WO2016062255A1 - 多轮会话交互方法、系统和计算机设备 - Google Patents

多轮会话交互方法、系统和计算机设备 Download PDF

Info

Publication number
WO2016062255A1
WO2016062255A1 PCT/CN2015/092462 CN2015092462W WO2016062255A1 WO 2016062255 A1 WO2016062255 A1 WO 2016062255A1 CN 2015092462 W CN2015092462 W CN 2015092462W WO 2016062255 A1 WO2016062255 A1 WO 2016062255A1
Authority
WO
WIPO (PCT)
Prior art keywords
requirement
node
user
information
standard
Prior art date
Application number
PCT/CN2015/092462
Other languages
English (en)
French (fr)
Inventor
焦增涛
汪冠春
Original Assignee
百度在线网络技术（北京）有限公司
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 百度在线网络技术（北京）有限公司 filed Critical 百度在线网络技术（北京）有限公司
Priority to KR1020167022267A priority Critical patent/KR101779387B1/ko
Priority to JP2016552346A priority patent/JP6234601B2/ja
Priority to EP15853398.4A priority patent/EP3211543A4/en
Priority to US15/119,348 priority patent/US10817786B2/en
Publication of WO2016062255A1 publication Critical patent/WO2016062255A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/901—Indexing; Data structures therefor; Storage structures
    • G06F16/9027—Trees
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/904—Browsing; Visualisation therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present invention relates to the field of communications technologies, and in particular, to a multi-round session interaction method, system, and computer device.
  • Embodiments of the present invention provide a multi-round session interaction method, system, and computer device to implement a session interaction of a task-type session system in which a system action has a tree-level dependency.
  • an embodiment of the present invention provides a multi-round session interaction method, including:
  • the standard requirement content included in the child node of the hit non-leaf node is output, or the output is hit. Session information corresponding to the leaf node.
  • An embodiment of the present invention further provides a multi-round session interaction system, including:
  • a receiving module configured to receive user demand information
  • mapping module configured to map the user requirement information to a standard requirement included in each node in a plurality of preset requirement structure trees, where a standard requirement included in each node is a child The upper requirements of the standard requirements contained in the node;
  • An output module if there is a node in the plurality of demand structure trees, and the standard requirement included by the user needs to be hit by the user requirement information, outputting a standard requirement content included in a child node of the hit non-leaf node, Or output the session information corresponding to the hit leaf node.
  • a computer apparatus comprising: one or more processors; a memory; one or more programs, the one or more programs being stored in the memory, and configured Executing, by the one or more processors, an instruction included in the one or more programs for executing a multi-round session interaction method: receiving user demand information; and the user demand information is preset to a plurality of local Mapping the standard requirements contained in each node in the requirements tree, in which the standard requirements contained in each node are the upper requirements of the standard requirements contained in the child nodes; if the plurality of demand trees exist A node that contains the standard requirement is hit by the user requirement information, outputs standard requirement content contained in the child node of the hit non-leaf node, or outputs session information corresponding to the hit leaf node.
  • the multi-round session interaction method, system and computer device provided by the embodiment of the present invention, by receiving user requirement information, mapping user requirement information to standard requirements included in each node of a plurality of local demand structure trees,
  • the standard requirement contained in each node is the upper requirement of the standard requirements contained in the child nodes; if there is one node in the multiple demand structure tree, and the standard requirements contained in it are hit by the user demand information, the output is
  • the sub-node of the hit non-leaf node contains the standard requirement content, or outputs the session information corresponding to the hit leaf node, thereby implementing a session interaction method for the task-type session system having the tree-level hierarchical dependency, and improving at the same time
  • the hit rate for user demand by receiving user requirement information, mapping user requirement information to standard requirements included in each node of a plurality of local demand structure trees,
  • the standard requirement contained in each node is the upper requirement of the standard requirements contained in the child nodes; if there is one node in the multiple demand structure tree,
  • FIG. 1 is a flowchart of a method for an embodiment of a multi-round session interaction method provided by the present invention
  • FIG. 3 is a flowchart of a method for another embodiment of a multi-round session interaction method according to the present invention.
  • FIG. 4 is a schematic structural diagram of an embodiment of a multi-round session interaction system provided by the present invention.
  • FIG. 5 is a schematic structural diagram of another embodiment of a multi-round session interaction system according to the present invention.
  • Figure 6 is a logic block diagram of a computer device in accordance with an exemplary embodiment of the present invention.
  • Embodiments of the present invention map a requirement input by a user to a standard requirement included in each node in the requirement structure tree by using a preset requirement structure tree; if such a node exists, the standard requirement contained by the node is hit by the demand information, Outputs the standard requirement content contained in the child node of the hit non-leaf node, or outputs the session information corresponding to the hit leaf node.
  • the technical solution of the embodiment of the present invention can be applied to a task-based session system having a tree-level hierarchical dependency.
  • FIG. 1 is a flowchart of a method for an embodiment of a multi-round session interaction method according to the present invention.
  • the execution body of the method may be an interactive system with multiple rounds of session functions.
  • the multi-round session interaction method specifically includes:
  • the user can select or directly input the demand information of interest from the demand candidate box provided by the interface through the interaction interface. For example, if the user wants to know the information about the Chihuahua, he can directly input the demand information such as "Chihuahua". These demand information can be regarded as user demand information.
  • each node includes a standard requirement that is a standard requirement of the child node. Superior demand;
  • each of the plurality of demand structure trees may be based on a large Most users' needs are in order, and a tree structure is constructed.
  • Each node on the tree represents a standard requirement type.
  • Figure 2 shows the partial structure of the demand tree for the Chihuahua.
  • each node on the demand structure tree contains information such as node requirements, standard requirements, and the level of the node (0 is the root node, which is gradually increased by 1), and the parent node index (no parent node is -1) and child node index (the node child node list, no child nodes are empty).
  • the node "purchase” in Figure 2 contains the node demand as "purchase” and the standard requirement is "Chihuahua purchase".
  • User behavior log A session log searched by means of large search, wise search, vertical domain search, etc., that is, a set of requirements that the user searches around a certain problem, and then digs out the demand sequence of most users to form the above demand structure tree. .
  • Entity upper and lower positions A task-based session system with entity requirements can construct a requirement structure tree through the upper and lower positions of entities in the entity library. If the user queries the "teacher” related questions, he or she can use the "subsidiary teacher", “high school teacher”, “primary school teacher” as the “teacher” according to the entity in the entity library with the entity's upper and lower relationship. The entity is recommended to the user.
  • the requirements structure tree can be constructed by artificial construction.
  • demand template mining In the above process of constructing the demand structure tree, further demand template mining of the similar requirements of the above mining needs to be performed to form the node requirements included in each node in the demand structure tree.
  • demand template mining it can be divided into three parts: demand clustering, template mining and template verification.
  • Demand clustering refers to clustering demand information searched in the target domain according to demand categories.
  • Template mining refers to extracting frequent patterns from a type of demand information through frequent pattern mining algorithms (such as prefixSpan).
  • the template verification is to verify the matching accuracy of the template outputted in the previous step, and discard the ambiguous template to finally use the retained requirement template as the requirement structure tree.
