Source code for pynlpir

# -*- coding: utf-8 -*-

"""Provides an easy-to-use Python interface to NLPIR/ICTCLAS.

The functions below are not as extensive as the full set of functions exported
by NLPIR (for that, see :mod:`pynlpir.nlpir`). A few design choices have been
made with these functions as well, e.g. they have been renamed and their output
is formatted differently.

The functions in this module all assume input is either unicode or encoded
using the encoding specified when :func:`open` is called.
These functions return unicode strings.

After importing this module, you must call :func:`open` in order to initialize
the NLPIR API. When you're done using the NLPIR API, call :func:`close` to exit
the API.


from __future__ import unicode_literals
import datetime as dt
import logging
import os
import sys

from . import nlpir, pos_map

__version__ = '0.5.2'

logger = logging.getLogger('pynlpir')

fopen = open

is_python3 = sys.version_info[0] > 2
if is_python3:
    unicode = str

#: The encoding configured by :func:`open`.
ENCODING = 'utf_8'

#: The encoding error handling scheme configured by :func:`open`.

[docs]class LicenseError(Exception): """A custom exception for missing/invalid license errors.""" pass
[docs]def open(data_dir=nlpir.PACKAGE_DIR, encoding=ENCODING, encoding_errors=ENCODING_ERRORS, license_code=None): """Initializes the NLPIR API. This calls the function :func:`~pynlpir.nlpir.Init`. :param str data_dir: The absolute path to the directory that has NLPIR's `Data` directory (defaults to :data:`pynlpir.nlpir.PACKAGE_DIR`). :param str encoding: The encoding that the Chinese source text will be in (defaults to ``'utf_8'``). Possible values include ``'gbk'``, ``'utf_8'``, or ``'big5'``. :param str encoding_errors: The desired encoding error handling scheme. Possible values include ``'strict'``, ``'ignore'``, and ``'replace'``. The default error handler is 'strict' meaning that encoding errors raise :class:`ValueError` (or a more codec specific subclass, such as :class:`UnicodeEncodeError`). :param str license_code: The license code that should be used when initializing NLPIR. This is generally only used by commercial users. :raises RuntimeError: The NLPIR API failed to initialize. Sometimes, NLPIR leaves an error log in the current working directory or NLPIR's ``Data`` directory that provides more detailed messages (but this isn't always the case). :raises LicenseError: The NLPIR license appears to be missing or expired. """ if license_code is None: license_code = '' global ENCODING if encoding.lower() in ('utf_8', 'utf-8', 'u8', 'utf', 'utf8'): ENCODING = 'utf_8' encoding_constant = nlpir.UTF8_CODE elif encoding.lower() in ('gbk', '936', 'cp936', 'ms936'): ENCODING = 'gbk' encoding_constant = nlpir.GBK_CODE elif encoding.lower() in ('big5', 'big5-tw', 'csbig5'): ENCODING = 'big5' encoding_constant = nlpir.BIG5_CODE else: raise ValueError("encoding must be one of 'utf_8', 'big5', or 'gbk'.") logger.debug("Initializing the NLPIR API: 'data_dir': '{}', 'encoding': " "'{}', 'license_code': '{}'".format( data_dir, encoding, license_code)) global ENCODING_ERRORS if encoding_errors not in ('strict', 'ignore', 'replace'): raise ValueError("encoding_errors must be one of 'strict', 'ignore', " "or 'replace'.") else: ENCODING_ERRORS = encoding_errors # Init in Python 3 expects bytes, not strings. if is_python3 and isinstance(data_dir, str): data_dir = _encode(data_dir) if is_python3 and isinstance(license_code, str): license_code = _encode(license_code) if not nlpir.Init(data_dir, encoding_constant, license_code): _attempt_to_raise_license_error(data_dir) raise RuntimeError("NLPIR function 'NLPIR_Init' failed.") else: logger.debug("NLPIR API initialized.")
[docs]def close(): """Exits the NLPIR API and frees allocated memory. This calls the function :func:`~pynlpir.nlpir.Exit`. """ logger.debug("Exiting the NLPIR API.") if not nlpir.Exit(): logger.warning("NLPIR function 'NLPIR_Exit' failed.") else: logger.debug("NLPIR API exited.")
def _attempt_to_raise_license_error(data_dir): """Raise an error if NLPIR has detected a missing or expired license. :param str data_dir: The directory containing NLPIR's `Data` directory. :raises LicenseError: The NLPIR license appears to be missing or expired. """ if isinstance(data_dir, bytes): data_dir = _decode(data_dir) data_dir = os.path.join(data_dir, 'Data') current_date ='%Y%m%d') timestamp ='[%Y-%m-%d %H:%M:%S]') data_files = os.listdir(data_dir) for f in data_files: if f == (current_date + '.err'): file_name = os.path.join(data_dir, f) with fopen(file_name) as error_file: for line in error_file: if not line.startswith(timestamp): continue if 'Not valid license' in line: raise LicenseError('Your license appears to have ' 'expired. Try running "pynlpir ' 'update".') elif 'Can not open License file' in line: raise LicenseError('Your license appears to be ' 'missing. Try running "pynlpir ' 'update".') def _decode(s, encoding=None, errors=None): """Decodes *s*.""" if encoding is None: encoding = ENCODING if errors is None: errors = ENCODING_ERRORS return s if isinstance(s, unicode) else s.decode(encoding, errors) def _encode(s, encoding=None, errors=None): """Encodes *s*.""" if encoding is None: encoding = ENCODING if errors is None: errors = ENCODING_ERRORS return s.encode(encoding, errors) if isinstance(s, unicode) else s def _to_float(s): """Converts *s* to a float if possible; if not, returns `False`.""" try: f = float(s) return f except ValueError: return False def _get_pos_name(code, name='parent', english=True, delimiter=':', pos_tags=pos_map.POS_MAP): """Gets the part of speech name for *code*. Joins the names together with *delimiter* if *name* is ``'all'``. See :func:``pynlpir.pos_map.get_pos_name`` for more information. """ pos_name = pos_map.get_pos_name(code, name, english, pos_tags=pos_tags) return delimiter.join(pos_name) if name == 'all' else pos_name
[docs]def segment(s, pos_tagging=True, pos_names='parent', pos_english=True, pos_tags=pos_map.POS_MAP): """Segment Chinese text *s* using NLPIR. The segmented tokens are returned as a list. Each item of the list is a string if *pos_tagging* is `False`, e.g. ``['我们', '是', ...]``. If *pos_tagging* is `True`, then each item is a tuple (``(token, pos)``), e.g. ``[('我们', 'pronoun'), ('是', 'verb'), ...]``. If *pos_tagging* is `True` and a segmented word is not recognized by NLPIR's part of speech tagger, then the part of speech code/name will be returned as :data:`None` (e.g. a space returns as ``(' ', None)``). This uses the function :func:`~pynlpir.nlpir.ParagraphProcess` to segment *s*. :param s: The Chinese text to segment. *s* should be Unicode or a UTF-8 encoded string. :param bool pos_tagging: Whether or not to include part of speech tagging (defaults to ``True``). :param pos_names: What type of part of speech names to return. This argument is only used if *pos_tagging* is ``True``. :data:`None` means only the original NLPIR part of speech code will be returned. Other than :data:`None`, *pos_names* may be one of ``'parent'``, ``'child'``, or ``'all'``. Defaults to ``'parent'``. ``'parent'`` indicates that only the most generic name should be used, e.g. ``'noun'`` for ``'nsf'``. ``'child'`` indicates that the most specific name should be used, e.g. ``'transcribed toponym'`` for ``'nsf'``. ``'all'`` indicates that all names should be used, e.g. ``'noun:toponym:transcribed toponym'`` for ``'nsf'``. :type pos_names: ``str`` or :data:`None` :param bool pos_english: Whether to use English or Chinese for the part of speech names, e.g. ``'conjunction'`` or ``'连词'``. Defaults to ``True``. This is only used if *pos_names* is ``True``. :param dict pos_tags: Custom part of speech tags to use. """ s = _decode(s) s = s.strip() logger.debug("Segmenting text with{} POS tagging: {}.".format( '' if pos_tagging else 'out', s)) result = nlpir.ParagraphProcess(_encode(s), pos_tagging) result = _decode(result) logger.debug("Finished segmenting text: {}.".format(result)) logger.debug("Formatting segmented text.") tokens = result.strip().replace(' ', ' ').split(' ') tokens = [' ' if t == '' else t for t in tokens] if pos_tagging: for i, t in enumerate(tokens): token = tuple(t.rsplit('/', 1)) if len(token) == 1: token = (token[0], None) if pos_names is not None and token[1] is not None: pos_name = _get_pos_name(token[1], pos_names, pos_english, pos_tags=pos_tags) token = (token[0], pos_name) tokens[i] = token logger.debug("Formatted segmented text: {}.".format(tokens)) return tokens
[docs]def get_key_words(s, max_words=50, weighted=False): """Determines key words in Chinese text *s*. The key words are returned in a list. If *weighted* is ``True``, then each list item is a tuple: ``(word, weight)``, where *weight* is a float. If it's *False*, then each list item is a string. This uses the function :func:`~pynlpir.nlpir.GetKeyWords` to determine the key words in *s*. :param s: The Chinese text to analyze. *s* should be Unicode or a UTF-8 encoded string. :param int max_words: The maximum number of key words to find (defaults to ``50``). :param bool weighted: Whether or not to return the key words' weights (defaults to ``True``). """ s = _decode(s) logger.debug("Searching for up to {}{} key words in: {}.".format( max_words, ' weighted' if weighted else '', s)) result = nlpir.GetKeyWords(_encode(s), max_words, weighted) result = _decode(result) logger.debug("Finished key word search: {}.".format(result)) logger.debug("Formatting key word search results.") fresult = result.strip('#').split('#') if result else [] if weighted: weights, words = [], [] for w in fresult: result = w.split('/') word, weight = result[0], result[2] weight = _to_float(weight) weights.append(weight or 0.0) words.append(word) fresult = zip(words, weights) if is_python3: # Return a list instead of a zip object in Python 3. fresult = list(fresult) logger.debug("Key words formatted: {}.".format(fresult)) return fresult