com.johnsnowlabs.nlp.annotators.TokenizerModel.scala Maven / Gradle / Ivy
/*
* Copyright 2017-2022 John Snow Labs
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.johnsnowlabs.nlp.annotators
import com.johnsnowlabs.nlp._
import com.johnsnowlabs.nlp.annotators.common._
import com.johnsnowlabs.nlp.serialization.StructFeature
import com.johnsnowlabs.nlp.util.regex.RuleFactory
import org.apache.spark.ml.param.{BooleanParam, IntParam, Param, StringArrayParam}
import org.apache.spark.ml.util.Identifiable
import scala.collection.mutable
import scala.util.matching.Regex
/** Tokenizes raw text into word pieces, tokens. Identifies tokens with tokenization open
* standards. A few rules will help customizing it if defaults do not fit user needs.
*
* This class represents an already fitted Tokenizer model.
*
* See the main class [[Tokenizer]] for more examples of usage.
*
* @param uid
* required uid for storing annotator to disk
* @groupname anno Annotator types
* @groupdesc anno
* Required input and expected output annotator types
* @groupname Ungrouped Members
* @groupname param Parameters
* @groupname setParam Parameter setters
* @groupname getParam Parameter getters
* @groupname Ungrouped Members
* @groupprio param 1
* @groupprio anno 2
* @groupprio Ungrouped 3
* @groupprio setParam 4
* @groupprio getParam 5
* @groupdesc param
* A list of (hyper-)parameter keys this annotator can take. Users can set and get the
* parameter values through setters and getters, respectively.
*/
class TokenizerModel(override val uid: String)
extends AnnotatorModel[TokenizerModel]
with HasSimpleAnnotate[TokenizerModel] {
import com.johnsnowlabs.nlp.AnnotatorType._
/** rules
*
* @group param
*/
val rules: StructFeature[RuleFactory] = new StructFeature[RuleFactory](this, "rules")
/** Words that won't be affected by tokenization rules
*
* @group param
*/
val exceptions: StringArrayParam =
new StringArrayParam(this, "exceptions", "Words that won't be affected by tokenization rules")
/** Whether to care for case sensitiveness in exceptions (Default: `true`)
*
* @group param
*/
val caseSensitiveExceptions: BooleanParam = new BooleanParam(
this,
"caseSensitiveExceptions",
"Whether to care for case sensitiveness in exceptions")
/** pattern to grab from text as token candidates. Defaults \\S+
*
* @group param
*/
val targetPattern: Param[String] = new Param(
this,
"targetPattern",
"pattern to grab from text as token candidates. Defaults \\S+")
/** Set the minimum allowed length for each token
*
* @group param
*/
val minLength = new IntParam(this, "minLength", "Set the minimum allowed length for each token")
/** Set the maximum allowed length for each token
*
* @group param
*/
val maxLength = new IntParam(this, "maxLength", "Set the maximum allowed length for each token")
/** character list used to separate from the inside of tokens
*
* @group param
*/
val splitChars: StringArrayParam = new StringArrayParam(
this,
"splitChars",
"character list used to separate from the inside of tokens")
/** pattern to separate from the inside of tokens. takes priority over splitChars.
*
* @group param
*/
val splitPattern: Param[String] = new Param(
this,
"splitPattern",
"pattern to separate from the inside of tokens. takes priority over splitChars.")
setDefault(targetPattern -> "\\S+", caseSensitiveExceptions -> true)
/** Output annotator type : TOKEN
*
* @group anno
*/
override val outputAnnotatorType: AnnotatorType = TOKEN
/** Input annotator type : DOCUMENT
*
* @group anno
*/
override val inputAnnotatorTypes: Array[AnnotatorType] = Array[AnnotatorType](
DOCUMENT
) // A Tokenizer could require only for now a SentenceDetector annotator
def this() = this(Identifiable.randomUID("REGEX_TOKENIZER"))
/** pattern to grab from text as token candidates. Defaults \\S+
*
* @group setParam
*/
def setTargetPattern(value: String): this.type = set(targetPattern, value)
/** pattern to grab from text as token candidates. Defaults \\S+
*
* @group getParam
*/
def getTargetPattern: String = $(targetPattern)
/** List of 1 character string to split tokens inside, such as hyphens. Ignored if using infix,
* prefix or suffix patterns.
