Handle ranges per guidelines

[i-be-snek]

Introduces:

• Ranges as discussed in the guidelines word document (copied directly from the document):

  • Greater than/more than/exceed 700: 701-799 (7 * scale + 1, (7+1) * scale – 1)
  • Greater/more than/over 640: 641-699 (6.4 * scale+1, (6+1) * scale – 1)
  • Less/lower/fewer than 700: 601-699 (6 * scale +1, (6+1) * scale – 1)
  • At least/a minimum of 630: 630-699 (6.3 * scale, (6+1) * scale – 1)
  • Up to 270: 200-270 (2 * scale – 2.7 * scale)
  • Approximately/around/nearly/about/almost/roughly/~ 700: ±5%: 665-735 for min and max (700 * 0.95, 700 * 1.05)
  • Dozens of, tens of, hundreds of, thousands of, etc: 2 * scale, 9 * scale; so, “thousands of injuries” becomes 2000-9000 injuries.
  • A number of/a group of/a few/several: 2-6
  • A few dozen: 24-72 (12 * 2, 12 * 6)
  • A dozen hundreds (if it ever appears!): 2400-7200 (12 * 2 * scale, 12 * 6*scale)
  • A few/several hundred/thousand/million, etc: 2 * scale-6 * scale; so “several millions” becomes 2000000-60000000
  • Many: 20-60
  • SPECIAL CASE: A couple/a couple hundred/thousand, etc...: 2 * scale – 3 * scale; so “a couple of deaths” becomes 2-3 deaths
  • SPECIAL CASE: If the number of human victims is reported as “family/families”, multiply by 5 to determine the number of human victims. For “families”, use 2 * 5-9 * 5 if there is no number before families. For example: “5 families are displaced” = 5 * 3-5 * 5 = 15-25 people displaced.
  • SPECIAL CASE: no causalities, no fatalities, no injuries, none, none reported, and similar phrases must be annotated as 0. If the information is missing, the annotator must enter NULL instead.

• Other guidelines:

  • m/mil -> million and similar
  • expand "approximation" synonym list

• Tests for normalizing numbers

[i-be-snek]

Note to self: add these as undesired location type matches:

[
"cemetery",
"church",
"orchard",
]

[liniiiiii]

Well checked for these normalization rules