Initial chekin post-discontinuity.
This commit is contained in:
commit
6556164879
|
@ -0,0 +1,61 @@
|
|||
*.py[cod]
|
||||
|
||||
# C extensions
|
||||
*.so
|
||||
|
||||
# Packages
|
||||
*.egg
|
||||
*.egg-info
|
||||
dist
|
||||
build
|
||||
eggs
|
||||
parts
|
||||
bin
|
||||
var
|
||||
sdist
|
||||
develop-eggs
|
||||
.installed.cfg
|
||||
lib
|
||||
lib64
|
||||
__pycache__
|
||||
|
||||
# Installer logs
|
||||
pip-log.txt
|
||||
|
||||
# Unit test / coverage reports
|
||||
.coverage
|
||||
.tox
|
||||
nosetests.xml
|
||||
|
||||
# Translations
|
||||
*.mo
|
||||
|
||||
# Mr Developer
|
||||
.mr.developer.cfg
|
||||
.project
|
||||
.pydevproject
|
||||
|
||||
# Emacs git ignore
|
||||
# -*- mode: gitignore; -*-
|
||||
*~
|
||||
\#*\#
|
||||
/.emacs.desktop
|
||||
/.emacs.desktop.lock
|
||||
*.elc
|
||||
auto-save-list
|
||||
tramp
|
||||
.\#*
|
||||
|
||||
# Org-mode
|
||||
.org-id-locations
|
||||
*_archive
|
||||
|
||||
# flymake-mode
|
||||
*_flymake.*
|
||||
|
||||
# eshell files
|
||||
/eshell/history
|
||||
/eshell/lastdir
|
||||
|
||||
# elpa packages
|
||||
/elpa/
|
|
@ -0,0 +1,11 @@
|
|||
Copyright 2021 Aldercone Studio
|
||||
|
||||
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
|
||||
|
||||
1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
|
||||
|
||||
2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
|
||||
|
||||
3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
|
||||
|
||||
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
@ -0,0 +1,56 @@
|
|||
# carkov #
|
||||
|
||||
This is a library for creating and walking simple markov chains. It is
|
||||
meant for things like text generators (such as ebooks bots and word
|
||||
generators) and thus is not 'mathetematically correct'. It has some
|
||||
tools for doing text analysis but more are planned in the future
|
||||
(stubs exist to illustrate some plans, see TODO.md).
|
||||
|
||||
## Command line interface ##
|
||||
|
||||
This library includes a command line interface to analyzing text and
|
||||
then walk the chain and generate text from the analysis.
|
||||
|
||||
To analyze a corpus of text files, thus:
|
||||
|
||||
`carkov analyze mychain.chain textfile1.txt textfile2.txt ... textfileN.txt`
|
||||
|
||||
To walk a chain and generate text form it, thus:
|
||||
|
||||
`carkov chain mychain.chain -c 10`
|
||||
|
||||
There are two analysis modes currently supported, `english` and
|
||||
`word`, which are passed to the analyze method with the `-m`
|
||||
argument. `english` mode analyzes the input in a word-wise method: the
|
||||
input is segmented into (English-style) sentences, each of which are
|
||||
analyzed as separate chains of words. `word` segments the input into
|
||||
tokens, each of which is analyzed as a series of characters
|
||||
separately.
|
||||
|
||||
Analysis also allows a window size to be specified, so that each item
|
||||
in the chain may be a fixed series of items of a specific length (for
|
||||
example, the word `foo` with a window of 2, would analyze to (_, _) ->
|
||||
'f', (_, f) -> o, (f, o) -> o, etc). The wider the window, the more
|
||||
similar or identical to the input stream the output becomes since
|
||||
there are fewer total options to follow any given token. This is
|
||||
specified with the analysis command line with the `-w` argument.
|
||||
|
||||
## About Library ##
|
||||
|
||||
The library itself exposes objects and interfaces to do the same as
|
||||
the command line above. A todo item on this project is to generate
|
||||
documentation and examples, but looking at the contents of __main__.py
|
||||
should be instructive. The library is written in such a way as to be
|
||||
pretty agnostic about the items that are chained, and hypothetically
|
||||
any sequential set of things could work for this. Some framework would
|
||||
have to be written to support displaying these sorts of things but it
|
||||
should be possible if non-textual data were desired.
