org.apache.ctakes.drugner.ae

Class DrugMentionAnnotator

java.lang.Object

JCasAnnotator_ImplBase

org.apache.ctakes.drugner.ae.DrugMentionAnnotator

public class DrugMentionAnnotator
extends JCasAnnotator_ImplBase

Finds tokens based on context. There are two major groupings or ranges that will be used to create the additional annotations needed to handle the drug mentions used to represent the status changes.

Field Summary

Fields

Modifier and Type	Field and Description
static java.lang.String	BOUNDARY_ANN_TYPE Annotation type that defines the boundary within which the dictionary hits should be present.
static java.lang.String	DISTANCE
static java.lang.String	DISTANCE_ANN_TYPE Annotation type that is used to count the distance.
private static boolean	handledRanges
private int	iAnnotationType
private int	iBoundaryAnnType
private int[]	intermediateTypesToRemove
private DecimalStrengthFSM	iv_decimalFSM
private DosagesFSM	iv_dosagesFSM
private DurationFSM	iv_durationFSM
private java.util.Set<java.lang.String>	iv_exclusionTagSet
private FormFSM	iv_formFSM
private FractionStrengthFSM	iv_fractionFSM
private FrequencyFSM	iv_frequencyFSM
private FrequencyUnitFSM	iv_frequencyUnitFSM
static Logger	iv_logger
private java.util.Set<java.lang.String>	iv_medicationRelatedSections
private RangeStrengthFSM	iv_rangeFSM
private RouteFSM	iv_routeFSM
private DrugChangeStatusFSM	iv_statusFSM
private StrengthFSM	iv_strengthFSM
private StrengthUnitFSM	iv_strengthUnitFSM
private SubSectionIndicatorFSM	iv_subMedSectionFSM
private SuffixStrengthFSM	iv_suffixFSM
private int	iWindowSize
private static int	NERTypeIdentifier
static int	NO_ANNOTATION_TYPE_SPECIFIED
static int	NO_WINDOW_SIZE_SPECIFIED
static java.lang.String	PARAM_SEGMENTS_MEDICATION_RELATED This identifies the section ids that will be considered in generating DrugMentionAnnotaitons

Constructor Summary

Constructors

Constructor and Description
DrugMentionAnnotator()