*
* @group setParam
*/
def setSplitPattern(value: String): this.type = set(splitPattern, value)
/** List of 1 character string to split tokens inside, such as hyphens. Ignored if using infix,
* prefix or suffix patterns.
*
* @group getParam
*/
def getSplitPattern: String = $(splitPattern)
/** Words that won't be affected by tokenization rules
*
* @group setParam
*/
def setExceptions(value: Array[String]): this.type = set(exceptions, value)
/** Words that won't be affected by tokenization rules
*
* @group getParam
*/
def getExceptions: Array[String] = $(exceptions)
/** Rules factory for tokenization
*
* @group setParam
*/
def setRules(ruleFactory: RuleFactory): this.type = set(rules, ruleFactory)
/** Whether to follow case sensitiveness for matching exceptions in text
*
* @group setParam
*/
def setCaseSensitiveExceptions(value: Boolean): this.type = set(caseSensitiveExceptions, value)
/** Whether to follow case sensitiveness for matching exceptions in text
*
* @group getParam
*/
def getCaseSensitiveExceptions(value: Boolean): Boolean = $(caseSensitiveExceptions)
/** Set the minimum allowed length for each token
*
* @group setParam
*/
def setMinLength(value: Int): this.type = set(minLength, value)
/** Set the minimum allowed length for each token
*
* @group getParam
*/
def getMinLength(value: Int): Int = $(minLength)
/** Set the maximum allowed length for each token
*
* @group setParam
*/
def setMaxLength(value: Int): this.type = set(maxLength, value)
/** Set the maximum allowed length for each token
*
* @group getParam
*/
def getMaxLength(value: Int): Int = $(maxLength)
/** List of 1 character string to split tokens inside, such as hyphens. Ignored if using infix,
* prefix or suffix patterns.
*
* @group setParam
*/
def setSplitChars(v: Array[String]): this.type = {
require(
v.forall(x => x.length == 1 || (x.length == 2 && x.substring(0, 1) == "\\")),
"All elements in context chars must have length == 1")
set(splitChars, v)
}
/** One character string to split tokens inside, such as hyphens. Ignored if using infix, prefix
* or suffix patterns.
*
* @group setParam
*/
def addSplitChars(v: String): this.type = {
require(
v.length == 1 || (v.length == 2 && v.substring(0, 1) == "\\"),
"Context char must have length == 1")
set(splitChars, get(splitChars).getOrElse(Array.empty[String]) :+ v)
}
/** List of 1 character string to split tokens inside, such as hyphens. Ignored if using infix,
* prefix or suffix patterns
*
* @group getParam
* .
*/
def getSplitChars: Array[String] = {
$(splitChars)
}
private val PROTECT_CHAR = "ↈ"
private val BREAK_CHAR = "ↇ"
private lazy val BREAK_PATTERN: Regex =
("[^(?:" + $(targetPattern) + ")" + PROTECT_CHAR + "]").r
private lazy val SPLIT_PATTERN: Regex =
("[^" + BREAK_CHAR + "]+").r
private var compiledExceptions: Option[Regex] = None
private def getOrCompileExceptionPattern(): Regex = compiledExceptions.getOrElse {
get(exceptions) match {
case None => "".r
case Some(patterns) =>
val compiledPattern: Regex = patterns
.map { pattern =>
val casedExceptionPattern =
if ($(caseSensitiveExceptions)) pattern else "(?i)" + pattern
casedExceptionPattern
}
.mkString("|")
.r
compiledExceptions = Some(compiledPattern)
compiledPattern
}
}
/** This func generates a Seq of TokenizedSentences from a Seq of Sentences.
*
* @param sentences
* to tag
* @return
* Seq of TokenizedSentence objects
*/
def tag(sentences: Seq[Sentence]): Seq[TokenizedSentence] = {
lazy val splitCharsExists = $(splitChars).map(_.last.toString)
sentences.map { text =>
/** Step 1, define breaks from non breaks */
val exceptionsDefined = get(exceptions).isDefined
var exceptionsWithoutBreak: Option[mutable.HashSet[String]] = None
val textContent = if (exceptionsDefined) {
/** If found, replace BREAK_PATTERN with PROTECT_CHAR, otherwise add to an exception list.
