Upload 3 files
Browse files- src/NTURegswa.parquet +3 -0
- src/biconWebStmParquetWa.py +462 -0
- src/stwebm_parquet_wa.py +247 -0
src/NTURegswa.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a69e6661df8d4df83e0d7525b7d425091f33fc95e6e6f5a0b06ede7b9ab686fe
|
| 3 |
+
size 2103648
|
src/biconWebStmParquetWa.py
ADDED
|
@@ -0,0 +1,462 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/home/rubentsui/anaconda3/bin/python
|
| 2 |
+
# coding: utf-8
|
| 3 |
+
|
| 4 |
+
import regex as re
|
| 5 |
+
import sys
|
| 6 |
+
#import time, datetime
|
| 7 |
+
import gzip, bz2, lzma
|
| 8 |
+
from collections import Counter
|
| 9 |
+
import itertools
|
| 10 |
+
from lemminflect import getAllInflections #, getAllLemmas
|
| 11 |
+
from opencc import OpenCC
|
| 12 |
+
import polars as pl
|
| 13 |
+
import struct
|
| 14 |
+
|
| 15 |
+
openCC = OpenCC('t2s')
|
| 16 |
+
|
| 17 |
+
def file_open(filepath):
|
| 18 |
+
#Function to allowing opening files based on file extension
|
| 19 |
+
if filepath.endswith('.gz'):
|
| 20 |
+
return gzip.open(filepath, 'rt', encoding='utf8')
|
| 21 |
+
elif filepath.endswith('.bz2'):
|
| 22 |
+
return bz2.open(filepath, 'rt', encoding='utf8')
|
| 23 |
+
elif filepath.endswith('.xz'):
|
| 24 |
+
return lzma.open(filepath, 'rt', encoding='utf8')
|
| 25 |
+
else:
|
| 26 |
+
return open(filepath, 'r', encoding='utf8')
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
class color:
|
| 30 |
+
PURPLE = '\033[95m'
|
| 31 |
+
CYAN = '\033[96m'
|
| 32 |
+
DARKCYAN = '\033[36m'
|
| 33 |
+
BLUE = '<font color="blue">'
|
| 34 |
+
GREEN = '<font color="#36f307">'
|
| 35 |
+
YELLOW = '\033[93m'
|
| 36 |
+
RED = '<font color="red">'
|
| 37 |
+
BOLD = '<b>'
|
| 38 |
+
UNDERLINE = '\033[4m'
|
| 39 |
+
END = '</b></font>'
|
| 40 |
+
CURSOR_UP = '<font color="blue"><b>' #+ "\033[F" #'\033[1;1H'
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def flatten(l): # flatten a nested list
|
| 44 |
+
def flatten0(l):
|
| 45 |
+
for i in l:
|
| 46 |
+
if isinstance(i,list):
|
| 47 |
+
yield from flatten0(i)
|
| 48 |
+
else:
|
| 49 |
+
yield i
|
| 50 |
+
return list(flatten0(l))
|
| 51 |
+
|
| 52 |
+
def getInflections(s):
|
| 53 |
+
'''
|
| 54 |
+
Get all inflections of the lemma s: Verb, Noun or Adjective
|
| 55 |
+
'''
|
| 56 |
+
infl = getAllInflections(s)
|
| 57 |
+
phr = []
|
| 58 |
+
for t in infl.values():
|
| 59 |
+
phr.extend(list(t))
|
| 60 |
+
if not phr:
|
| 61 |
+
return [s]
|
| 62 |
+
else:
|
| 63 |
+
return list(set(phr))
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
def mergeDicts(D1, D2):
|
| 67 |
+
'''
|
| 68 |
+
Input example:
|
| 69 |
+
D1 = {'a':2, 'b':3, 'c': 1}
|
| 70 |
+
D2 = {'b':5, 'c': 0, 'd': 7}
|
| 71 |
+
Output:
|
| 72 |
+
D = {'a':2, 'b': 3+5, 'c': 1+0, 'd': 7}
|
| 73 |
+
'''
|
| 74 |
+
return dict(Counter(D1) + Counter(D2))
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
def sortTuples(L):
|
| 78 |
+
'''
|
| 79 |
+
L: list of 2-tuples in the form [(1,2), (1,3), (4,3), (3,10), (4,5), (9,2)]
|
| 80 |
+
Sort by 1st number in tuple then by 2nd number, both in ascending order
|
| 81 |
+
'''
|
| 82 |
+
return None
|
| 83 |
+
|
| 84 |
+
def buildLexicon(matched_tokens, alignment, e, z):
|
| 85 |
+
'''
|
| 86 |
+
(1, 3), (1,4), (1, 5) becomes {e[1]: {' '.join([z[3], z[4], z[5]]): 1}}
|
| 87 |
+
'''
|
| 88 |
+
alignment = sorted(alignment)
