def getDatasetsOfDatasetDocumentFromDatabase(
dataset_document_name=DEFAULT_DATASET_DOCUMENT_NAME,
db_name=DEFAULT_STATISTICS_DB_NAME):
"""
Gets the given dataset from database.
:param dataset_document_name - Name of dataset to get from database
:param db_name: Name of database to get dataset from.
:return datasets - List of lists
[[DEFAULT_DATASET_LIST_INDEX_TRAINING],
[DEFAULT_DATASET_LIST_INDEX_TEST]]
"""
database = db.couch_database(db_name)
document = database.getDocumentFromDatabase(dataset_document_name)
datasets = []
datasets.insert(DEFAULT_DATASET_LIST_INDEX_TRAINING,
document[DATASET_DOCUMENT_FIELD_NAME_TRAINING_SET])
datasets.insert(DEFAULT_DATASET_LIST_INDEX_TEST,
document[DATASET_DOCUMENT_FIELD_NAME_TEST_SET])
return datasets
def addWordDictionariesToDocumentsOfDatabase(db_name, fields):
"""
Creates an word dictionary (word:occurences) from all documents in given
db_name of given fields.
:param db_name: The name of database to add word dictionaries
:param fields: The fields of documents to use for word dictionary
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
pattern = preprocessor.getWhiteListPattern()
for row in all_docs.rows:
document = row.doc
for field in fields:
counts = Counter()
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
for word in doc_field.split():
counts.update(word.lower()
for word in re.split(pattern, word))
word_dictionary = dict(counts)
new_field = getFieldForWordDictionaryByDocumentField(field)
document[new_field] = word_dictionary
database.updateDocumentInDatabase(document)
def cleanFieldsOfDocumentsFromDatabase(db_name, fields):
"""
Find and replace all (html-) tags from given fields in given database
:param db_name - The db in which tags should be replaced
:param fields - List fields from which html tags should be replaced
"""
pattern = getWhiteListPattern()
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
for row in all_docs.rows:
document = row.doc
for field in fields:
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
cleaned_doc_field = re.sub(pattern, ' ',doc_field)
document[field] = cleaned_doc_field
database.updateDocumentInDatabase(document)
def addWordDictionariesToDocumentsOfDatabase(db_name, fields):
"""
Creates an word dictionary (word:occurences) from all documents in given
db_name of given fields.
:param db_name: The name of database to add word dictionaries
:param fields: The fields of documents to use for word dictionary
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
pattern = preprocessor.getWhiteListPattern()
for row in all_docs.rows:
document = row.doc
for field in fields:
counts = Counter()
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
for word in doc_field.split():
counts.update(word.lower() for word in re.split(pattern,word))
word_dictionary = dict(counts)
new_field = getFieldForWordDictionaryByDocumentField(field)
document[new_field] = word_dictionary
database.updateDocumentInDatabase(document)
def perform_train_test_split(db_name=ds.DEFAULT_DB_NAME,
train_size=ds.DEFAULT_TRAININGSET_SIZE):
"""
Get all document_ids of given database and split's it according to given
train_size.
The tricky part is that we n
:param db_name: Name of database to split documents (default DEFAULT_DB_NAME)
:param train_size: Size in percentage [0,1] of the training set.
