n of child nodes of the "OR" node, copying the element list n times, removing the element list from the L, replacing the "OR" element with different child nodes for each copied element list, and finally adding the n list elements to the L. Second, the list L is queried in a loop.
And fourthly, exiting. After the traversal is finished, a combination mode of a plurality of word slot lists can be obtained.
After traversing the intention tree, two results are available, wherein the first result is that all attribute nodes of the intention tree and word slots associated with the intention tree are completely filled, and at the moment, only a preset external interface needs to be called or a pre-configured dialogue result associated with the intention tree needs to be returned; second, the property nodes or word slots associated with the intent tree are not filled, and the necessary word slots and property nodes are still missing, requiring the system to return dialog content to obtain the corresponding property values or slot values.
Step S308 is executed corresponding to the first case, and step S306 is executed corresponding to the second case.
And S306, inquiring system action corresponding to the vacant attribute node when the traversal result indicates that the vacant attribute node exists in the intention tree, and sending the dialogue content corresponding to the system action.
That is, when the traversal result indicates that there is a vacant attribute node in the intention tree, the system action needs to be queried, and optionally, querying the system action corresponding to the vacant attribute node includes: when the traversal result indicates that the intention tree has vacant attribute nodes, determining various word slot combinations according to the traversal result, wherein the word slot combinations indicate the combination mode of vacant word slots needing to be filled; acquiring a word slot combination with the least number of vacant word slots contained in the word slot combination; and selecting a vacant word slot from the word slot combination with the least number of vacant word slots, and determining the system action corresponding to the vacant word slot.
During the above-mentioned process, the unfilled word slot combination and attribute nodes are queried, and the combination mode with the least number of unfilled slots is selected as the word slot combination which needs to be finally inquired about the user, and when the combination with the least number of unfilled slots has multiple modes, one mode can be randomly selected. And finally, feeding back the user aiming at the unfilled word slots in the selected word slot combination, namely feeding back the system action corresponding to the word slot if the word slot is unfilled, and when the number of the unfilled word slots is zero, indicating that all necessary conditions of the intention tree are met, and triggering the intention action to feed back the user.
And step S308, after the traversal result indicates that all the attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree.
And the dialogue result indicates to call an external interface or call a text which is stored in the system in advance and corresponds to the intention tree, and corresponding information is returned to the user terminal. For example, if the user needs to inquire the weather of a certain day, after the word slots of the determined time and the city are completely filled, the weather inquiry interface can be called to call the specific content of the weather, and the content is returned to the user terminal.
By the method, a plurality of intention trees in the same field can be managed by using a field forest, word slot information sharing among different intention targets in the same field is realized, word slots are managed by using the intention trees, expression of complex word slot relation is realized, the defect of existing on-line associated word slot configuration is overcome, and the possibility of complex task configuration is improved.
The application is illustrated below by an example of an alternative dialog control.
Fig. 5 is a schematic diagram of another alternative intention tree according to an embodiment of the invention, as shown in fig. 5, the field forest of "flight field" is built, the field forest may be the flight field, and there is an intention tree in the flight field, namely, the "flight reservation" intention tree.
The "flight booking" intention tree shown in fig. 5 contains 8 attribute nodes in total, the system actions corresponding to the "slot" are listed in the tree, and the system actions corresponding to other attribute nodes are similar and are not listed in the application.
Optionally, the intention node in the intention tree in fig. 5 is "flight booking", the intention action corresponding to the intention node is "call booking flight API", and the intention action may be called after the node value of the attribute node is filled as necessary.
Fig. 6 is a schematic diagram of another alternative dialog control according to an embodiment of the present invention, as shown in fig. 6, when a user issues a first dialog text "help me start from hangzhou to beijing airplane on tomorrow", and after being parsed by the NLU module, structured data is obtained, as in this example: { "Domain": "flight," "intent target": "flight booking", "property node": [ { "date": "tomorrow" }, { "departure city": "Hangzhou" }, { "reach city": "Beijing" }, the above information can be obtained by the current natural language parsing mode.
