fv/lib/match_checkboxes.js

FormVision is a node.js library for extracting data from scanned forms

// Generated by CoffeeScript 2.3.1
var distance, estimateTransform, findClosestMark, findClosestShortWord, matchByMark, matchByWords, unpack, validate,
  indexOf = [].indexOf;

({unpack, validate} = require('./schema'));

({estimateTransform} = require('./estimate_transform'));

({distance} = require('./math'));

findClosestShortWord = function(words, box, maxDistance) {
  var closestIndex, dist, i, index, len, minDistance, word;
  minDistance = maxDistance;
  closestIndex = -1;
  for (index = i = 0, len = words.length; i < len; index = ++i) {
    word = words[index];
    if (!(word.text.length < 3)) {
      continue;
    }
    dist = distance(word.box, box);
    if (dist < minDistance) {
      minDistance = dist;
      closestIndex = index;
    }
  }
  return closestIndex;
};

matchByWords = function(formData, fields, words, schemaToPage, schemaToData) {
  var closeIndex, dataBox, field, fieldData, fieldIndex, i, index, j, len, len1, matchedFields, matches, wordUsage;
  matchedFields = [];
  matches = [];
  wordUsage = [];
  for (i = 0, len = fields.length; i < len; i++) {
    field = fields[i];
    // Find short words close to estimated locations (data and page transform).
    dataBox = schemaToData(field.box);
    closeIndex = findClosestShortWord(words, dataBox, dataBox.width);
    if (closeIndex === -1) {
      closeIndex = findClosestShortWord(words, schemaToPage(field.box), dataBox.width);
    }
    if (closeIndex === -1) {
      continue;
    }
    // Validate short words.
    if (validate(field, words[closeIndex].text, false)) {
      if (wordUsage[closeIndex] == null) {
        wordUsage[closeIndex] = [];
      }
      wordUsage[closeIndex].push(field.path);
      matchedFields.push(field);
      matches.push(closeIndex);
    }
  }
// Assign matching fields.
  for (fieldIndex = j = 0, len1 = matchedFields.length; j < len1; fieldIndex = ++j) {
    field = matchedFields[fieldIndex];
    index = matches[fieldIndex];
    fieldData = unpack(formData, field.path);
    fieldData.value = words[index].text;
    fieldData.confidence = words[index].confidence;
    fieldData.box = words[index].box;
    fieldData.conflicts = wordUsage[index].filter(function(path) {
      return path !== field.path;
    });
  }
  return matchedFields;
};

findClosestMark = function(marks, box, maxDistance) {
  var closestIndex, dist, i, index, len, mark, minDistance;
  minDistance = maxDistance;
  closestIndex = -1;
  for (index = i = 0, len = marks.length; i < len; index = ++i) {
    mark = marks[index];
    dist = distance(mark.box, box);
    if (dist < minDistance) {
      minDistance = dist;
      closestIndex = index;
    }
  }
  return closestIndex;
};

matchByMark = function(formData, fields, marks, schemaToPage) {
  var closeIndex, farDistance, field, fieldData, i, j, len, len1, markUsage, match, matches, nearDistance, pageBox;
  matches = {};
  markUsage = [];
  for (i = 0, len = fields.length; i < len; i++) {
    field = fields[i];
    // Find close marks using estimated locations (only page transform).
    pageBox = schemaToPage(field.box);
    nearDistance = pageBox.width * 0.95;
    farDistance = pageBox.width * 3.00;
    closeIndex = findClosestMark(marks, pageBox, farDistance);
    if (closeIndex === -1) {
      // No marks with less than far distance found, thus false.
      matches[field.path] = {
        index: -1,
        value: false,
        confidence: 100,
        box: pageBox
      };
    } else if (distance(marks[closeIndex].box, pageBox) > nearDistance) {
      // Mark between near and far distance found, thus false with reduced confidence.
      matches[field.path] = {
        index: -1,
        value: false,
        confidence: Math.round((distance(marks[closeIndex].box, pageBox) / farDistance) * 100),
        box: pageBox
      };
    } else {
      // Near mark found, thus use it.
      matches[field.path] = {
        index: closeIndex,
        value: marks[closeIndex].checked,
        confidence: marks[closeIndex].confidence,
        box: marks[closeIndex].box
      };
      if (markUsage[closeIndex] == null) {
        markUsage[closeIndex] = [];
      }
      markUsage[closeIndex].push(field.path);
    }
  }
// Assign matching fields.
  for (j = 0, len1 = fields.length; j < len1; j++) {
    field = fields[j];
    match = matches[field.path];
    fieldData = unpack(formData, field.path);
    fieldData.value = match.value;
    fieldData.confidence = match.confidence;
    fieldData.box = match.box;
    if (match.index === -1) {
      fieldData.conflicts = [];
    } else {
      fieldData.conflicts = markUsage[match.index].filter(function(path) {
        return path !== field.path;
      });
    }
  }
};

// Match checkboxes to form schema.

// This process is content- and location-sensitive. Short words are preferred over marks.
// XXX: false negatives are to be expected when words are invalidated, this hurts valid vs. positional matching.
module.exports.matchCheckboxes = function(formData, formSchema, marks, words, schemaToPage, schemaToData) {
  var assignedFields, checkboxFields, field, remainingFields;
  checkboxFields = formSchema.fields.filter(function(field) {
    return field.type === 'checkbox';
  });
  assignedFields = matchByWords(formData, checkboxFields, words, schemaToPage, schemaToData);
  remainingFields = (function() {
    var i, len, results;
    results = [];
    for (i = 0, len = checkboxFields.length; i < len; i++) {
      field = checkboxFields[i];
      if (indexOf.call(assignedFields, field) < 0) {
        results.push(field);
      }
    }
    return results;
  })();
  matchByMark(formData, remainingFields, marks, schemaToPage);
};