  • the node requirements contained in each node Of course, for less complex areas, you can also manually configure the requirements template.
  • the plurality of demand trees formed by the above method are pre-stored in a local memory.
  • the user demand information may be mapped to standard requirements included in each node of the demand structure tree to analyze The exact needs of current users.
  • the standard demand contained in the user demand information is known as "Chihuahua Purchase” by analysis.
  • the node "purchase” is hit, and the system can include the child nodes of the "purchase” node when determining the output action.
  • the standard demand content such as “price”, “good price”, “intelligence ranking”, “how to look pure impure” and “how long to live” are output as shown in Figure 2 to provide the user with the next demand option.
  • the hit node is a non-leaf node
  • the node hitting the user demand information is a leaf node in the demand structure tree, as shown in the “price” node in the demand structure tree in FIG. 2, the system directly The action corresponding to the hit leaf node is executed, that is, the session information corresponding to the "price" node is output.
  • the system can give a default output action according to the default setting, such as giving a recommendation for a certain field, or a dialogue. content.
  • the multi-round session interaction method provided by the embodiment of the present invention maps the user requirement information to the standard requirements included in each node in the requirement structure tree; if such a node exists, the standard requirement contained by the user is hit by the user demand information, and then the output is output.
  • the sub-node of the hit non-leaf node contains the standard requirement content, or outputs the session information corresponding to the hit leaf node.
  • FIG. 3 is a flowchart of a method for another embodiment of a multi-round session interaction method according to the present invention, which is a specific implementation manner of the method embodiment shown in FIG.
  • the round session interaction method specifically includes:
  • the standard requirements and user demand information are exactly the same, which means that the text information contained in the two is exactly the same, but the order of words can be different.
  • user demand information for "Chihuahua Buy” and “Buy Chihuahua” can be considered exactly the same as the standard requirement "Chihuahua Buy”.
  • the user requirement information may be semantically matched with the standard requirement corresponding to each node in each requirement structure tree, and The minimum standard requirement for semantic matching is determined as the standard requirement that is hit by the user's demand information.
  • the user demand information is "I want to buy a Chihuahua"
  • the user needs information can be physically decomposed first, such as "I want to buy Chihuahua” into three, I buy, Chihuahua three entities, and then three entities in a collection form and multiple demand trees in each
  • the standard requirements contained in the node are matched.
  • the so-called matching is to see whether the standard requirement is included in the entity set corresponding to the user requirement information. If it is included, the standard requirement is matched with the demand information.
  • the standard requirements and the user requirement information are exactly the same, and the user requirement information input by the user may not completely include the corresponding node in the requirement structure tree.
  • the system cannot identify the content of the current user's needs. In this case, the system can combine the user demand information input by the user with the current user demand information to determine the current content of the user. If the user enters “I want to buy Chihuahua” in the previous round, this time directly enter “How is the price”, and “How is the price” can not match a certain standard demand through the above matching, but the user asks the price demand for the purchase of Chihuahua is The session action that the system can provide.
  • the system can semantically combine the user demand information input by the previous user with the current user demand information, and then match the combined user demand information according to the above matching method, then the standard can be determined.
  • the demand "Chihuahua purchase price" is a standard requirement that matches the user's current needs.
  • the foregoing steps S302, S303, and S304 can be regarded as three specific implementation manners of the step S102.
  • the user requirements information is mapped to the standard requirements included in each node in the plurality of preset requirement structure trees.
  • the method of matching the current user demand information through semantic matching and utilizing the user demand information context can improve the matching rate, that is, increase the possibility that the standard demand is hit.
  • Step S305 is performed after step S302 or step S303 or step S304:
  • step S305 Output the standard requirement content included in the child node of the hit non-leaf node, or output the session information corresponding to the hit leaf node.
  • the specific execution process of the step refer to the corresponding content of step S103.
  • the sibling nodes of the nodes contain standard demand output to recommend more choices to the user.
  • the system outputs the standard requirements contained in each child node of the node "purchasing" in FIG. 2, the brother nodes that "purchase” the node may also include “feeding", “training”, “disease and health”,
  • the corresponding standard demand output such as "peripheral” is available for the user to select.
  • the method in this embodiment further includes constructing the foregoing requirement structure tree in different domains by training learning of the user's behavior log and various domain knowledge.
  • the specific construction basis and principle can be referred to the foregoing content, and details are not described herein.
  • the system can update or form nodes to be added or deleted according to requirements.
  • the multi-round session interaction method maps the user requirement information to the standard requirements included in each node in the requirement structure tree; if such a node exists, the standard requirement contained by the node is hit by the demand information, and the output is The child node of the hit non-leaf node contains the standard requirement content, or outputs the session information corresponding to the hit leaf node.
  • the judgment method of the standard requirement being hit is given separately, which not only realizes the task-type session task interaction with the tree-level hierarchical dependency, but also increases the possibility that the standard requirement is hit.
  • FIG. 4 is a schematic structural diagram of an embodiment of a multi-round session interaction system provided by the present invention, and the method steps shown in FIG. 1 can be performed.
  • the multi-round session interaction system specifically includes a receiving module 41, a mapping module 42 and an output module 43; wherein:
  • the receiving module 41 is configured to receive user demand information.
  • the mapping module 42 is configured to map user requirement information to standard requirements included in each node in a plurality of preset requirement structure trees, where each node includes a standard requirement of its child node The upper requirements of the standard requirements included;
  • the output module 43 is configured to: if there is a node in the multiple demand structure tree, the standard requirement contained by the user is required to be hit by the user requirement information, and the standard requirement content included in the child node of the hit non-leaf node is output, or the output is hit. Session information corresponding to the leaf node.
  • the multi-round session interaction system may further include:
  • the establishing module 44 is configured to construct the demand structure tree in different domains by training learning of the user's behavior log and various domain knowledge.
  • mapping module 42 in the multi-round session interaction system shown in FIG. 5 is specifically configured to: if there is one node in the multiple requirement structure tree, and the standard requirement and the user requirement information are identical, the node is included The standard requirements are hit by user demand information.
  • mapping module 42 in the multi-round session interaction system shown in FIG. 5 is further used to:
  • the user requirement information and the historical user requirement information previously input by the user are combined to semantically match the standard requirements included in each node in the plurality of demand structure trees, and the lowest standard that matches the semantics of the user demand information in each standard requirement is determined.
  • the demand is hit by the user demand information.
  • the output module 43 in the multi-round session interaction system shown in FIG. 5 is further configured to output the standard requirement content included in the sibling node of the hit node.