Optionally, the field forest receives the analysis result, and filters the analysis result, specifically: firstly, visiting a 'flight booking' intention tree in the 'flight' field, obtaining attribute nodes of the intention tree, then based on the analysis result, reserving the values obtained by analysis to prepare for subsequent attribute nodes and word slot filling, and marking the finally obtained slot value information as ti(i denotes the ith round of dialog).
Receiving slot value information t in analysis resultiCombining historical slot filling information Ti-1(T includes all slot value information under the current intention tree of the current field), and updating the total slot value information Ti＝f(Ti-1，ti) Wherein f represents a slot value update policy function, the most straightforward way is to update tiAll slot values in (1) are covered to Ti-1In, i.e. to Ti-1If each pair of slot values inThe groove being at tiIs over and the slot value and Ti-1If they are different, then T isi-1Replacing the value in (1) with tiValue of (1), otherwise, T is retainedi-1The slot value of (a) is unchanged. The result of DST in this example can be expressed as: { "Domain": "flight," "intent": "flight booking", "property node": [ { "date": "tomorrow" }, { "departure city": "Hangzhou" }, { "reach city": "Beijing" }, { "bay": 'air']And { "departure time": "empty" }, { "time to reach": "empty" }, { "airline": "empty" }, { "flight": "empty" }]}。
Traversing the intention tree of flight booking and the word slot associated with the intention tree by means of breadth-first search to obtain 4 word slot combination modes, including: first, the [ "cabin space", "date", "departure city", "arrival city", "departure time", "airline" ], second, the [ "cabin space", "date", "departure city", "arrival time", "airline" ], third, the [ "cabin space", "date", "departure city", "arrival city", "departure time", "flight" ], fourth, the [ "cabin space", "date", "departure city", "arrival city", "departure time", "flight" ].
And (3) counting the number of unfilled slots in each word slot combination mode by combining the DST results as follows: [3,3,3,3]. Because the number is the same, a word slot combination mode can be randomly selected as a final slot filling mode, for example, a first mode is selected: the optimal slot filling combination obtained by [ "cabin space", "date", "departure city", "arrival city", "departure time" and "airline" ] is combined with the result of the DST to obtain the word slot list of the not filled slot: [ "cabin space", "departure time", "airline" ].
In the word slot list determined to be filled, three word slot lists without filling slots are also provided: [ "cabin space", "departure time", "airline" ], and then a system action corresponding to a word slot can be randomly selected from the system actions as feedback, for example, a system action corresponding to the "cabin space" is selected, so that a user receives a feedback message of the system as "ask for which kind of cabin space you want to subscribe? ".
That is, after parsing the current conversation, a conversation content "ask for what kind of slots you want to subscribe? ", to obtain the node value or slot value of the corresponding bay.
Through the implementation mode, a dialogue can be realized by using an intention tree of ' flight reservation ' in the field of flights, word slots are managed in the form of the intention tree, the representation of the relation of complex word slots is realized, after the current dialogue is analyzed, texts corresponding to corresponding system actions are fed back to unfilled word slots, the texts are fed back to a user, the attribute nodes of the word slots and the intention tree can be filled through three queries, so that the intention actions are called, an ' API ' interface for calling the reserved flights ' is realized, the user is helped to realize the reservation of the flights, and the task of the flight reservation is completed.
The application is illustrated below by means of a further alternative embodiment.
Fig. 7 is a schematic diagram of a dialog control device according to an embodiment of the present invention, which, as shown in fig. 7, may include: an obtaining unit 71, a traversing unit 73, a querying unit 75, a sending unit 77, wherein,
an obtaining unit 71, configured to obtain an intention tree corresponding to a current dialog, where the intention tree indicates a task that needs to be completed by the current dialog, and a word slot is associated with a property node in the intention tree;
the traversal unit 73 is used for traversing the intention tree to obtain a traversal result;
the query unit 75 is configured to query a system action corresponding to a vacant attribute node when the traversal result indicates that the vacant attribute node exists in the intention tree, and send a dialog content corresponding to the system action;
a sending unit 77, configured to send the dialog result corresponding to the intention tree after the traversal result indicates that all the attribute nodes in the intention tree are filled.
