5 Matching Annotations
  1. Aug 2018
    1. Through spatio-temporalfiltering of messages we are able to observeevolution of topic signal that is consistent with rumor theory’s predictions. In the area immediately surrounding Moore wefind the strongest evidence that rumoring is an evolutionary process, as the topic and tone of messages shifts notably over thisthree-day span. Although these results may not necessarily generalize to all events, we dofind some initial support thatrumoring is a process marked by topical evolution over time. Once again, the spatio-temporalfiltering approach proves to berobust tool for measuring signal of hazard-related rumoring.

      Summary findings.

    2. While the counts of messages in the 100e200 and 700e800 mile bins are nearly identical, the latter binincludes the Atlanta metro area and much of the Chicago metro area, two of the top-ten largest metro areas in the country. Asillustrated inTable 1, the counts of messages vary by location based on the populations in those locations. Accordingly, rawcounts of“tornado”messages will be influenced by the population in each bin. In many contexts one would want to accountfor spatial heterogeneity in population by establishing a baseline measure of activity in each bin

      Explanation of the geographic bin filtering and some potential limitations.

    3. Using what is known about social responses to disaster events(impending or realized), we selectivelyfilter timestamped streams of geolocated, informal communication activity by timeand location in order to identify surges of rumoring activity in response to a disaster. Spatio-temporalfiltering enhances ourability to detect events by utilizing the signal produced by sources that are known (or expected) to produce reliable infor-mation, thereby enhancing our ability to detect distinct activity patterns above and beyond typical global signal (i.e. back-ground noise representing the array of signals irrelevant to our focus)

      Filtering technique relies on timestamp and geolocation.

      Per Sloan (2015), only 0.85% of tweets are geotagged (approx. 4M tweets per day)

      http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0142209

    4. Tobler’s First Law of Geography states that“Allthings are related, but nearby things are more related than distant things”(Tobler, 1970). Barring sudden, drastic changes tothe landscape, new information from a specific locale should be consistent with what we already know about that location aswell as what we know about nearby locations

      This is the first time I've seen this cited in crisis informatics research. Get this paper.

  2. Feb 2016
    1. Searches for tortoise food web and area meteorological data in the region at the DataONE portal. Searches for land-use histories, especially for former grazing lands. Searches for co-locality data for other animal species as possible signals for other ecological changes in the region.