Feedback in Conversation as Incremental Semantic Update
Feedback in Conversation as Incremental Semantic Update
复制标题
DOI：
10.3115/v1/w15-0130
复制
发表时间：
2015-04
期刊：
影响因子：
--
通讯作者：
Arash Eshghi;C. Howes;E. Gregoromichelaki;Julian Hough;Matthew Purver
中科院分区：
文献类型：
--
作者：
Arash Eshghi;C. Howes;E. Gregoromichelaki;Julian Hough;Matthew Purver
在交谈中，对话者通常会指出所说或所做的事情是否已经被处理和整合。这种反馈包括诸如‘OK’或‘MHM’之类的反向渠道，下一个相关回合的产生，以及通过澄清请求启动修复。重要的是，这样的反馈不仅可以在句子/话轮边界上产生，也可以在次句中产生。在本文中，我们基于动态句法(DS)语法框架，扩展了现有的对话中的增量语义处理模型，以提供对回溯通道、澄清请求及其响应的低层次、综合的描述；论证了它们可以被解释为语法的核心语义结构构建机制的一部分，而不是通过意图识别等更高层次的语用现象，或者被视为对话的一部分。最终结果是一个增量模型，在该模型中，单词而不是话轮被视为上下文更新的过程，反向通道用于调整参与者的语义处理上下文，从而简化后续会话动作的产生和解释。我们还展示了如何在同一DS框架内模拟澄清请求及其随后的响应和修复，其中参与者语义加工中的分歧和重新排列顺序推动对话向前推进。
In conversation, interlocutors routinely indicate whether something said or done has been processed and integrated. Such feedback includes backchannels such as ‘okay’ or ‘mhm’, the production of a next relevant turn, and repair initiation via clarification requests. Importantly, such feedback can be produced not only at sentence/turn boundaries, but also sub-sententially. In this paper, we extend an existing model of incremental semantic processing in dialogue, based around the Dynamic Syntax (DS) grammar framework, to provide a low-level, integrated account of backchannels, clarification requests and their responses; demonstrating that they can be accounted for as part of the core semantic structure-building mechanisms of the grammar, rather than via higher level pragmatic phenomena such as intention recognition, or treatment as an “uno cial” part of the conversation. The end result is an incremental model in which words, not turns, are seen as procedures for contextual update and backchannels serve to align participant semantic processing contexts and thus ease the production and interpretation of subsequent conversational actions. We also show how clarification requests and their following responses and repair can be modelled within the same DS framework, wherein the divergence and re-alignment e ort in participants’ semantic processing drives conversations forward.