  • the multi-round session interaction system maps the user requirement information to the standard requirements included in each node in the requirement structure tree; if such a node exists, the standard requirements contained therein are hit by the user demand information, and then the output is output.
  • the multi-round session interaction method adopted by the solution is in the specific mapping process. After judging whether the user demand information is exactly the same as the standard requirement, whether the user demand information matches the standard requirement, and whether the combination of the context requirement information including the current user demand information matches the standard requirement, respectively, the standard requirement is given.
  • the method of judging by hits not only realizes task-type session task interaction with tree-level hierarchical dependencies, but also increases the possibility that standard requirements are hit.
  • Figure 6 is a logic block diagram of a computer device in accordance with an exemplary embodiment of the present invention.
  • the computer device can be used to implement the multi-round session interaction method provided in the above embodiments. Specifically:
  • the computer device can include an input unit 610, a memory 620 including one or more computer readable storage media, a processor 630 including one or more processing cores, a display unit 640, a communication unit 650, a power source 660, and the like. component.
  • a processor 630 including one or more processing cores
  • a display unit 640 including one or more communication cores
  • a power source 660 and the like. component.
  • the computer device architecture illustrated in the figures does not constitute a limitation to a computer device, and may include more or fewer components than those illustrated, or a combination of certain components, or different component arrangements. among them:
  • the memory 620 can be used to store software programs and data, and the processor 630 executes various functional applications and data processing by running software programs and data stored in the memory 620.
  • the memory 620 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may be stored according to Data created by the use of the computer device (such as audio data, phone book, etc.), and the like.
  • memory 620 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, memory 620 can also include a memory controller to provide access to memory 620 by input unit 610 and processor 630.
  • Processor 630 is a control center of the computer device that connects various portions of the entire computer device using various interfaces and lines, by running or executing software programs and/or modules stored in memory 620, and for recalling storage in memory 620. The data, performing various functions and processing data of the computer device, thereby performing overall monitoring of the computer device.
  • Display unit 640 can be used to display information entered by a user or information provided to a user, as well as various graphical user interfaces of the computer device, which can be composed of graphics, text, icons, video, and any combination thereof.
  • the display unit 640 may include a display panel.
  • the display panel may be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), or the like.
  • LCD Liquid Crystal Display
  • OLED Organic Light-Emitting Diode
  • input unit 610 and display unit 640 are implemented as two separate components to implement input and output functions, in some embodiments, input unit 610 can be integrated with display unit 640 for input and output.
  • the communication unit 650 can be used for transmitting and receiving information or receiving and transmitting signals during a call.
  • the communication unit 650 can be a network communication device such as an RF (Radio Frequency) circuit, a router, a modem, or the like.
  • communication unit 650 can also communicate with the network and other devices via wireless communication.
  • the wireless communication may use any communication standard or protocol, including but not limited to GSM (Global System of Mobile communication), GPRS (General Packet Radio Service), CDMA (Code Division Multiple Access). , Code Division Multiple Access), WCDMA (Wideband Code Division Multiple Access), LTE (Long Term Evolution), e-mail, SMS (Short Messaging Service), and the like.
  • the computer device may also include a power source 660 (such as a battery) that supplies power to the various components.
  • the power source may be logically coupled to the processor 630 through a power management system to manage charging, discharging, and power management through the power management system.
  • the power supply 660 can also include any one or more of a DC or AC power source, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.
  • the computer device may further include a camera, a Bluetooth module, sensors (such as light sensors, motion sensors, and other sensors, etc.), audio circuits, wireless communication units, and the like, which are not described herein.
  • sensors such as light sensors, motion sensors, and other sensors, etc.
  • audio circuits such as audio signals, wireless communication units, and the like, which are not described herein.
  • the computer device includes one or more processors 630, a memory 620, and one or more programs, the one or more programs being stored in the memory 620 and configured to be
  • the one or more processors 630 execute instructions for executing the multi-round session interaction method included in the one or more programs: receiving a user demand letter Mapping the user requirement information to standard requirements contained in each of the plurality of local demand tree, in which the standard requirements of each node are included in the child nodes a higher-level requirement of the standard requirement; if there is a node in the plurality of demand structure trees, the standard requirement contained in the plurality of demand structure trees is hit by the user requirement information, and the standard requirement content included in the child node of the hit non-leaf node is output Or output the session information corresponding to the hit leaf node.
  • mapping, by the user requirement information, to a standard requirement included in each node of the preset multiple preset requirement tree includes: if there is one node in the plurality of demand structure trees, the The standard requirement is exactly the same as the user requirement information, and it is determined that the standard requirement included in the node is hit by the user requirement information.
  • the mapping the user requirement information to a standard requirement included in each node of the preset multiple preset requirement tree includes: if there is no node in the plurality of demand structure trees, the included The standard requirement is exactly the same as the user requirement information, and the user requirement information is semantically matched with the standard requirement included in each node of the plurality of requirement structure trees, and the user demand information is determined therein.
  • the lowermost standard requirement of the semantic matching is hit by the user requirement information; or the user requirement information and the historical user requirement information previously input by the user are combined together with the nodes included in the plurality of demand structure trees
  • the standard requirements are semantically matched, and it is determined that the lowest standard requirement in the standard requirements that is semantically matched with the user requirement information is hit by the user requirement information.
  • the method further includes: outputting the standard requirement content included in the sibling node of the hit node.
  • the memory 620 further includes instructions for performing the following process: constructing the demand structure tree in different fields by training learning of the user's behavior log and various domain knowledge.
  • the computer device provided by the present invention maps user requirement information to standard requirements included in each node in the demand structure tree; if there is such a node, the standard requirement contained therein is hit by the user demand information, and the output is hit by the non-leaf The standard requirement content contained in the child node of the node, or the session information corresponding to the leaf node that is hit, thereby implementing one A session interaction method for a task-based session system with tree-level hierarchical dependencies, while improving the hit rate for user requirements.
  • the multi-round session interaction method adopted by the solution determines whether the user requirement information is identical to the standard requirement, whether the user requirement information matches the standard requirement, and the context requirement including the current user requirement information in the specific mapping process. Whether the combination of information and the standard requirements match, the judgment methods of the standard requirements being hit are respectively given, which not only realizes the task-type session task interaction with tree-level hierarchical dependency, but also improves the possibility that the standard requirement is hit. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