Method Summary

Methods

Modifier and Type	Method and Description
private BaseToken	adaptToFSMBaseToken(BaseToken obj)
private void	addAnnotations(JCas jcas, java.util.Set annotations, int type) used by executeFSMs to add annotations for features for DrugMention
private void	addMedicationSpecificAttributes(JCas jcas, DrugMentionAnnotation fromAnnotation, MedicationMention medicationMention)
private float	alignDrugMentionAttributes(java.lang.String strengthText, DrugMention dm, DrugMentionAnnotation drugTokenAnt, DrugMentionAnnotation recurseNER, java.lang.String relatedStatus, boolean statusFound, boolean overrideStatus, boolean maxExists, java.lang.String doseText, java.lang.String frequencyText, java.lang.String frequencyUnitText)
private int	anchorEndofline(int[][] elementSpan, int elementSpanLength, int endPhrase, MedicationMention nea)
private void	executeFSMs(JCas jcas, java.util.List baseTokenList, java.util.List namedE, java.util.List wordTokenList) The namedE consists of a list of the Named Entities (NE) found in the prior method.
private boolean	findCoveredTextInSpan(JCas jcas, int annotationType, int beginOffset, int endOffset, java.lang.String[] searchStrs)
private void	findDrugAttributesInRange(JCas jcas, int begin, int end) finds drug attributes using the given range, this method uses FSM
private int[]	findElementRelativeToNE(int[][] elementSpan, int[][] parenthesisSpan, int elementSpanLength, int parenthesisSpanLength, int priorDrugEnd, int startWithParenNum, int endPhrase, MedicationMention nea)
private void	findFSMInRange(JCas jcas, int begin, int end)
private int	findInPattern(JCas jcas, int begin, int end, int elementType, int[][] location)
private boolean	findNextClosestRightParenRelativeToElement(int spanLength, int[][] elementSpan, Annotation nea, int beginSpan, int startOffset)
private int[]	findNextDrugEntityPost(int spanLength, int[][] elementSpan, MedicationMention nea, int endPhrase)
private int[]	findNextDrugEntityPre(int spanLength, int[][] elementSpan, MedicationMention nea, int priorDrugEnd)
private boolean	findNextParenRelativeToElement(int spanLength, int[][] elementSpan, Annotation nea, int parenEnd, int startOffset)
private boolean	findNextParenRelativeToNE(int spanLength, int[][] elementSpan, MedicationMention nea, int priorDrugEnd)
private boolean	findNextParenRelativeToNE(int spanLength, int[][] elementSpan, MedicationMention nea, int priorDrugEnd, int startOffset)
private int[]	findOffsetsInPattern(JCas jcas, int begin, int end, int elementType, int[][] location, boolean highest)
private int	findTextualStringOffset(java.lang.String text) Finds offset of string the represents end of numeric portion
private java.util.List	findUniqueMentions(java.lang.Object[] holdOutSet) finds unique annotations by their begin offsets
private int[]	generateAdditionalNER(JCas jcas, DrugMentionAnnotation tokenDrugNER, DrugChangeStatusAnnotation drugChangeStatus, int beginSpan, int endSpan, int count, java.util.List globalNER)
private void	generateDrugMentions(JCas jcas, Segment seg, boolean narrativeType)
private void	generateDrugMentionsAndAnnotations(JCas jcas, java.util.List<MedicationMention> nerTokenList, int begin, int end, DrugMentionAnnotation recurseNER, java.lang.String[] relatedStatus, int countNER, java.util.List<DrugMentionAnnotation> globalDrugNER)
private void	generateUidValues(JCas jcas) Generates UID values for all MedicationMention objects.
private int	generateUidValues(JCas jcas, int type, int firstId) Generates UID values for all MedicationMention objects.
private int[]	getAdjustedWindowSpan(JCas jcas, int begin, int end, boolean highestRange) return window span to find reasons for the given d
private java.util.List<Annotation>	getAnnotationsInSpan(JCas jcas, int type, int begin, int end)
private java.util.List	getAnnotationsInSpanWithAdaptToBaseTokenFSM(JCas jcas, int type, int begin, int end)
private int[]	getNarrativeSpansContainingGivenSpanType(JCas jcas, int begin, int annotType)
private Segment	getSegmentContainingOffsets(JCas jcas, int start, int end)
private int[]	getSentenceSpanContainingGivenSpan(JCas jcas, int begin, int end)
private int	getSubSectionAnnotationBodyEnd(Segment currSeg, SubSectionIndicator nextSsi) Decides where the SubSectionAnnotation should end.
private int[][]	getWindowSpan(JCas jcas, java.lang.String sectionType, int typeAnnotation, int begin, int end, boolean useBegin, int sizeArrays)
private boolean	hasMultipleDrugsInSpan(JCas jcas, int begin, int end) Return true if exists more than one drug and reason within the span, otherwise return false
private boolean	hasMultipleElementsInSpan(JCas jcas, int begin, int end) Return true if exists more than one drug and reason within the span, otherwise return false
void	initialize(UimaContext annotCtx)
private boolean	isDuplicate(java.lang.Object[] neArray, int curIdx, int checkIdx)
private int	lastInPattern(JCas jcas, int begin, int end, int elementType, int[][] location)
private boolean	multipleDrugsInSpan(JCas jcas, int begin, int end) Return true if exists more than one drug and reason within the span, otherwise return false
private boolean	multipleElementsInSpan(JCas jcas, int begin, int end) Return true if exists more than one drug and reason within the span, otherwise return false
private void	prepareSubSection(JCas jcas, JFSIndexRepository indexes, java.util.Set subSectionIndSet) Given the set of subSectionInds to parse (via SubSectionIndicatorFSM) create SubSectionAnnotation This method created SubsectionAnnotation and sets the header begin, end as well as body begin and end
void	process(JCas jcas)
private void	removeAnnotations(JCas jcas, int type)
private void	removeDrugNerTypes(JCas jcas, int[] intermediateTypesToRemove) Remove most extra annotation types that we created here but not all as downstream annotators might want to use some such as SubSectionAnnotation
private java.util.List	sortAnnotations(java.lang.Object[] holdOutSet) Sort annotations by begin offset
private int[][]	sortSignatureElements(JCas jcas, int begin, int end, int typeAnnotation, int[] senSpan, int sizeArray) The range defined by the begin and end is searched all discovered drugs, newline locations and related drug signature elements.
private java.util.List	sortSubSectionInd(java.lang.Object[] holdOutSet)
private int[]	statusChangePhraseGenerator(JCas jcas, int begin, int end, boolean maxExists, java.util.Iterator uniqueNER, java.util.Iterator orderedStatusChanges, java.util.List holdStatusChanges, java.lang.String[] relatedStatus, DrugMentionAnnotation drugTokenAnt, java.util.List globalDrugNER, int countNER)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

