File size: 9,663 Bytes
f4f7cf9 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 |
import streamlit as st
import pandas as pd
import transformers
import re
import zipfile
import postt
from postt import postcor , precor
from transformers import pipeline, TokenClassificationPipeline, BertForTokenClassification , AutoTokenizer , TextClassificationPipeline , AutoModelForSequenceClassification
st.set_page_config(layout="wide")
st.image("persEsi.png",width = 200)
st.title("Knowledge extraction: EDCs")
st.write("This tool lets you extract relation triples concerning interactions between: endocrine disrupting chemicals, hormones, receptors and cancers. It is the result of an end of studies project within ESI school and dedicated to biomedical researchers looking to extract precise information about the subject without digging into long publications.")
form = st.form(key='my-form')
x = form.text_area('Enter text', height=200)
submit = form.form_submit_button('Submit')
if submit and len(x) != 0:
#model.to("cpu")
st.write("Execution in progress... Results will be displayed below in a while or can be downloaded from the sidebar, please be patient.")
tokenizer = AutoTokenizer.from_pretrained("dmis-lab/biobert-large-cased-v1.1", truncation = True, padding=True, model_max_length=512,)
model_checkpoint = BertForTokenClassification.from_pretrained("dexay/Ner2HgF", )
st.caption("Downloading models")
model_re = AutoModelForSequenceClassification.from_pretrained("dexay/reDs3others", )
token_classifier = pipeline("token-classification", tokenizer = tokenizer,model=model_checkpoint, )
if x[-1] not in ".?:":
x += "."
biotext = precor(x)
#split document or text into sentences
lstbiotext = []
flag = 0
tempsen = ""
for e in biotext:
tempsen += e
if e=="(":
flag = 1
if e==")":
flag = 0
if (e =="." or e =="?" or e ==":" ) and flag == 0 :
lstbiotext += [tempsen.strip()]
tempsen = ""
ddata = lstbiotext
#tokenized_dat = tokenize_function(ddata)
az = token_classifier(ddata)
#code to convert NER output to RE input compatible format
#tg_inorder are decoding of labels on which the model was fine tuned on
tg_inorder = ['O',
'B-HORMONE',
'B-EXP_PER',
'I-HORMONE',
'I-CANCER',
'I-EDC',
'B-RECEPTOR',
'B-CANCER',
'I-RECEPTOR',
'B-EDC',
'PAD']
lstSentEnc = []
lstSentbilbl = []
lstSentEnt = []
for itsent in az:
sentaz = itsent
ph = []
phl = []
for e in sentaz:
if e["word"][0]=="#" and len(ph)!=0:
ph[-1]+= e["word"][2:]
else:
ph += [e["word"]]
phl += [e["entity"]]
phltr = []
for e in phl:
phltr += [tg_inorder[int(e[-1])] if len(e)==7 else tg_inorder[int(e[-2:])]]
nwph = []
nwphltr = []
flag = 0
for i in range(len(phltr)-2):
if phltr[i]=="O" and flag != 3 :
nwph += [ph[i]]
nwphltr += [phltr[i]]
continue
elif flag == 3:
nwph[-1] += " "+ph[i]
flag = 1
continue