一种多轮会话交互方法、系统和计算机设备，方法包括：接收用户需求信息(101)；将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射(102)，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息(103)。上述技术方案实现了一种针对有树形层次依赖关系的任务型会话系统的会话交互方法，同时提高了对用户需求的命中率。

Description

多轮会话交互方法、系统和计算机设备
本申请要求于2014年10月22日提交中国专利局、申请号为201410569388.7、发明名称为“多轮会话交互方法和系统”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及通信技术领域，尤其涉及一种多轮会话交互方法、系统和计算机设备。
背景技术
现有大部分的多轮会话交互系统多是预先定义好系统状态和系统动作集合；在系统运行时，根据当前系统的状态，通过一系列策略或者统计模型从系统动作集合中选择一个最佳的系统动作进行输出。
但是，对于各系统动作有树形层次依赖关系的任务型会话系统，现有大部分多轮会话交互系统的解决方案并不理想。例如：通过人工定义规则的多轮会话交互系统，在任务定制时较为复杂，且容易出现多条规则的冲突；基于增强学习的统计对话系统，在有充足训练语料的前提下，可以自动学习到这种树形依赖关系，但是语料获取难，且学习到的内容可理解性差，难于控制。
发明内容
本发明的实施例提供一种多轮会话交互方法、系统和计算机设备，以实现系统动作具有树形层次依赖关系的任务型会话系统的会话交互。
为达到上述目的，本发明的实施例提供了一种多轮会话交互方法，包括：
接收用户需求信息；
将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
本发明的实施例还提供了一种多轮会话交互系统，包括：
接收模块，用于接收用户需求信息；
映射模块，用于将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
输出模块，用于若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
根据本发明的另一方面，本发明提供一种计算机设备，包括：一个或多个处理器；存储器；一个或多个程序，所述一个或多个程序存储在所述存储器中，且经配置以由所述一个或者多个处理器执行所述一个或者多个程序包含的用于执行多轮会话交互方法的指令：接收用户需求信息；将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
本发明实施例提供的多轮会话交互方法、系统和计算机设备，通过接收用户需求信息；将用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；若多个需求结构树中存在一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息，由此实现一种针对有树形层次依赖关系的任务型会话系统的会话交互方法，同时提高了对用户需求的命中率。
附图说明
图1为本发明提供的多轮会话交互方法一个实施例的方法流程图；
图2为本发明提供的关于吉娃娃的需求结构树的部分结构；
图3为本发明提供的多轮会话交互方法另一个实施例的方法流程图；
图4为本发明提供的多轮会话交互系统一个实施例的结构示意图；
图5为本发明提供的多轮会话交互系统另一个实施例的结构示意图；
图6为本发明示例性实施例的计算机设备的逻辑框图。
具体实施方式
本发明的实施例利用预先设置的需求结构树，将用户输入的需求信息向需求结构树中各节点包含的标准需求进行映射；若存在这样一个节点，其包含的标准需求被需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。本发明实施例的技术方案可以适用于有树形层次依赖关系的任务型会话系统。
实施例一
图1为本发明提供的多轮会话交互方法一个实施例的方法流程图，该方法的执行主体可以为具有多轮会话功能的交互系统。如图1所示，该多轮会话交互方法具体包括：
S101，接收用户需求信息；
在多轮会话交互系统中，用户可通过交互界面从界面提供的需求候选框里逐级选择或者直接键入自己感兴趣的需求信息。例如，用户想了解吉娃娃的相关信息，则可直接键入如“吉娃娃”的需求信息。这些需求信息都可以视为用户需求信息。
S102，将用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
其中，上述多个需求结构树中各节点包含的标准需求可以是根据大 多数用户的需求习惯顺序，构建出的一种树形结构，树上每一个节点表示一种标准需求类型。各节点和其相应子节点包含的标准需求之间存在上下位关系。例如，如果某用户有某一节点对应的标准需求，则他有很大的概率也会有该节点子节点对应的标准需求(如果有子节点)，那么可以将子节点对应标准需求推荐给用户。图2给出了关于吉娃娃的需求结构树的部分结构。如图2所示，该需求结构树上每一个节点包含的信息有：节点需求、标准需求、节点所在层级(0为根节点，往下逐步加1)、父节点索引(无父节点则为-1)和子节点索引(该节点子节点列表，无子节点则为空)。例如图2中的节点“购买”，其包含的节点需求为“购买”，标准需求为“吉娃娃购买”。
上述需求结构树可以通过以下方式构建：
用户行为日志：通过大搜索、wise搜索、垂直领域搜索等方式搜索的session日志，即用户围绕某个问题搜索的一组需求集合，然后从中统计挖掘大部分用户的需求序列从而形成上述需求结构树。
领域知识：对于某些低频搜索的领域，通过统计用户行为日志很难获取置信度高的需求结构树，这时可以针对特定领域做定制化的需求挖掘。
实体上下位：有实体需求的任务型会话系统可以通过实体库中各实体的上下位关系构建需求结构树。如用户查询了“教师”的相关问题后，可以根据实体库中与之有实体上下位关系的实体，将如“大学教师”、“高中教师”、“小学教师”等作为“教师”的下位实体推荐给用户。
人工构建：对于特别简单的垂直领域，可以通过人工构建的方式构建需求结构树。
在上述构建需求结构树过程中，还需要对上述挖掘的同类需求进行进一步的需求模板挖掘，以形成需求结构树中各节点包含的节点需求。在需求模板挖掘中，可以分为需求聚类、模板挖掘和模板验证三部分。需求聚类是指将在目标领域内搜索的需求信息按需求类别进行聚类。模板挖掘指的是通过频繁模式挖掘算法(如prefixSpan)，从一类需求信息中抽出频繁模式。模板验证是对上步输出的模板进行匹配准确性验证，将有歧义的模板丢弃，以最终将保留的需求模板作为需求结构树中 各节点包含的节点需求。当然，对于不太复杂的领域，也可以通过人工配置需求模板。
利用上述方法形成的多个需求结构树被预先存储在本地的存储器中，当系统接收到用户需求信息后，可将用户需求信息向这些需求结构树中各节点包含的标准需求进行映射，以分析当前用户的准确需求。
S103，若多个需求结构树中存在一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息；
例如，当用户需求信息为“我想购买吉娃娃”，则通过分析获知该用户需求信息中蕴含的标准需求为“吉娃娃购买”。将该用户需求信息映射到需求结构树中后，可知以吉娃娃为根节点的需求结构树中，节点“购买”被命中，则系统在决定输出动作时，可将“购买”节点的子节点包括如图2中的“价格”、“好不好养”、“智商排名”、“怎看纯不纯”以及“能活多久”等标准需求内容输出以提供给用户下一步的需求选项。当然，这是针对被命中节点为非叶子节点的情况，针对用户需求信息命中的节点为需求结构树中的叶子节点的，如图2中需求结构树中的“价格”节点，则系统会直接执行被命中的叶子节点对应的动作，即将“价格”节点对应的会话信息输出。
