【24h】

A Reinforcement Learning Parser for Spoken Language

机译:口语强化学习解析器

获取原文
获取原文并翻译 | 示例

摘要

In this paper, we introduce a new parser that can manage spoken language. The parser consists of two parts: deep structure and surface structure following Chomskian philosophy. Candidates of predicted sentence are generated by the deep structure and tested on the surface structure with the incoming signal in the form of phoneme sequence. The result is fed back to the deep structure so that the best candidate is chosen and the grammar and the vocabulary can be adjusted via reinforcement learning. As an implementation, we realized a tagger for spoken Korean where morphological rules are abundant and the parts of speech may be freely ordered.
机译:在本文中,我们介绍了一种可以管理口语的新解析器。解析器由两部分组成:遵循乔姆斯基哲学的深层结构和表面结构。预测句子的候选由深层结构生成,并在表面结构上通过音素序列形式的传入信号进行测试。结果被反馈到深层结构,以便选择最佳候选者,并且可以通过强化学习来调整语法和词汇。作为一种实现,我们实现了一种针对朝鲜语的标记器,该标记器具有丰富的形态规则,并且可以自由地对词性进行排序。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号