Generating Distributed Representation of User Movement for Extracting Detour Spots
Generating Distributed Representation of User Movement for Extracting Detour Spots
复制标题
DOI：
10.1145/3297662.3365826
复制
发表时间：
2019-11
期刊：
影响因子：
--
通讯作者：
Masaharu Hirota;Tetsuya Oda;Masaki Endo;H. Ishikawa
中科院分区：
文献类型：
--
作者：
Masaharu Hirota;Tetsuya Oda;Masaki Endo;H. Ishikawa
由于内置全球定位系统传感器的移动的设备日益普及，大量包含空间信息的用户生成内容已被上传到Flickr或Twitter等社交媒体网站。从旅游景点发布的内容可以用于搜索和推荐其他旅游景点和路线。自然语言处理（NLP）领域的最新研究论文提出了使用嵌入算法学习单词的分布式表示。在本文中，我们使用Skip-gram模型来分析从社交媒体网站获得的用户移动。我们提出了一个新的跳过grammemodel学习的纬度和经度量化的一对位置之间的运动。我们的模型的嵌入向量表示用户在位置之间的移动。我们成功地证明了我们所提出的方法生成的嵌入式向量可以提取迂回点的旅游景点和路线。
Owing to the increasing popularity of mobile devices embedded with a Global Positioning System (GPS) sensor, large amounts of user-generated content containing spatial information have been uploaded to social media websites, such as Flickr or Twitter. That content posted from tourist spots can be used to search for and recommend other tourism spots and routes. Recent research papers in the field of Natural Language Processing (NLP) have proposed learning a distributed representation of words using embedding algorithms. In this paper, we use a Skip-gram model to analyze user movements obtained from social media websites. We propose a new Skip-gramgemodel to learn movements between a pair of locations quantified by the latitude and the longitude. The embedding vectors by our model represents user movements between the locations. We successfully demonstrated that the embedded vectors generated with our proposed method can extract detour spots for tourism spots and routes.