基于自然语言空间关系描述的地图近似表达方法
作者简介:曹 青(1992-),女,硕士生,主要从事空间认知与地图可视化。E-mail: 2454295083@qq.com
收稿日期: 2018-06-19
要求修回日期: 2018-10-03
网络出版日期: 2018-11-20
基金资助
国家自然科学基金项目(41571382、61472191)
江苏省高校自然科学研究重大项目(15KJA420001)
Map Approximate Expression Method Based on Spatial Relationship Description in Natural Language
Received date: 2018-06-19
Request revised date: 2018-10-03
Online published: 2018-11-20
Supported by
National Natural Science Foundation of China, No.41571382, 61472191
College Natural Science Research Key Program of Jiang Su Province, No.15KJA420001.
Copyright
自然语言和地图都具备表达地理实体空间关系的能力,自然语言使用方便、抽象化程度高,而地图更为直观,从自然语言转换到地图,有助于人们更深入地了解自然语言描述的地理实体空间关系。然而,如何让计算机具有从自然语言构建图形信息的能力,使计算机具有智能化空间认知思维是当前研究的难点。本文总结了自然语言空间关系描述的类型及特点,提出了基于自然语言描述的地理实体抽象表达方法以及空间关系近似转换方法,建立了一种基于自然语言空间关系描述的地图近似表达策略。实验结果表明,本文方法有效可行,能够实现定性描述的自然语言空间关系向定量(或近似定量)的图形空间关系的转换,为自然语言到地图的转换研究奠定了基础。
曹青 , 洪必文 , 张翎 , 阮陵 , 龙毅 . 基于自然语言空间关系描述的地图近似表达方法[J]. 地球信息科学学报, 2018 , 20(11) : 1541 -1549 . DOI: 10.12082/dqxxkx.2018.180288
With the further development of mobile GIS, intelligent GIS and socialized GIS, the geospatial information service based on natural language processing is an inevitable trend in the field of geographical information science. The intelligent conversion from text to map is one of the important research directions. Both natural language and maps have the ability to express spatial relationship of geographical entities. Natural language has the natural characteristic of usability and is highly abstract, while map language is more intuitive and revealing. The ubiquitous natural language contains a great deal of geographic information. Converting natural language to map language can help people intuitively understand the geographic space environment and bring out new discoveries. The current research difficulties focus on that how to make a computer construct graphical information from natural language and have the intelligent spatial cognitive thinking ability. This paper proposes a method that using point coordinated pairs, straight line segments and rectangular/circular shapes to quantitatively represent point, polyline and polygon geographical entities in natural language respectively. First the spatial relations description types in natural language between point and point, point and line, point and surface, line and line, line and surface, surface and surface geographic entities are summarized. Second, approximate transformation model of spatial relationships in natural language which considering the geometric types of geographical entities is constructed, and an approximate expression strategy based on spatial relationships description in natural language is proposed. Third, a prototype system is designed to implement "text-map" conversion, and scenic spot travel notes are selected as the experimental text to finish the experiment. The experimental results showed that the method mentioned above was feasible, the goal that converting qualitative spatial relationships in natural language to quantitative (or approximately quantitative) graphical spatial relationships could be achieved. This paper lays a foundation for the study of the conversion from natural language to map.
Fig. 1 Relationship model of “Between reference object A and B”图1 “在参照物A和B之间”关系模型 |
Fig. 2 “Direction Relation” model图2 “方向关系”模型 |
Fig. 3 “Distance Relation” model图3 “距离关系”模型 |
Fig. 4 “Direction and Distance Relation” model图4 “方向+距离关系”模型 |
Fig. 5 “Inclusion Relation” model图5 “包含关系”模型 |
Fig. 6 "Meet Relation" model图6 “相接关系”模型 |
Fig. 7 “Intersection Relation” model图7 “相交关系”模型 |
Tab. 1 Map approximate expression case表1 地图近似表达案例 |
类型 | 示例 | 解析结果 | 模拟表 达结果 | ||
---|---|---|---|---|---|
参照物 | 空间关系 | 目标物 | |||
拓扑关系 | 长廊位于昆明湖和万寿山之间 | 昆明湖、万寿山 | 在……之间 | 长廊 | ![]() |
方向关系 | 宜芸馆位于玉澜堂北面 | 玉澜堂 | 北面 | 宜芸馆 | ![]() |
距离关系 | 仁寿殿距离东宫门很近 | 东宫门 | 很近 | 仁寿殿 | ![]() |
多种空间关系组合 | 昆明湖中有一座南湖岛,由一座美丽的十七孔桥和东岸相连,十七孔桥东侧有个廓如亭,它与十七孔桥及南湖岛连接在一起。 | 昆明湖 | 包含+南 | 南湖岛 | ![]() |
南湖岛 | 相接+东 | 十七孔桥 | |||
十七孔桥 | 相接+东 | 廊如亭 |
Fig. 8 Comparison between simulated results and guide map图8 模拟表达结果与导览图的对比 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
[
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
[
|
[14] |
|
[15] |
|
[16] |
|
[17] |
[
|
[18] |
[
|
[19] |
[
|
[20] |
|
[21] |
|
[22] |
|
[23] |
|
[24] |
|
/
〈 |
|
〉 |