|
||||
|
||||
The library also provides a few mechanisms for serializing a ready to
|
||||
use chain for reuse in other projects. The command line makes use of
|
||||
the binary serialization mechanism (which uses `msgpack`) to save
|
||||
chains from the analysis step for re-use in the chain step. There is
|
||||
also a mechanism which produces a python source file tthat can be
|
||||
embedded in a target project so that a python project can use the
|
||||
chain without having to include an extra data file. It should be noted
|
||||
that this of course is extremely inefficient for large chains.
|
|
@ -0,0 +1,9 @@
|
|||
* Implement text filters
|
||||
** implement abstractize number filter which will take any number as imput and return a NUMBER abstract.
|
||||
** Implement the abstractize roman numeral filter which will take a token that looks like a roman numeral (except for I)
|
||||
return a NUMBER abstract.
|
||||
** implement punctuation stripper / abstractizer which will take a punctuation token and return an abstract or abort the
|
||||
token.
|
||||
** Implement asciifier token which will take a unicode string and return an ascii approximation.
|
||||
* Implement some example code.
|
||||
* Complete documentation and introductions to actually be useful to users.
|
|
@ -0,0 +1,7 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
version = '0.1.2'
|
|
@ -0,0 +1,171 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
# This module provides a command line interface to doing some common operations.
|
||||
#
|
||||
|
||||
import argparse
|
||||
import enum
|
||||
import pathlib
|
||||
import sys
|
||||
import traceback
|
||||
from random import Random
|
||||
from typing import cast
|
||||
|
||||
from .analyze.abstract import AbstractAnalyzer
|
||||
from .analyze.english import English
|
||||
from .analyze.words import Words
|
||||
from .chain import Chain, from_analyzer
|
||||
from .serializer import dump_chainer, load_chainer
|
||||
from .utils import make_sent, make_word
|
||||
|
||||
ERROR_NO_FILE = 1
|
||||
ERROR_EXISTING_OUTPUT = 2
|
||||
ERROR_WRONG_ORDER = 3
|
||||
ERROR_WRONG_CLASS = 4
|
||||
|
||||
|
||||
class AnalyzeMode(enum.Enum):
|
||||
english = "english"
|
||||
word = "word"
|
||||
|
||||
def __str__(self) -> str:
|
||||
return self.value
|
||||
|
||||
|
||||
JOINERS = {"Words": make_word, "English": make_sent}
|
||||
|
||||
|
||||
def parse_arguments() -> argparse.Namespace:
|
||||
parser = argparse.ArgumentParser(prog="python -mcarkov",
|
||||
description=("Process a text corpus in a markov chain fashion and/or output from"
|
||||
"an analysis."))
|
||||
subparsers = parser.add_subparsers(dest='command')
|
||||
analyze_sub = subparsers.add_parser('analyze', help="Analyze a corpus")
|
||||
analyze_sub.add_argument('output', help="Output chain to specified destination", type=pathlib.Path)
|
||||
analyze_sub.add_argument('input', help="The corpus to analyze", type=pathlib.Path, nargs='+')
|
||||
overappend = analyze_sub.add_mutually_exclusive_group()
|
||||
overappend.add_argument('-o', '--overwrite', help='Overwrite output file.', action='store_true')
|
||||
overappend.add_argument('-a', '--append', help='Append output file.', action='store_true')
|
||||
analyze_sub.add_argument('-w', '--window', help='Select length of analysis window', type=int, default=2)
|
||||
analyze_sub.add_argument('-m',
|
||||
'--mode',
|
||||
help='Select analyzis mode',
|
||||
type=AnalyzeMode,
|
||||
choices=list(AnalyzeMode),
|
||||
default=AnalyzeMode.english)
|
||||
|
||||
analyze_sub.add_argument('-t', '--test', help="Output a sample from the chainer generated", action='store_true')
|
||||
|
||||
chain_sub = subparsers.add_parser('chain', help="Output from a chainer")
|
||||
chain_sub.add_argument('input', help="The chain file to load", type=pathlib.Path)
|
||||
chain_sub.add_argument('-c', '--count', help="Number of chain outputs to output", type=int, default=10)
|
||||
chain_sub.add_argument('-m', '--maxlen', help="Maximum length in tokens of output (0 is unlimited)", type=int, default=0)
|
||||
|
||||
return parser.parse_args()
|
||||
|
||||
|
||||
def print_chainer_output(chainer: Chain, random_state: Random):
|
||||
if chainer.analyzer_class in JOINERS:
|
||||
print(JOINERS[cast(str, chainer.analyzer_class)](chainer.walk(random_state, True)))
|
||||
else:
|
||||
print(chainer.walk(random_state, True))
|
||||
|
||||
|
||||
def command_analyze(args: argparse.Namespace) -> int:
|
||||
if not any(x.exists() for x in args.input):
|
||||
print("Must specify an existing file as input for the analyzer.")