|
| 89 |
+
L = dict()
|
| 90 |
+
for (i, j) in alignment:
|
| 91 |
+
if e[i] in matched_tokens:
|
| 92 |
+
s = e[i]; t = z[j]
|
| 93 |
+
if s not in L:
|
| 94 |
+
L[s] = [t]
|
| 95 |
+
else:
|
| 96 |
+
L[s].append(t)
|
| 97 |
+
for k in L:
|
| 98 |
+
v = ' '.join(L[k])
|
| 99 |
+
L[k] = {v: 1}
|
| 100 |
+
|
| 101 |
+
return L
|
| 102 |
+
|
| 103 |
+
|
| 104 |
+
|
| 105 |
+
def sentAlignHighlight(s, alignment, e, z):
|
| 106 |
+
# s = list of matches (each "match" is a tuple (i, j) where e[i] is mapped to z[j])
|
| 107 |
+
# e = list of tokens
|
| 108 |
+
# z = list of tokens
|
| 109 |
+
sep = ' '
|
| 110 |
+
e_marked = list(e)
|
| 111 |
+
z_marked = list(z)
|
| 112 |
+
for (i, j) in alignment:
|
| 113 |
+
if e[i] in s:
|
| 114 |
+
src = e[i]; tgt = z[j]
|
| 115 |
+
e_marked[i] = f'{color.RED}{color.BOLD}{e[i]}{color.END}'
|
| 116 |
+
z_marked[j] = f'{color.GREEN}{color.BOLD}{z[j]}{color.END}'
|
| 117 |
+
#e_marked[i] = f'{color.CYAN}{color.BOLD}{e[i]}{color.END}'
|
| 118 |
+
#z_marked[j] = f'{color.YELLOW}{color.BOLD}{z[j]}{color.END}'
|
| 119 |
+
|
| 120 |
+
return sep.join(e_marked), sep.join(z_marked)
|
| 121 |
+
|
| 122 |
+
|
| 123 |
+
def bindata2tuplelist(binary_data):
|
| 124 |
+
'''
|
| 125 |
+
The reverse of the above.
|
| 126 |
+
'''
|
| 127 |
+
# Unpack from bytes
|
| 128 |
+
unpacked_list = []
|
| 129 |
+
num_tuples = len(binary_data) // 4 # Each tuple is 4 bytes (H + H)
|
| 130 |
+
for i in range(num_tuples):
|
| 131 |
+
offset = i * 4
|
| 132 |
+
tup = struct.unpack('>HH', binary_data[offset:offset+4])
|
| 133 |
+
unpacked_list.append(tup)
|
| 134 |
+
|
| 135 |
+
return unpacked_list
|
| 136 |
+
|
| 137 |
+
|
| 138 |
+
corpora = """
|
| 139 |
+
[0] TWP
|
| 140 |
+
[1] Patten
|
| 141 |
+
[2] UNPC
|
| 142 |
+
[3] FIN
|
| 143 |
+
[4] QING
|
| 144 |
+
[5] TWL
|
| 145 |
+
[6] NTURegs
|
| 146 |
+
[7] FTV
|
| 147 |
+
[8] SAT
|
| 148 |
+
[9] CIA
|
| 149 |
+
[10] NEJM
|
| 150 |
+
[11] VOA
|
| 151 |
+
[12] NYT
|
| 152 |
+
[13] BBC
|
| 153 |
+
[14] Quixote
|
| 154 |
+
[15] Wiki
|
| 155 |
+
[16] TEST
|
| 156 |
+
""".strip().split("\n")
|
| 157 |
+
|
| 158 |
+
C = {k: c.split()[-1]+".xz" for k, c in enumerate(corpora)}
|
| 159 |
+
C2 = {k: c.split()[-1] for k, c in enumerate(corpora)}
|
| 160 |
+
|
| 161 |
+
|
| 162 |
+
def sz(s, c=0, max_matches=50, stats_only=False):
|
| 163 |
+
'''
|
| 164 |
+
s: Chinese search phrase
|
| 165 |
+
'''
|
| 166 |
+
|
| 167 |
+
corpus = C[c]
|
| 168 |
+
corpus_code = C2[c]
|
| 169 |
+
html_text = []
|
| 170 |
+
Lexicon = dict()
|
| 171 |
+
|
| 172 |
+
# buit search phrase (sp) regexp
|
| 173 |
+
sp = s.split() # split string into a list of tokens by white spaces
|
| 174 |
+
regex_phr = []
|
| 175 |
+
for s0 in sp:
|
| 176 |
+
inflections = getInflections(s0)
|
| 177 |
+
regex_phr.append(re.compile(fr"\b({'|'.join(inflections)})\b"))
|
| 178 |
+
#html_text += f"regex_phr = {regex_phr}<br>"
|
| 179 |
+
|
| 180 |
+
#sys.exit(0)
|
| 181 |
+
if not regex_phr:
|
| 182 |
+
html_text.append(f"Sorry, zero matches found for search phrase [{s}].\n")
|
| 183 |
+
return None
|
| 184 |
+
|
| 185 |
+
cnt = 0
|
| 186 |
+
raw_cnt = 0
|
| 187 |
+
num_matches = 0
|
| 188 |
+
with file_open(corpus) as fi:
|
| 189 |
+
for line in fi:
|
| 190 |
+
raw_cnt += 1
|
| 191 |
+
if line.strip().count('\t') < 2:
|
| 192 |
+
continue
|
| 193 |
+
