NLP false positives about parsedatetime HOT 6 CLOSED

probably - looking at the code we could introduce a list of strings to check at around

https://github.com/bear/parsedatetime/blob/master/parsedatetime/__init__.py#L1672

that is where "$300" becomes "3:00" and the chunk1 variable contains "$" - so yea, if chunk1 was found in the exception list then fail the match and proceed as if nothing happend.

from parsedatetime.

It seems that at https://github.com/bear/parsedatetime/blob/master/parsedatetime/__init__.py#L1672
the value that comes through from the regex is 300 (without the $), so not particularly easy to filter it out at that point.

from parsedatetime.

what version are you using? I did a test parse for "$300" and this is the debug output:

DEBUG:parsedatetime:parse()
DEBUG:parsedatetime:parse (top of loop): [$300][]
DEBUG:parsedatetime:parse 1 [][][]
DEBUG:parsedatetime:parse 2 [][][]
DEBUG:parsedatetime:parse 3 [][][]
DEBUG:parsedatetime:parse 4 [][][]
DEBUG:parsedatetime:parse 5 [][][]
DEBUG:parsedatetime:parse 6 [][][]
DEBUG:parsedatetime:parse 7 [][][]
DEBUG:parsedatetime:parse 8 [][][]
DEBUG:parsedatetime:parse 9 [][][]
DEBUG:parsedatetime:parse A [][][]
DEBUG:parsedatetime:parse (bottom) [$ ][3:00][$][]
DEBUG:parsedatetime:weekday False, dateStd False, dateStr False, time True, timeStr False, meridian False
DEBUG:parsedatetime:dayStr False, modifier False, modifier2 False, units False, qunits False
DEBUG:parsedatetime:_evalString(3:00, None)

the debug line for "parse (bottom)" shows that chunk1 has the value of "$" at that point

from parsedatetime.

I'm running from head.
Debug for me is as follows. Have I possibly setup something wrong?

DEBUG:parsedatetime:parse()
DEBUG:parsedatetime:parse (top of loop): [300][]
DEBUG:parsedatetime:parse 1 [][][]
DEBUG:parsedatetime:parse 2 [][][]
DEBUG:parsedatetime:parse 3 [][][]
DEBUG:parsedatetime:parse 4 [][][]
DEBUG:parsedatetime:parse 5 [][][]
DEBUG:parsedatetime:parse 6 [][][]
DEBUG:parsedatetime:parse 7 [][][]
DEBUG:parsedatetime:parse 8 [][][]
DEBUG:parsedatetime:parse 9 [][][]
DEBUG:parsedatetime:parse A [][][]
DEBUG:parsedatetime:parse (bottom) [ ][3:00][][]
DEBUG:parsedatetime:weekday False, dateStd False, dateStr False, time True, timeStr False, meridian False
DEBUG:parsedatetime:dayStr False, modifier False, modifier2 False, units False, qunits False
DEBUG:parsedatetime:_evalString(3:00, None)
DEBUG:parsedatetime:attempt to parse as rfc822 - None
DEBUG:parsedatetime:parse (top of loop): [ ][]
DEBUG:parsedatetime:parse 1 [][][]
DEBUG:parsedatetime:parse 2 [][][]
DEBUG:parsedatetime:parse 3 [][][]
DEBUG:parsedatetime:parse 4 [][][]

DEBUG:parsedatetime:parse 5 [][][]
DEBUG:parsedatetime:parse 6 [][][]
DEBUG:parsedatetime:parse 7 [][][]
DEBUG:parsedatetime:parse 8 [][][]
DEBUG:parsedatetime:parse 9 [][][]
DEBUG:parsedatetime:parse A [][][]
DEBUG:parsedatetime:parse (bottom) [][][][]
DEBUG:parsedatetime:weekday False, dateStd False, dateStr False, time False, timeStr False, meridian False
DEBUG:parsedatetime:dayStr False, modifier False, modifier2 False, units False, qunits False
DEBUG:parsedatetime:parse() return dateFlag 0 timeFlag 2 totalTime (2014, 10, 29, 3, 0, 0, 2, 302, 0)

Code running from testNlp class:
result = self.cal.nlp("$300") #dollar amounts shouldn't interfere with recognition self.assertTrue(result [0][4] != "300")

from parsedatetime.

ah - your running with nlp() and I ran with parse() - look at the output difference for the "(top of loop)" debug line.

i'll rerun that test in a bit and see why it's different.

from parsedatetime.

Cool, thanks. As I said, not sure it's essential. One could always pre-parse text for a given word e.g. $300 and strip that before sending to nlp for parsing.

from parsedatetime.