访问MALLET'通过Gensim的诊断文件

Question

是否有方法通过Python中的Gensim使用提供的API访问MALLET的诊断文件或其内容？

Answer 1

似乎没有任何可能性。我通过Python的子流程模块在命令行中运行MALLET来解决这个问题：

import subprocess
from pathlib import Path
MALLET_PATH = r"C:mallet"  # set to where your "bin/mallet" path is
seglen = 500
topic_count = 20
start = 0
iterations = 20
num_threads = 10  # determines threads used for parallel training
# remember to change backslashes if needed
wdir = Path("../..")
corpusdir = wdir.joinpath("5_corpus", f"seglen-{seglen}")
corpusdir.mkdir(exist_ok=True, parents=True)
mallet_dir = wdir.joinpath("6_evaluation/models/mallet", f"seglen-{seglen}")
topic_dir = mallet_dir.joinpath(f"topics-{topic_count}")

def create_input_files():
# create MALLETs input files
for file in corpusdir.glob("*.txt"):
output = mallet_dir.joinpath(f"{file.stem}.mallet")
# doesn't need to happen more than once -- usually.
if output.is_file(): continue
print(f"--{file.stem}")
cmd = f"bin\mallet import-file " 
f"--input {file.absolute()} " 
f"--output {output.absolute()} " 
f"--keep-sequence"
subprocess.call(cmd, cwd=MALLET_PATH, shell=True)
print("import finished")

def modeling():
# start modeling
for file in mallet_dir.glob("*.mallet"):
for i in range(start, iterations):
print("iteration ", str(i))
print(f"--{file.stem}")
# output directory
formatdir = topic_dir.joinpath(f"{file.stem.split('-')[0]}")
outputdir = formatdir.joinpath(f"iteration-{i}")
outputdir.mkdir(parents=True, exist_ok=True)
outputdir = str(outputdir.absolute())
# output files
statefile = outputdir + r"topic-state.gz"
keysfile = outputdir + r"keys.txt"
compfile = outputdir + r"composition.txt"
diagnostics_xml = outputdir + r"diagnostics.xml"
# building cmd string
cmd = f"bin\mallet train-topics " 
f"--input {file.absolute()} " 
f"--num-topics {topic_count} " 
f"--output-state {statefile} " 
f"--output-topic-keys {keysfile} " 
f"--output-doc-topics {compfile} " 
f"--diagnostics-file {diagnostics_xml} " 
f"--num-threads {num_threads}"
# call mallet
subprocess.call(cmd, cwd=MALLET_PATH, shell=True)
print("models trained")
#create_input_files()
modeling()