The dialog control device may acquire, by the acquisition unit 71, an intention tree corresponding to a current dialog, where the intention tree indicates a task that needs to be completed by the current dialog, the attribute nodes in the intention tree are associated with word slots, the traversal unit 73 traverses the intention tree to obtain a traversal result, the query unit 75 queries a system action corresponding to a vacant attribute node when the traversal result indicates that the vacant attribute node exists in the intention tree, and transmits a dialog content corresponding to the system action, and the transmission unit 77 transmits a dialog result corresponding to the intention tree after the traversal result indicates that all the attribute nodes in the intention tree are filled. In the embodiment, multiple rounds of dialogs can be completed by using the intention tree, complex relationships can be dealt with, vacant attribute nodes can be filled, word slots are filled, associated word slots of complex relationships can be filled quickly and efficiently, and the technical problem that effective filling of associated word slots of complex relationships cannot be achieved in task-based multiple rounds of dialogs in the related technology is solved.
Optionally, the obtaining unit includes: the system comprises a construction unit, a tree management unit and a tree management unit, wherein the construction unit is used for constructing a plurality of field forests corresponding to different fields, and each field forest at least comprises an intention tree; and the determining unit is used for determining a field forest corresponding to the current dialogue and an intention tree under the field forest.
In this application, the intention tree includes a plurality of nodes, the plurality of nodes including at least one of: the intention node is used for indicating an intention target and corresponds to a task which needs to be completed by the current conversation; a relationship node for indicating a logical relationship between each of the intent nodes and the attribute nodes in the intent tree; and the attribute nodes correspond to each word slot and are used for indicating the attributes of the word slots, wherein one word slot corresponds to one attribute node, and each attribute node corresponds to a system action.
As an optional example of the present application, there is one intention node corresponding to each intention tree, and the intention node is an intention graph node identifier; the relationship node has a relationship node identifier or an attribute node identifier.
As an alternative example of the present application, attribute nodes are shared between target intent trees in a domain forest.
As an optional example of the present application, the obtaining unit includes: the analysis module is used for analyzing the current conversation to obtain an analysis result, wherein the analysis result at least comprises: domain forest, intent target, slot value; and the first query module is used for querying an intention tree corresponding to the field forest according to the analysis result, corresponding the intention target to the intention node and filling the slot value into the corresponding word slot and the attribute node.
As an optional example of the present application, the obtaining unit further includes: the first acquisition module is used for acquiring historical slot filling information, wherein the historical slot filling information comprises slot filling information of each word slot associated with the intention tree in a historical time period; and the first updating module is used for updating the attribute nodes in the intention tree and the slot values of the word slots related to the intention tree according to the historical slot filling information.
As an alternative example of the present application, the traversal unit includes: the initialization module is used for initializing a multi-layer empty list; an access module, configured to access, starting from an intention node, a child node of the intention node, and add the accessed child node to an empty list to obtain an element list, where the child node includes: a relationship node and/or an attribute node; the judging module is used for judging whether the element list only contains attribute nodes; a traversing module, configured to traverse each element in the element list to obtain a traversal result when the element list does not include only the attribute node, where the element includes at least one of the following: and element, and or element, in the case of element and element, remove the child node from the list of elements, and expand the list of elements through the child node; under the condition that an element is an OR element, counting the number n of child nodes corresponding to the OR element, copying the element list n times, removing the element list from an empty list, replacing the OR node with different child nodes for each copied element list, and adding the n element lists into the empty list, wherein n is a natural number.
As an optional example of the present application, the query unit includes: the determining module is used for determining various word groove combinations according to the traversal result when the traversal result indicates that the intention tree has vacant attribute nodes, wherein the word groove combinations indicate the combination mode of vacant word grooves needing to be filled; the second acquisition module is used for acquiring the word slot combination with the least number of the vacant word slots in the word slot combination; and the selection module is used for selecting a vacant word slot from the word slot combination with the least number of vacant word slots and determining the system action corresponding to the vacant word slot.
As an optional example of the application, each intention tree corresponds to an intention action, and the intention action is used to indicate a dialog text corresponding to a task, or the intention action is used to indicate that query information corresponding to the task is acquired after a slot value corresponding to a word slot is passed to a preset invoking interface, where a dialog result includes the dialog text and the query information.
The above-mentioned dialog control device may further include a processor and a memory, the above-mentioned acquiring unit 71, the traversing unit 73, the querying unit 75, the sending unit 77, and the like are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions.
