2017-03-08 21:33:55 +03:00
|
|
|
-- Copyright (c) 2016-present, Facebook, Inc.
|
|
|
|
|
-- All rights reserved.
|
|
|
|
|
--
|
|
|
|
|
-- This source code is licensed under the BSD-style license found in the
|
|
|
|
|
-- LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
|
-- of patent rights can be found in the PATENTS file in the same directory.
|
|
|
|
|
|
|
|
|
|
-----------------------------------------------------------------
|
|
|
|
|
-- Auto-generated by regenClassifiers
|
|
|
|
|
--
|
|
|
|
|
-- DO NOT EDIT UNLESS YOU ARE SURE THAT YOU KNOW WHAT YOU ARE DOING
|
|
|
|
|
-- @generated
|
|
|
|
|
-----------------------------------------------------------------
|
|
|
|
|
{-# LANGUAGE OverloadedStrings #-}
|
|
|
|
|
module Duckling.Ranking.Classifiers.ES (classifiers) where
|
|
|
|
|
import Prelude
|
|
|
|
|
import Duckling.Ranking.Types
|
|
|
|
|
import qualified Data.HashMap.Strict as HashMap
|
|
|
|
|
import Data.String
|
|
|
|
|
|
|
|
|
|
classifiers :: Classifiers
|
|
|
|
|
classifiers
|
|
|
|
|
= HashMap.fromList
|
|
|
|
|
[("midnight",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<time> timezone",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -0.6931471805599453),
|
|
|
|
|
("hour", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("integer (numeric)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.5839478885949533, unseen = -4.0943445622221,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 58},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.8157495026522777, unseen = -3.871201010907891,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 46}}),
|
|
|
|
|
("the day before yesterday",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("hh(:|.|h)mm (time-of-day)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.3978952727983707,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 9},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<named-month|named-day> past",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("el <time>", -1.2992829841302609),
|
|
|
|
|
("day", -0.7884573603642702),
|
|
|
|
|
("named-day", -1.2992829841302609)],
|
|
|
|
|
n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("dd[/-]mm",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.3025850929940455,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 8},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("intersect by `de`",
|
|
|
|
|
Classifier{okData =
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
ClassData{prior = -0.16362942378180204,
|
|
|
|
|
unseen = -4.812184355372417,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("el <day-of-month> (non ordinal)intersect by `de`",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-3.7054087560651467),
|
|
|
|
|
("day of month (1st)named-month", -3.1945831322991562),
|
|
|
|
|
("daymonth", -1.7594986070098335),
|
|
|
|
|
("monthyear", -3.7054087560651467),
|
|
|
|
|
("named-dayel proximo <cycle> ", -4.110873864173311),
|
|
|
|
|
("named-dayla <cycle> pasado", -4.110873864173311),
|
|
|
|
|
("el <time>named-month", -3.7054087560651467),
|
|
|
|
|
("dayyear", -2.164963715117998),
|
|
|
|
|
("dd-dd <month>(interval)year", -4.110873864173311),
|
|
|
|
|
("el <time>el <cycle> (proximo|que viene)", -4.110873864173311),
|
2017-03-08 21:33:55 +03:00
|
|
|
("two time tokens separated by \",\"named-month",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-4.110873864173311),
|
|
|
|
|
("named-monthyear", -3.7054087560651467),
|
|
|
|
|
("el <time>year", -3.7054087560651467),
|
|
|
|
|
("el <time>la <cycle> pasado", -4.110873864173311),
|
|
|
|
|
("el <day-of-month> de <named-month>year", -3.7054087560651467),
|
|
|
|
|
("el <time>este|en un <cycle>", -3.7054087560651467),
|
|
|
|
|
("named-dayel <cycle> (proximo|que viene)", -4.110873864173311),
|
2017-03-08 21:33:55 +03:00
|
|
|
("el <day-of-month> (non ordinal)named-month",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-2.406125771934886),
|
|
|
|
|
("<day-of-month> de <named-month>year", -3.7054087560651467),
|
|
|
|
|
("two time tokens separated by \",\"year", -3.417726683613366),
|
2017-03-08 21:33:55 +03:00
|
|
|
("two time tokens separated by \",\"intersect by `de`",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-4.110873864173311),
|
|
|
|
|
("dayweek", -2.406125771934886),
|
|
|
|
|
("intersect by `de`year", -3.417726683613366),
|
|
|
|
|
("named-dayeste|en un <cycle>", -3.417726683613366)],
|
|
|
|
|
n = 45},
|
2017-03-08 21:33:55 +03:00
|
|
|
koData =
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
ClassData{prior = -1.890850371872286, unseen = -3.891820298110627,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
[("monthyear", -3.1780538303479458),
|
|
|
|
|
("hourmonth", -2.772588722239781),
|
|
|
|
|
("monthmonth", -3.1780538303479458),
|
|
|
|
|
("a las <time-of-day>named-month", -2.772588722239781),
|
|
|
|
|
("dayyear", -3.1780538303479458),
|
|
|
|
|
("minutemonth", -2.772588722239781),
|
|
|
|
|
("named-monthyear", -3.1780538303479458),
|
2017-03-08 21:33:55 +03:00
|
|
|
