Re:Show HN: We've developed a Natural Language API
by pknight on Wednesday November 21, 2012 @05:12PM
(#4816598)
Would be great if this were a library that can be integrated rather than an api. An api makes things problematic because it provides the kind of functionality that would break an app if it for some reason isn't working. I don't see how it can be viable service for anything other than non-commercial/non-critical projects. Also...no php?
Re:Show HN: We've developed a Natural Language API
by apetresc on Tuesday November 20, 2012 @12:32PM
(#4810084)
They do work, they just need to be authenticated with the apikey parameter -- given the experimental stage it's in, we didn't really want to publish a publicly-available key =/
Re:Show HN: We've developed a Natural Language API
by ithcy2 on Tuesday November 20, 2012 @12:47PM
(#4810154)
Very cool, but I find it strange that all of the example phrases on the demo page are much longer than recommended by the API docs (<10 words). Why is this?
Re:Show HN: We've developed a Natural Language API
by drewch on Tuesday November 20, 2012 @01:01PM
(#4810231)
9 or 10 words is about the average sentence that we receive, it is just a guideline so that users understand that we aren't going to handle multiple sentences or paragraphs.
Re:Show HN: We've developed a Natural Language API
by ithcy2 on Tuesday November 20, 2012 @01:30PM
(#4810419)
I see. Thanks for your responses.
What happens when you encounter phrases with multiple intents? Is it something you can detect but is difficult to deal with, or is it a best-guess kind of thing where you attempt to obey the "first intent"? Or something else?
Re:Show HN: We've developed a Natural Language API
by drewch on Tuesday November 20, 2012 @12:28PM
(#4810060)
It's 1000 api calls a day, right now we are in an early phrase and it is free. If you require more calls we can accomodate more calls based on your needs, and it would still be at no cost.
Sorry for the inconvenience!
What happens when you encounter phrases with multiple intents? Is it something you can detect but is difficult to deal with, or is it a best-guess kind of thing where you attempt to obey the "first intent"? Or something else?
Python: https://github.com/Maluuba/napi-python/blob/master/tests/tes...
Java: https://github.com/Maluuba/napi-java/blob/master/src/test/ja...
Here is an example of a sentence and response:
>> client.interpret phrase: 'Set up a meeting with Bob tomorrow night at 7 PM to discuss the TPS reports'
response:
{:entities=> { :daterange=> [{ :start=>"2012-11-15", :end=>"2012-11-16" }], :title=> ["meeting to discuss the tps reports"], :timerange=> [{ :start=>"12:00:00AM", :end=>"12:00:00AM" }], :contacts=>[{ :name=>"bob" }] }, :action=>:CALENDAR_CREATE_EVENT, :category=>:CALENDAR }