:return splitted_dataset - List of lists
[[DEFAULT_DATASET_LIST_INDEX_TRAINING],
[DEFAULT_DATASET_LIST_INDEX_TEST]]
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
doc_ids_list = []
all_tag_list = []
i = 0
for row in all_docs.rows:
document = row.doc
#append the document id to doc_ids_list
doc_ids_list.append(document[cp.COUCHDB_DOCUMENT_FIELD_ID])
tag_list = []
#if document has tags than split and add them
if pp.STACKEXCHANGE_TAGS_COLUM in document.keys():
document_tags = document[pp.STACKEXCHANGE_TAGS_COLUM]
tags_list = document_tags.split(sep=dtm_provider.TAG_SPLIT_separator)
for tag in tags_list:
#remove the closing tag (last item)
tag_list.append(tag[:-1])
#append the list of document tags to all_tag_list
all_tag_list.append(tag_list)
i += 1
if i > 10000:
break
mlb = MultiLabelBinarizer()
tags_encoded = mlb.fit_transform(all_tag_list)
print(len(doc_ids_list))
splitted_dataset = cross_validation.train_test_split(doc_ids_list,tags_encoded,
train_size=0.8, random_state=42,
stratify=tags_encoded)
def insertWordDictionaryIntoDatabase(db_name, document):
"""
Insert a a worddictionary document into db with given name
:param db_name: The name of database to insert document to
:param document: The document to insert into database
"""
database = db.couch_database(db_name)
database.insertDocumentIntoDatabase(document)
def readJsonFromFileAndImportIntoDatabase(db_name, file):
"""
Reads json from given file and imports it into given db_name
:param db_name: Name of database to import json content of file to
:param file: File incl. name and path to import json content from
"""
database = db.couch_database(db_name)
database.bulkInsertDocumentsFromFile(file)
def insertWordDictionaryIntoDatabase(db_name, document):
"""
Insert a a worddictionary document into db with given name
:param db_name: The name of database to insert document to
:param document: The document to insert into database
"""
database = db.couch_database(db_name)
database.insertDocumentIntoDatabase(document)
def performFeatureSelection():
#1.) replicate our current clean database to new database
database = db.couch_database(cp.COUCHDB_CLEANED_NAME)
#2.) Now we work with new clean database
db_name = cp.COUCHDB_CLEANED_WD_NAME
database.replicateCouchDBToNewCouchDB(db_name)
#3.) Define the fields we want to create word dictionaries of
fields = [pp.STACKEXCHANGE_TITLE_COLUMN, pp.STACKEXCHANGE_BODY_COLUMN]
#4) Add the word dictionaries of given fields to documents
wd.addWordDictionariesToDocumentsOfDatabase(db_name, fields)
def performFeatureSelection():
#1.) replicate our current clean database to new database
database = db.couch_database(cp.COUCHDB_CLEANED_NAME)
#2.) Now we work with new clean database
db_name = cp.COUCHDB_CLEANED_WD_NAME
database.replicateCouchDBToNewCouchDB(db_name)
#3.) Define the fields we want to create word dictionaries of
fields = [pp.STACKEXCHANGE_TITLE_COLUMN, pp.STACKEXCHANGE_BODY_COLUMN]
#4) Add the word dictionaries of given fields to documents
wd.addWordDictionariesToDocumentsOfDatabase(db_name, fields)
def performPreprocessor():
"""
Equivalent to the main method, but it needed to be called separate
Performs all relevant preprocessing steps
"""
import_file = pp.RESOURCES_FOLDER + pp.STACKEXCHANGE_JSON_FILE
readJsonFromFileAndImportIntoDatabase(cp.COUCHDB_RAW_NAME,import_file)
importDesignDocumentIntoGivenDatabase(cp.COUCHDB_RAW_NAME)
db_name = cp.COUCHDB_CLEAN_HTML_NAME
FIELDS = [pp.STACKEXCHANGE_TITLE_COLUMN, pp.STACKEXCHANGE_BODY_COLUMN]
database = db.couch_database(cp.COUCHDB_RAW_NAME)
database.replicateCouchDBToNewCouchDB(db_name)
cleanHTMLTagsFromDocumentsInDatabase(db_name, FIELDS)
database = db.couch_database(db_name)
database.replicateCouchDBToNewCouchDB(cp.COUCHDB_CLEANED_NAME)
db_name = cp.COUCHDB_CLEANED_NAME
cleanFieldsOfDocumentsFromDatabase(db_name, FIELDS)
def importDesignDocumentIntoGivenDatabase(dbname):
"""
Imports the default design document into given database
:param db_name: Name of database to import design document to
"""
#path to design_document_file
design_document_file = pp.RESOURCES_FOLDER + pp.STACKEXCHANGE_DESIGN_DOCUMENT
database = db.couch_database(dbname=dbname)
stored_id = database.insertDesignDocumentFromGivenPath(design_document_file)
print("ID of stored doc: " + str(stored_id))
def getDatasetContentDocumentFromDatabase(dataset_document_name,
dataset_name,
used_fields,
doc_type=DEFAULT_DSCD_TYPE,
db_name=DEFAULT_STATISTICS_DB_NAME):
"""
Gets the required dataset content list from database
:param dataset_document_name - The name of dataset_document the dataset
content document was created of.