当然，对于用户需求信息，当前所有需求结构树各节点对应的标准需求均未被命中的情况，系统可根据默认设置给出默认输出动作，如给出某一领域的需求推荐，或者是一段对话内容。
本发明实施例提供的多轮会话交互方法，通过将用户需求信息向需求结构树中各节点包含的标准需求进行映射；若存在这样一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。本发明实施例的技术方案可以适用于有树形层次依赖关系的任务型会话系统。
实施例二
图3为本发明提供的多轮会话交互方法另一个实施例的方法流程图，是如图1所示方法实施例的一种具体实现方式，如图3所示，该多 轮会话交互方法具体包括：
S301，接收用户需求信息；该步骤具体执行过程可参见步骤101的相应内容。
S302，若多个需求结构树中存在一个节点，其包含的标准需求与用户需求信息完全相同，则确定该节点包含的标准需求被用户需求信息命中；
其中，标准需求与用户需求信息完全相同是指二者所包含的文字信息完全相同，但词语顺序可以不相同。例如，用户需求信息为“吉娃娃购买”和“购买吉娃娃”均可以视为与标准需求“吉娃娃购买”完全相同。当上述多个需求结构树中存在一个节点，其包含的标准需求与用户需求信息完全相同时，可确定该节点对应的标准需求被该用户需求信息命中。
或者，
S303，若多个需求结构树中不存在一个节点，其包含的标准需求与用户需求信息完全相同，则：
将用户需求信息与多个需求结构树中各节点包含的标准需求进行语义匹配，并确定其中与用户需求信息语义匹配的最下位标准需求被用户需求信息命中；
当上述多个需求结构树中不存在一个节点，其包含的标准需求与用户需求信息完全相同时，则可将该用户需求信息与各需求结构树中各节点对应的标准需求进行语义匹配，并将语义匹配的最下位标准需求确定为被用户需求信息命中的标准需求。
例如，当用户需求信息为“我想购买吉娃娃”时，可获知没有一个节点包含的标准需求与该用户需求信息完全相同。对于这种情况，可先对用户需求信息进行实体分解，如将“我想购买吉娃娃”分解为我、购买、吉娃娃三个实体，然后将三个实体以集合形式与多个需求结构树中各节点包含的标准需求进行匹配，所谓匹配就是看标准需求是否包含在用户需求信息对应的实体集合中，如果包含则认为该标准需求与需求信息匹配。例如对用户需求信息“我想购买吉娃娃”分析后形成的实体集合，标准需求“吉娃娃”和“吉娃娃购买”都与该实体集合匹配，那么 将与该实体集合匹配的最下位的标准需求确定为该用户需求信息命中的标准需求，即“吉娃娃购买”被命中。
或者，
S304，若多个需求结构树中不存在一个节点，其包含的标准需求与用户需求信息完全相同，则将用户需求信息以及用户之前输入的历史用户需求信息组合共同与多个需求结构树中各节点包含的标准需求进行语义匹配，并确定各标准需求中与用户需求信息语义匹配的最下位标准需求被用户需求信息命中；
当上述多个需求结构树中不存在一个节点，其包含的标准需求与用户需求信息完全相同，且由于用户每次输入的用户需求信息中，可能不完全包括需求结构树中某一节点对应的标准需求时，系统无法辨认当前用户需求的内容。这种情况下，系统可通过该用户之前输入的几条用户需求信息与当前的用户需求信息进行组合，整体判断该用户当前的需求内容。如用户上一轮输入“我想购买吉娃娃”，这次直接输入“价格怎么样”，而“价格怎么样”通过上述匹配无法匹配到某一标准需求，但用户询问购买吉娃娃的价格需求又是系统可以提供的会话动作，此时，系统可根据上一条用户输入的用户需求信息与当前的用户需求信息进行语义组合，然后将组合后的用户需求信息按上述匹配方法匹配，则就可以确定标准需求“吉娃娃购买价格”为与用户当前需求匹配的标准需求。
上述步骤S302、S303和S304可视为步骤S102的三种具体实现方式，本实施例中，在将用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射以确定被命中的标准需求时，通过语义匹配以及利用用户需求信息上下文来对当前用户需求信息进行匹配的方法可以提高匹配率，即提高标准需求被命中的可能性。
在步骤S302或步骤S303或步骤S304之后执行步骤S305：
S305，输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息；该步骤具体执行过程可参见步骤S103的相应内容。
为了给用户当前需求提供更丰富的推荐需求，可以将上述被命中的 节点(包括非叶子节点和叶子节点)的兄弟节点包含的标准需求输出以推荐给用户更多需求选择。例如，当系统将如图2中的节点“购买”的各子节点包含的标准需求输出时，还可将节点“购买”的兄弟节点包括“喂养”、“训练”、“疾病与健康”、“周边”等对应的标准需求输出，以供用户选择。
这里说明，本实施例所述方法中还包括通过对用户的行为日志和各领域知识的训练学习构建不同领域下的上述需求结构树。具体构建依据和原理可参见前述内容，在此不作赘述。系统可根据需求对形成的需求结构树进行更新或节点的增加和删减。
本发明实施例提供的多轮会话交互方法，通过将用户需求信息向需求结构树中各节点包含的标准需求进行映射；若存在这样一个节点，其包含的标准需求被需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。在具体映射过程中，通过判断用户需求信息与标准需求是否完全相同、用户需求信息与标准需求是否匹配，以及包含当前用户需求信息在内的上下文需求信息的组合与标准需求是否匹配这三种情况，分别给出标准需求被命中的判断方法，不但实现了有树形层次依赖关系的任务型会话任务交互，而且提高了标准需求被命中的可能性。
以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应以所述权利要求的保护范围为准。
实施例三
图4为本发明提供的多轮会话交互系统一个实施例的结构示意图，可执行如图1所示方法步骤。如图4所示，该多轮会话交互系统具体包括接收模块41，映射模块42和输出模块43；其中：
接收模块41，用于接收用户需求信息；
映射模块42，用于将用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
输出模块43，用于若多个需求结构树中存在一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
进一步的，在如图4所示实施例的基础上，如图5所示，上述多轮会话交互系统还可以包括：
建立模块44，用于通过对用户的行为日志和各领域知识的训练学习构建不同领域下的所述需求结构树。
进一步的，如图5所示的多轮会话交互系统中映射模块42具体用于：若多个需求结构树中存在一个节点，其包含的标准需求与用户需求信息完全相同，则确定该节点包含的标准需求被用户需求信息命中。
进一步的，如图5所示的多轮会话交互系统中映射模块42具体还用于：
若多个需求结构树中不存在一个节点，其包含的标准需求与用户需求信息完全相同，则：
将用户需求信息与多个需求结构树中各节点包含的标准需求进行语义匹配，并确定其中与用户需求信息语义匹配的最下位标准需求被用户需求信息命中；
或者，将用户需求信息以及用户之前输入的历史用户需求信息组合共同与多个需求结构树中各节点包含的标准需求进行语义匹配，并确定各标准需求中与用户需求信息语义匹配的最下位标准需求被用户需求信息命中。
进一步的，如图5所示的多轮会话交互系统中输出模块43还用于输出被命中节点的兄弟节点包含的标准需求内容。