("de <datetime> - <datetime> (interval)named-month",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-3.1780538303479458),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> <integer> (as relative minutes)named-month",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-3.1780538303479458),
|
|
|
|
|
("<day-of-month> de <named-month>year", -3.1780538303479458),
|
|
|
|
|
("intersect by `de`year", -3.1780538303479458),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> <integer> (as relative minutes)intersect by `de`",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-3.1780538303479458),
|
|
|
|
|
("minuteyear", -3.1780538303479458)],
|
2017-03-08 21:33:55 +03:00
|
|
|
n = 8}}),
|
|
|
|
|
("n pasados <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.1972245773362196,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("number (0..15)a\241o (grain)", -1.3862943611198906),
|
|
|
|
|
("year", -1.3862943611198906),
|
|
|
|
|
("number (0..15)mes (grain)", -1.3862943611198906),
|
|
|
|
|
("month", -1.3862943611198906)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<hour-of-day> and half",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.772588722239781,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.3217558399823195),
|
|
|
|
|
("time-of-day (latent)", -1.3217558399823195),
|
|
|
|
|
("hour", -0.7621400520468967)],
|
|
|
|
|
n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("pasados n <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.2188758248682006,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)dia (grain)", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)mes (grain)", -2.4849066497880004),
|
|
|
|
|
("second", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)minutos (grain)", -2.4849066497880004),
|
|
|
|
|
("day", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)segundo (grain)", -2.4849066497880004),
|
|
|
|
|
("year", -2.4849066497880004), ("month", -2.4849066497880004),
|
|
|
|
|
("minute", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)a\241o (grain)", -2.4849066497880004),
|
|
|
|
|
("number (0..15)semana (grain)", -2.4849066497880004)],
|
|
|
|
|
n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("semana (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.833213344056216,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 15},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("two time tokens separated by \",\"",
|
|
|
|
|
Classifier{okData =
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
ClassData{prior = 0.0, unseen = -3.295836866004329,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
[("named-dayel <time>", -2.159484249353372),
|
|
|
|
|
("dayday", -0.8602012652231115),
|
2017-03-08 21:33:55 +03:00
|
|
|
("named-dayel <day-of-month> (non ordinal)",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-2.5649493574615367),
|
|
|
|
|
("named-dayintersect by `de`", -1.466337068793427),
|
2017-03-08 21:33:55 +03:00
|
|
|
("named-day<day-of-month> de <named-month>",
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
-2.5649493574615367),
|
|
|
|
|
("named-dayel <day-of-month> de <named-month>",
|
|
|
|
|
-2.5649493574615367)],
|
|
|
|
|
n = 10},
|
2017-03-08 21:33:55 +03:00
|
|
|
koData =
|
Optimize simple time predicates
Summary:
This is the next step for:
https://fb.facebook.com/groups/527352907463243/permalink/600056483526218/
This:
* changes the time language to be able to track contradictions (`EmptyPredicate`)
* changes the time language to be able to collect non-contradicting pieces, like month and hour and unify them
* provides an efficient way to convert those pieces into (past,future) time series
* adds AMPM predicate runner - there's a bit of overlap with is12H, but it basically works
* changes a test case that was wrong before
* regenerates classifiers, I'm not sure why they changed exactly
Before:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(15.50 secs, 6,171,188,928 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(110.82 secs, 44,031,569,512 bytes)
```
After:
```
res <- H.io $ let sentence = "10am thurs 4.30 thurs 12pm sat" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(1.24 secs, 703,020,912 bytes)
res <- H.io $ let sentence = "I have 9 am 12 pm 1 pm 2pm 4 pm 3 pm on Saturday" in (debugTokens sentence $ analyze sentence (testContext {lang = EN}) HashSet.empty)
(9.51 secs, 5,891,109,592 bytes)
```
Reviewed By: JonCoens
Differential Revision: D4676812
fbshipit-source-id: 9810203
2017-03-14 02:49:47 +03:00
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<dim time> de la manana",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.15415067982725836,
|
|
|
|
|
unseen = -2.772588722239781,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.3217558399823195),
|
|
|
|
|
("time-of-day (latent)", -1.3217558399823195),
|
|
|
|
|
("hour", -0.7621400520468967)],
|
|
|
|
|
n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.9459101490553135, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)", -0.916290731874155),
|
|
|
|
|
("hour", -0.916290731874155)],
|
|
|
|
|
n = 1}}),
|
|
|
|
|
("del mediod\237a",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<day-of-month> de <named-month>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -9.53101798043249e-2,