:param dataset_name - The dataset to retrieve (training|test)
:param used_fields - The used fields for creating the dataset content doc
:parm doc_type - Default set to DEFAULT_DSCD_TYPE
:param db_name - Default set to DEFAULT_STATISTICS_DB_NAME
:return dataset_content_document
"""
database = db.couch_database(db_name)
#we need all documents from the statistics db
all_docs = database.getAllDocumentsFromDatabase()
#we only need the one with used fields
for row in all_docs.rows:
document = row.doc
document_keys = document.keys()
#check if all needed fields exist
if DSCD_FIELD_DATASET_DOCUMENT_USED in document_keys \
and DSCD_FIELD_TYPE in document_keys \
and DSCD_FIELD_DATASET_NAME in document_keys \
and DSCD_FIELD_USED_FIELDS in document_keys:
#the document we want needs to have the matching
#dataset_document_name type, dataset_name and the correct
#used fields
if dataset_document_name == document[DSCD_FIELD_DATASET_DOCUMENT_USED]\
and doc_type == document[DSCD_FIELD_TYPE] \
and dataset_name == document[DSCD_FIELD_DATASET_NAME] \
and used_fields == document[DSCD_FIELD_USED_FIELDS]:
return document
def perform_train_test_split(db_name=DEFAULT_DB_NAME,
train_size=DEFAULT_TRAININGSET_SIZE):
"""
Get all document_ids of given database and split's it according to given
train_size.
:param db_name: Name of database to split documents (default DEFAULT_DB_NAME)
:param train_size: Size in percentage [0,1] of the training set.
:return splitted_dataset - List of lists
[[DEFAULT_DATASET_LIST_INDEX_TRAINING], [DEFAULT_DATASET_LIST_INDEX_TEST]]
"""
database = db.couch_database(db_name)
all_doc_ids = database.getAllDocumentIdsFromDatabase()
splitted_dataset = cross_validation.train_test_split(all_doc_ids,
train_size=train_size,
random_state=42)
return splitted_dataset
def cleanHTMLTagsFromDocumentsInDatabase(db_name, fields):
"""
Find and replace all (html-) tags from given fields in given database
:param db_name - The db in which tags should be replaced
:param fields - List fields from which html tags should be replaced
"""
#see source:
#http://kevin.deldycke.com/2008/07/python-ultimate-regular-expression-to-catch-html-tags/
ultimate_regexp = """(?i)<\/?\w+((\s+\w+(\s*=\s*(?:\".*?\"|'.*?'|[^'\">\s]+))?)+\s*|\s*)\/?>"""
html_tag_pattern = re.compile(ultimate_regexp)
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
for row in all_docs.rows:
#print("ID of document: " + str(row.id))
document = row.doc
for field in fields:
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
if re.findall(html_tag_pattern, doc_field):
cleaned_doc_field = re.sub(html_tag_pattern,'', doc_field)
document[field] = cleaned_doc_field
database.updateDocumentInDatabase(document)
def getAllStoredWordDictionaryDocumentsFromDatabase(db_name):
"""
Loads the word statistic documents from given database and returns list
of word dictionaries
:param db_name: The database to load the word statistics from
:return [list] of {word_dictionaries}
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
word_dictionary_list = []
for row in all_docs.rows:
document = row.doc
if row.doc[cp.COUCHDB_DOCUMENT_FIELD_ID] != cp.DESIGN_DOCUMENT_ID:
document = row.doc
word_dictionary_list.append(document)
return word_dictionary_list
def insertSplittedDatasetToDatabase(dataset_name, splitted_dataset,
db_name=DEFAULT_STATISTICS_DB_NAME):
"""
Stores the given dataset with given dataset_name into given database.
:param dataset_name - Name of dataset to store
:param splitted_dataset - The dataset to store as list of lists
[[DEFAULT_DATASET_LIST_INDEX_TRAINING][DEFAULT_DATASET_LIST_INDEX_TEST]]
:param db_name: Name of database to split documents.
"""
database = db.couch_database(db_name)
dataset_document = {}
dataset_document[DATASET_DOCUMENT_FIELD_NAME] = dataset_name
dataset_document[DATASET_DOCUMENT_FIELD_TYPE] = DATASET_DOCUMENT_TYPE
dataset_document[DATASET_DOCUMENT_FIELD_NAME_TRAINING_SET] = \
splitted_dataset[DEFAULT_DATASET_LIST_INDEX_TRAINING]
dataset_document[DATASET_DOCUMENT_FIELD_NAME_TEST_SET] = \
splitted_dataset[DEFAULT_DATASET_LIST_INDEX_TEST]
database.insertDocumentIntoDatabase(dataset_document)
def getAllStoredWordDictionaryDocumentsFromDatabase(db_name):
"""
Loads the word statistic documents from given database and returns list
of word dictionaries
:param db_name: The database to load the word statistics from
:return [list] of {word_dictionaries}
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
word_dictionary_list = []
for row in all_docs.rows:
document = row.doc
if row.doc[cp.COUCHDB_DOCUMENT_FIELD_ID] != cp.DESIGN_DOCUMENT_ID:
document = row.doc
word_dictionary_list.append(document)
return word_dictionary_list
def createWordDictionaryFromDatabaseContent(db_name, fields):
"""
Creates an word dictionary (word:occurences) from all documents in given
db_name of given fields.