#html_text += f"raw_cnt = [{raw_cnt}]; line fewer than 2 tabs<br>"
|
| 194 |
+
#html_text += line.strip() + '<br>'
|
| 195 |
+
#html_text += '-'*80 + '<br>'
|
| 196 |
+
score, en, zh = line.strip().split('\t', maxsplit=2)
|
| 197 |
+
en = en.replace('``', '‘‘').replace("''", '’’')
|
| 198 |
+
e = en.split()
|
| 199 |
+
z = zh.split()
|
| 200 |
+
en_marked, zh_marked = None, None
|
| 201 |
+
|
| 202 |
+
MATCHED_ALL = True
|
| 203 |
+
matches_list = []
|
| 204 |
+
for r in regex_phr:
|
| 205 |
+
matches = r.findall(zh)
|
| 206 |
+
if matches:
|
| 207 |
+
matches_list.extend(matches)
|
| 208 |
+
MATCHED_ALL &= (len(matches)>0)
|
| 209 |
+
matches_list = list(set(matches_list))
|
| 210 |
+
if MATCHED_ALL:
|
| 211 |
+
#print(f"All words matched!")
|
| 212 |
+
cnt += 1
|
| 213 |
+
#alignments = myaligner.get_word_aligns(en, zh)
|
| 214 |
+
#a = alignments[align_method]
|
| 215 |
+
#a = align_word(en, zh)
|
| 216 |
+
a = align_word(zh, en)
|
| 217 |
+
#print(f"alignment = {a}")
|
| 218 |
+
#print(f"matches_list = {matches_list}")
|
| 219 |
+
zh_marked, en_marked = sentAlignHighlight(flatten(matches_list), a, z, e)
|
| 220 |
+
if stats_only:
|
| 221 |
+
pass
|
| 222 |
+
else:
|
| 223 |
+
lineOut = f'[{corpus_code}]\t{score}\t<p class="chinese">{zh_marked}</p>\t<p class="europe">{en_marked}</p>'
|
| 224 |
+
html_text.append(lineOut)
|
| 225 |
+
L = buildLexicon(flatten(matches_list), a, z, e)
|
| 226 |
+
for k in L:
|
| 227 |
+
if k in Lexicon:
|
| 228 |
+
Lexicon[k] = mergeDicts(Lexicon[k], L[k])
|
| 229 |
+
else:
|
| 230 |
+
Lexicon[k] = L[k]
|
| 231 |
+
#print(f"Lexicon per match: {L}")
|
| 232 |
+
#print()
|
| 233 |
+
|
| 234 |
+
if cnt >= max_matches: break
|
| 235 |
+
|
| 236 |
+
summary = f"No. of matches: {cnt}\n<br>\n"
|
| 237 |
+
#print(Lexicon)
|
| 238 |
+
for k1 in Lexicon:
|
| 239 |
+
v1 = Lexicon[k1]
|
| 240 |
+
#html_text += f"[{k1}]\n"
|
| 241 |
+
s = [(k2, v1[k2]) for k2 in sorted(v1, key=v1.get, reverse=True)]
|
| 242 |
+
for k2, v2 in s:
|
| 243 |
+
if k2:
|
| 244 |
+
summary += f"{v2}\t{k2}\n<br>\n"
|
| 245 |
+
#html_text.append(summary)
|
| 246 |
+
|
| 247 |
+
return html_text, summary
|
| 248 |
+
|
| 249 |
+
|
| 250 |
+
def se(s, c=0, max_matches=50, stats_only=False):
|
| 251 |
+
'''
|
| 252 |
+
s: English search phrase
|
| 253 |
+
'''
|
| 254 |
+
corpus = C[c]
|
| 255 |
+
corpus_code = C2[c]
|
| 256 |
+
html_text = []
|
| 257 |
+
Lexicon = dict()
|
| 258 |
+
|
| 259 |
+
# buit search phrase (sp) regexp
|
| 260 |
+
sp = s.split() # split string into a list of tokens by white spaces
|
| 261 |
+
# regexp
|
| 262 |
+
regex_phr = None
|
| 263 |
+
if len(sp) == 1: # regexp for single-word search phrase
|
| 264 |
+
inflections = getInflections(s)
|
| 265 |
+
num = len(inflections)
|
| 266 |
+
if num > 0:
|
| 267 |
+
regex_phr = [re.compile(fr"\b({'|'.join(inflections)})\b", flags=re.IGNORECASE)]
|
| 268 |
+
else: # multi-word search phrase
|
| 269 |
+
regex_phr = []
|
| 270 |
+
for s0 in sp:
|
| 271 |
+
inflections = getInflections(s0)
|
| 272 |
+
regex_phr.append(re.compile(fr"\b({'|'.join(inflections)})\b", flags=re.IGNORECASE))
|
| 273 |
+
|
| 274 |
+
#sys.exit(0)
|
| 275 |
+
|
| 276 |
+
if not regex_phr:
|
| 277 |
+
html_text.append(f"Sorry, zero matches found for search phrase [{s}].\n<br>\n")
|
| 278 |
+
return None
|
| 279 |
+
|
| 280 |
+
cnt = 0
|
| 281 |
+
raw_cnt = 0
|
| 282 |
+