*/
getOrCompileExceptionPattern().replaceAllIn(
text.content,
{ m: Regex.Match =>
{
val breakReplaced = BREAK_PATTERN.replaceAllIn(m.matched, PROTECT_CHAR)
if (breakReplaced == m.matched)
exceptionsWithoutBreak.getOrElse({
exceptionsWithoutBreak = Some(new mutable.HashSet[String])
exceptionsWithoutBreak.get
}) += m.matched
breakReplaced
}
})
} else {
text.content
}
val protectedText = BREAK_PATTERN.replaceAllIn(textContent, BREAK_CHAR)
/** Step 2, Return protected tokens back into text and move on */
val tokens = SPLIT_PATTERN
.findAllMatchIn(protectedText)
.flatMap { candidate =>
/** If exceptions are defined, check for candidate whether PROTECT_CHAR present or in
* exception list.
*/
if (exceptionsDefined &&
(candidate.matched.contains(PROTECT_CHAR) ||
(exceptionsWithoutBreak.isDefined && exceptionsWithoutBreak.get.contains(
candidate.matched)))) {
/** Put back character and move on */
Seq(
IndexedToken(
text.content.slice(candidate.start, candidate.end),
text.start + candidate.start,
text.start + candidate.end - 1))
} else {
/** Step 3, If no exception found, find candidates through the possible general rule
* patterns
*/
val rr = $$(rules)
.findMatchFirstOnly(candidate.matched)
.map { m =>
var curPos = m.content.start
(1 to m.content.groupCount)
.flatMap(i => {
val target = m.content.group(i)
val applyPattern = isSet(splitPattern) && (target
.split($(splitPattern))
.length > 1)
val applyChars = isSet(splitChars) && splitCharsExists.exists(target.contains)
def defaultCandidate = {
val it = IndexedToken(
target,
text.start + candidate.start + curPos,
text.start + candidate.start + curPos + target.length - 1)
curPos += target.length
Seq(it)
}
if (target.nonEmpty && (applyPattern || applyChars)) {
try {
val strs =
if (applyPattern) target.split($(splitPattern))
else target.split($(splitChars).mkString("|"))
strs.map { str =>
curPos = m.content.matched.indexOf(str, curPos)
val indexedToken = IndexedToken(
str,
text.start + candidate.start + curPos,
text.start + candidate.start + curPos + str.length - 1)
curPos += str.length
indexedToken
}
} catch {
case e: Throwable =>
logWarning(s"Tokenizer: Could not apply custom pattern or char. " +
s"Using default split pattern for \'$target\'. Exception: ${e.toString}")
defaultCandidate
}
} else defaultCandidate
})
/** Step 4, If rules didn't match, return whatever candidate we have and leave it as
* is
*/
}
.getOrElse(Seq(IndexedToken(
candidate.matched,
text.start + candidate.start,
text.start + candidate.end - 1)))
rr
}
}
.filter(t =>
t.token.nonEmpty && t.token.length >= $(minLength) && get(maxLength).forall(m =>
t.token.length <= m))
.toArray
TokenizedSentence(tokens, text.index)
}
}
/** one to many annotation */
override def annotate(annotations: Seq[Annotation]): Seq[Annotation] = {
val sentences = SentenceSplit.unpack(annotations)
val tokenized = tag(sentences)
TokenizedWithSentence.pack(tokenized)
}
}
trait ReadablePretrainedTokenizer
extends ParamsAndFeaturesReadable[TokenizerModel]
with HasPretrained[TokenizerModel] {
override val defaultModelName: Option[String] = Some("token_rules")
/** Java compliant-overrides */
override def pretrained(): TokenizerModel = super.pretrained()
override def pretrained(name: String): TokenizerModel = super.pretrained(name)
override def pretrained(name: String, lang: String): TokenizerModel =
super.pretrained(name, lang)
override def pretrained(name: String, lang: String, remoteLoc: String): TokenizerModel =
super.pretrained(name, lang, remoteLoc)
}
/** This is the companion object of [[TokenizerModel]]. Please refer to that class for the
* documentation.
*/
object TokenizerModel extends ReadablePretrainedTokenizer
© 2015 - 2024 Weber Informatics LLC | Privacy Policy