:param db_name: The name of database to create word dictionary from
: param fields: The fields of documents to use for word dictionary
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
counts = Counter()
#TODO replace
pattern = preprocessor.getWhiteListPattern()
for row in all_docs.rows:
document = row.doc
for field in fields:
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
for word in doc_field.split():
counts.update(word.lower()
for word in re.split(pattern, word))
word_dictionary = dict(counts)
return word_dictionary
def insertDatasetContentDocumentInDatabase(dataset_document_used,
dataset_name,
used_fields,
index_dictionary,
document_content,
tag_index_dictionary,
document_tags,
tags_list,
db_name=DEFAULT_STATISTICS_DB_NAME):
document = {}
document[DSCD_FIELD_DATASET_DOCUMENT_USED] = dataset_document_used
document[DSCD_FIELD_DATASET_NAME] = dataset_name
document[DSCD_FIELD_USED_FIELDS] = used_fields
document[DSCD_FIELD_TYPE] = DEFAULT_DSCD_TYPE
document[DSCD_FIELD_INDEX_DICTIONARY] = index_dictionary
document[DSCD_FIELD_CONTENT] = document_content
document[DSCD_FIELD_TAG_INDEX_DICTIONARY] = tag_index_dictionary
document[DSCD_FIELD_CONTENT_TAGS] = document_tags
document[DSCD_FIELD_TAGS_LIST] = tags_list
database = db.couch_database(db_name)
id = database.insertDocumentIntoDatabase(document)
print(id)
def createWordDictionaryFromDatabaseContent(db_name, fields):
"""
Creates an word dictionary (word:occurences) from all documents in given
db_name of given fields.
:param db_name: The name of database to create word dictionary from
: param fields: The fields of documents to use for word dictionary
"""
database = db.couch_database(db_name)
all_docs = database.getAllDocumentsFromDatabase()
counts = Counter()
#TODO replace
pattern = preprocessor.getWhiteListPattern()
for row in all_docs.rows:
document = row.doc
for field in fields:
#sadly view _all_docs also gives design docs, so we have to check
if field in row.doc.keys():
doc_field = document[field]
for word in doc_field.split():
counts.update(word.lower() for word in re.split(pattern,word))
word_dictionary = dict(counts)
return word_dictionary
def buildDatasetContentListsOfDataset(dataset, document_fields):
"""
Builds the dataset content list of given dataset. This is the structure we
use later for the sklearn Vectorizer (for document term matrix)
:param dataset - Dataset to build dataset content list from
:parum document_fields - List of Lists [[]] for the fields of document
to use.
:return index_dictionary - Dictionary{} key:docId, values:index in
dataset_content_list
:return dataset_content - The content (from document fields #document_fields
of all documents as list
:return tag_index_dictionary - Dictionary{] keys:tag, values: index in
taglist
:return dataset_content_tags - All tags of given documents
:return tags_list - List[] of all tags
"""
database = db.couch_database(DEFAULT_DATABASE)
all_doc_ids = database.getDocumentsForGivenIds(dataset)
#stores an item for content of document
dataset_content = []
dataset_content_tags = []
#key=document_id, value=index in content list
index_dictionary = {}
#key=tag, value=index in tag_list
tag_index_dictionary = {}
#list of tags
dataset_tags = []
for row in all_doc_ids.rows:
document = row.doc
document_id = document.id
document_content = ''
#"sum" the text for an document over all fields
#workaround to check that title and body exist
if pp.STACKEXCHANGE_TITLE_COLUMN in document.keys():
for doc_field in document_fields:
if doc_field in document.keys():
document_content = document_content + document[
doc_field] + ' - '
dataset_content.append(document_content)
index_dictionary[document_id] = len(dataset_content) - 1
if pp.STACKEXCHANGE_TAGS_COLUM in document.keys():
document_tags = document[pp.STACKEXCHANGE_TAGS_COLUM]
dataset_content_tags.append(document_tags)
tags_list = provide_tag_list_of_tag_content(document_tags)
for tag in tags_list:
# tag does not exist in dict, add it to list and dictionary
if tag not in tag_index_dictionary.keys():
dataset_tags.append(tag)
tag_index_dictionary[tag] = len(dataset_tags) - 1
return index_dictionary, dataset_content, tag_index_dictionary, \
dataset_content_tags, dataset_tags