num_matches = 0
|
| 283 |
+
with file_open(corpus) as fi:
|
| 284 |
+
for line in fi:
|
| 285 |
+
raw_cnt += 1
|
| 286 |
+
if line.strip().count('\t') < 2:
|
| 287 |
+
continue
|
| 288 |
+
score, en, zh = line.strip().split('\t', maxsplit=2)
|
| 289 |
+
en = en.replace('``', '‘‘').replace("''", '’’')
|
| 290 |
+
e = en.split()
|
| 291 |
+
z = zh.split()
|
| 292 |
+
en_marked, zh_marked = None, None
|
| 293 |
+
|
| 294 |
+
MATCHED_ALL = True
|
| 295 |
+
matches_list = []
|
| 296 |
+
for r in regex_phr:
|
| 297 |
+
matches = r.findall(en)
|
| 298 |
+
matches_list.append(matches)
|
| 299 |
+
MATCHED_ALL &= (len(matches)>0)
|
| 300 |
+
if MATCHED_ALL:
|
| 301 |
+
#print(f"All words matched!")
|
| 302 |
+
cnt += 1
|
| 303 |
+
#alignments = myaligner.get_word_aligns(en, zh)
|
| 304 |
+
#a = alignments[align_method]
|
| 305 |
+
a = align_word(en, zh)
|
| 306 |
+
en_marked, zh_marked = sentAlignHighlight(flatten(matches_list), a, e, z)
|
| 307 |
+
if stats_only:
|
| 308 |
+
pass
|
| 309 |
+
else:
|
| 310 |
+
lineOut = f'[{corpus_code}]\t{score}\t<p class="europe">{en_marked}</p>\t<p class="chinese">{zh_marked}</p>'
|
| 311 |
+
html_text.append(lineOut)
|
| 312 |
+
L = buildLexicon(flatten(matches_list), a, e, z)
|
| 313 |
+
for k in L:
|
| 314 |
+
if k in Lexicon:
|
| 315 |
+
Lexicon[k] = mergeDicts(Lexicon[k], L[k])
|
| 316 |
+
else:
|
| 317 |
+
Lexicon[k] = L[k]
|
| 318 |
+
|
| 319 |
+
if cnt >= max_matches: break
|
| 320 |
+
|
| 321 |
+
summary = f"No. of matches: {cnt}\n<br>\n"
|
| 322 |
+
#print(Lexicon)
|
| 323 |
+
for k1 in Lexicon:
|
| 324 |
+
v1 = Lexicon[k1]
|
| 325 |
+
#html_text += f"[{k1}]<br>"
|
| 326 |
+
s = [(k2, v1[k2]) for k2 in sorted(v1, key=v1.get, reverse=True)]
|
| 327 |
+
for k2, v2 in s:
|
| 328 |
+
if k2:
|
| 329 |
+
summary += f"{v2}\t{k2}\n<br>\n"
|
| 330 |
+
#html_text.append(summary)
|
| 331 |
+
return html_text, summary
|
| 332 |
+
|
| 333 |
+
|
| 334 |
+
regex_zh = re.compile(r"[一-龥]")
|
| 335 |
+
|
| 336 |
+
def s(ss, c=0, max_matches=100, stats_only=False):
|
| 337 |
+
|
| 338 |
+
actual_search_function = None
|
| 339 |
+
if regex_zh.findall(ss): # Chinese characters found
|
| 340 |
+
actual_search_function = sz
|
| 341 |
+
if c in [99]: # ROCLaws has en, zh reversed
|
| 342 |
+
actual_search_function = se
|
| 343 |
+
print(f"Search by Chinese: actual search function = [{actual_search_function}]")
|
| 344 |
+
else: # Non-Chinese
|
| 345 |
+
actual_search_function = se
|
| 346 |
+
if c in [99]: # ROCLaws has en, zh reversed
|
| 347 |
+
actual_search_function = sz
|
| 348 |
+
|
| 349 |
+
return actual_search_function(ss, c=c, max_matches=max_matches, stats_only=stats_only)
|
| 350 |
+
|
| 351 |
+
|
| 352 |
+
def tokenIndices(ss, i, j):
|
| 353 |
+
'''
|
| 354 |
+
Given: input indices (i, j) of the string ss (tokens separated by single spaces),
|
| 355 |
+
Return: the list indices of ss.split() that correspond to the substring ss[i:j]
|
| 356 |
+
'''
|
| 357 |
+
L = ss.split()
|
| 358 |
+
part1 = ss[:i].split()
|
| 359 |
+
part2 = ss[i:j].split()
|
| 360 |
+
part3 = ss[j:].split()
|
| 361 |
+
return len(part1), len(L) - len(part3) # these are the list indices
|
| 362 |
+
|
| 363 |
+
|
| 364 |
+
def regex_search(ss, c=0, max_matches=100, stats_only=False, literal=False):
|
| 365 |
+
|
| 366 |
+
zhSearch = False
|
| 367 |
+
if regex_zh.findall(ss): # Chinese characters found