|
||||
return ERROR_NO_FILE
|
||||
|
||||
if args.output.exists() and not (args.overwrite or args.append):
|
||||
print("Output file exists, pass --overwrite to overwrite or --append to add to exsiting analysis.")
|
||||
return ERROR_EXISTING_OUTPUT
|
||||
|
||||
analyzer: AbstractAnalyzer
|
||||
if args.mode == AnalyzeMode.english:
|
||||
analyzer = English(args.window)
|
||||
# we just dump a whole file into the english analyzer
|
||||
for inp in args.input:
|
||||
if not inp.exists():
|
||||
print(f"warning {inp} does not exist")
|
||||
continue
|
||||
print(f"analyze: {inp}")
|
||||
analyzer.analyze(inp.read_text('utf-8'))
|
||||
else:
|
||||
analyzer = Words(args.window)
|
||||
# we do line-by-line single word dumps into the word analyzer
|
||||
for inp in args.input:
|
||||
if not inp.exists():
|
||||
print(f"warning {inp} does not exist")
|
||||
continue
|
||||
print(f"analyze: {inp}")
|
||||
with inp.open('r') as words:
|
||||
for line in words:
|
||||
analyzer.analyze(line.strip().lower())
|
||||
|
||||
if args.append:
|
||||
# in append mode we load an existing chain file, and then run the analyzer and merge the contents.
|
||||
with open(args.output, 'rb') as old:
|
||||
chainer = load_chainer(old)
|
||||
if chainer.order != analyzer.order:
|
||||
print("Append chainer order isn't the same as the analyzer order.")
|
||||
return ERROR_WRONG_ORDER
|
||||
if chainer.analyzer_class != analyzer.__class__.__name__:
|
||||
print("Append chainer class isn't the same as analyzer class.")
|
||||
return ERROR_WRONG_CLASS
|
||||
chainer.integrate(analyzer.chain_counts)
|
||||
else:
|
||||
chainer = from_analyzer(analyzer)
|
||||
|
||||
with open(args.output, 'wb') as output:
|
||||
dump_chainer(chainer, output)
|
||||
print(f"Wrote chainer to {args.output}")
|
||||
|
||||
if args.test:
|
||||
r = Random()
|
||||
for _ in range(0, 5):
|
||||
print_chainer_output(chainer, r)
|
||||
|
||||
return 0
|
||||
|
||||
|
||||
def command_chain(args: argparse.Namespace) -> int:
|
||||
r = Random()
|
||||
|
||||
if not args.input.exists():
|
||||
print("Must specify a chain file to load.")
|
||||
return ERROR_NO_FILE
|
||||
|
||||
with args.input.open('rb') as inp:
|
||||
chainer = load_chainer(inp)
|
||||
|
||||
if args.count < 1:
|
||||
args.count = 1
|
||||
|
||||
for _ in range(0, args.count):
|
||||
print_chainer_output(chainer, r)
|
||||
|
||||
return 0
|
||||
|
||||
|
||||
def main() -> int:
|
||||
args = parse_arguments()
|
||||
if args.command == 'analyze':
|
||||
return command_analyze(args)
|
||||
elif args.command == 'chain':
|
||||
return command_chain(args)
|
||||
else:
|
||||
print("Expect a command `analyze` or `chain`. See --help for details.")