iv_logger

public static Logger iv_logger

PARAM_SEGMENTS_MEDICATION_RELATED

public static final java.lang.String PARAM_SEGMENTS_MEDICATION_RELATED

This identifies the section ids that will be considered in generating DrugMentionAnnotaitons

See Also:

Constant Field Values

DISTANCE

public static java.lang.String DISTANCE

DISTANCE_ANN_TYPE

public static java.lang.String DISTANCE_ANN_TYPE

Annotation type that is used to count the distance.

BOUNDARY_ANN_TYPE

public static java.lang.String BOUNDARY_ANN_TYPE

Annotation type that defines the boundary within which the dictionary hits should be present.

NO_WINDOW_SIZE_SPECIFIED

public static int NO_WINDOW_SIZE_SPECIFIED

NO_ANNOTATION_TYPE_SPECIFIED

public static int NO_ANNOTATION_TYPE_SPECIFIED

iWindowSize

private int iWindowSize

iAnnotationType

private int iAnnotationType

iBoundaryAnnType

private int iBoundaryAnnType

iv_fractionFSM

private FractionStrengthFSM iv_fractionFSM

iv_rangeFSM

private RangeStrengthFSM iv_rangeFSM

iv_subMedSectionFSM

private SubSectionIndicatorFSM iv_subMedSectionFSM

iv_dosagesFSM

private DosagesFSM iv_dosagesFSM

iv_suffixFSM

private SuffixStrengthFSM iv_suffixFSM

iv_durationFSM

private DurationFSM iv_durationFSM

iv_routeFSM

private RouteFSM iv_routeFSM

iv_frequencyFSM

private FrequencyFSM iv_frequencyFSM

iv_statusFSM

private DrugChangeStatusFSM iv_statusFSM

iv_decimalFSM

private DecimalStrengthFSM iv_decimalFSM

iv_strengthFSM

private StrengthFSM iv_strengthFSM

iv_strengthUnitFSM

private StrengthUnitFSM iv_strengthUnitFSM

iv_frequencyUnitFSM

private FrequencyUnitFSM iv_frequencyUnitFSM

iv_formFSM

private FormFSM iv_formFSM

NERTypeIdentifier

private static final int NERTypeIdentifier

See Also:

Constant Field Values

handledRanges

private static boolean handledRanges

iv_exclusionTagSet

private java.util.Set<java.lang.String> iv_exclusionTagSet

iv_medicationRelatedSections

private java.util.Set<java.lang.String> iv_medicationRelatedSections

intermediateTypesToRemove

private int[] intermediateTypesToRemove

Constructor Detail

DrugMentionAnnotator

public DrugMentionAnnotator()

Method Detail

initialize

public void initialize(UimaContext annotCtx)
                throws ResourceInitializationException

Throws:

ResourceInitializationException

process

public void process(JCas jcas)
             throws AnalysisEngineProcessException

Throws:

AnalysisEngineProcessException

removeDrugNerTypes

private void removeDrugNerTypes(JCas jcas,
                      int[] intermediateTypesToRemove)

