「地方」的探究─網路相片的分析取徑

Abstract

傳統區域地理學的研究，一般將區域視為「先驗」的存在，在已知的「容器」中闡述區域內部的各種資訊，因而簡化了區域的空間關係，到了新區域地理學，開始強調區域內「人」的活動以及對於地方的認知如何建構出區域的過程。然而近年來一些標榜新區域地理的作品，沿用傳統區域地理學將區域視為「容器」的做法，不探討區域邊界的問題，更談不上區域是一個建構的過程 (王秋原等，2007：6)。因此本研究嘗試藉由網路相片的標籤分析方法來擷取區域內部人群活動與地方認知的角色，並且重拾新區域地理學中邊界的議題與區域建構的過程。在研究方法上，網路相簿中帶有地名性質的地理標籤 (Geotag) 反映了人們對於此地方的認知概念，近來相關研究 (如Deng et al., 2009、Grothe and Schaab, 2009以及Wilske, 2008等) 認為可藉由分析這樣的集體情報 (collective intelligence) 來界定出非正式地名區域的可能範圍，而有助於進行地方性的研究；且若分析網路相片中的標籤則可以擷取在地特色與人群活動的相關訊息。因此，本研究應用幾何演算法─Alpha Shape，以地名「公館」為關鍵詞作為試驗對象，嘗試利用網路蒐集而來的地理標籤點位來界定出其空間範圍、建構出區域邊界。此外，本研究更近一步使用網絡分析 (Network Analysis) 模式，針對相片標籤進行屬性分析。研究結果顯示，Alpha Shape可根據相片記錄空間點位的群聚現象來劃界出記錄者對「公館」的認知範圍；另一方面，分析相片不同時間的點位分布，例如白天與夜晚的差異以及各年度的認知範圍變遷，則可呈現「公館」熱點分布及活動行為在時序上的變化；最後，網絡分析方式則是藉由標籤的共現帶出了地方上人地互動的特性與空間結構。本研究提供了一個新的分析方法來釐清人們的地方認知，進而展現地方的自明性，也提供新區域地理學另一種研究操作與整合空間分析傳統的基礎。

In the past, regional geographers used to think region as a “container”, and regional research had been simplified as the description of place. Therefore, new regional geographers embedded human activities and place cognition into the definition of the spatial relationship and the process of place’s construction. Recently, however, most of the new regional geographical studies focus on theories and constructs, neglecting the basic issue of boundaries problems and the process of region’s construction. This paper tries to use the analysis of geotags to solve the problem of issue. Recently web-techniques allow worldwide users to upload and share their photos, and the users can put photos’ geotags (photos’ descriptions and predefined geographic coordinates) onto public cyberspaces. Studies have considered that a collection of geotags not only reflects people’s “spatial cognition”, but also involves the concept of “collective intelligence”, which is helpful to meaningfully define physical boundaries for places. The challenge to delineate those physical boundaries is the complication of geotags’ topology in geographic extent. This study integrated Alpha Shape, an algorithm for topological analysis, and web-photos to delineate the boundary of a place. The test was performed by collecting web-photos and geotags of a custom name, “Gongguan”, through internet. In addition, Network analysis was also applied to extract local-characteristics of Gongguan based on a set of predefined attribute categories. The result shows that the boundary of Gongguan can be successfully generated by integrating Alpha Shape and collecting geotags through calibration of Alpha Shape, and most frequently identified attributes are food, NTU…etc. However, derived boundaries and attributes vary with time, such as day and night. This study provides an approach to explore the identity of a place and its dynamics of spatial cognition and interaction with local-characteristics in time and space, and it also provides the new regional geography to integrate the regional and spatial analyses.