|
||||
return 1
|
||||
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
try:
|
||||
sys.exit(main())
|
||||
except Exception:
|
||||
print("Unexpected exception!")
|
||||
traceback.print_exc()
|
|
@ -0,0 +1,54 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
This module provides a few utility objects, especially the Abstract object which is used for terminals
|
||||
and other abstract tokens.
|
||||
"""
|
||||
|
||||
|
||||
class CarkovFilterException(BaseException):
|
||||
"""
|
||||
Base exception for filter stages.
|
||||
"""
|
||||
|
||||
|
||||
class Abort(CarkovFilterException):
|
||||
"""
|
||||
This exception is intended for a protocol by which filters can abort a particular token from being added to the
|
||||
stream.
|
||||
"""
|
||||
|
||||
|
||||
class AbortSegment(CarkovFilterException):
|
||||
"""
|
||||
This exception is intended for a protocol by which filters can abort an entire segment if a particular token would
|
||||
be rejected.
|
||||
"""
|
||||
|
||||
|
||||
class Abstract:
|
||||
"""
|
||||
This is used as a way to indicate abstract tokens in a stream of tokens.
|
||||
"""
|
||||
def __init__(self, name: str):
|
||||
self.name = name
|
||||
|
||||
def __repr__(self) -> str:
|
||||
if self == NUMBER:
|
||||
return 'NUMBER'
|
||||
elif self == TERMINAL:
|
||||
return 'TERMINAL'
|
||||
|
||||
return f"carkov.abstracts.Abstract({self.name})"
|
||||
|
||||
|
||||
NUMBER = Abstract("NUMBER")
|
||||
"""A universal Number abstract."""
|
||||
|
||||
|
||||
TERMINAL = Abstract("TERMINAL")
|
||||
"""A Universal Terminal abostract."""
|
|
@ -0,0 +1,5 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
|
@ -0,0 +1,120 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
from abc import ABC, abstractmethod
|
||||
from collections import deque, defaultdict
|
||||
|
||||
from ..abstracts import TERMINAL
|
||||
from ..utils import merge_dict
|
||||
|
||||
"""
|
||||
This module defines the base class for analyzers which do basic statistical analysis on a corpus.
|
||||
"""
|
||||
|
||||
|
||||
class AbstractAnalyzer(ABC):
|
||||
def __init__(self, order, filters=None):
|
||||
"""
|
||||
Initialize the analyzer.
|
||||
|
||||
Arguments:
|
||||
order (int): Defines the window size this analyzer uses.
|
||||
filters: A list of callables to apply to each token before processing.
|
||||
"""
|
||||
if filters is None:
|
||||
filters = []
|
||||
self.order = order
|
||||
self.filters = filters
|
||||
self.tokens = {}
|
||||
self.chain_counts = {}
|
||||
|
||||
def analyze(self, corpus):
|
||||
"""
|
||||
Analyze a corpus and integrate the data into the internal state.
|
||||
|
||||
Arguments:
|
||||
corpus (abstract): This could be any type that the class is prepared to process.
|
||||
|
||||
Retruns:
|
||||
self.chain_counts after processing.
|
||||
"""
|
||||
counts = self.analyze_corpus(corpus)
|
||||
merge_dict(self.chain_counts, counts)
|
||||
return self.chain_counts
|
||||
|
||||
def analyze_corpus(self, corpus):
|
||||
"""
|
||||
Do the actual analysis of corpus, and return a count dictionary.
|
||||
|
||||
Arguments:
|
||||
corpus (abstract): This could be any type that the class is prepared to process.
|
||||
|
||||
Returns:
|
||||
(dict) a count dictionary of just this corpus
|
||||
"""
|
||||
segments = self.segmentize_corpus(corpus)
|
||||
counts = {}
|
||||
for segment in segments:
|
||||
merge_dict(counts, self.analyze_segment(segment))
|
||||
|
||||
return counts
|
||||
|
||||
@abstractmethod
|
||||
def segmentize_corpus(self, corpus):
|
||||
"""
|
||||
Convert a corpus into a series of segments.
|
||||
|
||||
This must be overloaded by child class.