Remove most extra annotation types that we created here but not all as downstream annotators might want to use some such as SubSectionAnnotation

Parameters:

jcas -

removeAnnotations

private void removeAnnotations(JCas jcas,
                     int type)

addMedicationSpecificAttributes

private void addMedicationSpecificAttributes(JCas jcas,
                                   DrugMentionAnnotation fromAnnotation,
                                   MedicationMention medicationMention)

generateUidValues

private void generateUidValues(JCas jcas)

Generates UID values for all MedicationMention objects.

generateUidValues

private int generateUidValues(JCas jcas,
                    int type,
                    int firstId)

Generates UID values for all MedicationMention objects.

sortSubSectionInd

private java.util.List sortSubSectionInd(java.lang.Object[] holdOutSet)

sortAnnotations

private java.util.List sortAnnotations(java.lang.Object[] holdOutSet)

Sort annotations by begin offset

Parameters:

holdOutSet -

Returns:

findUniqueMentions

private java.util.List findUniqueMentions(java.lang.Object[] holdOutSet)

finds unique annotations by their begin offsets

Parameters:

holdOutSet -

Returns:

isDuplicate

private boolean isDuplicate(java.lang.Object[] neArray,
                  int curIdx,
                  int checkIdx)

prepareSubSection

private void prepareSubSection(JCas jcas,
                     JFSIndexRepository indexes,
                     java.util.Set subSectionIndSet)
                        throws java.lang.Exception

Given the set of subSectionInds to parse (via SubSectionIndicatorFSM) create SubSectionAnnotation This method created SubsectionAnnotation and sets the header begin, end as well as body begin and end

Parameters:

jcas -

indexes -

subSectionIndSet -

Throws:

java.lang.Exception

getSubSectionAnnotationBodyEnd

private int getSubSectionAnnotationBodyEnd(Segment currSeg,
                                 SubSectionIndicator nextSsi)

Decides where the SubSectionAnnotation should end. This is based on the following: If the next SubSectionIndicator starts before the end of current segment then, use the next SubSectionIndicator.begin - 1 else if the next SubSectionIndicator starts after the end of the current segment then, use the end of the current Segment.end

getSegmentContainingOffsets

private Segment getSegmentContainingOffsets(JCas jcas,
                                  int start,
                                  int end)

addAnnotations

private void addAnnotations(JCas jcas,
                  java.util.Set annotations,
                  int type)

used by executeFSMs to add annotations for features for DrugMention

Parameters:

jcas -

annotations -

type -

findTextualStringOffset

private int findTextualStringOffset(java.lang.String text)

Finds offset of string the represents end of numeric portion

executeFSMs

private void executeFSMs(JCas jcas,
               java.util.List baseTokenList,
               java.util.List namedE,
               java.util.List wordTokenList)
                  throws AnnotatorProcessException

The namedE consists of a list of the Named Entities (NE) found in the prior method. wordTokenList is included in case there are tokens representing drug names not yet discovered. If no NE is found in a provided window size and the provided confidence level has been met then the word token is marked as a NE and added to a cache to hold undiscovered drug names.

Parameters:

jcas -

baseTokenList -

namedE -

wordTokenList -

Throws:

AnnotatorProcessException

generateDrugMentions

private void generateDrugMentions(JCas jcas,
                        Segment seg,
                        boolean narrativeType)
                           throws java.lang.Exception

Throws:

java.lang.Exception

findCoveredTextInSpan

private boolean findCoveredTextInSpan(JCas jcas,
                            int annotationType,
                            int beginOffset,
                            int endOffset,
                            java.lang.String[] searchStrs)

getAnnotationsInSpanWithAdaptToBaseTokenFSM

private java.util.List getAnnotationsInSpanWithAdaptToBaseTokenFSM(JCas jcas,
                                                         int type,
                                                         int begin,
                                                         int end)
                                                            throws java.lang.Exception

