DGRP Session II Response

Mike Dobson (Telemapics)

  • you don't always get what you want
  • interest: gazetteers from the point of view of local search
  • how should a gazetteer function?
    • mixtures of semantic patterns occur geographically, but we don't always account for them meaningfully
    • text string matching vs. geographically-aware search
      • targeting ad revenue
    • word tracker (context: search engine optimization)
      • top 1000 terms searched on internet daily - no geographical name in this group
      • beyond website optimization, frequency of keywords on internet may have some meaning for gazetteers?
    • surrogate measures
      • number of competing websites
      • frequency of use of key (geographic) words may be a useful feature
    • why geographical errors are often overlooked in local search
      • these pale in comparison to failures on other axes
        • try a "local search" for fish tacos in zip 92653
    • search companies are buying gazetteer companies
      • definition of neighborhood names and how do you find and exploit them
    • easy access to names tends to be limited to administrative areas
      • the most important name for marketers (and geo researchers?) are functional, not administrative, names - only reinsurers, title insurers and real estate companies know them

May Yuan (University of Oklahoma)

  • personal names vs. placenames
    • personal names have meaning, but these are well defined in most cases
    • place names are complicated ... they change temporally, spatially, name itself
      • identity is much more slippery with place than with person
      • global ids attach to concepts -- how do we know that two concepts are duplication, or variants, or ... ?
      • linguistic difficulties
    • what really constitutes place names
      • geographical knowledge embedding the placenames; geo knowledge is implicit in the cultural, historical and locative contexts
      • spatial component
      • temporal component
      • all 3 of these
    • place names are composite constructs
    • we don't just want to know about place names, or just their associated locations, but we want access to stories etc. behind them
    • when we build gazetteers
      • names
      • types -- ontology - build linkages to work in this field
      • footprints
    • wikipedia
      • a good way to solicit local knowledge ... but
      • should not used as a mean for building for gazetteers; rather, as a means for data collection
      • we still need an organized research group to put the harvested results into a more structured and consistent format
    • geocaching as a model
      • why don't we have similar types of games for placenames to get people more involved in a more active way in raw data collection
    • need knowledge model that is adaptable to different forms of placename knowledge
    • controversial points for sake of discussion:
      • georeferencing placenames: accuracy more important than precision
        • better to direct user to a vague, general location than to very precisely point them to the wrong place entirely
        • even if we don't have coordinates for the particular name in our gazetteer, but we do know the relationship to a broader feature, highlight that feature
      • relative location more important than absolute location
        • most users do not care about exact lat/lon, but they are interested in whether the location is near a river, or atop a mountain
      • implicit local meaning and stories for a placename of interest are mor important than the explicit ones
        • which country this place is in is less important than the location of major cities along a highway
      • place information provided by the gazetteer needs to be scale dependent
      • objects should be used as metadata for fields and images
        • bounding boxes and polygons can be used as metadata to refer to location of a field or item of interest or images to retrieve; vector data as metadata for image or field data
      • when does folk knowledge (e.g., web questionnaire results or field data) become common knowledge to be considered acceptable; should everyone's opinions be added to the "core" gazetteer?

summary: Bodenhamer

  • things are always more complex than they seem
  • web = sandbox, but what are the methods for using that sand
  • if building for everyone, are we building for no-one? what's our purpose in creating gazetteers?
  • do we need a scale-dependent knowledge model?