|
|
|
|
|
unseen = -3.784189633918261,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("number (0..15)named-month", -2.151762203259462),
|
|
|
|
|
("integer (numeric)named-month", -0.9279867716373464),
|
|
|
|
|
("month", -0.7166776779701395)],
|
|
|
|
|
n = 20},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -2.3978952727983707,
|
|
|
|
|
unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)named-month", -0.8472978603872037),
|
|
|
|
|
("month", -0.8472978603872037)],
|
|
|
|
|
n = 2}}),
|
|
|
|
|
("<time-of-day> <part-of-day>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.3528213746227423, unseen = -4.31748811353631,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<day-of-month> de <named-month>in the <part-of-day>",
|
|
|
|
|
-3.6109179126442243),
|
|
|
|
|
("dayhour", -2.2246235515243336),
|
|
|
|
|
("tomorrowin the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("a las <time-of-day>del mediod\237a", -3.6109179126442243),
|
|
|
|
|
("el <day-of-month> de <named-month>in the <part-of-day>",
|
|
|
|
|
-3.6109179126442243),
|
|
|
|
|
("hourhour", -2.2246235515243336),
|
|
|
|
|
("a las <time-of-day>in the <part-of-day>",
|
|
|
|
|
-2.3581549441488563),
|
|
|
|
|
("intersectin the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("minutehour", -1.739115735742633),
|
|
|
|
|
("intersect by `de`in the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("<hour-of-day> and halfin the <part-of-day>",
|
|
|
|
|
-3.20545280453606),
|
|
|
|
|
("named-dayin the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("<hour-of-day> and <relative minutes>del mediod\237a",
|
|
|
|
|
-3.20545280453606),
|
|
|
|
|
("el <time>in the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("<hour-of-day> and quarterin the <part-of-day>",
|
|
|
|
|
-2.6946271807700692),
|
|
|
|
|
("yesterdayin the <part-of-day>", -3.6109179126442243),
|
|
|
|
|
("time-of-day (latent)in the <part-of-day>",
|
|
|
|
|
-2.917770732084279)],
|
|
|
|
|
n = 26},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.213022639845854, unseen = -3.8066624897703196,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("yearhour", -1.4816045409242156),
|
|
|
|
|
("year (latent)del mediod\237a", -3.0910424533583156),
|
|
|
|
|
("monthhour", -3.0910424533583156),
|
|
|
|
|
("hourhour", -3.0910424533583156),
|
|
|
|
|
("named-monthin the <part-of-day>", -3.0910424533583156),
|
|
|
|
|
("year (latent)in the <part-of-day>", -1.5869650565820417),
|
|
|
|
|
("time-of-day (latent)in the <part-of-day>",
|
|
|
|
|
-3.0910424533583156)],
|
|
|
|
|
n = 11}}),
|
|
|
|
|
("de <datetime> - <datetime> (interval)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("monthday", -1.791759469228055),
|
|
|
|
|
("monthyear", -1.791759469228055),
|
|
|
|
|
("monthhour", -1.791759469228055),
|
|
|
|
|
("named-monthyear (latent)", -1.791759469228055),
|
|
|
|
|
("named-monthtime-of-day (latent)", -1.791759469228055),
|
|
|
|
|
("named-month<day-of-month> de <named-month>",
|
|
|
|
|
-1.791759469228055)],
|
|
|
|
|
n = 3}}),
|
|
|
|
|
("<time-of-day> horas",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -1.0986122886681098,
|
|
|
|
|
unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.252762968495368),
|
|
|
|
|
("time-of-day (latent)", -1.252762968495368),
|
|
|
|
|
("hour", -0.8472978603872037)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.40546510810816444,
|
|
|
|
|
unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)", -0.7884573603642702),
|
|
|
|
|
("hour", -0.7884573603642702)],
|
|
|
|
|
n = 4}}),
|
|
|
|
|
("intersect",
|
|
|
|
|
Classifier{okData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -0.2578291093020998,
|
|
|
|
|
unseen = -4.6443908991413725,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<day-of-month> de <named-month>in the <part-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.9415818076696905),
|
|
|
|
|
("dayhour", -1.9956716586143772),
|
|
|
|
|
("tomorrowin the <part-of-day>", -3.9415818076696905),
|
2017-03-08 21:33:55 +03:00
|
|
|
("el <day-of-month> de <named-month>in the <part-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.9415818076696905),
|
|
|
|
|
("hourhour", -3.0252910757955354),
|
2017-03-08 21:33:55 +03:00
|
|
|
("now<hour-of-day> minus quarter (as relative minutes)",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.9415818076696905),
|
|
|
|
|
("dayyear", -3.0252910757955354),
|
2017-03-08 21:33:55 +03:00
|
|
|
("a las <time-of-day>in the <part-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-2.6888188391743224),
|
|
|
|
|
("intersectin the <part-of-day>", -3.9415818076696905),
|
|
|
|
|
("<named-month> <day-of-month>del <year>", -3.9415818076696905),
|
|
|
|
|
("minutehour", -2.33214389523559),
|
|
|
|
|
("intersect by `de`in the <part-of-day>", -3.9415818076696905),
|
2017-03-08 21:33:55 +03:00
|
|
|
("now<hour-of-day> and <relative minutes>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.9415818076696905),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> and halfin the <part-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.536116699561526),
|
|
|
|
|
("named-dayin the <part-of-day>", -3.9415818076696905),