elif phltr[i][2:]==phltr[i+1][2:] and phltr[i+1][0]=="I" and flag == 0:
nwph += [ph[i]]
nwphltr += [phltr[i]]
flag = 1
continue
elif phltr[i][2:]==phltr[i+1][2:] and phltr[i+1][0]=="I" and flag == 1:
nwph[-1] += " "+ph[i]
continue
# xox with flag == 3
elif phltr[i][2:]==phltr[i+2][2:] and phltr[i+1]=="O" and phltr[i+2][0]=="I" and flag == 0:
nwph += [ph[i]]
nwphltr += [phltr[i]]
flag = 3
continue
elif phltr[i][2:]==phltr[i+2][2:] and phltr[i+1]=="O" and phltr[i+2][0]=="I" and flag == 1:
nwph[-1] += " "+ph[i]
flag = 3
continue
#\ xox
elif flag == 1:
nwph[-1] += " "+ph[i]
flag = 0
continue
else :
nwph += [ph[i]]
nwphltr += [phltr[i]]
continue
# nwph,nwphltr,len(nwph),len(nwphltr)
if nwphltr.count("O") <= len(nwphltr)-2:
for i in range(len(nwph)-1):
if nwphltr[i] != "O":
for j in range(i,len(nwph)):
if nwphltr[j] != "O" and nwphltr[j] != nwphltr[i] and {nwphltr[j], nwphltr[i]} != {"B-CANCER","B-RECEPTOR"}:
sen2ad = ""
for g in range(i):
sen2ad += nwph[g]+" "
sen2ad += "<e1>"+nwph[i]+"</e1> "
for t in range(i+1,j):
sen2ad += nwph[t]+" "
sen2ad += "<e2>"+nwph[j]+"</e2>"
if j<len(nwph):
for l in range(j+1,len(nwph)):
sen2ad += " "+nwph[l]
lstSentEnc += [sen2ad]
lstSentbilbl += [[nwphltr[i],nwphltr[j]]]
lstSentEnt += [[nwph[i],nwph[j]]]
#lstSentEnc,lstSentEnt,lstSentbilbl
st.caption("Entities detected.")
st.caption("Next: Relation detection ...")
# Relation extraction part
token_classifier = pipeline("text-classification", tokenizer = tokenizer,model=model_re, )
rrdata = lstSentEnc
outre = token_classifier(rrdata)
trLABELS = ['INCREASE_RISK(e1,e2)',
'SPEED_UP(e2,e1)',
'DECREASE_ACTIVITY(e1,e2)',
'NO_ASSOCIATION(e1,e2)',
'DECREASE(e1,e2)',
'BLOCK(e1,e2)',
'CAUSE(e1,e2)',
'ACTIVATE(e2,e1)',
'DEVELOP(e2,e1)',
'ALTER(e1,e2)',
'INCREASE_RISK(e2,e1)',
'SPEED_UP(e1,e2)',
'INTERFER(e1,e2)',
'DECREASE(e2,e1)',
'NO_ASSOCIATION(e2,e1)',
'INCREASE(e2,e1)',
'INTERFER(e2,e1)',
'ACTIVATE(e1,e2)',
'INCREASE(e1,e2)',
'MIMIC(e1,e2)',
'MIMIC(e2,e1)',
'BLOCK(e2,e1)',
'other',
'BIND(e2,e1)',
'INCREASE_ACTIVITY(e2,e1)',
'ALTER(e2,e1)',
'CAUSE(e2,e1)',
'BIND(e1,e2)',
'DEVELOP(e1,e2)',
'DECREASE_ACTIVITY(e2,e1)']
outrelbl = []
for e in outre:
outrelbl += [trLABELS[int(e['label'][-1])] if len(e["label"])==7 else trLABELS[int(e['label'][-2:])] ]
for i in range(len(outrelbl)):
if "(e2,e1)" in outrelbl[i]:
lstSentbilbl[i][0],lstSentbilbl[i][1] = lstSentbilbl[i][1],lstSentbilbl[i][0]
lstSentEnt[i][0],lstSentEnt[i][1] = lstSentEnt[i][1],lstSentEnt[i][0]
edccan = []
edccanbis = []
for i in range(len(outrelbl)):
if outrelbl[i] != "other":
edccanbis += [[lstSentEnt[i][0], lstSentEnt[i][1], outrelbl[i][:-7], lstSentEnc[i], lstSentbilbl[i]]]
#edccan += [[lstSentEnc[i],lstSentEnt[i][0]+" ["+lstSentbilbl[i][0][2:]+"]", lstSentEnt[i][1]+" ["+lstSentbilbl[i][1][2:]+"]",outrelbl[i][:-7]]]
