Empirical Methods in Natural Language Generation
Empirical Methods in Natural Language Generation
复制标题
自然语言生成中的经验方法
DOI：
10.1007/978-3-642-15573-4_4
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
Janarthanam S
中科院分区：
文献类型：
--
作者：
Janarthanam S
我们解决的问题，不同的用户有不同的词汇知识的问题域，使自动对话系统需要适应他们的一代选择在线用户的领域知识，因为它遇到他们。我们使用马尔可夫决策过程（MDP）中的强化学习来解决这个问题。我们提出了一个强化学习框架来学习自适应引用表达式生成（REG）策略，可以动态地适应不同领域知识水平的用户。在相关工作的对比，我们还提出了一个新的统计用户模型，它结合了不同用户的词汇知识。我们评估这个框架表明，它使我们能够学习对话的政策，自动适应他们的选择，指表达在线不同的用户，这些政策是显着优于手工编码的自适应政策，这个问题。学习的策略始终比一系列不同的手工编码但自适应的基线REG策略短2到8圈。
We address the problem that different users have different lexical knowledge about problem domains, so that automated dialogue systems need to adapt their generation choices online to the users’ domain knowledge as it encounters them. We approach this problem using Reinforcement Learning in Markov Decision Processes (MDP). We present a reinforcement learning framework to learn adaptive referring expression generation (REG) policies that can adapt dynamically to users with different domain knowledge levels. In contrast to related work we also propose a new statistical user model which incorporates the lexical knowledge of different users. We evaluate this framework by showing that it allows us to learn dialogue policies that automatically adapt their choice of referring expressions online to different users, and that these policies are significantly better than hand-coded adaptive policies for this problem. The learned policies are consistently between 2 and 8 turns shorter than a range of different hand-coded but adaptive baseline REG policies.