|
|
|
|
|
("tomorrowa las <time-of-day>", -3.536116699561526),
|
|
|
|
|
("named-daya las <time-of-day>", -3.9415818076696905),
|
|
|
|
|
("named-dayintersect", -3.9415818076696905),
|
|
|
|
|
("dayminute", -3.0252910757955354),
|
|
|
|
|
("<named-month> <day-of-month>el <time>", -3.9415818076696905),
|
|
|
|
|
("named-day<dim time> de la manana", -3.9415818076696905),
|
|
|
|
|
("el <time>in the <part-of-day>", -3.9415818076696905),
|
|
|
|
|
("named-day<time-of-day> <part-of-day>", -3.9415818076696905),
|
|
|
|
|
("dd[/-]mmyear", -3.536116699561526),
|
|
|
|
|
("nowa las <time-of-day>", -3.536116699561526),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> and quarterin the <part-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.0252910757955354),
|
|
|
|
|
("yesterdayin the <part-of-day>", -3.9415818076696905)],
|
2017-03-08 21:33:55 +03:00
|
|
|
n = 34},
|
|
|
|
|
koData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -1.4816045409242156, unseen = -4.02535169073515,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
2017-04-06 20:50:34 +03:00
|
|
|
[("hourday", -3.3141860046725258),
|
|
|
|
|
("dayhour", -2.6210388241125804),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<day-of-month> de <named-month>a las <time-of-day>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.3141860046725258),
|
|
|
|
|
("<time-of-day> am|pm<day-of-month> de <named-month>",
|
|
|
|
|
-3.3141860046725258),
|
|
|
|
|
("monthhour", -2.908720896564361),
|
2017-03-08 21:33:55 +03:00
|
|
|
("now<hour-of-day> and <relative minutes>",
|
2017-04-06 20:50:34 +03:00
|
|
|
-3.3141860046725258),
|
|
|
|
|
("dayminute", -2.908720896564361),
|
|
|
|
|
("named-monthin the <part-of-day>", -3.3141860046725258),
|
|
|
|
|
("named-montha las <time-of-day>", -3.3141860046725258),
|
|
|
|
|
("nowa las <time-of-day>", -2.6210388241125804),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> <integer> (as relative minutes)year",
|
2017-04-06 20:50:34 +03:00
|
|
|
-2.908720896564361),
|
|
|
|
|
("minuteyear", -2.908720896564361)],
|
|
|
|
|
n = 10}}),
|
2017-03-08 21:33:55 +03:00
|
|
|
("a las <time-of-day>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -8.338160893905101e-2,
|
|
|
|
|
unseen = -4.634728988229636,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<hour-of-day> and half", -3.2386784521643803),
|
|
|
|
|
("<time-of-day> horas", -3.9318256327243257),
|
|
|
|
|
("<hour-of-day> and quarter", -3.0155349008501706),
|
|
|
|
|
("time-of-day (latent)", -1.2237754316221157),
|
|
|
|
|
("<hour-of-day> and <relative minutes>", -2.833213344056216),
|
|
|
|
|
("<time-of-day> am|pm", -3.9318256327243257),
|
|
|
|
|
("<hour-of-day> minus <integer> (as relative minutes)",
|
|
|
|
|
-3.9318256327243257),
|
|
|
|
|
("<hour-of-day> minus quarter (as relative minutes)",
|
|
|
|
|
-3.5263605246161616),
|
|
|
|
|
("hour", -1.1592369104845446), ("minute", -1.8523840910444898)],
|
|
|
|
|
n = 46},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -2.5257286443082556,
|
|
|
|
|
unseen = -2.9444389791664407,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)", -1.791759469228055),
|
|
|
|
|
("<hour-of-day> and <relative minutes>", -1.791759469228055),
|
|
|
|
|
("hour", -1.791759469228055), ("minute", -1.791759469228055)],
|
|
|
|
|
n = 4}}),
|
|
|
|
|
("season",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("minutos (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.1972245773362196,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 7},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("this <day-of-week>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("day", -0.6931471805599453),
|
|
|
|
|
("named-day", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("year (latent)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.639057329615259,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)", -1.1786549963416462),
|
|
|
|
|
("number (0..15)", -0.6190392084062235),
|
|
|
|
|
("number (20..90)", -1.8718021769015913)],
|
|
|
|
|
n = 10}}),
|
|
|
|
|
("el <cycle> (proximo|que viene)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.70805020110221,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.540445040947149),
|
|
|
|
|
("semana (grain)", -1.540445040947149),
|
|
|
|
|
("mes (grain)", -1.9459101490553135),
|
|
|
|
|
("year", -1.9459101490553135),
|
|
|
|
|
("a\241o (grain)", -1.9459101490553135),
|
|
|
|
|
("month", -1.9459101490553135)],
|
|
|
|
|
n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("yesterday",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<hour-of-day> and quarter",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.0910424533583156,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.4350845252893227),
|
|
|
|
|
("time-of-day (latent)", -1.252762968495368),
|
|
|
|
|
("hour", -0.7419373447293773)],
|
|
|
|
|
n = 9},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("dentro de <duration>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<integer> <unit-of-duration>", -0.6931471805599453),
|
|
|
|
|
("hour", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("hace <duration>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.639057329615259,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.8718021769015913), ("year", -1.8718021769015913),
|
|
|
|
|
("<integer> <unit-of-duration>", -0.9555114450274363),
|
|
|
|
|
("hour", -1.8718021769015913), ("month", -1.8718021769015913)],
|
|
|
|
|
n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("el <day-of-month> de <named-month>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.1780538303479458,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("number (0..15)named-month", -1.749199854809259),
|
|
|
|
|
("integer (numeric)named-month", -1.0560526742493137),
|
|
|
|
|
("month", -0.7375989431307791)],
|
|
|
|
|
n = 10},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("named-month",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.295836866004329,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 25},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("el <time>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -5.406722127027582e-2,
|
|
|
|
|
unseen = -4.465908118654584,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<named-month|named-day> past", -3.355735007585398),
|
|
|
|
|
("dd[/-]mm", -3.068052935133617),
|
|
|
|
|
("intersect by `de`", -2.257122718917288),
|
|
|
|
|
("<day-of-month> de <named-month>", -2.056452023455137),
|
|
|
|
|
("<time-of-day> <part-of-day>", -3.7612001156935624),
|
|
|
|
|
("intersect", -3.7612001156935624),
|
|
|
|
|
("day", -0.9279867716373464), ("year", -3.7612001156935624),
|
|
|
|
|
("named-day", -2.374905754573672),
|
|
|
|
|
("day of month (1st)", -3.355735007585398),
|
|
|
|
|
("<named-month|named-day> next", -3.7612001156935624),
|
|
|
|
|
("hour", -3.355735007585398)],
|
|
|
|
|
n = 36},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -2.9444389791664407, unseen = -2.995732273553991,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("noon", -2.2512917986064953), ("hour", -2.2512917986064953),
|
|
|
|
|
("minute", -2.2512917986064953),
|
|
|
|
|
("<hour-of-day> <integer> (as relative minutes)",
|
|
|
|
|
-2.2512917986064953)],
|
|
|
|
|
n = 2}}),
|
|
|
|
|
("now",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("ordinals (primero..10)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 10}}),
|
|
|
|
|
("this <part-of-day>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("hour", -0.6931471805599453),
|
|
|
|
|
("evening", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("number (0..15)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.9318256327243257,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 49},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("este|en un <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.70805020110221,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.0296194171811581),
|
|
|
|
|
("semana (grain)", -1.0296194171811581),
|
|
|
|
|
("year", -1.9459101490553135),
|
|
|
|
|
("a\241o (grain)", -1.9459101490553135)],
|
|
|
|
|
n = 5},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("numbers prefix with -, negative or minus",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)", -0.2006706954621511),
|
|
|
|
|
("number (0..15)", -1.7047480922384253)],
|
|
|
|
|
n = 9}}),
|
|
|
|
|
("dd-dd <month>(interval)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("named-month", -0.6931471805599453),
|
|
|
|
|
("month", -0.6931471805599453)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("tomorrow",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -1.0116009116784799, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.45198512374305727,
|
|
|
|
|
unseen = -2.1972245773362196,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 7}}),
|
|
|
|
|
("mes (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 10},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("number (20..90)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<day-of-week> <day-of-month>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("named-dayinteger (numeric)", -0.6931471805599453),
|
|
|
|
|
("day", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("afternoon",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.3364722366212129,
|
|
|
|
|
unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 5},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.252762968495368, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2}}),
|
|
|
|
|
("time-of-day (latent)",
|
|
|
|
|
Classifier{okData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -0.3272129112084162,
|
2017-03-08 21:33:55 +03:00
|
|
|
unseen = -3.5263605246161616,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)", -1.1939224684724346),
|
|
|
|
|
("number (0..15)", -0.3610133455373305)],
|
|
|
|
|
n = 31},
|
|
|
|
|
koData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -1.276293465905562, unseen = -2.70805020110221,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
2017-04-06 20:50:34 +03:00
|
|
|
[("integer (numeric)", -0.3364722366212129),
|
|
|
|
|
("number (0..15)", -1.252762968495368)],
|
|
|
|
|
n = 12}}),
|
2017-03-08 21:33:55 +03:00
|
|
|