|
| 368 |
+
zhSearch = True
|
| 369 |
+
|
| 370 |
+
corpus = C[c]
|
| 371 |
+
corpus_code = C2[c]
|
| 372 |
+
results = []
|
| 373 |
+
|
| 374 |
+
df = pl.read_parquet(f"{corpus_code}wa.parquet")
|
| 375 |
+
search_str = r"(?i)"
|
| 376 |
+
if zhSearch:
|
| 377 |
+
column = 'zh'
|
| 378 |
+
else:
|
| 379 |
+
column = 'en'
|
| 380 |
+
res = df.filter(
|
| 381 |
+
pl.col(column).str.contains(fr"({ss})", literal=literal)
|
| 382 |
+
)
|
| 383 |
+
query_results = res.to_dict(as_series=False)
|
| 384 |
+
length = len(query_results['en'])
|
| 385 |
+
#length = 5
|
| 386 |
+
p = re.compile(fr"({ss})")
|
| 387 |
+
cnt = 0
|
| 388 |
+
for i in range(length):
|
| 389 |
+
cnt += 1
|
| 390 |
+
en = query_results['en'][i]
|
| 391 |
+
zh = query_results['zh'][i]
|
| 392 |
+
enList = en.split()
|
| 393 |
+
zhList = zh.split()
|
| 394 |
+
was = bindata2tuplelist(query_results['word_alignments'][i])
|
| 395 |
+
#print('was = ', was)
|
| 396 |
+
#enSub, zhSub = en, zh
|
| 397 |
+
if zhSearch:
|
| 398 |
+
#a = align_word(openCC.convert(zh), en)
|
| 399 |
+
#a = [(zhList[j], enList[i]) for (i, j) in was]
|
| 400 |
+
a = was
|
| 401 |
+
for m in p.finditer(zh):
|
| 402 |
+
ii = m.start()
|
| 403 |
+
jj = m.end()
|
| 404 |
+
k, q = tokenIndices(zh, ii, jj)
|
| 405 |
+
for idx in range(k, q):
|
| 406 |
+
zhList[idx] = f"{color.RED}{color.BOLD}{zhList[idx]}{color.END}"
|
| 407 |
+
|
| 408 |
+
idxT = [e for (e, z) in a if z == idx] # target indices
|
| 409 |
+
#print(f"idxT = {idxT}")
|
| 410 |
+
for iT in idxT:
|
| 411 |
+
enList[iT] = f"{color.GREEN}{color.BOLD}{enList[iT]}{color.END}"
|
| 412 |
+
|
| 413 |
+
zhSub = ' '.join(zhList)
|
| 414 |
+
enSub = ' '.join(enList)
|
| 415 |
+
score = '_score_'
|
| 416 |
+
lineOut = f'[{corpus_code}]\t{score}\t<p class="chinese">{zhSub}</p>\t<p class="europe">{enSub}</p>'
|
| 417 |
+
results.append(lineOut)
|
| 418 |
+
else:
|
| 419 |
+
#a = align_word(en, openCC.convert(zh))
|
| 420 |
+
#a = [(enList[i], zhList[j]) for (i, j) in was]
|
| 421 |
+
a = was
|
| 422 |
+
|
| 423 |
+
for m in p.finditer(en):
|
| 424 |
+
ii = m.start()
|
| 425 |
+
jj = m.end()
|
| 426 |
+
k, q = tokenIndices(en, ii, jj)
|
| 427 |
+
for idx in range(k, q):
|
| 428 |
+
enList[idx] = f"{color.RED}{color.BOLD}{enList[idx]}{color.END}"
|
| 429 |
+
idxT = [z for (e, z) in a if e == idx] # target indices
|
| 430 |
+
for iT in idxT:
|
| 431 |
+
zhList[iT] = f"{color.GREEN}{color.BOLD}{zhList[iT]}{color.END}"
|
| 432 |
+
|
| 433 |
+
zhSub = ' '.join(zhList)
|
| 434 |
+
enSub = ' '.join(enList)
|
| 435 |
+
|
| 436 |
+
score = 1
|
| 437 |
+
lineOut = f'[{corpus_code}]\t{score}\t<p class="europe">{enSub}</p>\t<p class="chinese">{zhSub}</p>'
|
| 438 |
+
results.append(lineOut)
|
| 439 |
+
|
| 440 |
+
if cnt > max_matches: break
|
| 441 |
+
|
| 442 |
+
return results
|
| 443 |
+
|
| 444 |
+
#'''
|
| 445 |
+
#EXAMPLES of REGEX RESEARCH
|
| 446 |
+
#(take[sn]|taking|took) .{1,20} for granted
|
| 447 |
+
#'''
|
| 448 |
+
|
| 449 |
+
if __name__ == '__main__':
|
| 450 |
+
|
| 451 |
+
print('\n\n'+'='*100)
|
| 452 |
+
print("""Usage:
|
| 453 |
+
Chinese search phrase
|
| 454 |
+
s('打擊 犯罪', c=0)
|
| 455 |
+
English search phrase
|
| 456 |
+
s('preemptive strike', c=0, mac_matches=200)
|
| 457 |
+
Type C (Capital "C") followed by the <Enter> key to see a list of corpora available.
