def createDiseaseLabelHash():
"""
Create and save a hash that connects every PMID with one or more diseases.
"""
t1 = time.time()
files = IOmodule.getSortedFilelist(_subMatrixDir+'/')
labelHash={}
fileCount=0
for file in files:
subMatrix=IOmodule.readInTDM(_subMatrixDir, file)
colMatrix=subMatrix.tocsc()
pmids=colMatrix.getcol(0)[1:].data
for pmid in pmids:
try:
labelHash[pmid].append(file[:-4])
except:
labelHash[pmid]=[]
labelHash[pmid].append(file[:-4])
fileCount+=1
print "Remaining:",(len(files)-fileCount),"Completed",file[:-4]
t2 = time.time()
print 'Created disease label hash in:',str(t2-t1)
IOmodule.pickleOut(_hashTablesDir, _labelHash,"btd", labelHash)
def createVLHash(M_lil):
VLHash={}
for pmidHash in range(M_lil.shape[0]):
VLHash[pmidHash]=linalg.norm((M_lil.getrow(pmidHash).data[0])[1:])
IOmodule.pickleOut("/root/The_Hive/term_doc/hashTables", "VLHash","btd", VLHash)
def createCLHash(M_coo,filename,save_file=True):
"""
Precompute and save the length of each column vector in the term-doc matrix.
Here the length refers to the number of elements.
"""
t1=time.time()
if not os.path.isdir(_hashTablesDir):
os.mkdir(_hashTablesDir)
M_lil=(M_coo.transpose()).tolil()
CLHash={}
count=0
for termHash in range(1,M_coo.shape[1]):
termVectorLength=len((M_lil.getrow(termHash).nonzero()[0])[1:])
CLHash[termHash]=termVectorLength
count+=1
if save_file: print "Hashes created: "+str(count)+". Length:"+str(termVectorLength)
t2=time.time()
if save_file:
print "Created and saved ColumnLength-hash in: "+str(t2-t1)
IOmodule.pickleOut(_hashTablesDir, filename,"btd", CLHash)
else:
return CLHash
def createRLHash(M_lil,filename,save_file=True):
"""
Precompute and save the norm of each row vector in the term-doc matrix.
"""
t1=time.time()
if not os.path.isdir(_hashTablesDir):
os.mkdir(_hashTablesDir)
RLHash={}
count=0
for pmidHash in range(1,M_lil.shape[0]):
RLHash[pmidHash]=linalg.norm((M_lil.getrow(pmidHash).data[0])[1:])
count+=1
if save_file: print "Hashes created: "+str(count)
if save_file:
IOmodule.pickleOut(_hashTablesDir, filename,"btd", RLHash)
else:
return RLHash
t2=time.time()
print "Created and saved RowLength-hash in: "+str(t2-t1)
def createDiseaseHash(dir,output=False):
"""
Recieves a directory containing files to be hashed. It uses the
filename as a key. It requires the files to be in .mtx format. The
hashes starts from 1 ... number_of_files
"""
diseaseHashes={}
files = IO.getSortedFilelist(dir)
counter=0
for f in files:
diseaseName=f[0:f.find('.mtx')]
stdm=IO.readInTDM(dir,f)
if stdm.shape[0]==1:
continue
if diseaseName not in diseaseHashes.keys():
counter+=1
if output:
print 'Created', diseaseName, 'with hash', counter
diseaseHashes[diseaseName]=counter
IO.pickleOut(_hashTablesDir, diseaseHash,"btd", diseaseHashes)
def createTermAndPmidHashes():
"""
This function creates two hash tables of the PMID's and terms to be used
for the term-doc matrix.
Note that the terms are sanitized for any non-alphanumerical characters.
And it is default to remove stop words.
"""
medlineDir = _medlineDir
hashTables = _hashTablesDir
termHashTable={}
pmidHashTable={}
termCounter = 0
pmidCounter = 0
files = IOmodule.getSortedFilelist(medlineDir+'/')
# files = sorted([f for f in os.listdir(medlineDir+"/") if os.path.isfile(medlineDir+"/"+f)])
# Get the regex pattern that sanitizeses strings.
sanitizer = TextCleaner.sanitizeString()
for file in files:
records = RecordHandler.loadMedlineRecords(medlineDir, file)
# *Note*
# Parts of the following loops could be optimized by using dictionaries
# for direct loopkups instead of linear lookups, but since it's not
# important, optimization will have to wait for another day.
# Hash PMID's
for diseaseRecords in records.values():
for record in diseaseRecords:
pmid=record[0]
if pmid not in pmidHashTable:
pmidCounter+=1
pmidHashTable[pmid]=pmidCounter
information=''
# Get the abstract
try:
information=' '+record[1]['AB']
except:
print 'Unable to get abstract', record[0]
try:
information+=' '+record[1]['TI']
except:
print 'Unable to get title for', record[0]
if 'MH' in record[1]:
for meshterm in record[1]['MH']:
information+=' '+meshterm
# We do not want to print this, as most of the
# records do not have MeSH.
# print 'Unable to get MeSH terms for', record[0]
# Sanitize the information
information=sanitizer.sub(' ', information)
# remove stopwords from the abstract
information=FilterInterface.stopwordRemover(information)
# OPTIONAL:
# Stem the abstract
if _stemmer: information=FilterInterface.porterStemmer(information)
termList = [word for word in information.split(' ') if word != '']
for term in termList:
if term not in termHashTable:
termCounter+=1
termHashTable[term]=termCounter
else: continue
print str(termCounter)+" terms hashed. "+str(pmidCounter)+" pmids hashed."
IOmodule.pickleOut(hashTables, _termHash,"btd", termHashTable)
IOmodule.pickleOut(hashTables, _pmidHash,"btd", pmidHashTable)
return termHashTable, pmidHashTable
