src.dackar.causal.CausalBase

Created on April, 2024

@author: wangc, mandd

Attributes

logger
_corefAvail
ver

Classes

CausalBase

Base Class for Causal Analysis

Module Contents

src.dackar.causal.CausalBase.logger

src.dackar.causal.CausalBase._corefAvail = False

src.dackar.causal.CausalBase.ver

class src.dackar.causal.CausalBase.CausalBase(nlp, entID='SSC', causalKeywordID='causal', *args, **kwargs)

Bases: object

Base Class for Causal Analysis

type = 'CausalBase'

name = 'CausalBase'

nlp

_causalFile

_causalPOS

_causalKeywords

_statusFile

_statusKeywords

_updateStatusKeywords = False

_updateCausalKeywords = False

_conjectureFile

_conjectureKeywords

_doc = None

entityRuler = None

_entityRuler = False

_entityRulerMatches = []

_matchedSents = []

_matchedSentsForVis = []

_visualizeMatchedSents = True

_coref = False

_entityLabels

_entID = 'SSC'

_causalKeywordID = 'causal'

_causalNames = ['cause', 'cause health status', 'causal keyword', 'effect', 'effect health status', 'sentence',...

_relationNames = ['subject', 'relation', 'object']

_extractedCausals = []

_causalSentsNoEnts = []

_rawCausalList = []

_causalSentsOneEnt = []

_entHS = None

_entStatus = None

_screen = False

dataframeRelations = None

dataframeEntities = None

_causalRelation = None

_causalRelationGeneral = None

_relationGeneral = None

_allRelPairs = []

_textProcess

reset()

Reset rule-based matcher

getAttribute(name)

Get self attribute data

Parameters:

name (str) – name of protected variable

Returns:

attribute data

Return type:

pandas.DataFrame

textProcess()

Function to clean text

Parameters:

None

Returns:

procObj, DACKAR.Preprocessing object

getKeywords(filename, columnNames=None)

Get the keywords from given file

Parameters:

filename – str, the file name to read the keywords

Returns:

dict, dictionary contains the keywords

Return type:

kw

extractLemma(varList)

Lammatize the variable list

Parameters:

varList – list, list of variables

Returns:

list, list of lammatized variables

Return type:

lemmaList

addKeywords(keywords, ktype)

Method to update self._causalKeywords or self._statusKeywords

Parameters:

• keywords – dict, keywords that will be add to self._causalKeywords or self._statusKeywords

• ktype – string, either ‘status’ or ‘causal’

addEntityPattern(name, patternList)

Add entity pattern, to extend doc.ents, similar function to self.extendEnt

Parameters:

• name – str, the name for the entity pattern.

• patternList – list, the pattern list, for example:

• {"label" – “GPE”, “pattern”: [{“LOWER”: “san”}, {“LOWER”: “francisco”}]}

__call__(text, extract=True, screen=False)

Find all token sequences matching the supplied pattern

Parameters:

text – string, the text that need to be processed

Returns:

None

abstract extractInformation()

extract information

Parameters:

None

Returns:

None

visualize()

Visualize the processed document

Parameters:

None

Returns:

None

isPassive(token)

Check the passiveness of the token

Parameters:

token – spacy.tokens.Token, the token of the doc

Returns:

True, if the token is passive

Return type:

isPassive

isConjecture(token)

Check the conjecture of the token

Parameters:

token – spacy.tokens.Token, the token of the doc, the token should be the root of the Doc

Returns:

True, if the token/sentence indicates conjecture

Return type:

isConjecture

isNegation(token)

Check negation status of given token

Parameters:

token – spacy.tokens.Token, token from spacy.tokens.doc.Doc

Returns:

tuple, the negation status and the token text

Return type:

(neg, text)

findVerb(doc)

Find the first verb in the doc

Parameters:

doc – spacy.tokens.doc.Doc, the processed document using nlp pipelines

Returns:

spacy.tokens.Token, the token that has VERB pos

Return type:

token

getCustomEnts(ents, labels)

Get the custom entities

Parameters:

• ents – list, all entities from the processed doc

• labels – list, list of labels to be used to get the custom entities out of “ents”

Returns:

list, the customEnts associates with the “labels”

Return type:

customEnts

getPhrase(ent, start, end, include=False)

Get the phrase for ent with all left children

Parameters:

• ent – Span, the ent to amend with all left children

• start – int, the start index of ent

• end – int, the end index of ent

• include – bool, include ent in the returned expression if True

Returns:

Span or Token, the identified status

Return type:

status

getAmod(ent, start, end, include=False)

Get amod tokens for ent

Parameters:

• ent – Span, the ent to amend with all left children

• start – int, the start index of ent

• end – int, the end index of ent

• include – bool, include ent in the returned expression if True

Returns:

Span or Token, the identified status

Return type:

status

getAmodOnly(ent)

Get amod tokens texts for ent

Parameters:

ent – Span, the ent to amend with all left children

Returns:

list, the list of amods for ent

Return type:

amod

getCompoundOnly(headEnt, ent)

Get the compounds for headEnt except ent

Parameters:

headEnt – Span, the head entity to ent

Returns:

list, the list of compounds for head ent

Return type:

compDes

getNbor(token)

Method to get the nbor from token, return None if nbor is not exist

Parameters:

token – Token, the provided Token to request nbor

Returns:

Token, the requested nbor

Return type:

nbor

validSent(sent)

Check if the sentence has valid structure, either contains subject or object

Parameters:

sent – Span, sentence from user provided text

Returns:

bool, False if the sentence has no subject and object.