|
||||
|
||||
Arguments:
|
||||
corpus (abstract): This could be any type that the class is prepared to process.
|
||||
|
||||
Returns:
|
||||
(array of abstract): An array of segments that this class is prepared to process.
|
||||
"""
|
||||
|
||||
def analyze_segment(self, segment):
|
||||
"""
|
||||
Process a segment into a series of tokens.
|
||||
|
||||
Arguments:
|
||||
segment (abstract): This could be of any type that this class is prepared to process.
|
||||
|
||||
Returns:
|
||||
(counts dictionary): A dictionary keyed by windowed token keys with counts of each following token
|
||||
"""
|
||||
tokens = self.tokenize_segment(segment) + [TERMINAL]
|
||||
token = deque([None] * self.order, self.order)
|
||||
counts = defaultdict(lambda: defaultdict(int))
|
||||
for raw_token in tokens:
|
||||
raw_token = self.process_token(raw_token)
|
||||
tkey = tuple(token)
|
||||
counts[tkey][raw_token] += 1
|
||||
|
||||
token.append(raw_token)
|
||||
return counts
|
||||
|
||||
@abstractmethod
|
||||
def tokenize_segment(self, segment):
|
||||
"""
|
||||
Convert a segment into a series of tokens.
|
||||
|
||||
This must be overloaded by child class.
|
||||
|
||||
Arguments:
|
||||
segment (abstract): This could be of any type that this class is prepared to process.
|
||||
|
||||
Returns:
|
||||
(array of tokens): The format and type of tokens is defined by the child class.
|
||||
|
||||
"""
|
||||
|
||||
...
|
||||
|
||||
def process_token(self, raw_token):
|
||||
for filter in self.filters:
|
||||
raw_token = filter(raw_token)
|
||||
return raw_token
|
|
@ -0,0 +1,26 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
import nltk
|
||||
|
||||
from .abstract import AbstractAnalyzer
|
||||
|
||||
|
||||
class English(AbstractAnalyzer):
|
||||
def __init__(self, order, filters=None):
|
||||
if filters is None:
|
||||
filters = []
|
||||
super().__init__(order, filters)
|
||||
|
||||
def segmentize_corpus(self, corpus):
|
||||
chunks = corpus.split('\n\n')
|
||||
ret = []
|
||||
for chunk in chunks:
|
||||
ret = ret + nltk.sent_tokenize(chunk)
|
||||
return ret
|
||||
|
||||
def tokenize_segment(self, segment):
|
||||
return list(nltk.word_tokenize(segment))
|
|
@ -0,0 +1,20 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
from .abstract import AbstractAnalyzer
|
||||
|
||||
|
||||
class Words(AbstractAnalyzer):
|
||||
def __init__(self, order, filters=None):
|
||||
if filters is None:
|
||||
filters = []
|
||||
super().__init__(order, filters)
|
||||
|
||||
def segmentize_corpus(self, corpus):
|
||||
return corpus.split(' ')
|
||||
|
||||
def tokenize_segment(self, segment):
|
||||
return list(segment)
|
|
@ -0,0 +1,166 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
This module defines a chainer class which can process a count dictionary from an analyzer
|
||||
and provides convenience functions for walking the chain.
|
||||
"""
|
||||
|
||||
|
||||
from collections import UserDict, deque
|
||||
from typing import Any, Dict, Tuple, Optional
|
||||
|
||||
from .abstracts import TERMINAL
|
||||
from .analyze.abstract import AbstractAnalyzer
|
||||
from .utils import merge_stats, convert_counts, weighted_stat_choice
|
||||
|
||||
|
||||
ChainType = Dict[Tuple[Any], Any]
|
||||
|
||||
|
||||
def from_analyzer(analyzer: AbstractAnalyzer):
|
||||
"""
|
||||
Static initializer: Return a chainer with parameters and contents based on an analyzer instance.
|
||||
"""
|
||||
chainer = Chain(analyzer.order, analyzer.__class__.__name__)
|
||||
chainer.integrate(analyzer.chain_counts)
|
||||
return chainer
|
||||
|
||||
|
||||
class Chain(UserDict):
|
||||
def __init__(self, order: int, analyzer_class: Optional[str] = None):
|
||||
"""
|
||||
Initialize Chain class
|
||||
|
||||
Arguments:
|
||||
order: The window size of this chainer.