本发明实施例提供的多轮会话交互系统，通过将用户需求信息向需求结构树中各节点包含的标准需求进行映射；若存在这样一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息，由此实现一种针对有树形层次依赖关系的任务型会话系统的会话交互方法，同时提高了对用户需求的命中率。
进一步的，本方案采用的多轮会话交互方法在具体映射过程中，通 过判断用户需求信息与标准需求是否完全相同、用户需求信息与标准需求是否匹配，以及包含当前用户需求信息在内的上下文需求信息的组合与标准需求是否匹配这三种情况，分别给出标准需求被命中的判断方法，不但实现了有树形层次依赖关系的任务型会话任务交互，而且提高了标准需求被命中的可能性。
实施例四
图6为本发明示例性实施例的计算机设备的逻辑框图。
参照图6，所述计算机设备可用于实施上述实施例中提供的多轮会话交互方法。具体来讲：
所述计算机设备可包括输入单元610、包括有一个或一个以上计算机可读存储介质的存储器620、包括有一个或者一个以上处理核心的处理器630、显示单元640、通信单元650、以及电源660等部件。本领域技术人员可以理解，图中示出的计算机设备结构并不构成对计算机设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。其中：
存储器620可用于存储软件程序以及数据，处理器630通过运行存储在存储器620的软件程序以及数据，从而执行各种功能应用以及数据处理。存储器620可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据所述计算机设备的使用所创建的数据(比如音频数据、电话本等)等。此外，存储器620可包括高速随机存取存储器，还可包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。相应地，存储器620还可包括存储器控制器，以提供输入单元610和处理器630对存储器620的访问。
处理器630是所述计算机设备的控制中心，利用各种接口和线路连接整个计算机设备的各个部分，通过运行或执行存储在存储器620内的软件程序和/或模块，以及调用存储在存储器620内的数据，执行所述计算机设备的各种功能和处理数据，从而对所述计算机设备进行整体监控。
显示单元640可用于显示由用户输入的信息或提供给用户的信息以及所述计算机设备的各种图形用户接口，这些图形用户接口可以由图形、文本、图标、视频和其任意组合来构成。显示单元640可包括显示面板，可选的，可以采用LCD(Liquid Crystal Display，液晶显示器)、OLED(Organic Light-Emitting Diode，有机发光二极管)等形式来配置显示面板。虽然在图示中，输入单元610与显示单元640是作为两个独立的部件来实现输入和输出功能，但是在某些实施例中，可将输入单元610与显示单元640集成而实现输入和输出功能。
通信单元650可用于收发信息或通话过程中，信号的接收和发送，通信单元650可以为RF(Radio Frequency，射频)电路、路由器、调制解调器、等网络通信设备。此外，通信单元650还可通过无线通信与网络和其他设备通信。所述无线通信可使用任一通信标准或协议，包括但不限于GSM(Global System of Mobile communication，全球移动通讯系统)、GPRS(General Packet Radio Service，通用分组无线服务)、CDMA(Code Division Multiple Access，码分多址)、WCDMA(Wideband Code Division Multiple Access，宽带码分多址)、LTE(Long Term Evolution，长期演进)、电子邮件、SMS(Short Messaging Service，短消息服务)等。
所述计算机设备还可包括给各个部件供电的电源660(比如电池)，优选的，电源可通过电源管理系统与处理器630逻辑相连，从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。电源660还可包括一个或一个以上的直流或交流电源、再充电系统、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。
尽管未示出，所述计算机设备还可以包括摄像头、蓝牙模块、传感器(比如光传感器、运动传感器以及其他传感器等)、音频电路和无线通信单元等，在此不再赘述。
在本实施例中，所述计算机设备包括一个或者多个处理器630、存储器620，以及一个或多个程序，所述一个或多个程序存储在所述存储器620中，且经配置以由所述一个或者多个处理器630执行所述一个或者多个程序包含的用于执行多轮会话交互方法的指令：接收用户需求信 息；将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
此外，所述将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射包括：若所述多个需求结构树中存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则确定该节点包含的所述标准需求被所述用户需求信息命中。
此外，所述将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射包括：若所述多个需求结构树中不存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则：将所述用户需求信息与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定其中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中；或者，将所述用户需求信息以及用户之前输入的历史用户需求信息组合共同与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定各所述标准需求中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中。
此外，所述若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息的过程中，还包括：输出所述被命中节点的兄弟节点包含的标准需求内容。
此外，存储器620中还包括执行以下处理的指令：通过对用户的行为日志和各领域知识的训练学习构建不同领域下的所述需求结构树。
本发明提供的计算机设备，通过将用户需求信息向需求结构树中各节点包含的标准需求进行映射；若存在这样一个节点，其包含的标准需求被用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息，由此实现一 种针对有树形层次依赖关系的任务型会话系统的会话交互方法，同时提高了对用户需求的命中率。
进一步的，本方案采用的多轮会话交互方法在具体映射过程中，通过判断用户需求信息与标准需求是否完全相同、用户需求信息与标准需求是否匹配，以及包含当前用户需求信息在内的上下文需求信息的组合与标准需求是否匹配这三种情况，分别给出标准需求被命中的判断方法，不但实现了有树形层次依赖关系的任务型会话任务交互，而且提高了标准需求被命中的可能性。
最后应说明的是：以上各实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述各实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的范围。

Claims (11)

  1. 一种多轮会话交互方法，其特征在于，包括：
    接收用户需求信息；
    将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
    若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
  2. 根据权利要求1所述的方法，其特征在于，所述将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射包括：
    若所述多个需求结构树中存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则确定该节点包含的所述标准需求被所述用户需求信息命中。
  3. 根据权利要求1或2所述的方法，其特征在于，所述将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射包括：
    若所述多个需求结构树中不存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则：
    将所述用户需求信息与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定其中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中；
    或者，将所述用户需求信息以及用户之前输入的历史用户需求信息组合共同与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定各所述标准需求中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，所述若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内 容，或者输出被命中的叶子节点对应的会话信息的过程中，还包括：
    输出所述被命中节点的兄弟节点包含的标准需求内容。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，还包括：
    通过对用户的行为日志和各领域知识的训练学习构建不同领域下的所述需求结构树。
  6. 一种多轮会话交互系统，其特征在于，包括：
    接收模块，用于接收用户需求信息；
    映射模块，用于将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
    输出模块，用于若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
  7. 根据权利要求6所述的系统，其特征在于，所述映射模块具体用于：若所述多个需求结构树中存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则确定该节点包含的所述标准需求被所述用户需求信息命中。
  8. 根据权利要求6或7所述的系统，其特征在于，所述映射模块具体还用于：
    若所述多个需求结构树中不存在一个节点，其包含的所述标准需求与所述用户需求信息完全相同，则：
    将所述用户需求信息与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定其中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中；
    或者，将所述用户需求信息以及用户之前输入的历史用户需求信息组合共同与所述多个需求结构树中各节点包含的所述标准需求进行语义匹配，并确定各所述标准需求中与所述用户需求信息语义匹配的最下位所述标准需求被所述用户需求信息命中。
  9. 根据权利要求6-8任一项所述的系统，其特征在于，所述输出模 块还用于输出所述被命中节点的兄弟节点包含的标准需求内容。
  10. 根据权利要求6-9任一项所述的系统，其特征在于，还包括：
    建立模块，用于通过对用户的行为日志和各领域知识的训练学习构建不同领域下的所述需求结构树。
  11. 一种计算机设备，其特征在于，包括：
    一个或多个处理器；
    存储器；
    一个或多个程序，所述一个或多个程序存储在所述存储器中，且经配置以由所述一个或者多个处理器执行所述一个或者多个程序包含的用于执行多轮会话交互方法的指令：
    接收用户需求信息；
    将所述用户需求信息向预置在本地的多个需求结构树中各节点包含的标准需求进行映射，在所述需求结构树中，每个节点包含的标准需求是其子节点包含的标准需求的上位需求；
    若所述多个需求结构树中存在一个节点，其包含的所述标准需求被所述用户需求信息命中，则输出被命中的非叶子节点的子节点包含的标准需求内容，或者输出被命中的叶子节点对应的会话信息。
PCT/CN2015/092462 2014-10-22 2015-10-21 多轮会话交互方法、系统和计算机设备 WO2016062255A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
KR1020167022267A KR101779387B1 (ko) 2014-10-22 2015-10-21 멀티라운드 세션 인터랙션 방법, 시스템 및 컴퓨터 기기
JP2016552346A JP6234601B2 (ja) 2014-10-22 2015-10-21 マルチラウンドセッションインタラクション方法、システム及びコンピュータ装置
EP15853398.4A EP3211543A4 (en) 2014-10-22 2015-10-21 Multi-round session interaction method and system, and computer device
US15/119,348 US10817786B2 (en) 2014-10-22 2015-10-21 Multi-round session interaction method and system, and computer device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410569388.7A CN104408639A (zh) 2014-10-22 2014-10-22 多轮会话交互方法和系统
CN201410569388.7 2014-10-22

Publications (1)

Publication Number Publication Date
WO2016062255A1 true WO2016062255A1 (zh) 2016-04-28

Family

ID=52646268

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/092462 WO2016062255A1 (zh) 2014-10-22 2015-10-21 多轮会话交互方法、系统和计算机设备

Country Status (6)

Country Link
US (1) US10817786B2 (zh)
EP (1) EP3211543A4 (zh)
JP (1) JP6234601B2 (zh)
KR (1) KR101779387B1 (zh)
CN (1) CN104408639A (zh)
WO (1) WO2016062255A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107911443A (zh) * 2017-11-14 2018-04-13 北京小度信息科技有限公司 一种会话信息处理方法、装置、服务器和可读存储介质
CN115168565A (zh) * 2022-07-07 2022-10-11 北京数美时代科技有限公司 一种垂直领域语言模型冷启动方法、装置、设备及存储介质

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408639A (zh) 2014-10-22 2015-03-11 百度在线网络技术（北京）有限公司 多轮会话交互方法和系统
CN106155978B (zh) * 2015-06-03 2019-11-01 上海红神信息技术有限公司 可重构系统的构建方法和装置
CN106326307A (zh) * 2015-06-30 2017-01-11 芋头科技（杭州）有限公司 一种语言交互方法
CN106528613B (zh) * 2016-05-26 2018-11-27 中科鼎富（北京）科技发展有限公司 智能问答方法及装置
CN106844499A (zh) * 2016-12-26 2017-06-13 网易（杭州）网络有限公司 多轮会话交互方法及装置
US11200501B2 (en) * 2017-12-11 2021-12-14 Adobe Inc. Accurate and interpretable rules for user segmentation
CN107992610A (zh) * 2017-12-15 2018-05-04 泰康保险集团股份有限公司 一种自动问答方法、装置、设备及存储介质
CN108153902A (zh) * 2018-01-16 2018-06-12 和美（深圳）信息技术股份有限公司 多轮会话交互方法、装置、计算机设备和存储介质
CN108804643A (zh) * 2018-06-05 2018-11-13 北京玄科技有限公司 多轮会话的流程配置单元、方法及智能机器人交互系统
CN109241256B (zh) * 2018-08-20 2022-09-27 百度在线网络技术（北京）有限公司 对话处理方法、装置、计算机设备和可读存储介质
JP2020052857A (ja) * 2018-09-28 2020-04-02 大和ハウス工業株式会社 推定システム及び推定方法
US11032217B2 (en) 2018-11-30 2021-06-08 International Business Machines Corporation Reusing entities in automated task-based multi-round conversation
CN109783627B (zh) * 2019-01-07 2023-06-30 平安科技（深圳）有限公司 自动应答的方法、装置、计算机设备及存储介质
US11164562B2 (en) 2019-01-10 2021-11-02 International Business Machines Corporation Entity-level clarification in conversation services
US11704591B2 (en) 2019-03-14 2023-07-18 Adobe Inc. Fast and accurate rule selection for interpretable decision sets
US20220138170A1 (en) * 2020-10-29 2022-05-05 Yext, Inc. Vector-based search result generation
CN112650854B (zh) * 2020-12-25 2022-09-27 平安科技（深圳）有限公司 基于多知识图谱的智能答复方法、装置及计算机设备
CN113239164B (zh) * 2021-05-13 2023-07-04 杭州摸象大数据科技有限公司 多轮对话流程构建方法、装置、计算机设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101639831A (zh) * 2008-07-29 2010-02-03 华为技术有限公司 一种搜索方法、装置及系统
CN103020052A (zh) * 2011-09-20 2013-04-03 北京百度网讯科技有限公司 一种识别搜索需求的方法和装置
CN103885947A (zh) * 2012-12-19 2014-06-25 北京百度网讯科技有限公司 一种搜索需求的挖掘方法、智能搜索方法及其装置
US20140258285A1 (en) * 2002-11-11 2014-09-11 Steven David Lavine User interface for search method and system
CN104408639A (zh) * 2014-10-22 2015-03-11 百度在线网络技术（北京）有限公司 多轮会话交互方法和系统

Family Cites Families (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5331554A (en) * 1992-12-10 1994-07-19 Ricoh Corporation Method and apparatus for semantic pattern matching for text retrieval
JP2001167087A (ja) 1999-12-14 2001-06-22 Fujitsu Ltd 構造化文書検索装置，構造化文書検索方法，構造化文書検索用プログラム記録媒体および構造化文書検索用インデックス作成方法
US20050234973A1 (en) 2004-04-15 2005-10-20 Microsoft Corporation Mining service requests for product support
US20060074980A1 (en) * 2004-09-29 2006-04-06 Sarkar Pte. Ltd. System for semantically disambiguating text information
US10002325B2 (en) * 2005-03-30 2018-06-19 Primal Fusion Inc. Knowledge representation systems and methods incorporating inference rules
GB0512435D0 (en) * 2005-06-17 2005-07-27 Queen Mary & Westfield College An ontology-based approach to information management for semantic music analysis systems
KR100820746B1 (ko) * 2007-01-22 2008-04-11 조선대학교산학협력단 온톨로지를 이용한 정보의 브라우징 시스템 및 방법
JP4521417B2 (ja) * 2007-03-16 2010-08-11 インターナショナル・ビジネス・マシーンズ・コーポレーション コンピュータの処理対象のオブジェクトが各ノードを成す、木構造により表現される、構造化電子文書を編集するためのコンピュータプログラム、方法及び装置
CN101639840A (zh) * 2008-07-29 2010-02-03 华天清 网络信息语义结构识别方法和装置
JP5153601B2 (ja) 2008-12-10 2013-02-27 ヤフー株式会社 検索結果分類装置、方法及びプログラム
CN101655862A (zh) * 2009-08-11 2010-02-24 华天清 信息对象搜索的方法和装置
US8538982B2 (en) * 2009-08-31 2013-09-17 Google Inc. Framework for selecting and presenting answer boxes relevant to user input as query suggestions
US8341099B2 (en) 2010-03-12 2012-12-25 Microsoft Corporation Semantics update and adaptive interfaces in connection with information as a service
US8738627B1 (en) * 2010-06-14 2014-05-27 Amazon Technologies, Inc. Enhanced concept lists for search
KR101192439B1 (ko) 2010-11-22 2012-10-17 고려대학교 산학협력단 디지털 콘텐츠 검색 장치 및 방법
CN102096717B (zh) * 2011-02-15 2013-01-16 百度在线网络技术（北京）有限公司 搜索方法及搜索引擎
CN102368246A (zh) * 2011-09-15 2012-03-07 张德长 一种自动应答机器人系统
KR20120088632A (ko) 2012-06-26 2012-08-08 고려대학교 산학협력단 디지털 콘텐츠 검색 장치 및 방법
US9355157B2 (en) * 2012-07-20 2016-05-31 Intertrust Technologies Corporation Information targeting systems and methods
CN102800006B (zh) * 2012-07-23 2016-09-14 姚明东 基于客户购物意图挖掘的实时商品推荐方法
CN103870506B (zh) * 2012-12-17 2017-02-08 中国科学院计算技术研究所 一种网页信息的抽取方法和系统
CN103593412B (zh) * 2013-10-24 2017-10-03 北京京东尚科信息技术有限公司 一种基于树形结构问题的应答方法及系统
WO2015078529A1 (en) 2013-11-29 2015-06-04 Huawei Technologies Co., Ltd. Transmission and receiving method in a wireless communication system
US9785705B1 (en) * 2014-10-16 2017-10-10 Google Inc. Generating and applying data extraction templates

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140258285A1 (en) * 2002-11-11 2014-09-11 Steven David Lavine User interface for search method and system
CN101639831A (zh) * 2008-07-29 2010-02-03 华为技术有限公司 一种搜索方法、装置及系统
CN103020052A (zh) * 2011-09-20 2013-04-03 北京百度网讯科技有限公司 一种识别搜索需求的方法和装置
CN103885947A (zh) * 2012-12-19 2014-06-25 北京百度网讯科技有限公司 一种搜索需求的挖掘方法、智能搜索方法及其装置
CN104408639A (zh) * 2014-10-22 2015-03-11 百度在线网络技术（北京）有限公司 多轮会话交互方法和系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3211543A4 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107911443A (zh) * 2017-11-14 2018-04-13 北京小度信息科技有限公司 一种会话信息处理方法、装置、服务器和可读存储介质
CN107911443B (zh) * 2017-11-14 2021-01-26 北京星选科技有限公司 一种会话信息处理方法、装置、服务器和可读存储介质
CN115168565A (zh) * 2022-07-07 2022-10-11 北京数美时代科技有限公司 一种垂直领域语言模型冷启动方法、装置、设备及存储介质

Also Published As

Publication number Publication date
KR101779387B1 (ko) 2017-09-18
JP6234601B2 (ja) 2017-11-22
US10817786B2 (en) 2020-10-27
CN104408639A (zh) 2015-03-11
EP3211543A4 (en) 2018-06-20
US20170308791A1 (en) 2017-10-26
JP2017511524A (ja) 2017-04-20
KR20160124765A (ko) 2016-10-28
EP3211543A1 (en) 2017-08-30

Similar Documents

Publication Publication Date Title
WO2016062255A1 (zh) 多轮会话交互方法、系统和计算机设备
US10877979B2 (en) Determining explanations for predicted links in knowledge graphs
CN108038183B (zh) 结构化实体收录方法、装置、服务器和存储介质
WO2021139701A1 (zh) 一种应用推荐方法、装置、存储介质及电子设备
WO2018072663A1 (zh) 数据处理方法及装置、分类器训练方法及系统、存储介质
CN110276075A (zh) 模型训练方法、命名实体识别方法、装置、设备及介质
CN109145303A (zh) 命名实体识别方法、装置、介质以及设备
US20150186535A1 (en) Determining an Active Persona of a User Device
CN110334347A (zh) 基于自然语言识别的信息处理方法、相关设备及存储介质
US20210350084A1 (en) Intention Identification Model Learning Method, Apparatus, and Device
CN107992585A (zh) 通用标签挖掘方法、装置、服务器及介质
CN111797222B (zh) 课程知识图谱构建方法、装置、终端及存储介质
CN109214674B (zh) 客服系统管理方法、客服系统及电子设备
US7840549B2 (en) Updating retrievability aids of information sets with search terms and folksonomy tags
WO2021017734A1 (zh) 一种实体消歧方法、装置、计算机设备及存储介质
WO2023040516A1 (zh) 一种事件整合方法、装置、电子设备、计算机可读存储介质及计算机程序产品
CN111159382B (zh) 会话系统知识模型的构建和使用方法及装置
US20230237277A1 (en) Aspect prompting framework for language modeling
US20230237787A1 (en) Techniques for dynamic time-based custom model generation
CN110196833A (zh) 应用程序的搜索方法、装置、终端及存储介质
CN111104436B (zh) 一种标签维值自调整方法及系统
CN113609379B (zh) 标签体系构建方法、装置、电子设备及存储介质
WO2023018479A1 (en) System and method for implementing federated learning engine for integration of vertical and horizontal ai
US20150193683A1 (en) Assisting an expert to answer preexisting questions in a time efficient manner
CN110781274A (zh) 一种问答对生成的方法与装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15853398

Country of ref document: EP

Kind code of ref document: A1

REEP Request for entry into the european phase

Ref document number: 2015853398

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2016552346

Country of ref document: JP

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 20167022267

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 15119348

Country of ref document: US