Throws:

java.lang.Exception

getAnnotationsInSpan

private java.util.List<Annotation> getAnnotationsInSpan(JCas jcas,
                                              int type,
                                              int begin,
                                              int end)

findDrugAttributesInRange

private void findDrugAttributesInRange(JCas jcas,
                             int begin,
                             int end)
                                throws java.lang.Exception

finds drug attributes using the given range, this method uses FSM

Parameters:

jcas -

begin -

end -

Throws:

java.lang.Exception

generateDrugMentionsAndAnnotations

private void generateDrugMentionsAndAnnotations(JCas jcas,
                                      java.util.List<MedicationMention> nerTokenList,
                                      int begin,
                                      int end,
                                      DrugMentionAnnotation recurseNER,
                                      java.lang.String[] relatedStatus,
                                      int countNER,
                                      java.util.List<DrugMentionAnnotation> globalDrugNER)
                                         throws java.lang.Exception

Throws:

java.lang.Exception

alignDrugMentionAttributes

private float alignDrugMentionAttributes(java.lang.String strengthText,
                               DrugMention dm,
                               DrugMentionAnnotation drugTokenAnt,
                               DrugMentionAnnotation recurseNER,
                               java.lang.String relatedStatus,
                               boolean statusFound,
                               boolean overrideStatus,
                               boolean maxExists,
                               java.lang.String doseText,
                               java.lang.String frequencyText,
                               java.lang.String frequencyUnitText)

statusChangePhraseGenerator

private int[] statusChangePhraseGenerator(JCas jcas,
                                int begin,
                                int end,
                                boolean maxExists,
                                java.util.Iterator uniqueNER,
                                java.util.Iterator orderedStatusChanges,
                                java.util.List holdStatusChanges,
                                java.lang.String[] relatedStatus,
                                DrugMentionAnnotation drugTokenAnt,
                                java.util.List globalDrugNER,
                                int countNER)
                                   throws java.lang.Exception

Throws:

java.lang.Exception

generateAdditionalNER

private int[] generateAdditionalNER(JCas jcas,
                          DrugMentionAnnotation tokenDrugNER,
                          DrugChangeStatusAnnotation drugChangeStatus,
                          int beginSpan,
                          int endSpan,
                          int count,
                          java.util.List globalNER)
                             throws java.lang.Exception

Throws:

java.lang.Exception

adaptToFSMBaseToken

private BaseToken adaptToFSMBaseToken(BaseToken obj)
                               throws java.lang.Exception

Throws:

java.lang.Exception

findNextDrugEntityPost

private int[] findNextDrugEntityPost(int spanLength,
                           int[][] elementSpan,
                           MedicationMention nea,
                           int endPhrase)

findNextDrugEntityPre

private int[] findNextDrugEntityPre(int spanLength,
                          int[][] elementSpan,
                          MedicationMention nea,
                          int priorDrugEnd)

findNextParenRelativeToNE

private boolean findNextParenRelativeToNE(int spanLength,
                                int[][] elementSpan,
                                MedicationMention nea,
                                int priorDrugEnd,
                                int startOffset)

findNextParenRelativeToElement

private boolean findNextParenRelativeToElement(int spanLength,
                                     int[][] elementSpan,
                                     Annotation nea,
                                     int parenEnd,
                                     int startOffset)

hasMultipleDrugsInSpan

private boolean hasMultipleDrugsInSpan(JCas jcas,
                             int begin,
                             int end)

Return true if exists more than one drug and reason within the span, otherwise return false

Parameters:

jcas -

begin -

end -

Returns:

hasMultipleElementsInSpan

private boolean hasMultipleElementsInSpan(JCas jcas,
                                int begin,
                                int end)

Return true if exists more than one drug and reason within the span, otherwise return false

Parameters:

jcas -

begin -

end -

Returns:

getSentenceSpanContainingGivenSpan

private int[] getSentenceSpanContainingGivenSpan(JCas jcas,
                                       int begin,
                                       int end)

Parameters:

jcas -

begin -

end -

Returns:

int[] - int[0] is begin offset and int[1] is end offset

getNarrativeSpansContainingGivenSpanType

private int[] getNarrativeSpansContainingGivenSpanType(JCas jcas,
                                             int begin,
                                             int annotType)