edccanbis = postcor(edccanbis)
edccann = []
edchorm = []
edcrecep = []
hormrecep = []
hormcan = []
for e in edccanbis:
if e[-1]== ["B-EDC","B-CANCER"] and e[2] in ["CAUSE","INCREASE_RISK","SPEED_UP","NO_ASSOCIATION"]:
edccann += [[e[0],e[1],e[2]]]
edccan += [[e[3],e[0]+" ["+e[-1][0][2:]+"]", e[1]+" ["+e[-1][1][2:]+"]",e[2]]]
elif e[-1]== ["B-EDC","B-HORMONE"] and e[2] in ["ALTER", "INCREASE", "DECREASE", "MIMIC", "INCREASE_ACTIVITY","DECREASE_ACTIVITY"]:
edchorm += [[e[0],e[1],e[2]]]
edccan += [[e[3],e[0]+" ["+e[-1][0][2:]+"]", e[1]+" ["+e[-1][1][2:]+"]",e[2]]]
elif e[-1]== ["B-EDC","B-RECEPTOR"] and e[2] in ["BLOCK", "ACTIVATE", "INTERFER"] :
edcrecep += [[e[0],e[1],e[2]]]
edccan += [[e[3],e[0]+" ["+e[-1][0][2:]+"]", e[1]+" ["+e[-1][1][2:]+"]",e[2]]]
elif e[-1]== ["B-HORMONE","B-RECEPTOR"] and e[2] in ["BIND"] :
hormrecep += [[e[0],e[1],e[2]]]
edccan += [[e[3],e[0]+" ["+e[-1][0][2:]+"]", e[1]+" ["+e[-1][1][2:]+"]",e[2]]]
elif e[-1]== ["B-HORMONE","B-CANCER"] and e[2] in ["DEVELOP"]:
hormcan += [[e[0],e[1],e[2]]]
edccan += [[e[3],e[0]+" ["+e[-1][0][2:]+"]", e[1]+" ["+e[-1][1][2:]+"]",e[2]]]
edcrecepdf = pd.DataFrame(edcrecep, columns=["EDC", "RECEPTOR", "RELATION"])
edccanndf = pd.DataFrame(edccann, columns= ["EDC", "CANCER", "RELATION"] )
edchormdf = pd.DataFrame(edchorm , columns = ["EDC", "HORMONE", "RELATION"])
hormrecepdf = pd.DataFrame(hormrecep, columns = ["HORMONE", "RECEPTOR", "RELATION"])
hormcandf = pd.DataFrame(hormcan, columns = ["HORMONE", "CANCER", "RELATION"])
edccancsv = edccanndf.to_csv('edccan.csv')
edcrecepcsv = edcrecepdf.to_csv('edcrecep.csv')
edchormcsv = edchormdf.to_csv('edchorm.csv')
hormcancsv = hormcandf.to_csv('hormcan.csv')
hormrecepcsv = hormrecepdf.to_csv('hormrecep.csv')
edccandf = pd.DataFrame(edccan, columns= ["Sentence", "Entity 1", "Entity 2", "Relation"] )
edccandf.to_csv("table.csv")
with zipfile.ZipFile("allcsvs.zip", "w") as zipf:
if len(edccan)!=0:
zipf.write('table.csv')
if len(edccann)!=0:
zipf.write('edccan.csv')
if len(edcrecep)!=0:
zipf.write('edcrecep.csv')
if len(edchorm)!=0:
zipf.write('edchorm.csv')
if len(hormcan)!=0:
zipf.write('hormcan.csv')
if len(hormrecep)!=0:
zipf.write('hormrecep.csv')
zipf.close()
st.table(edccandf)
csv = edccandf.to_csv(index=False).encode('utf-8')
with st.sidebar:
st.subheader("You can only choose one download !!")
st.caption("we recommed ZIP file.")
st.write("Download table only :")
st.download_button(
label="Download CSV",
data=csv,
file_name='Relation_triples_table.csv',
mime='text/csv',
)
st.write("Download table plus separate csvs for each family of pairs :")
with open("allcsvs.zip", "rb") as fp:
btn = st.download_button(
label="Download ZIP",
data=fp,
file_name="SeparateCsvs.zip",
mime="application/zip"
)
|