The processor comprises a kernel, and the kernel calls a corresponding program unit from the memory. The kernel can set one or more parameters, and sends the dialogue result corresponding to the intention tree after the traversal result indicates that all the attribute nodes in the intention tree are filled completely by adjusting the kernel parameters.
The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.
According to another aspect of the embodiments of the present invention, there is also provided an electronic device, and fig. 8 is a schematic diagram of an alternative electronic device according to the embodiments of the present invention, as shown in fig. 8, the electronic device 810 includes a memory 820 and a processor 830, where the memory is used for storing executable instructions of the processor, and the processor is configured to execute a program of the following steps by executing the executable instructions: acquiring an intention tree corresponding to the current conversation, wherein the intention tree indicates tasks required to be completed by the current conversation, and attribute nodes in the intention tree are associated with word slots; traversing the intention tree to obtain a traversal result; when the traversal result indicates that the intention tree has a vacant attribute node, inquiring the system action corresponding to the vacant attribute node, and sending the dialogue content corresponding to the system action; and after the traversal result indicates that all the attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree.
According to another aspect of the embodiments of the present invention, there is also provided a storage medium including a stored program, wherein when the program runs, a device on which the storage medium is located is controlled to execute any one of the above-mentioned dialog control methods.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (11)
1. A dialog control method, applied to a context of a task-based multi-turn dialog, comprising:
acquiring an intention tree corresponding to a current conversation, wherein the intention tree indicates tasks required to be completed by the current conversation, and a word slot is associated with a property node in the intention tree;
traversing the intention tree to obtain a traversal result;
when the traversal result indicates that the intention tree has a vacant attribute node, querying a system action corresponding to the vacant attribute node, and sending dialog content corresponding to the system action, wherein querying the system action corresponding to the vacant attribute node comprises: when the traversal result indicates that the intention tree has vacant attribute nodes, determining various word slot combinations according to the traversal result, wherein the word slot combinations indicate the combination mode of vacant word slots needing to be filled; acquiring a word slot combination with the least number of vacant word slots contained in the word slot combination; selecting a vacant word slot from the word slot combination with the least number of vacant word slots, and determining the system action corresponding to the vacant word slot;
and after the traversal result indicates that all attribute nodes in the intention tree are filled, sending a dialogue result corresponding to the intention tree, wherein each intention tree corresponds to an intention action, the intention action is used for indicating a dialogue text corresponding to the task, or the intention action is used for indicating that after a groove value corresponding to a word groove is transferred to a calling preset interface, inquiry information corresponding to the task is obtained, and the dialogue result comprises the dialogue text and the inquiry information.
2. The method of claim 1, wherein obtaining the intent tree corresponding to the current dialog comprises:
constructing a plurality of domain forests corresponding to different domains, wherein each domain forest at least comprises an intention tree;
and determining a domain forest corresponding to the current dialogue and an intention tree under the domain forest.
3. The method of claim 2, wherein the intent tree comprises a plurality of nodes, the plurality of nodes comprising at least one of:
the intention node is used for indicating an intention target and corresponds to the task which needs to be completed by the current conversation;
a relationship node for indicating a logical relationship between each intent node and attribute node in the intent tree;
and the attribute nodes correspond to each word slot and are used for indicating the attributes of the word slots, wherein one word slot corresponds to one attribute node, and each attribute node corresponds to a system action.
4. The method of claim 3,
each intention tree corresponds to an intention node, and the intention node is identified by an intention node;
the relationship node corresponds to a relationship node identifier or an attribute node identifier.
5. The method of claim 3, wherein the attribute nodes are shared among target intent trees in the domain forest.
6. The method of claim 3, wherein obtaining the intent tree corresponding to the current dialog comprises:
analyzing the current conversation to obtain an analysis result, wherein the analysis result at least comprises: domain forest, intent target, slot value;
and inquiring an intention tree corresponding to the field forest according to the analysis result, corresponding an intention target to the intention node, and filling the slot value into the corresponding word slot and the attribute node.
7. The method of claim 6, wherein obtaining the intent tree corresponding to the current dialog further comprises:
acquiring historical slot filling information, wherein the historical slot filling information comprises slot filling information of each word slot associated with the intention tree in a historical time period;
and updating the attribute nodes in the intention tree and the slot values of the word slots associated with the intention tree according to the historical slot filling information.