Return type:

valid

findLeftSubj(pred, passive)

Find closest subject in predicates left subtree or predicates parent’s left subtree (recursive). Has a filter on organizations.

Parameters:

• pred – spacy.tokens.Token, the predicate token

• passive – bool, True if passive

Returns:

spacy.tokens.Token, the token that represent subject

Return type:

subj

findRightObj(pred, deps=['dobj', 'pobj', 'iobj', 'obj', 'obl', 'oprd'], exclPrepos=[])

Find closest object in predicates right subtree. Skip prepositional objects if the preposition is in exclude list. Has a filter on organizations.

Parameters:

• pred – spacy.tokens.Token, the predicate token

• exclPrepos – list, list of the excluded prepositions

findRightKeyword(pred, exclPrepos=[])

Find Skip prepositional objects if the preposition is in exclude list. Has a filter on organizations.

Parameters:

• pred – spacy.tokens.Token, the predicate token

• exclPrepos – list, list of the excluded prepositions

findHealthStatus(root, deps)

Return first child of root (included) that matches dependency list by breadth first search. Search stops after first dependency match if firstDepOnly (used for subject search - do not “jump” over subjects)

Parameters:

• root – spacy.tokens.Token, the root token

• deps – list, the dependency list

Returns:

token, the token represents the health status

Return type:

child

isValidCausalEnts(ent)

Check the entity if it belongs to the valid causal entities

Args:

ent: list, list of entities

Returns:

valid: bool, valid cansual ent if True

getIndex(ent, entList)

Get index for ent in entList

Parameters:

• ent – Span, ent that is used to get index

• entList – list, list of entities

Returns:

int, the index for ent

Return type:

idx

getConjuncts(entList)

Get a list of conjuncts from entity list

Parameters:

entList – list, list of entities

Returns:

list, list of conjuncts

Return type:

conjunctList

collectSents(doc)

collect data of matched sentences that can be used for visualization

Args:

doc: spacy.tokens.doc.Doc, the processed document using nlp pipelines

extract(sents, predSynonyms=[], exclPrepos=[])

General extraction method

Parameters:

• sents – list, the list of sentences

• predSynonyms – list, the list of predicate synonyms

• exclPrepos – list, the list of exlcuded prepositions

Returns:

generator, the extracted causal relation

Return type:

(subject tuple, predicate, object tuple)

bfs(root, deps)

Return first child of root (included) that matches entType and dependency list by breadth first search. Search stops after first dependency match if firstDepOnly (used for subject search - do not “jump” over subjects)

Parameters:

• root – spacy.tokens.Token, the root token

• deps – list, list of dependency

Returns:

spacy.tokens.Token, the matched token

Return type:

child

findSubj(pred, passive)

Find closest subject in predicates left subtree or predicates parent’s left subtree (recursive). Has a filter on organizations.

Parameters:

• pred – spacy.tokens.Token, the predicate token

• passive – bool, True if the predicate token is passive

Returns:

spacy.tokens.Token, the token that represents subject

Return type:

subj

findObj(pred, deps=['dobj', 'pobj', 'iobj', 'obj', 'obl'], exclPrepos=[])

Find closest object in predicates right subtree. Skip prepositional objects if the preposition is in exclude list. Has a filter on organizations.

Parameters:

• pred – spacy.tokens.Token, the predicate token

• exclPrepos – list, the list of prepositions that will be excluded

Returns:

spacy.tokens.Token,, the token that represents the object

Return type:

obj

isValidKeyword(var, keywords)

Parameters:

• var – token

• keywords – list/dict

Returns: True if the var is a valid among the keywords

getStatusForSubj(ent, include=False)

Get the status for nsubj/nsubjpass ent

Parameters:

• ent – Span, the nsubj/nsubjpass ent that will be used to search status

• include – bool, include ent in the returned expression if True

Returns:

Span or Token, the identified status

Return type:

status

getStatusForObj(ent, include=False)

Get the status for pobj/dobj ent

Parameters:

• ent – Span, the pobj/dobj ent that will be used to search status

• include – bool, include ent in the returned expression if True

Returns:

Span or Token, the identified status

Return type:

status

getStatusForPobj(ent, include=False)

Get the status for ent root pos pobj

Parameters:

• ent – Span, the span of entity

• include – bool, ent will be included in returned status if True

Returns:

Span or Token, the identified health status

extractRelDep(matchedSents)

Extract general relationship between entities :param matchedSents: list, the list of matched sentences

Returns:

generator, the extracted causal relation

Return type:

(subject tuple, predicate, object tuple)