suzanne.soy/zotero
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Tweak recognizePDF to avoid false positives: only count lines with more than 3 words (since we drop first and last later) as cleaned lines, increase number of pages read and number of lines away from median length as good lines (due to more stringent selection criteria), do not attempt to query google with a short query if we run out of lines, try to pick lines for a query that are not next to each other (to avoid finding articles that quote the article we're looking for)

Browse Source
This commit is contained in:
aurimasv 2012-12-21 14:56:08 -06:00
parent 11a83e5df3
commit 74c9967b95
1 changed files with 18 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
chrome/content/zotero/recognizePDF.js
View File

@ -243,7 +243,7 @@ Zotero_RecognizePDF.Recognizer = function () {}
* (function will be passed image as URL and must return text of CAPTCHA)
*/
Zotero_RecognizePDF.Recognizer.prototype.recognize = function(file, libraryID, callback, captchaCallback) {
const MAX_PAGES = 3;
const MAX_PAGES = 5;
this._libraryID = libraryID;
this._callback = callback;
@ -315,9 +315,9 @@ Zotero_RecognizePDF.Recognizer.prototype.recognize = function(file, libraryID, c
// Use only first column from multi-column lines
const lineRe = /^\s*([^\s]+(?: [^\s]+)+)/;
for(var i=0; i<lines.length; i++) {
for(var i=0; i<lines.length && cleanedLines.length<30; i++) {
var m = lineRe.exec(lines[i]);
if(m) {
if(m && m[1].split(' ').length > 3) {
cleanedLines.push(m[1]);
cleanedLineLengths.push(m[1].length);
}
@ -334,8 +334,8 @@ Zotero_RecognizePDF.Recognizer.prototype.recognize = function(file, libraryID, c
// pick lines within 4 chars of the median (this is completely arbitrary)
this._goodLines = [];
var uBound = medianLength + 4;
var lBound = medianLength - 4;
var uBound = medianLength + 6;
var lBound = medianLength - 6;
for (var i=0; i<lineLengthsLength; i++) {
if(cleanedLineLengths[i] > lBound && cleanedLineLengths[i] < uBound) {
// Strip quotation marks so they don't mess up search query quoting
@ -344,7 +344,7 @@ Zotero_RecognizePDF.Recognizer.prototype.recognize = function(file, libraryID, c
}
}
this._startLine = this._iteration = 0;
this._nextLine = this._iteration = 0;
this._queryGoogle();
}
}
@ -354,7 +354,7 @@ Zotero_RecognizePDF.Recognizer.prototype.recognize = function(file, libraryID, c
* @private
*/
Zotero_RecognizePDF.Recognizer.prototype._queryGoogle = function() {
if(this._iteration > 3 || this._startLine >= this._goodLines.length) {
if(this._iteration > 3 || !this._goodLines.length) {
try {
if(this._hiddenBrowser) Zotero.Browser.deleteHiddenBrowser(me._hiddenBrowser);
} catch(e) {}
@ -385,8 +385,17 @@ Zotero_RecognizePDF.Recognizer.prototype._queryGoogle = function() {
} else {
// take the relevant parts of some lines (exclude hyphenated word)
var queryStringWords = 0;
while(queryStringWords < 25 && this._startLine < this._goodLines.length) {
var words = this._goodLines[this._startLine].split(/\s+/);
while(queryStringWords < 25) {
/**a bit of a hack. We're relying on the same test being applied above.
* But this way we don't have to rewrite the same error reporting code
*/
if(!this._goodLines.length) this._queryGoogle();
var words = this._goodLines.splice(this._nextLine,1)[0].split(/\s+/);
//Try to avoid picking adjacent strings so the odds of them appearing in another
// document quoting our document is low. Every 7th line is a magic value
this._nextLine = (this._nextLine + 7) % this._goodLines.length;
// get rid of first and last words
words.shift();
words.pop();
@ -403,7 +412,6 @@ Zotero_RecognizePDF.Recognizer.prototype._queryGoogle = function() {
queryStringWords += words.length;
queryString += '"'+words.join(" ")+'" ';
}
this._startLine++;
}
Zotero.debug("RecognizePDF: Query string "+queryString);