8. The method of claim 6, wherein traversing the intent tree, resulting in a traversal result, comprises:
initializing a multi-layer empty list;
starting from the intention node, accessing child nodes of the intention node, and adding the accessed child nodes into the empty list to obtain an element list, wherein the child nodes comprise: a relationship node and/or an attribute node;
judging whether the element list only contains attribute nodes;
under the condition that the element list not only contains attribute nodes, traversing each element in the element list to obtain a traversal result, wherein the element comprises at least one of the following elements: and element, and or element, in the case of the element being an and element, removing the child node from the element list and expanding the element list by the child node; and under the condition that the element is the OR element, counting the number n of child nodes corresponding to the OR element, copying the element list n times, removing the element list from an empty list, replacing the OR node with different child nodes for each copied element list, and adding the n element lists into the empty list, wherein n is a natural number.
9. A dialog control device, which is applied to an environment of a task-based multi-turn dialog, comprising:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring an intention tree corresponding to the current conversation, the intention tree indicates tasks required to be completed by the current conversation, and attribute nodes in the intention tree are associated with word slots;
the traversal unit is used for traversing the intention tree to obtain a traversal result;
a query unit, configured to query, when the traversal result indicates that there is a vacant attribute node in the intent tree, a system action corresponding to the vacant attribute node, and send dialog content corresponding to the system action, where the query unit, when querying the system action corresponding to the vacant attribute node, includes: when the traversal result indicates that the intention tree has vacant attribute nodes, determining various word slot combinations according to the traversal result, wherein the word slot combinations indicate the combination mode of vacant word slots needing to be filled; acquiring a word slot combination with the least number of vacant word slots contained in the word slot combination; selecting a vacant word slot from the word slot combination with the least number of vacant word slots, and determining the system action corresponding to the vacant word slot;
and a sending unit, configured to send a dialog result corresponding to the intent tree after the traversal result indicates that all attribute nodes in the intent tree are filled, where each of the intent trees corresponds to an intent action, and the intent action is used to indicate a dialog text corresponding to the task, or the intent action is used to indicate that query information corresponding to the task is obtained after a slot value corresponding to a word slot is transferred to a calling preset interface, and the dialog result includes the dialog text and the query information.
10. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the dialog control method of any of claims 1 to 8 via execution of the executable instructions.
11. A storage medium characterized by comprising a stored program, wherein an apparatus in which the storage medium is located is controlled to execute the conversation control method according to any one of claims 1 to 8 when the program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910067893.4A CN109815326B (en) | 2019-01-24 | 2019-01-24 | Conversation control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910067893.4A CN109815326B (en) | 2019-01-24 | 2019-01-24 | Conversation control method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109815326A CN109815326A (en) | 2019-05-28 |
| CN109815326B true CN109815326B (en) | 2021-09-10 |
Family
ID=66603171
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910067893.4A Active CN109815326B (en) | 2019-01-24 | 2019-01-24 | Conversation control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109815326B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110297616B (en) * | 2019-05-31 | 2023-06-02 | 百度在线网络技术（北京）有限公司 | Method, device, equipment and storage medium for generating speech technology |
| CN110188185A (en) * | 2019-06-05 | 2019-08-30 | 网易（杭州）网络有限公司 | Processing method, device, equipment and the storage medium of more wheel dialogues |
| CN110532361A (en) * | 2019-08-09 | 2019-12-03 | 深圳追一科技有限公司 | Recognition methods, device, computer equipment and the storage medium that user is intended to |
| CN110609618B (en) * | 2019-08-26 | 2023-06-20 | 杭州城市大数据运营有限公司 | Man-machine conversation method and device, computer equipment and storage medium |
