-
Notifications
You must be signed in to change notification settings - Fork 5
/
etl_transcribe_stats.py
174 lines (146 loc) · 6.02 KB
/
etl_transcribe_stats.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import os
import argparse
import datetime
import time
import json
import re
import logging
logging.basicConfig(format='%(asctime)s %(message)s', level=logging.INFO)
from collections import defaultdict
from speech2text import log_kv, make_dir
"""
What it does:
Merges and formats the output generated by two other stats scripts,
and writes the results to a TSV file.
"""
IBM_LOG_STATS_FILEPATH = "processed.json"
GOOGLE_LOG_STATS_FILEPATH = "gprocessed.json"
IBM_WORD_COUNT_FILEPATH = "file_stats.json"
GOOGLE_WORD_COUNT_FILEPATH = "gfile_stats.json"
IBM_BASE = "ibm_stt"
GOOGLE_BASE = "google_stt"
TRANSCRIBED_STATS_TSV = "transcribed_stats.tsv"
API_META = {"ibm" : {"base" : "ibm_stt", "transcripts" : ["hypotheses.txt", "hypotheses.txt.dictated"]},
"google": {"base": "google_stt", "transcripts": ["transcript.txt", "transcript.txt.dictated"]}}
def etl_transcripts(log_stats_path, word_counts_path, api, ext=".out"):
"""
Merges the two datasets using a canonicalized key.
If the two share a filed then the value in <word_counts_path> will overwrite the value from <log_stats_path>
:param log_stats_path: gives processing time
:param word_counts_path: gives word counts
:param api: "ibm" or "google"
:param ext: extension stripped from basename
:return: dict
"""
result = {}
loaded = {}
if os.path.isfile(log_stats_path):
with open(log_stats_path) as fp:
log_kv("Loading(%s log stats)" % api, log_stats_path)
loaded = json.load(fp)
if loaded and type(loaded) is dict:
log_kv("Count (%s log stats)" % api, len(loaded))
else:
logging.error("Expected log stats data.")
for key, row in loaded.items():
id = key
if key.startswith(API_META[api]["base"]+"/"):
id = key.replace(API_META[api]["base"]+"/", "")
result[id] = row
counts = {}
if os.path.isfile(word_counts_path):
with open(word_counts_path) as fp:
log_kv("Loading(%s word counts)" % api, word_counts_path)
counts = json.load(fp)
if counts and type(counts) is dict:
log_kv("Count (%s word counts)" % api, len(counts))
else:
logging.error("Expected word counts data.")
for key, row in counts.items():
id = key
if key.startswith(API_META[api]["base"]+"/"):
id = key.replace(API_META[api]["base"]+"/", "")
for suffix in API_META[api]["transcripts"]:
if id.endswith("/"+suffix):
id = re.sub(re.escape("/"+suffix) + r"$", "", id)
if id.endswith(ext):
id = re.sub(re.escape(ext) + r"$", "", id)
if id in result:
for x, y in row.items():
result[id][x] = y
else:
result[id] = row
return result
if __name__ == '__main__':
start_time = time.time()
parser = argparse.ArgumentParser(description='Analyze transcribe rate')
parser.add_argument('--infolder','-i', action='store', default='.', help='folder containing previous ETL files')
parser.add_argument('--outfolder','-o', action='store', default='./output', help='output directory')
args = parser.parse_args()
log_kv("Running", __file__)
log_kv("From", os.path.dirname(os.path.realpath(__file__)))
print
inpath = os.path.realpath(args.infolder if args.infolder else os.getcwd())
log_kv("inpath", inpath)
outpath = os.path.realpath(args.outfolder if args.outfolder else u'./output')
log_kv("outpath", outpath)
make_dir(outpath)
log_kv("")
log_kv("IBM log stats", IBM_LOG_STATS_FILEPATH)
log_kv("IBM transcript stats", IBM_WORD_COUNT_FILEPATH)
log_kv("Google log stats", GOOGLE_LOG_STATS_FILEPATH)
log_kv("Google transcript stats", GOOGLE_WORD_COUNT_FILEPATH)
log_kv("")
log_kv("IBM base", API_META["ibm"]["base"])
log_kv("Google base", API_META["google"]["base"])
log_kv("")
log_kv("tsv filename", TRANSCRIBED_STATS_TSV)
print
log_kv("Loading IBM")
ibm_data = etl_transcripts(os.path.join(inpath, IBM_LOG_STATS_FILEPATH),
os.path.join(inpath,IBM_WORD_COUNT_FILEPATH),
"ibm")
log_kv("Count(IBM)", len(ibm_data))
print
log_kv("Loading Google")
google_data = etl_transcripts(os.path.join(inpath, GOOGLE_LOG_STATS_FILEPATH),
os.path.join(inpath, GOOGLE_WORD_COUNT_FILEPATH),
"google")
log_kv("Count(Google)", len(google_data))
print
keys = set(ibm_data.keys()).union(set(google_data.keys()))
log_kv("Number of keys in both", len(keys))
tsv_fields = set(["etl_time"])
tsv_data = defaultdict(dict)
for key in keys:
if key in ibm_data:
for x,y in ibm_data[key].items():
tsv_data[key]["ibm_" + x] = y
tsv_fields.add("ibm_" + x)
if key in google_data:
for x, y in google_data[key].items():
tsv_data[key]["google_" + x] = y
tsv_fields.add("google_" + x)
tsv_data[key]["etl_time"] = datetime.datetime.now().isoformat()
tsv_keys = sorted(tsv_data.keys())
tsv_fields = sorted(tsv_fields)
print "Fields:\npath,", ",".join(tsv_fields)
# Writes line data
tsv_path = os.path.join(args.outfolder,TRANSCRIBED_STATS_TSV)
log_kv("Writing", tsv_path)
with open(tsv_path, 'w') as fp:
fp.write("path\t")
for field in tsv_fields:
fp.write("\t"+field if tsv_fields.index(field) else field)
fp.write("\n")
for key in tsv_keys:
fp.write('"%s"\t' % key)
for field in tsv_fields:
value = tsv_data[key][field] if field in tsv_data[key] else None
value = "" if not value else '"%s"' % value if type(value) is str else "%s" % value
fp.write("\t"+value if tsv_fields.index(field) else value)
fp.write("\n")
log_kv("Done: ", __file__)
print("(%.1f min)" % ((time.time() - start_time) / 60.0))