BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.osgeo.org//foss4g-2026//talk//LGWHNG BEGIN:VTIMEZONE TZID:JST BEGIN:STANDARD DTSTART:20000101T000000 RRULE:FREQ=YEARLY;BYMONTH=1 TZNAME:JST TZOFFSETFROM:+0900 TZOFFSETTO:+0900 END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-2026-LGWHNG@talks.osgeo.org DTSTART;TZID=JST:20260902T163000 DTEND;TZID=JST:20260902T170000 DESCRIPTION:The Kaii-Yokai Densho Database\, maintained by the Internationa l Research Center for Japanese Studies (Nichibunken)\, a\n Japanese natio nal research institute for Japanese studies\, is one of the largest struct ured archives of Japanese yokai\n folklore. Each record contains a yokai or phenomenon name\, a prefecture field\, source metadata\, and a short te xtual\n summary. These fields make the archive searchable by name or regi on\, but they do not directly support geographic\n interpretation. Folklo re records often refer to place in ways that are difficult to reduce to a single coordinate. A\n summary may contain a formal place name\, but it m ay also describe a riverbank\, road\, house\, grave\, pass\, shrine\,\n v illage edge\, or boundary without giving a mappable toponym. This creates a methodological problem for geographic\n analysis: treating every record as a point can hide the uncertainty and evidence structure that make the record\n spatially interpretable.\n\n This study proposes a Folklore Geo spatial Support Model for representing vague folklore place descriptions i n the\n Kaii-Yokai Densho Database. The model shifts the analytical unit from an asserted event coordinate to the set of\n evidence channels that support spatial interpretation. These channels include prefecture or munic ipality support\,\n formal place mentions\, place-description terms\, coo rdinate-derived terrain evidence\, text-derived terrain cues\, human-\n c ondition terms\, boundary-interface types\, source metadata\, resolution l evel\, and confidence reason. In this\n representation\, a coordinate rem ains useful for display and spatial computation\, but it is not treated as the whole\n geographic meaning of the record. The model keeps administra tive metadata\, gazetteer-resolvable place names\, vague\n place descript ions\, and GIS-derived polygon context separate so that each record can be inspected according to the\n evidence it actually supplies.\n\n The wor kflow transforms 33\,378 records from the Kaii-Yokai Densho Database into resolution-aware geospatial features.\n First\, all records receive prefe cture-level support from the structured prefecture field. This creates com plete\n national coverage and provides a baseline for later refinement. S econd\, Japanese named entity recognition and rule-\n based extraction ar e applied to the summary field. Formal place mentions are stored separatel y from terrain and place-\n description terms because many folklore summa ries describe settings without naming a municipality or landmark. Third\,\ n a conservative local GADM level 2 gazetteer refinement searches for mun icipality-like names within each record’s\n prefecture. A record is upg raded only when exactly one local municipality candidate is found. Fourth\ , GIS enrichment\n adds coastline\, lake\, and river context using GADM\, Natural Earth\, and National Land Numerical Information River Data\n (W0 5). The workflow also computes coordinate-only and text-aware terrain labe ls separately\, so that GIS thresholds and\n summary-derived terrain cues are not merged into a single evidential category.\n\n The resulting cove rage shows why folklore geography cannot rely only on formal toponym resol ution. All 33\,378 records\n receive prefecture-polygon support. Formal p lace mentions occur in 9\,361 records\, or 28.1% of the archive. Place-\n description terms occur in 25\,023 records\, or 75.0%. Terrain terms occu r in 20\,242 records\, human-condition terms in\n 16\,066 records\, and s trict boundary-interface terms in 9\,047 records. Local GADM admin2 refine ment upgrades 1\,231\n records\, or 3.7% of the archive\, to municipality -level representative points. Although this subset is small\, it is\n ana lytically useful because it provides a diagnostic set for measuring how mu ch prefecture-centroid representation\n changes environmental interpretat ion.\n\n The admin2 subset reduces support area by a median 96.4% relativ e to prefecture support. It also shows that coarser\n coordinates can cha nge coordinate-derived terrain interpretation. Among the 1\,231 locally re fined records\, 528\n records\, or 42.9%\, change coordinate-only terrain class when the record is represented by an admin2 representative\n point rather than by the prefecture centroid. The most common changes are coast al to inland-water\, coastal to plain\,\n plain to inland-water\, valley to inland-water\, and plain to coastal. This result is not an archive-wide accuracy\n estimate\, because the refined subset is produced by successf ul local name matching. Its value is diagnostic: it\n demonstrates that p refecture-centroid maps can change the environmental reading of records th at contain enough\n information for municipality-level support.\n\n The study also evaluates place-function vocabulary under a prefecture-stratifi ed permutation model. Category labels\n are shuffled within prefectures\, preserving regional composition while breaking category-specific associat ion with\n extracted place-function terms. This tests whether category la bels are associated with particular kinds of setting\n descriptions beyon d prefectural background composition. The strongest associations include K appa\, or water spirits\,\n with water-setting vocabulary\; Yurei\, or gh osts\, with death-ritual vocabulary\; and Snake/Dragon\, or serpentine bei ngs\,\n with boundary-interface evidence. Observed-to-expected ratios pro vide interpretable effect summaries: Yurei/death-\n Dragon/water-setting vocabulary at 2.23 times\, and Snake/Dragon/boundary-interface vocabulary at 1.47 times. These\n results describe archive language and evidence str ucture rather than verified physical event locations.\n\n 0.371 km. A pre fecture-stratified category-shuffle null with 10\,000 permutations expects a median of 0.295 km. The\n observed median is larger than the shuffle e xpectation\, so the Kappa result is treated as summary evidence rather tha n\n a demonstrated physical water-proximity effect. This separation is ce ntral to the model: text-derived place-function\n associations and GIS-de rived proximity measurements are related but distinct forms of evidence.\n \n Robustness diagnostics further test whether the strongest associations depend on category-name leakage or dictionary\n design. Category-specifi c surface forms such as kappa\, tengu\, kitsune\, yurei\, and snake/dragon terms are masked from\n summaries before extraction. Dictionary sensitiv ity is tested by randomly removing 10%\, 20%\, and 30% of terms from\n ea ch dictionary group across seeded trials. Boundary-interface definitions a re also varied across broad\, strict\, and\n conservative versions. The f ocal associations remain positive across these diagnostics\, although thei r magnitudes\n change. This supports the use of place-function and bounda ry-interface channels as inspectable evidence fields within\n the archive representation.\n\n The main contribution is a resolution-aware GIS repr esentation for vague folklore places. The model preserves\n uncertainty i nstead of forcing all records into single asserted coordinates. It also sh ows that coarse spatial support\n is not only imprecise\; in the locally refined subset\, it can alter coordinate-derived environmental interpretat ion. By\n separating formal toponyms\, place-description vocabulary\, coo rdinate-derived terrain\, text-derived terrain\, and\n boundary-interface evidence\, the workflow makes the geographic structure of a large folklor e archive inspectable while\n keeping the evidential basis of each record visible. DTSTAMP:20260717T225758Z LOCATION:Cosmos2 SUMMARY:Geographic Visualization of the Kaii-Yokai Folklore Database Using Open-Source GIS and NLP - Kosuke Shimizu\, Hiroki Ichikura URL:https://talks.osgeo.org/foss4g-2026/talk/LGWHNG/ END:VEVENT END:VCALENDAR