mecab_sample.py

# -*- coding: utf-8 -*-
import re
from natto import MeCab

patt = re.compile(r'''(?x)
# Ticker symbols
[0-9\uFF10-\uFF19]{2,}(\s|\.)[A-Z\uFF21-\uFF3A]{1,2}  |

# short-from contractions
[\'|\u2019](d|ll|m|s|re|ve)                           |

# short-form negative (preserve in their entirety
n[\'|\u2019]t                                         |

# hyphen- and forward-slash delimited words
([A-Z]+(\-|\/))+[A-Z]+                                |

# extract text from single- and double-quotes
[\"\u2018\u2019\u201C\u201D]
''', re.IGNORECASE| re.MULTILINE)

txt = """
Wouldn't you know? I've had it! This's a fine how-do-you-do, isn't it, Jor-El! 

She said "That'll be the day, uh-huh, that I die."

1099.T は銘柄コードの例です。

注意：
　今季はやはり 1079 JP および AAPL も気になる存在である。
　P/E の観点からいうと、双方は「ばちぐぅ～！」という感じです。

GOOGおよびAAPL、または1234.JPか１０９９ ＪＰも何れも買いである。

An example with fancy double quotes: \u201CThe Sun Doesn't Rise Always, Don't You know?\u201D.

An example with fancy single quotes: What an \u2018exquisite\u2019 bouquet!
"""

nm = MeCab()

for n in nm.parse(txt, boundary_constraints=patt, as_nodes=True):
    print(n.surface)