|
| 458 |
+
""")
|
| 459 |
+
for c in C:
|
| 460 |
+
print(f"c={c}: {C[c][:-3]}")
|
| 461 |
+
|
| 462 |
+
|
src/stwebm_parquet_wa.py
ADDED
|
@@ -0,0 +1,247 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# -*- coding: utf-8 -*-
|
| 2 |
+
"""
|
| 3 |
+
Created on Sun Dec 11 19:51:02 2022
|
| 4 |
+
|
| 5 |
+
@author: ruben
|
| 6 |
+
"""
|
| 7 |
+
|
| 8 |
+
import streamlit as st
|
| 9 |
+
#from streamlit.components.v1 import html
|
| 10 |
+
#import streamlit.components.v1 as components
|
| 11 |
+
import pandas as pd
|
| 12 |
+
#import numpy as np
|
| 13 |
+
#from datetime import datetime
|
| 14 |
+
|
| 15 |
+
from biconWebStmParquetWa import s as search, regex_search as rs
|
| 16 |
+
|
| 17 |
+
# Session variables
|
| 18 |
+
if 'queryresults' not in st.session_state:
|
| 19 |
+
st.session_state.queryresults = None
|
| 20 |
+
|
| 21 |
+
if 'page' not in st.session_state:
|
| 22 |
+
st.session_state.page = 1 # starts at 1, not 0
|
| 23 |
+
|
| 24 |
+
if 'maxpage' not in st.session_state:
|
| 25 |
+
st.session_state.maxpage = 0
|
| 26 |
+
|
| 27 |
+
if 'minpage' not in st.session_state:
|
| 28 |
+
st.session_state.minpage = 1
|
| 29 |
+
|
| 30 |
+
if 'datasize' not in st.session_state:
|
| 31 |
+
st.session_state.datasize = 0
|
| 32 |
+
|
| 33 |
+
if 'chunksize' not in st.session_state:
|
| 34 |
+
st.session_state.chunksize = 3
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
#if 'table' not in st.session_state:
|
| 38 |
+
# st.session_state.table = ''
|
| 39 |
+
table = 'Empty Table'
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
def buildTable(page):
|
| 43 |
+
# Build table
|
| 44 |
+
slices = st.session_state.queryresults
|
| 45 |
+
datasize = st.session_state.datasize
|
| 46 |
+
table = '<table width="100%">\n'
|
| 47 |
+
n = st.session_state.chunksize
|
| 48 |
+
for j in range(n):
|
| 49 |
+
index = (page-1)*n + j
|
| 50 |
+
if index >= datasize: break
|
| 51 |
+
try:
|
| 52 |
+
corpus, score, en, zh = slices[page-1][j].split('\t')
|
| 53 |
+
except:
|
| 54 |
+
continue
|
| 55 |
+
table += '<tr>\n'
|
| 56 |
+
table += f'<td>{corpus}</td><td colspan=2>{score}</td>\n'
|
| 57 |
+
table += '</tr>\n'
|
| 58 |
+
table += '<tr>\n'
|
| 59 |
+
table += f'<td>{index+1}</td><td width="45%" valign="top">{en}</td><td width="50%" valign="top">{zh}</td>'
|
| 60 |
+
table += '</tr>\n'
|
| 61 |
+
table += '</table>'
|
| 62 |
+
return table
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
appTitle= '國教院華英雙語索引典系統2.0β版'
|
| 67 |
+
appTitle= '華英雙語索引典系統2.0β版'
|
| 68 |
+
sources = ('TWP', 'Patten', 'UNPC', 'FIN', 'QING', 'TWL', 'NTURegs', 'FTV', 'SAT', 'CIA',
|
| 69 |
+
'NEJM', 'VOA', 'NYT', 'BBC', 'Quixote', 'Wiki', 'TEST')
|
| 70 |
+
corpus_labels = ('光華雜誌', '彭定康', '聯合國平行語料庫', '清史', '台灣法律(全國法規資料庫)', '臺大法規', '民視英語新聞', '科學人', '美國華人史', '新英格蘭醫學期刊', '美國之音', '紐約時報中文網', 'BBC', '唐吉柯德', '維基百科', '測試')
|
| 71 |
+
|
| 72 |
+
files = [c + '.xz' for c in sources]
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
st.set_page_config(
|
| 76 |
+
page_title=appTitle,
|
| 77 |
+
#page_icon=icon,
|
| 78 |
+
layout='wide',
|
| 79 |
+
initial_sidebar_state='auto',
|
| 80 |
+
menu_items={
|
| 81 |
+
'Get Help': 'https://streamlit.io/',
