class Morph(object):

self.pos1 = pos1

class Chunk(object):

self.srcs = srcs

doc = []

skip_empty_line = False

with open("neko.txt.cabocha") as f:

lines = f.readlines()

chunks = []

chunk = Chunk()

morphs = []

for line in lines:

line = line[:-1]

# Enter new chunk

if line.startswith("*"):

if len(morphs) > 0:

chunk.morphs = morphs

chunks.append(chunk)

morphs = []

# parse chunk

idx, dst, _, score = line[2:].split(" ")

dst = dst[:-1] # remove "D"

idx, dst, score = int(idx), int(dst), float(score)

chunk = Chunk(idx=idx, morphs=[], dst=dst, srcs=[])

continue

elif line == "EOS":

# todo: remove dup

if len(morphs) > 0:

chunk.morphs = morphs

chunks.append(chunk)

morphs = []

if not skip_empty_line or len(chunks) > 0:

if len(chunks) > 0:

assert chunks[0].idx == 0

for chunk in chunks:

chunk.srcs = (

list(map(lambda c: c.idx, filter(lambda c: chunk.idx == c.dst, chunks))))

doc.append(chunks)

chunks = []

morphs = []

continue

# parse POS tagging result

surface, rest = line.split("\t")

rest = rest.split(",")

assert len(rest) >= 6

pos, pos1, base = rest[0], rest[1], rest[6]

m = Morph(surface, base, pos, pos1)

morphs.append(m)

import pydot

def vis_dag(chunks):

g.write_jpeg("44.jpeg", prog="dot")

vis_dag(doc[7])