Recently, I am doing some tasks about NLP. LTP provides a series of Chinese natural language processing tools. Users can use these tools to divide Chinese texts, segmentation, word -based marking, syntax analysis, and semantic marking.

Here is mainly rightLanguage Technology PlantForm | LTP)The online document records the test of Python.

Version:

ltp4.1.5.post2

python3.9

One, clause

from ltp import LTP
ltp = LTP()
input_file_dir = R'd: \ Graduation thesis \ Data processing \ text example.txt '
lines = [line.strip() for line in open(input_file_dir, 'r', encoding='utf-8').readlines()]
for line in lines:
    print(line)
    sents = ltp.sent_split(line)
    print(sents)
    exit()

result:

There were no clauses, but a single character.

Word

from ltp import LTP
ltp = LTP()
 input_file_dir = R'd: \ Graduation thesis \ Data processing \ text example.txt '
lines = [line.strip() for line in open(input_file_dir, 'r', encoding='utf-8').readlines()]
for line in lines:
    print(line)
    sents, hidden = ltp.seg([line])
    print(sents)
    exit()

result:

Successful word

Third, word marks

print(line)
sents, hidden = ltp.seg([line])
pos = ltp.pos(hidden)
print(sents)
print(pos)

result

Label correctly

Fourth, naming entity identification

ner = ltp.ner(hidden)
print(sents)
print(ner)
for i in range(len(ner[0])):
    tag, start, end = ner[0][i]
    print(tag, ":", "".join(sents[0][start:end + 1]))

result:

LTP's naming entity recognition can only identify common people's names, place names, and institutional names, which is not applicable to proprietary entities in the professional field.

5. Semantic character labeling

srl = ltp.srl(hidden,keep_empty=False)
print(srl)
for i in range(len(srl[0])):
tag = srl[0][i][0]
for j in range(len(srl[0][i][1])):
    lable_,start,end=srl[0][i][1][j]
    print(sents[0][tag], ":",lable_, "".join(sents[0][start:end + 1]))

result:

Semantic character labeling effect is very good

6. Analysis of dependent clauses

dep = ltp.dep(hidden)

result:

Successful labeling, because there is no sentence, the marking is too long and inaccurate, and you need to divide the sentence in advance.

seven,

  #Svanic dependence analysis (tree)
 sdp1 = ltp.sdp(hidden, mode='tree')
   # Semantic dependence analysis (Figure)
 sdp2 = ltp.sdp(hidden, mode='graph')

result: