stanford_ner.prop

#
# NOTE: most of this features were gathered by looking at the source code, of some files:
#
# https://github.com/stanfordnlp/CoreNLP/blob/master/src/edu/stanford/nlp/ie/NERFeatureFactory.java
# https://github.com/stanfordnlp/CoreNLP/blob/master/src/edu/stanford/nlp/ie/crf/CRFClassifier.java
# https://github.com/stanfordnlp/CoreNLP/blob/master/src/edu/stanford/nlp/sequences/SeqClassifierFlags.java
#
#

#################
# Files locations
#################
# serializeTo = CINTIL-BIO-CRF-model.ser.gz
# trainFile =
# testFile =

# mapping of columns meanings in test/train file
map = word=0,answer=1



########################
# General/Misc. Features
########################

# The number of words on each side of the current word that are included in the disjunction features
disjunctionWidth = 3

# Include in features giving disjunctions of words anywhere in the left or right disjunctionWidth
# words (preserving direction but not position)
useDisjunctive = true

# Gives you feature for w
useWord = true

# Include word and tag pair features
useWordTag = true

# Gives you feature for (pw,c), and together with other options enables other previous features, such as (pt,c) [with useTags)
usePrev = true

# Gives you feature for (nw,c), and together with other options enables other next features, such as (nt,c) [with useTags)
useNext = true

# Gives you features for (t,c), (pt,c) [if usePrev], (nt,c) [if useNext]
useTags = true

# Gives you features for (pw, w, c) and (w, nw, c)
useWordPairs = true



# Gives you features (pt, t, nt, c), (t, nt, c), (pt, t, c)
#useSymTags = false
#useSymWordPairs = false

# Use extra second order class sequence features when previous is CoNLL boundary, so entity knows it can span boundary.
#useBoundarySequences = false

# As an override to whatever other options are in effect, deletes all features other than
# C and CpC clique features when building the classifier
#strictlyFirstOrder = false

# Include in features giving disjunctions of words anywhere in the left or
# right wideDisjunctionWidth words
# (preserving direction but not position)
#useWideDisjunctive = false
#wideDisjunctionWidth = 8

# Use combination of initial position in sentence and class (and word shape)
# as a feature. (Doesn't seem to help.)
#useBeginSent = true

# Iff the prev word is of length 3 or less, add an extra feature that combines
# the word two back and the current word's shape. Weird!
#useLastRealWord = false
#useNextRealWord = false



###########################
# Sequence-related features
###########################

# Does not use any class combination features using previous classes if this is false
#usePrevSequences = true

# Does not use any class combination features using next classes if this is false
#useNextSequences = true

# Use plain higher-order state sequences out to minimum of length or maxLeft
#useLongSequences = false

# Does not use any class combination features if this is false
#useSequences = true

# Use first, second, and third order class and tag sequence interaction features
#useTaggySequences = false

# Add in sequences of tags with just current class features
#useExtraTaggySequences = false

# Add in terms that join sequences of 2 or 3 tags with the current shape
#useTaggySequencesShapeInteraction = false

# Don't extend the range of useTaggySequences when maxLeft is increased.
# dontExtendTaggy = false



########################
# Shape-related features
########################

# Either "none" for no wordShape use, or the name of a word shape function recognized by
# WordShapeClassifier.lookupShaper(String)
#
# https://github.com/stanfordnlp/CoreNLP/blob/master/src/edu/stanford/nlp/process/WordShapeClassifier.java
wordShape = "chris2useLC"

# Include features giving disjunctions of word shapes anywhere in the left or right disjunctionWidth
# words (preserving direction but not position)
useDisjShape = true

# Conjoin shape with tag or position
useShapeConjunctions = true

# Include in features giving disjunctions of words anywhere in the left or right disjunctionWidth
# words (preserving direction but not position) interacting with the word shape of the current word
useDisjunctiveShapeInteraction = false

# Some first order word shape patterns.
# useTypeySequences = true



################
# Entity Schema
################

# If set, convert the labeling of classes (but not the background) into one of several alternate encodings
# (IO, IOB1, IOB2, IOE1, IOE2, SBIEO, with a S(ingle), B(eginning), E(nding), I(nside) 4-way classification for each class.
# By default, we either do no re-encoding, or the CoNLLDocumentIteratorFactory does a lossy encoding as IO.
# Note that this is all CoNLL-specific, and depends on their way of prefix encoding classes, and is only implemented
# by the CoNLLDocumentIteratorFactory.
entitySubclassification = "IOB2"

# If true, rather than undoing a recoding of entity tag subtypes (such as BIO variants), just leave them in the output.
retainEntitySubclassification = true



#########
# N-Grams
#########

# Make features from letter n-grams, i.e., substrings of the word
#useNGrams = false

# Make features from letter n-grams only lowercase
#lowercaseNGrams = false

# Remove hyphens before making features from letter n-grams
#dehyphenateNGrams = false

# Conjoin word shape and n-gram features
#conjoinShapeNGrams = false

# Use letter n-grams for the previous and current words in the CpC clique. This feature helps languages such as Chinese, but not so much for English
#useNeighborNGrams = false

# If true, record the NGram features that correspond to a String (under the current option settings) and reuse rather than recalculating if the String is seen again.
#cacheNGrams = false

# Do not include character n-gram features for n-grams that contain neither the beginning or end of the word
#noMidNGrams = true

# If this number is positive, n-grams above this size will not be used in the model
#maxNGramLeng = -1



##########################
# Gazetteer and Name lists
##########################
useGazettes = true

# one or more filenames (names separated by a comma, semicolon or space).
gazette=resources/DBPedia-pt-per-gazette.txt;resources/DBPedia-pt-loc-gazette.txt;resources/DBPedia-pt-org-gazette.txt

# If true, a gazette feature fires when all tokens of a gazette entry match
cleanGazette = true

# If true, a gazette feature fires when any token of a gazette entry matches
sloppyGazette = false

# undocumented features, but part of the featuresC() method in NERFeatureFactory.java
checkNameList=true
lastNameList=resources/gazettes/lastNames.txt
maleNameList=resources/gazettes/all-first-names.txt
femaleNameList=resources/gazettes/all-first-names.txt

# Match a word against a list of name titles (Mr, Mrs, etc.). Doesn't really seem to help.
# useTitle = true

# Match a word against a better list of English name titles (Mr, Mrs, etc.). Still doesn't really seem to help.
# useTitle2 = true

# This is a very engineered feature designed to capture multiple references to names
useOccurrencePatterns = true



################
# Word Clusters
################

# Load a file of distributional similarity classes (specified by distSimLexicon) and use it for features
useDistSim = true

# Files should be formatted as tab separated rows where each row is a word/class pair. alexclark=word first, terrykoo=class first
# distSimFileFormat = alexclark

# The file to be loaded for distsim classes.
distSimLexicon = resources/word_cluster.txt

# keep word case when reading from the lexicon
casedDistSim = true



#########################
# Optimization parameters
#
# Note:
#   - see getMinimizer() method on CRFClassifier.java for much more flags and different optimization methods
#   - https://nlp.stanford.edu/nlp/javadoc/javanlp/edu/stanford/nlp/sequences/SeqClassifierFlags.html

###########################################################################
# parameters for Orthant-Wise Limited-memory Quasi-Newton algorithm (OWL-QN)
useOWLQN = true

# L1-prior used in QNMinimizer's OWLQN
priorLambda = 0.1

# terminate on maximum number of iteration
maxQNItr = 125

# Convergence tolerance in optimization
tolerance = 1e-4


###########################################
# parameters for Stochastic Gradient Descent
#useInPlaceSGD = false

# If this number is greater than 0, specifies the number of SGD passes over entire training set) to
# do before giving up (default is 50). Can be smaller if sample size is very large.
#SGDPasses = 100

# If this number is greater than 0, specifies the number of samples to use for tuning (default is 1000).
#tuneSampleSize = 1000

# ???
#sigma = 20

# Stochastic Batch Gradient Descent Size
#stochasticBatchSize = 15


# Use SGD (SGD version selected by useInPlaceSGD or useSGD) for a certain number of passes (SGDPasses)
# and then switches to QN. Gives the quick initial convergence of SGD, with the desired convergence
# criterion of QN (there is some ramp up time for QN). NOTE: Remember to set useQN to false
#useSGDtoQN = false
#useQN = true



#######################################
# Applies only to English (I suppose..)
#######################################

# If this is true, capitalization of day and month names is normalized to lowercase
#normalizeTerms = false

# If this is true, capitalization of day and month names is normalized to lowercase
#normalizeTimex = false

# Include the lemma of a word as a feature.
#useLemmas = false

# Include the previous/next lemma of a word as a feature.
#usePrevNextLemmas = false

# Include the lemma of a word as a feature.
#useLemmaAsWord = false



########################################
# Save features and extra stuff to file
########################################

# A parameter to the printing style, which may give, for example the number of parameters to print
#printClassifierParam = 100

# Print out all feature/class pairs and their weight, and then for each input data point,
# print justification (weights) for active features
#justify = true

# Print out all the features generated by the classifier for a dataset to a file based on this
# name (starting with "features-", suffixed "-1" and "-2" for train and test).
# This simply prints the feature names, one per line.
#printFeatures = portuguese

# Print out features for only the first this many datums, if the value is positive.
#printFeaturesUpto = -1

# Style in which to print the classifier. One of: HighWeight, HighMagnitude, Collection, AllWeights, WeightHistogram
#printClassifier = "AllWeights"