|
||||
"""
|
||||
self.order = order
|
||||
self.data: ChainType = {}
|
||||
self.start_token = (None, ) * self.order
|
||||
self.analyzer_class = analyzer_class
|
||||
|
||||
def integrate(self, counts: ChainType):
|
||||
"""
|
||||
Accept a counts dictionary and merge it with local data and recalculate statistical relationships between
|
||||
outcomes. The counts must be from an analyzer of the same order.
|
||||
|
||||
Arguments:
|
||||
counts: A counts dictionary as contained in the analyzer's chain_counts
|
||||
"""
|
||||
for key, count in counts.items():
|
||||
stat = convert_counts(count)
|
||||
if key in self.data:
|
||||
merge_stats(self.data[key], stat)
|
||||
else:
|
||||
self.data[key] = stat
|
||||
self.update_stats()
|
||||
|
||||
def merge(self, other):
|
||||
"""
|
||||
Merge a separate chainer's data into this chainer. They must be of the same order.
|
||||
|
||||
Arguments:
|
||||
other (Chain): Another chain of the same order.
|
||||
"""
|
||||
for key, stat in other.items():
|
||||
if key in self.data:
|
||||
merge_stats(self.data[key], stat)
|
||||
else:
|
||||
self.data[key] = stat
|
||||
self.update_stats()
|
||||
|
||||
def update_stats(self):
|
||||
"""
|
||||
Update all of the statistical ratios in the chain.
|
||||
"""
|
||||
for token in self.data:
|
||||
self.update_stat(token)
|
||||
|
||||
def update_stat(self, parent_token):
|
||||
"""
|
||||
Update one specific set of statistical ratios in the chain.
|
||||
|
||||
Arguments:
|
||||
parent_token: A windowed token tuple which points at the part of the chain to update
|
||||
"""
|
||||
stat = self.data[parent_token]
|
||||
total = sum([s[0] for s in stat.values()])
|
||||
for it in stat.values():
|
||||
it[1] = int((it[0] / total) * 100)
|
||||
|
||||
def add(self, parent_token, token):
|
||||
"""
|
||||
Add a new count to the chain.
|
||||
|
||||
Arguments:
|
||||
parent_token: A windowed token tuple which points to the location to add the new token.
|
||||
token: The token to add.
|
||||
"""
|
||||
if parent_token not in self.data:
|
||||
self.data[parent_token] = {}
|
||||
|
||||
if token in self.data[parent_token]:
|
||||
self.data[parent_token][token][0] += 1
|
||||
self.update_stat(parent_token)
|
||||
else:
|
||||
self.data[parent_token][token] = [1, 100]
|
||||
|
||||
def select(self, parent_token, random_generator, weighted=False):
|
||||
"""
|
||||
Select a token from a given parent token.
|
||||
|
||||
Arguments:
|
||||
parent_token: A windowed token tuple
|
||||
random_generator: A random.Random instance
|
||||
weighted (bool, default=False): Whether to do a weighted select or a random select.
|
||||
|
||||
Returns:
|
||||
A token
|
||||
"""
|
||||
if parent_token not in self.data:
|
||||
return None
|
||||
if weighted:
|
||||
return weighted_stat_choice(random_generator, self.data[parent_token])
|
||||
else:
|
||||
return random_generator.choice(list(self.data[parent_token].keys()))
|
||||
|
||||
def walk(self, random_generator, weighted=False, maximum=1000):
|
||||
"""
|
||||
Return a list of tokens by walking the chain.
|
||||
|
||||
Arguments:
|
||||
random_generator: A random.Random instance
|
||||
weighted: Whether to do a weighted select at each step.
|
||||
maximum: THe maximum number of walks to perform.
|
||||
|
||||
Returns:
|
||||
A list of tokens
|
||||
"""
|
||||
token = self.start_token
|
||||
item = None
|
||||
output = []
|
||||
while (len(output) < maximum):
|
||||
item = self.select(token, random_generator, weighted)
|
||||
if item == TERMINAL:
|
||||
return output
|
||||
output.append(item)
|
||||
token = self.next_token(token, item)
|
||||
|
||||
def next_token(self, parent_token, token):
|
||||
"""
|
||||
Given a windowed token tuple and a token, return the next windowed token tuple.