|
| 82 |
+
'Report a bug': 'https://github.com',
|
| 83 |
+
'About': f'**{appTitle}**\nCopyright (c) Ruben G. Tsui'
|
| 84 |
+
}
|
| 85 |
+
)
|
| 86 |
+
|
| 87 |
+
page_style = '''
|
| 88 |
+
<style>
|
| 89 |
+
.css-o18uir.e16nr0p33 {
|
| 90 |
+
margin-top: -125px;
|
| 91 |
+
}
|
| 92 |
+
.reportview-container .css-1lcbmhc .css-1outpf7 {{
|
| 93 |
+
padding-top: -125px;
|
| 94 |
+
}}
|
| 95 |
+
.reportview-container .main .block-container{{
|
| 96 |
+
padding-top: 0rem;
|
| 97 |
+
padding-right: 0rem;
|
| 98 |
+
padding-left: 0rem;
|
| 99 |
+
padding-bottom: 0rem;}}
|
| 100 |
+
.europe {
|
| 101 |
+
font-family: Consolas, Menlo, Courier New, Arial;
|
| 102 |
+
font-size: 14px;
|
| 103 |
+
}
|
| 104 |
+
.chinese {
|
| 105 |
+
/* font-family: Xingkai TC; */
|
| 106 |
+
font-family: Microsoft Jhenghei, Source Han Sans, Hiragino Sans CNS, LantingHei TC, Source Han Serif;
|
| 107 |
+
font-size: 36px;
|
| 108 |
+
font-weight: lighter;
|
| 109 |
+
}
|
| 110 |
+
</style>
|
| 111 |
+
'''
|
| 112 |
+
st.markdown(page_style, unsafe_allow_html=True)
|
| 113 |
+
|
| 114 |
+
# Sidebar
|
| 115 |
+
st.sidebar.subheader(appTitle)
|
| 116 |
+
|
| 117 |
+
with st.sidebar:
|
| 118 |
+
|
| 119 |
+
#query = st.sidebar.text_input('輸入搜尋字串').strip()
|
| 120 |
+
query = st.sidebar.text_area('輸入搜尋字串').strip()
|
| 121 |
+
multicorpora = st.multiselect('選擇語料庫(可複選)', sources, ['TWP', 'FTV'])
|
| 122 |
+
|
| 123 |
+
colc, cold = st.sidebar.columns([1, 1])
|
| 124 |
+
with colc:
|
| 125 |
+
submit_button = st.button('搜尋')
|
| 126 |
+
with cold:
|
| 127 |
+
regex_search = st.radio(
|
| 128 |
+
"Regex search",
|
| 129 |
+
["No", "Yes"], horizontal=True
|
| 130 |
+
)
|
| 131 |
+
|
| 132 |
+
|
| 133 |
+
cola, colb = st.sidebar.columns([1, 1])
|
| 134 |
+
with cola:
|
| 135 |
+
size = st.selectbox('筆數上限', [10,20,50,100,200,500,5000], index=2)
|
| 136 |
+
with colb:
|
| 137 |
+
stats_only = st.radio(
|
| 138 |
+
"Stats only",
|
| 139 |
+
["No", "Yes"], horizontal=True
|
| 140 |
+
)
|
| 141 |
+
|
| 142 |
+
st.session_state.chunksize = st.slider("每頁筆數", 1, 20, 20)
|
| 143 |
+
|
| 144 |
+
# Build user interface
|
| 145 |
+
col1, col2, col3, col4 = st.columns([1, 1, 1, 1])
|
| 146 |
+
with col1:
|
| 147 |
+
first_button = st.button('First')
|
| 148 |
+
with col2:
|
| 149 |
+
prev_button = st.button('Prev')
|
| 150 |
+
with col3:
|
| 151 |
+
next_button = st.button('Next')
|
| 152 |
+
with col4:
|
| 153 |
+
last_button = st.button('Last')
|
| 154 |
+
|
| 155 |
+
# Navigation
|
| 156 |
+
if next_button:
|
| 157 |
+
if st.session_state.page < st.session_state.maxpage:
|
| 158 |
+
st.session_state.page += 1
|
| 159 |
+
|
| 160 |
+
if prev_button:
|
| 161 |
+
if st.session_state.page > st.session_state.minpage:
|
| 162 |
+
st.session_state.page -= 1
|
| 163 |
+
|
| 164 |
+
if first_button:
|
| 165 |
+
st.session_state.page = st.session_state.minpage
|
| 166 |
+
|
| 167 |
+
if last_button:
|
| 168 |
+
st.session_state.page = st.session_state.maxpage
|
| 169 |
+
|
| 170 |
+
page = st.session_state.page
|
| 171 |
+
|
| 172 |
+
|
| 173 |
+
n = st.session_state.chunksize # chunk size (no. of rows per chunk)
|
| 174 |
+
|
| 175 |
+
# Logic to query database when button is pressed
|