("<hour-of-day> and <relative minutes>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.3364722366212129, unseen = -3.332204510175204,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)number (21..29 31..39 41..49 51..59 61..69 71..79 81..89 91..99)",
|
|
|
|
|
-2.1972245773362196),
|
|
|
|
|
("time-of-day (latent)number (0..15)", -2.6026896854443837),
|
|
|
|
|
("a las <time-of-day>number (20..90)", -2.1972245773362196),
|
|
|
|
|
("a las <time-of-day>number (21..29 31..39 41..49 51..59 61..69 71..79 81..89 91..99)",
|
|
|
|
|
-2.1972245773362196),
|
|
|
|
|
("hour", -0.8979415932059586),
|
|
|
|
|
("a las <time-of-day>number (0..15)", -2.6026896854443837),
|
|
|
|
|
("time-of-day (latent)number (20..90)", -2.1972245773362196)],
|
|
|
|
|
n = 10},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.252762968495368, unseen = -2.772588722239781,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>number (20..90)", -1.6094379124341003),
|
|
|
|
|
("hour", -1.0986122886681098),
|
|
|
|
|
("time-of-day (latent)number (20..90)", -1.6094379124341003)],
|
|
|
|
|
n = 4}}),
|
|
|
|
|
("year",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.40546510810816444,
|
|
|
|
|
unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods = HashMap.fromList [("integer (numeric)", 0.0)],
|
|
|
|
|
n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.0986122886681098,
|
|
|
|
|
unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [("integer (numeric)", 0.0)],
|
|
|
|
|
n = 3}}),
|
|
|
|
|
("en <duration>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.6635616461296463,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -2.9444389791664407), ("second", -2.9444389791664407),
|
|
|
|
|
("day", -2.538973871058276), ("year", -2.538973871058276),
|
|
|
|
|
("<integer> <unit-of-duration>", -0.8649974374866046),
|
|
|
|
|
("hour", -2.2512917986064953), ("month", -2.9444389791664407),
|
|
|
|
|
("minute", -1.845826690498331)],
|
|
|
|
|
n = 15},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -2.1972245773362196,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("n <cycle> (proximo|que viene)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.791759469228055),
|
|
|
|
|
("integer (numeric)semana (grain)", -1.791759469228055),
|
|
|
|
|
("year", -1.791759469228055),
|
|
|
|
|
("number (0..15)mes (grain)", -1.791759469228055),
|
|
|
|
|
("month", -1.791759469228055),
|
|
|
|
|
("integer (numeric)a\241o (grain)", -1.791759469228055)],
|
|
|
|
|
n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<integer> <unit-of-duration>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.587786664902119, unseen = -4.143134726391533,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -3.028522096376982),
|
|
|
|
|
("integer (numeric)hora (grain)", -3.4339872044851463),
|
|
|
|
|
("integer (numeric)dia (grain)", -3.4339872044851463),
|
|
|
|
|
("number (0..15)segundo (grain)", -3.4339872044851463),
|
|
|
|
|
("second", -3.4339872044851463),
|
|
|
|
|
("number (0..15)a\241o (grain)", -3.028522096376982),
|
|
|
|
|
("integer (numeric)minutos (grain)", -2.740840023925201),
|
|
|
|
|
("day", -3.028522096376982), ("year", -2.740840023925201),
|
|
|
|
|
("number (0..15)mes (grain)", -3.028522096376982),
|
|
|
|
|
("number (0..15)hora (grain)", -2.740840023925201),
|
|
|
|
|
("hour", -2.3353749158170367), ("month", -3.028522096376982),
|
|
|
|
|
("number (0..15)dia (grain)", -3.4339872044851463),
|
|
|
|
|
("number (0..15)minutos (grain)", -3.028522096376982),
|
|
|
|
|
("number (16..19 21..29)hora (grain)", -3.4339872044851463),
|
|
|
|
|
("minute", -2.3353749158170367),
|
|
|
|
|
("integer (numeric)a\241o (grain)", -3.4339872044851463),
|
|
|
|
|
("number (0..15)semana (grain)", -3.028522096376982)],
|
|
|
|
|
n = 20},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.8109302162163288, unseen = -4.007333185232471,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -2.890371757896165),
|
|
|
|
|
("integer (numeric)hora (grain)", -2.890371757896165),
|
|
|
|
|
("integer (numeric)dia (grain)", -2.890371757896165),
|
|
|
|
|
("integer (numeric)mes (grain)", -3.295836866004329),
|
|
|
|
|
("second", -2.890371757896165),
|
|
|
|
|
("number (0..15)a\241o (grain)", -3.295836866004329),
|
|
|
|
|
("integer (numeric)semana (grain)", -3.295836866004329),
|
|
|
|
|
("integer (numeric)minutos (grain)", -2.890371757896165),
|
|
|
|
|
("day", -2.890371757896165),
|
|
|
|
|
("integer (numeric)segundo (grain)", -2.890371757896165),
|
|
|
|
|
("year", -2.6026896854443837),
|
|
|
|
|
("number (0..15)mes (grain)", -2.890371757896165),
|
|
|
|
|
("hour", -2.890371757896165), ("month", -2.6026896854443837),
|
|
|
|
|
("minute", -2.890371757896165),
|
|
|
|
|
("integer (numeric)a\241o (grain)", -2.890371757896165),
|
|
|
|
|
("number (0..15)semana (grain)", -3.295836866004329)],
|
|
|
|
|
n = 16}}),
|
|
|
|
|
("proximas n <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.2188758248682006,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)hora (grain)", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)dia (grain)", -2.4849066497880004),
|
|
|
|
|
("second", -2.4849066497880004),
|
|
|
|
|
("number (0..15)a\241o (grain)", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)minutos (grain)", -2.4849066497880004),
|
|
|
|
|
("day", -2.4849066497880004),
|
|
|
|
|
("integer (numeric)segundo (grain)", -2.4849066497880004),
|
|
|
|
|
("year", -2.4849066497880004),
|
|
|
|
|
("number (0..15)mes (grain)", -2.4849066497880004),
|
|
|
|
|
("hour", -2.4849066497880004), ("month", -2.4849066497880004),
|
|
|
|
|
("minute", -2.4849066497880004)],
|
|
|
|
|
n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<time-of-day> am|pm",
|
|
|
|
|
Classifier{okData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -0.40546510810816444,
|
|
|
|
|
unseen = -2.0794415416798357,
|
2017-03-08 21:33:55 +03:00
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.252762968495368),
|
|
|
|
|
("time-of-day (latent)", -1.252762968495368),
|
|
|
|
|
("hour", -0.8472978603872037)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
2017-04-06 20:50:34 +03:00
|
|
|
ClassData{prior = -1.0986122886681098, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)", -0.916290731874155),
|
|
|
|
|
("hour", -0.916290731874155)],
|
|
|
|
|
n = 1}}),
|
2017-03-08 21:33:55 +03:00
|
|
|
("n proximas <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.791759469228055),
|
|
|
|
|
("integer (numeric)mes (grain)", -1.791759469228055),
|
|
|
|
|
("integer (numeric)semana (grain)", -1.791759469228055),
|
|
|
|
|
("year", -1.791759469228055), ("month", -1.791759469228055),
|
|
|
|
|
("integer (numeric)a\241o (grain)", -1.791759469228055)],
|
|
|
|
|
n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("ano nuevo",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("la <cycle> pasado",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.3978952727983707,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.2039728043259361),
|
|
|
|
|
("semana (grain)", -1.2039728043259361),
|
|
|
|
|
("year", -1.6094379124341003),
|
|
|
|
|
("a\241o (grain)", -1.6094379124341003)],
|
|
|
|
|
n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("la pasado <cycle>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("mes (grain)", -0.6931471805599453),
|
|
|
|
|
("month", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("named-day",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.4657359027997265,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 30},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("number (21..29 31..39 41..49 51..59 61..69 71..79 81..89 91..99)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList [("number (20..90)number (0..15)", 0.0)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("day of month (1st)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("a\241o (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.639057329615259,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 12},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<hour-of-day> minus <integer> (as relative minutes)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)number (0..15)", -1.252762968495368),
|
|
|
|
|
("hour", -0.8472978603872037),
|
|
|
|
|
("a las <time-of-day>number (0..15)", -1.252762968495368)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<named-month|named-day> next",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.3025850929940455,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("el <time>", -1.5040773967762742),
|
|
|
|
|
("day", -0.8109302162163288),
|
|
|
|
|
("named-day", -1.0986122886681098)],
|
|
|
|
|
n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("number (16..19 21..29)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<dim time> de la tarde",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -3.713572066704308,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("<hour-of-day> and half", -2.5902671654458267),
|
|
|
|
|
("a las <time-of-day>", -1.6094379124341003),
|
|
|
|
|
("<hour-of-day> and quarter", -2.0794415416798357),
|
|
|
|
|
("time-of-day (latent)", -2.0794415416798357),
|
|
|
|
|
("hour", -1.491654876777717), ("minute", -1.3862943611198906)],
|
|
|
|
|
n = 17},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("el proximo <cycle> ",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.70805020110221,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("week", -1.540445040947149),
|
|
|
|
|
("semana (grain)", -1.540445040947149),
|
|
|
|
|
("mes (grain)", -1.9459101490553135),
|
|
|
|
|
("year", -1.9459101490553135),
|
|
|
|
|
("a\241o (grain)", -1.9459101490553135),
|
|
|
|
|
("month", -1.9459101490553135)],
|
|
|
|
|
n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.9459101490553135,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<hour-of-day> minus quarter (as relative minutes)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.4849066497880004,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("a las <time-of-day>", -1.2992829841302609),
|
|
|
|
|
("time-of-day (latent)", -1.2992829841302609),
|
|
|
|
|
("hour", -0.7884573603642702)],
|
|
|
|
|
n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("the day after tomorrow",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("del <year>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("integer (numeric)", 0.0)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("dd[/-.]mm[/-.]yyyy",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("noon",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.6931471805599453,