|
||||
|
||||
Arguments:
|
||||
parent_token: A windowed token tuple
|
||||
token: A token
|
||||
|
||||
Returns:
|
||||
A windowed token tuple which would be the next step in the chain after the token.
|
||||
"""
|
||||
q = deque(parent_token, self.order)
|
||||
q.append(token)
|
||||
return tuple(q)
|
|
@ -0,0 +1,35 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
Various filter functions that may be useful for processing certain kinds of corpora.
|
||||
"""
|
||||
|
||||
from typing import Optional
|
||||
|
||||
# from unidecode import unidecode # fixme asciifying filter
|
||||
|
||||
|
||||
# All of these filters operate on string tokens
|
||||
|
||||
def str_abstractize_numbers(token: str) -> Optional[str]:
|
||||
"""Replace all numbers with a Number abstract."""
|
||||
return None
|
||||
|
||||
|
||||
def str_abstractize_roman(token: str) -> Optional[str]:
|
||||
"""Replace roman numerals with a Number abstract."""
|
||||
return None
|
||||
|
||||
|
||||
def str_strip_punct(token: str) -> Optional[str]:
|
||||
"""Remove any punctuation characters."""
|
||||
return None
|
||||
|
||||
|
||||
def str_asciify(token: str) -> Optional[str]:
|
||||
"""Convert all characters to an ascii approximation."""
|
||||
return None
|
|
@ -0,0 +1,37 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
Serialize chain as a python structure.
|
||||
"""
|
||||
|
||||
from io import TextIOBase
|
||||
from . import version
|
||||
from .chain import Chain
|
||||
|
||||
template = """
|
||||
# serialized from version {version}
|
||||
def get_chainer():
|
||||
from carkov.chain import Chain
|
||||
from carkov.abstracts import NUMBER, TERMINAL, Abstract
|
||||
chain = Chain({order}, "{analyzer}")
|
||||
chain.data = {data}
|
||||
return chain
|
||||
"""
|
||||
|
||||
|
||||
def dump_chainer(chain: Chain, outfile: TextIOBase):
|
||||
"""
|
||||
Serialize a chainer to an open IO stream
|
||||
|
||||
Arguments:
|
||||
chain: A Chain object
|
||||
outfile: An open IO stream in text mode that will be writen to
|
||||
"""
|
||||
outfile.write(template.format(version=version,
|
||||
order=chain.order,
|
||||
analyzer=chain.analyzer_class,
|
||||
data=repr(chain.data).replace(']},', ']},\n')))
|
|
@ -0,0 +1,77 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
Use msgpack to serialize a chainer to disk and then reread it from a serialized file.
|
||||
"""
|
||||
|
||||
from typing import Any, BinaryIO, Dict, Tuple, cast
|
||||
|
||||
import msgpack
|
||||
|
||||
from . import version
|
||||
from .abstracts import NUMBER, TERMINAL, Abstract
|
||||
from .chain import Chain
|
||||
|
||||
|
||||
def _unserialize_encode_helper(obj: Dict) -> Any:
|
||||
"""
|
||||
This is a helper function which handles Abstract objects for serialization.
|
||||
"""
|
||||
if '$$####$$' in obj:
|
||||
val: Abstract
|
||||
if obj['n'] == 'TERMINAL':
|
||||
val = TERMINAL
|
||||
elif obj['n'] == 'NUMBER':
|
||||
val = NUMBER
|
||||
else:
|
||||
val = Abstract(obj['n'])
|
||||
return val
|
||||
return obj
|
||||
|
||||
|
||||
def _serialize_encode_helper(obj: Any) -> Any:
|
||||
"""
|
||||
This is a helper function which handles Abstract objects for serialization.
