Wit.ai not parsing data for pence -
wit trained extract wit/number , wit/amountofmoney correctly interprets following sentences 3.03 pounds: 1)spent 3 pound , 3 pence 2) spent 3 cent 3) spent £3:
but fails : 1) spent 3 pence 2) spent 3p
questions: 1) bug/ new feature ? 2) there workaround ? reference : https://en.wikipedia.org/wiki/pound_sterling wit.ai api response spent 3 pounds , 3 cent
{ "msg_id": "xxxxxxxxxxxxxxxxxxx", "_text": "spent 3 pounds , 3 cent", "entities": { "amount_of_money": [ { "_body": "3 pounds , 3 cent", "confidence": 1, "_start": 6, "_end": 25, "_entity": "amount_of_money", "type": "value", "value": 3.03, "unit": "£" } ], "intent": [ { "confidence": 0.9873680621152465, "_entity": "intent", "value": "expense" } ] } }
wit.ai api response "spent 3p" :
{ "msg_id": "xxxxxxxxxxxxxxxxxxx", "_text": "spent 3p", "entities": { "number": [ { "_body": "3", "confidence": 1, "_start": 6, "_end": 7, "_entity": "number", "type": "value", "value": 3 } ], "intent": [ { "confidence": 0.9928105550766221, "_entity": "intent", "value": "expense" } ] } }
the wit/number
, wit/amount-of-money
entities rely on duckling parsing.
here rules defining amounts of money english: https://github.com/wit-ai/duckling/blob/master/resources/languages/en/rules/finance.clj
as can see, there no rule "pence" nor "p". duckling being open source, can contribute , pull request.
Comments
Post a Comment