|
|
|
|
|
unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.6931471805599453,
|
|
|
|
|
unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1}}),
|
|
|
|
|
("evening",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<datetime> - <datetime> (interval)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -1.252762968495368, unseen = -2.70805020110221,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("minuteminute", -1.9459101490553135),
|
|
|
|
|
("dayday", -1.9459101490553135),
|
|
|
|
|
("hh(:|.|h)mm (time-of-day)hh(:|.|h)mm (time-of-day)",
|
|
|
|
|
-1.9459101490553135),
|
|
|
|
|
("<day-of-month> de <named-month><day-of-month> de <named-month>",
|
|
|
|
|
-1.9459101490553135)],
|
|
|
|
|
n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -0.3364722366212129, unseen = -3.044522437723423,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("monthday", -2.3025850929940455),
|
|
|
|
|
("dayyear", -1.8971199848858813),
|
|
|
|
|
("named-month<day-of-month> de <named-month>",
|
|
|
|
|
-2.3025850929940455),
|
|
|
|
|
("dd[/-]mmyear", -1.8971199848858813),
|
|
|
|
|
("<hour-of-day> <integer> (as relative minutes)year",
|
|
|
|
|
-1.8971199848858813),
|
|
|
|
|
("minuteyear", -1.8971199848858813)],
|
|
|
|
|
n = 5}}),
|
|
|
|
|
("el <day-of-month> (non ordinal)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -6.453852113757118e-2,
|
|
|
|
|
unseen = -2.890371757896165,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("integer (numeric)", -0.2682639865946794),
|
|
|
|
|
("number (0..15)", -1.4469189829363254)],
|
|
|
|
|
n = 15},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -2.772588722239781, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList [("integer (numeric)", -0.40546510810816444)],
|
|
|
|
|
n = 1}}),
|
|
|
|
|
("segundo (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 3},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<named-month> <day-of-month>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.6094379124341003,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("named-monthinteger (numeric)", -0.6931471805599453),
|
|
|
|
|
("month", -0.6931471805599453)],
|
|
|
|
|
n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("dia (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("in the <part-of-day>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.3101549283038396, unseen = -3.295836866004329,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("afternoon", -1.466337068793427),
|
|
|
|
|
("hour", -0.7731898882334817), ("evening", -2.159484249353372),
|
|
|
|
|
("morning", -1.6486586255873816)],
|
|
|
|
|
n = 11},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.3217558399823195,
|
|
|
|
|
unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("afternoon", -1.3862943611198906),
|
|
|
|
|
("hour", -0.8754687373538999),
|
|
|
|
|
("morning", -1.3862943611198906)],
|
|
|
|
|
n = 4}}),
|
|
|
|
|
("morning",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.40546510810816444,
|
|
|
|
|
unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.0986122886681098,
|
|
|
|
|
unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2}}),
|
|
|
|
|
("week-end",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("Nochevieja",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("hora (grain)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -0.15415067982725836,
|
|
|
|
|
unseen = -2.0794415416798357,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 6},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -1.9459101490553135,
|
|
|
|
|
unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 1}}),
|
|
|
|
|
("right now",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 4},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("Navidad",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -1.3862943611198906,
|
|
|
|
|
likelihoods = HashMap.fromList [("", 0.0)], n = 2},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -0.6931471805599453,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}}),
|
|
|
|
|
("<hour-of-day> <integer> (as relative minutes)",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.0986122886681098,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.5649493574615367,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("time-of-day (latent)integer (numeric)", -0.6931471805599453),
|
|
|
|
|
("hour", -0.6931471805599453)],
|
|
|
|
|
n = 5}}),
|
|
|
|
|
("ce <time>",
|
|
|
|
|
Classifier{okData =
|
|
|
|
|
ClassData{prior = 0.0, unseen = -2.772588722239781,
|
|
|
|
|
likelihoods =
|
|
|
|
|
HashMap.fromList
|
|
|
|
|
[("season", -1.6094379124341003), ("day", -1.3217558399823195),
|
|
|
|
|
("named-day", -2.0149030205422647),
|
|
|
|
|
("hour", -1.6094379124341003),
|
|
|
|
|
("week-end", -1.6094379124341003)],
|
|
|
|
|
n = 5},
|
|
|
|
|
koData =
|
|
|
|
|
ClassData{prior = -infinity, unseen = -1.791759469228055,
|
|
|
|
|
likelihoods = HashMap.fromList [], n = 0}})]
|
