def merge_data(data, infile, outfile=None):
"""
Merge the data generated from parsing ELAN files with an existing CSV file
-Arguments:
* data -- data dictionary to be added to the existing file
* infile -- filename of the existing CSV file
* outfile -- filename for the new file.
Default: None -- will update the infile in place
-Return: nothing
"""
if VERBOSE:
print "Merging with existing file: ", infile
tmpfile = ".elanfmt-tmp"
with open(infile, "rb") as fin:
with open(tmpfile, "wb") as fout:
#initialize the input file reader
csvin = csv.DictReader(fin, delimiter=",", quotechar='"',
quoting=csv.QUOTE_ALL, lineterminator="\n")
#generate the complete field list for output file
fields = copy.deepcopy(csvin.fieldnames)
for field in headers:
if field not in fields:
fields.append(field)
#initialize the output file writer and output file
csvout = csv.DictWriter(fout, fields, delimiter=",", quotechar='"',
quoting=csv.QUOTE_ALL, lineterminator="\n")
csvout.writerow(dict(zip(fields,fields)))
#for each line in the input file, update with the corresponding line
#of data then, write to Excel-readable CSV file
for line in csvin:
#if there is data for a line, update the line and output.
try:
line.update(data[line[COLUMN_SUBJECT]+line[COLUMN_STIMULUS]])
csvout.writerow(line)
#if not, copy over the existing data, leaving the new columns blank
except KeyError:
datum = filldict(fields, "")
for key, value in line.iteritems():
datum[key] = value
csvout.writerow(datum)
if outfile is not None:
if VERBOSE:
print "Outputting to new file: ", outfile
if os.path.isfile(outfile):
os.remove(outfile)
os.rename(tmpfile, outfile)
else:
if VERBOSE:
print "Outputting to same file: ", infile
if os.path.isfile(infile):
os.remove(infile)
os.rename(tmpfile, infile)
def parse_elan(elanfile, subjectID, stimID, filterdic=None, checkTranscript=False,
includeNonWord=False):
"""
Parse an ELAN generated annotation file
-Arguments:
* elanfile -- the filename of the ELAN (XML) file to parse
* subjectID -- identifying info for the current participant
* stimID -- identifier for the current stimulus; together with subjectID
identifies a row of output
* filter -- filter to apply to determine what info to extract.
Default=None : return only latency information
* checkTranscript -- if true, consult an external transcript file to
supply new information
Default=False
* includeNonWord -- if true, include non word sounds as words (for
purposes of indexing)
-Return
* data -- dictionary containing an entry for each of the desired fields
given by filter.
"""
#Parse the XML ELAN file and gather the pertinent parts
doc = minidom.parse(elanfile)
times = doc.getElementsByTagName("TIME_SLOT")
tiers = doc.getElementsByTagName("TIER")
#write row identifying information
dataline = filldict(headers, "-") #(make sure all datalines have same fields
dataline.update({COLUMN_SUBJECT: subjectID, \
COLUMN_STIMULUS: stimID, \
COLUMN_LATENCY: times[0].attributes["TIME_VALUE"].value,})
#Default behavior, return only latency
if filterdic is None:
return dataline
for tier in tiers:
#if filter contains entries for tier, process them
key = tier.attributes["TIER_ID"].value.upper()
if key in filterdic.keys():
#check if each annotation entry is in the filter
#(saves searching the annotations multiple times)
nwOffset = 0
for i, item in enumerate(tier.getElementsByTagName("ALIGNABLE_ANNOTATION")):
itemdata = item.childNodes[1].firstChild.data.lower()
#Determine if either the word/phoneme/region/index/etc is a match
if itemdata in filterdic[key] or str(i+1-nwOffset) in filterdic[key]:
try:
#extract timing information
#get time indices
timeindex1 = int(item.attributes["TIME_SLOT_REF1"].value[2:])-1
timeindex2 = int(item.attributes["TIME_SLOT_REF2"].value[2:])-1
#use indices to get onset and duration
onset = times[timeindex1].attributes["TIME_VALUE"].value
duration = int(times[timeindex2].attributes["TIME_VALUE"].value) - int(onset)
except ValueError:
print "WARNING: bad timing info at: \n\tsubject %s,\n\tstimulus %s,\n\ttier %s,\n\titem# %s" % (subjectID, stimID, tier.attributes["TIER_ID"].value, i+1)
continue
#Remember that both keyword and index may match at the same time
#for keyword search
if itemdata in filterdic[key]:
filterpair = key.capitalize()+itemdata.capitalize()
#write the data fields
dataline["Onset"+filterpair] = int(onset)
dataline["Duration"+filterpair] = int(duration)
#for index search
#prevent ++SOUND++ entries from matching if includeNonWord is false
if not includeNonWord and itemdata == "++" + itemdata[2:-2] + "++":
nwOffset += 1
elif str(i+1-nwOffset) in filterdic[key]:
filterpair = key.capitalize()+str(i+1-nwOffset)
#write the data fields
dataline["Onset"+filterpair] = int(onset)
dataline["Duration"+filterpair] = int(duration)
return dataline