| CN111209381B (en) * | 2020-01-03 | 2023-12-12 | 北京搜狗科技发展有限公司 | Time management method and device in dialogue scene |
| CN111914073A (en) * | 2020-07-15 | 2020-11-10 | 中国联合网络通信集团有限公司 | Customer service response method, device, equipment and storage medium |
| CN113190668A (en) * | 2021-05-19 | 2021-07-30 | 平安国际智慧城市科技股份有限公司 | Man-machine interaction method, device and equipment based on multi-turn conversation and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108920434A (en) * | 2018-06-06 | 2018-11-30 | 武汉酷犬数据科技有限公司 | A kind of general Web page subject method for extracting content and system |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2001293343A1 (en) * | 2000-04-06 | 2001-10-23 | Paul R. Bindler | Automated and intelligent networked-based psychological services |
| CN105513593B (en) * | 2015-11-24 | 2019-09-17 | 南京师范大学 | A kind of intelligent human-machine interaction method of voice driven |
| CN106528613B (en) * | 2016-05-26 | 2018-11-27 | 中科鼎富（北京）科技发展有限公司 | Intelligent answer method and device |
| CN107169034B (en) * | 2017-04-19 | 2020-08-04 | 畅捷通信息技术股份有限公司 | Multi-round human-computer interaction method and system |
| CN107688614B (en) * | 2017-08-04 | 2018-08-10 | 平安科技（深圳）有限公司 | It is intended to acquisition methods, electronic device and computer readable storage medium |
| CN108052547B (en) * | 2017-11-27 | 2019-09-27 | 华中科技大学 | Natural language question-answering method and system based on question sentence and knowledge graph structural analysis |
| CN108416051A (en) * | 2018-03-20 | 2018-08-17 | 杭州声讯网络科技有限公司 | A kind of Intelligent dialogue logical algorithm for sector application |
| CN111813961B (en) * | 2020-08-25 | 2020-12-18 | 腾讯科技（深圳）有限公司 | Data processing method and device based on artificial intelligence and electronic equipment |
| CN112131360A (en) * | 2020-09-04 | 2020-12-25 | 交通银行股份有限公司太平洋信用卡中心 | Intelligent multi-turn dialogue customization method and system |
-
2019
- 2019-01-24 CN CN201910067893.4A patent/CN109815326B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108920434A (en) * | 2018-06-06 | 2018-11-30 | 武汉酷犬数据科技有限公司 | A kind of general Web page subject method for extracting content and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109815326A (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109815326B (en) | Conversation control method and device | |
| KR102189855B1 (en) | Parameter collection and automatic dialog generation in dialog systems | |
| CN109948017B (en) | Information processing method and device | |
| CN108984157B (en) | Skill configuration and calling method and system for voice conversation platform | |
| US9600228B2 (en) | Enhanced system and method for static query generation and entry | |
| US8346563B1 (en) | System and methods for delivering advanced natural language interaction applications | |
| US10417567B1 (en) | Learning user preferences in a conversational system | |
| US20210166693A1 (en) | Man- machine interaction system and multi-task processing method in the man-machine interaction system | |
| US20190354594A1 (en) | Building and deploying persona-based language generation models | |
| CN116737900A (en) | Man-machine interaction processing system and method, storage medium and electronic equipment | |
| CN109543010A (en) | The interactive method and system of fused data library inquiry | |
| CN109992338B (en) | Method and system for exposing virtual assistant services across multiple platforms | |
| CN111639168A (en) | Multi-turn conversation processing method and device, electronic equipment and storage medium | |
| CN110489198A (en) | A kind of method and system of worksheet | |
| CN111680144A (en) | Method and system for multi-turn dialogue voice interaction, storage medium and electronic equipment | |
| US10326863B2 (en) | Speed and accuracy of computers when resolving client queries by using graph database model | |
| CN111831795A (en) | Multi-turn conversation processing method and device, electronic equipment and storage medium | |
| CN112131885A (en) | Semantic recognition method and device, electronic equipment and storage medium | |
| CN111813900A (en) | Multi-turn conversation processing method and device, electronic equipment and storage medium | |
| CN111753061A (en) | Multi-turn conversation processing method and device, electronic equipment and storage medium | |
| CN106446265A (en) | Question inquiry display method for intelligent terminal | |
| US10847141B2 (en) | Dialogue system and a dialogue method | |
| CN107609073A (en) | A kind of information methods of exhibiting, device, electronic equipment and storage medium | |
| CN109597739A (en) | Voice log services method and system in human-computer dialogue | |
| Fuckner et al. | Using a personal assistant for exploiting service interfaces |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |