RavenClaw介绍_005_对话任务规范_层次化表示

总结

层次化表示的优点

the advantage of hierarachical representation,

1、子组件独立:设计容易、维护容易、扩展容易
2、动态构建:运行时动态扩展
3、捕获上下文:父子关系、动作左右顺序
4、简化对话渲染引擎设计

这种方法来自于robotics,就是机器人学,被借鉴到了对话管理领域;


对话树对话任务
对话agent对话子任务

RavenClaw
Agenda Communicatortree-of-handlers:处理器树
WITAsactivity trees:活动树
Collagenrecipes:配方

2.1. The Dialog Task Specification - 对话任务规范

The domain-specific dialog control is represented in the Dialog Task Specification level using a tree of dialog agents, with each agent handling a certain part of the dialog task.

特定领域的对话控制对话任务规范级别中通过对话代理的树形结构来表示,每个代理处理对话任务的某个部分

对话树当中有很多的对话agents,对话agents代表对话子任务

对话树对话任务
对话agent对话子任务

To exemplify, Figure 1 illustrates the top portion of the dialog task tree for RoomLine, a spoken dialog system for conference room reservation and scheduling.

为了举例说明图1展示了RoomLine的对话任务树的顶部部分,RoomLine是一个用于会议室预订和调度的口语对话系统。

The root node subsumes several children: Login, which identifies the user to the system, GetQuery, which obtains the time and room constraints from the user, GetResults, which executes the query against the backend, and DiscussResults which presents the obtained results and handles the forthcoming negotiation for selecting the conference room that best matches the user’s needs.

根节点包含几个子节点

登录,用于将用户识别给系统;

获取查询,用于从用户处获取时间和房间约束;

获取结果,用于将查询执行到后端

以及讨论结果,用于展示获得的结果并处理即将进行的协商,以选择最符合用户需求会议室

Moving one level deeper in the tree, Login decomposes into Welcome, which introduces the user to the system, AskRegistered and AskName, which identify the user, and finally GreetUser, which sends a greeting to the user.

在树中再深入一层,登录分解为

  • 欢迎,用于向用户介绍系统;
  • 询问是否已注册询问姓名,用于识别用户;
  • 最后是向用户致意,用于向用户发送问候。

image-20240821115744583


Hierarchical task decompositions, traditionally used for task execution in robotics, have gained popularity in the dialog management community.

层次化任务分解,传统上用于机器人学中的任务执行,现已在对话管理领域中广受欢迎。

Examples include the use of a tree-of-handlers in Agenda Communicator 1, activity trees in WITAS 2 and recipes in Collagen 3.

示例包括在Agenda Communicator1中使用的处理器树、在WITAS2中使用的活动树和在Collagen3中使用的配方

Xu, W. and Rudnicky, A.,

“Task-based Dialog Management Using an Agenda”,

ANLP/NAACL 2000 Workshop on Conversational Systems, May 2000.

############################

徐伟(Xu, W.) 和 鲁德尼克(Rudnicky, A.),

“使用议程的任务型对话管理”,

ANLP/NAACL 2000 对话系统研讨会,2000年5月。

Lemon, O.,

Gruenstein, A.,

Battle, A.,

and Peters, S.

“Multi-tasking and Collaborative Activities in Dialogue Systems”,

Proceedings of 3rd SIGDIAL Workshop on Discourse and Dialogue, Philadelphia, p.113-124, 2002.

########################

莱蒙(Lemon, O.),

格鲁斯坦因(Gruenstein, A.),

巴特尔(Battle, A.),

彼得斯(Peters, S)。

“对话系统中的多任务处理和协作活动”,

第三届SIGDIAL话语与对话研讨会论文集,费城,第113-124页,2002年。

Rich, C.,

and Sidner, C.,

“Collagen: A Collaboration Agent for Software Interface Agents”,

Journal of User Modeling and User-Adapted Interaction, vol.8, 1998

###############################################

里奇(Rich, C.),

西德纳(Sidner, C.),

“Collagen:一种用于软件界面代理的协作代理”,

《用户建模与用户适应交互》杂志,第8卷,1998年


The hierarchical representation has several advantages.

层次化表示具有多个优点

Most (goal-oriented) dialog tasks have an identifiable structure which naturally lends itself to a hierarchical description.

大多数(面向目标的)对话任务具有可识别的结构,这种结构自然适用于层次化描述

The subcomponents are typically independent, leading to ease in design and maintenance, as well as good scalability properties.

子组件通常是独立的,这使得设计和维护变得更加容易,同时也具有良好的可扩展性

Moreover, the tree structure can be easily extended at run-time, allowing for the dynamic construction of dialog structure, a very useful feature in certain types of tasks.

此外,树形结构可以在运行时轻松扩展,允许动态构建对话结构,这对于某些类型的任务而言是非常有用的特性。

Finally, the tree representation implicitly captures the notion of context (via the parent relationship), as well as a default chronological ordering of the actions (i.e. left-to-right traversal); these elements significantly simplify the design of a dialog rendering engine operating over this type of representation.

最后,树形表示隐式地捕获了上下文的概念(通过父子关系),以及动作的默认时间顺序(即从左到右的遍历);这些元素极大地简化了操作这种类型表示的对话渲染引擎的设计

As a prerequisite to a more detailed presentation of the Dialog Engine control mechanisms (Section 2.2), we first turn our attention to the structure and functionality of the dialog task agents.

作为对话引擎控制机制(第2.2节)的更详细介绍先决条件,我们首先将注意力转向对话任务agents结构和功能。


  1. 第一篇引用; ↩︎ ↩︎

  2. 第二篇引用; ↩︎ ↩︎

  3. 第三篇引用; ↩︎ ↩︎

  • 10
    点赞
  • 6
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值