Kinda failures

[everythingability]

Hello, I'm parsing an academic book (with Flair NER) that contains "dates" such as:

• more than 500 years ago
• the fourteenth century
• the Middle Ages
• the sixth century
• the year 105 CE
• 1962
• 1967
• 1952
• the eighteenth century
• 1887
• 1993
• 1925
• the century
• today
• 1952
• 1979
• around 5000 years old
• the thirty-second century
• a thousand or more years
• the first century
• the sixth century BCE
• 1996
• 1998
• 1999
• 2017
• 1542
• 6th century
• the 2nd century
• the 5th century BCE
• the 7th century CE
• the last decade
• 1473
• 1500
• 1599
• the 14th century
• 1894
• 1989
• 1990
• 9 April 1945
• July
• March 1944
• May 1933
• the first months
• Today
• fifteen hundred years ago
• 1925
• at least 3000 years
• the 3rd century BCE in Qing Dynasty
• 5000 years ago
• seventy years
• 2015
• the 21st century
• 1814
• 2000 BCE
• 1815
• Christmas Eve, 1851
• today
• 1998
• 2014
• 2025
• those days
• 1983
• 1991
• 1996
• twenty years
• 1555
• the twelfth century
• today
• 1998
• 2017
• twenty years from 1998
• 1969
• 1974
• 1998
• 2001
• April 2002
• every year
• monthly
• 1991
• 2002
• 2003
• 2004
• 2014
• four years
• 1269
• a hundred years
• seventh century
• the 9th century
• the Yuan dynasty
• 2015
• today

and ideally want a "close enough for jazz" year (or not) from a parse... but many of the dates like these fail... even ones that seem pretty easy to get like "2000 BCE".

I even tried...

theDate = theDate.replace("BCE", "BC")
theDate = theDate.replace("th century", "00")
theDate = theDate.replace("rd century", "00")
theDate = theDate.replace("st century", "00")
theDate = theDate.replace("nd century", "00")

..to help it out a bit... but it also fails on items like "the fourteenth century" .

Thanks for the work though. I found a python 2.7 module called dataparse that might be worth a peek? Couldn't install it through.

Tom

[rufuspollock]

@everythingability i am surprised that 2000 BCE does not work - could you confirm that.

On other items I don't flexidate supports a lot of those yet. Not hard to add them so if you are interested in a PR please go for it ...

[everythingability]

@rufuspollock Yeah, not sure how that happened but it did/may have :-) I can't reproduce it in the Terminal, all seems OK.

Not sure how to, or I can contribute, I'm no great coder and even worse with git :-)

Attached is a whole heap of "challenging dates", many of which like "2 days after" can't be parsed without a contextual date, or "two thousand years" which definitely shouldn't be <2000 AD>.

But anyway, I've attached it hoping it may help you or someone.

Just a thought, but years ago I used a chatbot tool called Alicebot that read AIML files, which were essentially just long lists of search_and_replace items, that aimed to get the input down to something more managable... or parseable. I think that that approach might work well (But don't know).

Keep up the good work.

[everythingability]

Oh, it doesn't seem to have worked.

https://docs.google.com/document/d/1jdbn_m4viKbKgpckkUTwChkC_RuPwPjiMWAZ5HDDmtY/edit?usp=sharing