| 176 |
+
divider = '-'*80
|
| 177 |
+
if submit_button:
|
| 178 |
+
|
| 179 |
+
# reset certin parameters
|
| 180 |
+
st.session_state.queryresults = None
|
| 181 |
+
st.session_state.page = 1 # starts at 1, not 0
|
| 182 |
+
st.session_state.maxpage = 0
|
| 183 |
+
st.session_state.minpage = 1
|
| 184 |
+
st.session_state.datasize = 0
|
| 185 |
+
#st.session_state.chunksize = 3
|
| 186 |
+
|
| 187 |
+
|
| 188 |
+
#selectedCorpus = corpora.index(True)
|
| 189 |
+
selectedCorpora = multicorpora
|
| 190 |
+
|
| 191 |
+
if regex_search == 'Yes':
|
| 192 |
+
|
| 193 |
+
#st.write('Regex search selected!!')
|
| 194 |
+
all_results = [] # results from all corpora selected
|
| 195 |
+
for c in selectedCorpora:
|
| 196 |
+
selectedCorpus = sources.index(c)
|
| 197 |
+
results = rs(query, c=selectedCorpus, max_matches=size, stats_only=(stats_only=="Yes"), literal=False)
|
| 198 |
+
#st.write(f'No. of matches found in [{c}]: {len(results)}')
|
| 199 |
+
st.success(f'No. of matches found in [{c}]: {len(results)}')
|
| 200 |
+
all_results.extend(results)
|
| 201 |
+
|
| 202 |
+
datasize = len(all_results)
|
| 203 |
+
slices = [all_results[i:i+n] for i in range(0, datasize, n)]
|
| 204 |
+
pagesize = len(slices) # total no. of pages available
|
| 205 |
+
|
| 206 |
+
st.session_state.datasize = datasize
|
| 207 |
+
st.session_state.maxpage = pagesize
|
| 208 |
+
st.session_state.queryresults = slices
|
| 209 |
+
|
| 210 |
+
else:
|
| 211 |
+
|
| 212 |
+
all_results = [] # results from all corpora selected
|
| 213 |
+
all_summaries = []
|
| 214 |
+
for c in selectedCorpora:
|
| 215 |
+
selectedCorpus = sources.index(c)
|
| 216 |
+
results, summary = search(query, c=selectedCorpus, max_matches=size, stats_only=(stats_only=="Yes"))
|
| 217 |
+
#results = regex_search(query, c=selectedCorpus, max_matches=size, stats_only=(stats_only=="Yes"), literal=true)
|
| 218 |
+
st.markdown(f'Corpus [{c}]: {len(results)} matches')
|
| 219 |
+
all_results.extend(results)
|
| 220 |
+
all_summaries.append(summary)
|
| 221 |
+
|
| 222 |
+
datasize = len(all_results)
|
| 223 |
+
slices = [all_results[i:i+n] for i in range(0, datasize, n)]
|
| 224 |
+
pagesize = len(slices) # total no. of pages available
|
| 225 |
+
|
| 226 |
+
st.session_state.datasize = datasize
|
| 227 |
+
st.session_state.maxpage = pagesize
|
| 228 |
+
st.session_state.queryresults = slices
|
| 229 |
+
|
| 230 |
+
|
| 231 |
+
#st.write(st.session_state)
|
| 232 |
+
|
| 233 |
+
if st.session_state.queryresults != None:
|
| 234 |
+
#col1a, col2a = st.columns([1, 2])
|
| 235 |
+
#with col1a:
|
| 236 |
+
st.markdown(f"page {st.session_state.page} of {st.session_state.maxpage}")
|
| 237 |
+
#with col2a:
|
| 238 |
+
# st.slider("pages", 1, st.session_state.maxpage, st.session_state.page)
|
| 239 |
+
table = buildTable(st.session_state.page)
|
| 240 |
+
st.markdown(table, unsafe_allow_html=True)
|
| 241 |
+
#st.markdown(all_summaries, unsafe_allow_html=True)
|
| 242 |
+
#table = buildTable(st.session_state.page)
|
| 243 |
+
#st.markdown(table, unsafe_allow_html=True)
|
| 244 |
+
|
| 245 |
+
#st.write("That's all, folks!")
|
| 246 |
+
#st.write(table)
|
| 247 |
+
|