基于非受限路径自然语言处理中的机器人导航

PDF下载HTML

Title:

Robot navigation based on non-restricted route natural language processing

作者:

张珂, 陈奇

华北电力大学 电子与通信工程系, 河北 保定 071003

Author(s):

ZHANG Ke, CHEN Qi

Department of Electronic and Communication, North China Electric Power University Code, Baoding 071003, China

关键词:

语义角色标注; 路径自然语言; 语块分析; 依存句法分析; 机器人问路导航

Keywords:

semantic role labeling; route natural language; chunking; dependency parsing; robot navigation by asking the way

分类号:

TP18

DOI:

10.11992/tis.201607016

摘要:

为了实现使用自然语言控制机器人完成自主导航任务，提出一种基于语义角色标注（SRL）的语义提取方法，用于提高机器人对路径自然语言理解的准确率。首先，收集了一个非受限的路径自然语言语料库，在深入研究路径自然语言语料库的基础上，提出了8个语块对语料进行语块分析，完成语义角色标注；然后，对语料进行依存句法分析，完成语义角色标注；接着，结合语块分析和依存句法分析，提出了一种基于语块分析和依存句法分析的语义角色标注方法，实验结果得到的准确率、召回率、F₁-值分别达到了98.22%、98.48%和98.35%；最后，基于语义提取结果在机器人Nao平台上完成了机器人问路导航任务。

Abstract:

In order to use a natural-language controlled robot to complete autonomous navigation, this paper proposes a semantic extraction method based on semantic role labeling (SRL), which is used to enhance the understanding of route natural language for robots. First, a non-restricted route natural language corpus is constructed. Then, based on thorough study of this corpus, eight chunks are put forward to analyze the data, which lays the foundation for SRL. Finally, an SRL based on chunking and dependency parsing is proposed. The experimental results of precision, recall, and F₁-measurements were 98.22%, 98.48%, and 98.35%, respectively. Using the proposed SRL method, the robot Nao was able to accomplish navigation tasks by asking the way.