Parameters:

jcas -

begin -

end -

Returns:

int[] - int[0] is begin offset and int[1] is end offset of subsequent sentence end (if available)

findFSMInRange

private void findFSMInRange(JCas jcas,
                  int begin,
                  int end)
                     throws java.lang.Exception

Throws:

java.lang.Exception

findOffsetsInPattern

private int[] findOffsetsInPattern(JCas jcas,
                         int begin,
                         int end,
                         int elementType,
                         int[][] location,
                         boolean highest)

Parameters:

jcas -

begin -

end -

Returns:

getAdjustedWindowSpan

private int[] getAdjustedWindowSpan(JCas jcas,
                          int begin,
                          int end,
                          boolean highestRange)
                             throws java.lang.Exception

return window span to find reasons for the given d

Parameters:

jcas -

Returns:

int[0] is begin offset and int[1] is end offset

Throws:

java.lang.Exception

findInPattern

private int findInPattern(JCas jcas,
                int begin,
                int end,
                int elementType,
                int[][] location)

Parameters:

jcas -

begin -

end -

Returns:

getWindowSpan

private int[][] getWindowSpan(JCas jcas,
                    java.lang.String sectionType,
                    int typeAnnotation,
                    int begin,
                    int end,
                    boolean useBegin,
                    int sizeArrays)
                       throws java.lang.Exception

Throws:

java.lang.Exception

multipleDrugsInSpan

private boolean multipleDrugsInSpan(JCas jcas,
                          int begin,
                          int end)

Return true if exists more than one drug and reason within the span, otherwise return false

Parameters:

jcas -

begin -

end -

Returns:

multipleElementsInSpan

private boolean multipleElementsInSpan(JCas jcas,
                             int begin,
                             int end)

Return true if exists more than one drug and reason within the span, otherwise return false

Parameters:

jcas -

begin -

end -

Returns:

sortSignatureElements

private int[][] sortSignatureElements(JCas jcas,
                            int begin,
                            int end,
                            int typeAnnotation,
                            int[] senSpan,
                            int sizeArray)

The range defined by the begin and end is searched all discovered drugs, newline locations and related drug signature elements. Each drug is assigned an span which is used later assign the drug mention elements. The range begins with the start of the drug mention and ends with the EOL or beginning of the subsequent drug mention. The end range value can be overridden by the following cases: A) The subsequent drug mention is contained in brackets or parenthesis w/out other drug signature items and there are no other signature elments that exist between the mentions; e.g. "ibuprofen [Motrin]". In this case the drug span will end with the right most signature element or EOL (which ever is farthest, but before a subsequent drug mention). B) If there are multiple occurrences of signature elements between mentions find the second positive offset in the series Up to 100 of each type of signature item and drug mentions are allowed. If more than that are available the ArrayIndexOutOfBoundsException is thrown and alternate algorithm is run

Parameters:

jcas -

begin -

end -

typeAnnotation -

senSpan -

Returns:

findNextParenRelativeToNE

private boolean findNextParenRelativeToNE(int spanLength,
                                int[][] elementSpan,
                                MedicationMention nea,
                                int priorDrugEnd)

lastInPattern

private int lastInPattern(JCas jcas,
                int begin,
                int end,
                int elementType,
                int[][] location)

Parameters:

jcas -

begin -

end -

Returns:

findElementRelativeToNE

private int[] findElementRelativeToNE(int[][] elementSpan,
                            int[][] parenthesisSpan,
                            int elementSpanLength,
                            int parenthesisSpanLength,
                            int priorDrugEnd,
                            int startWithParenNum,
                            int endPhrase,
                            MedicationMention nea)

anchorEndofline

private int anchorEndofline(int[][] elementSpan,
                  int elementSpanLength,
                  int endPhrase,
                  MedicationMention nea)

findNextClosestRightParenRelativeToElement

private boolean findNextClosestRightParenRelativeToElement(int spanLength,
                                                 int[][] elementSpan,
                                                 Annotation nea,
                                                 int beginSpan,
                                                 int startOffset)