|
||||
"""
|
||||
if isinstance(obj, Abstract):
|
||||
obj = {'$$####$$': True, 'n': cast(Abstract, obj).name}
|
||||
return obj
|
||||
|
||||
|
||||
def load_chainer(infile: BinaryIO) -> Chain:
|
||||
"""
|
||||
Unserialize a chainer from an open IO stream
|
||||
|
||||
Arguments:
|
||||
infile: An open IO stream in binary mode pointing at a messagepack stream
|
||||
|
||||
Returns:
|
||||
a new Chain object initialized with the contents of the stream.
|
||||
"""
|
||||
serialdict = msgpack.unpackb(infile.read(), object_hook=_unserialize_encode_helper, raw=False, strict_map_key=False)
|
||||
if serialdict['version'] != version:
|
||||
import warnings
|
||||
warnings.warn(f"Version mismatch while loading chain expect: [{version}] got: [{serialdict['version']}]")
|
||||
chain = Chain(serialdict['order'], serialdict['analyzer_class'])
|
||||
chain.data = {cast(Tuple[Any], tuple(x)): y for x, y in serialdict['data']}
|
||||
return chain
|
||||
|
||||
|
||||
def dump_chainer(chain: Chain, outfile: BinaryIO):
|
||||
"""
|
||||
Serialize a chainer to an open IO stream
|
||||
|
||||
Arguments:
|
||||
chain: A Chain object
|
||||
outfile: An open IO stream in binary mode that will be writen to
|
||||
"""
|
||||
serialdict: Dict[str, Any] = {}
|
||||
serialdict['version'] = version
|
||||
serialdict['order'] = chain.order
|
||||
serialdict['analyzer_class'] = chain.analyzer_class
|
||||
serialdict['data'] = [(k, v) for k, v in chain.items()]
|
||||
outfile.write(msgpack.packb(serialdict, use_bin_type=True, default=_serialize_encode_helper))
|
|
@ -0,0 +1,129 @@
|
|||
#
|
||||
# carkov markov chain library
|
||||
# © Copyright 2021 by Aldercone Studio <alderconestudio@gmail.com>
|
||||
# This is free software, see the included LICENSE for terms and conditions.
|
||||
#
|
||||
|
||||
"""
|
||||
Various utilities the chainers and analyzers use.
|
||||
"""
|
||||
|
||||
import string
|
||||
|
||||
from bisect import bisect
|
||||
from typing import Dict, Callable, List, Sequence
|
||||
from random import Random
|
||||
|
||||
|
||||
def merge_dict(into: Dict, outof: Dict, mergefunction: Callable = lambda x, y: x + y) -> Dict:
|
||||
"""
|
||||
Given two dictionries of dictionaries, merge them together by applying the mergefunction to the
|
||||
values of the second level dictionary.
|
||||
|
||||
Arguments:
|
||||
into: The dictionary that is being operated on which gets modified.
|
||||
outof: The dictionary to merge into into.
|
||||
mergefunction: A function applied to every value in the second level dictionries, defaults to
|
||||
adding the two values together.
|
||||
|
||||
Returns:
|
||||
into dictionary after modification.
|
||||
"""
|
||||
for key in outof.keys():
|
||||
if key in into:
|
||||
for innerkey in outof[key].keys():
|
||||
if innerkey in into[key]:
|
||||
into[key][innerkey] = mergefunction(into[key][innerkey], outof[key][innerkey])
|
||||
else:
|
||||
into[key][innerkey] = outof[key][innerkey]
|
||||
else:
|
||||
into[key] = outof[key]
|
||||
return into
|
||||
|
||||
|
||||
def convert_counts(ind: Dict) -> Dict:
|
||||
"""
|
||||
Convert counts produced by analyzers into the statistics counts used by chainers.
|
||||
|
||||
Arguments:
|
||||
ind (dict): The second level dictionary of a counts dictionary
|
||||
|
||||
Returns:
|
||||
dict: A copy of ind with the values updated for chainer use.
|
||||
"""
|
||||
out = {}
|
||||
for k in ind:
|
||||
out[k] = [ind[k], 0]
|
||||
|
||||
return out
|
||||
|
||||
|
||||
def merge_stats(into: Dict, outof: Dict) -> Dict:
|
||||
"""
|
||||
Perform a merge_dict in a way safe for the statistics dictionaries used by chainers.
|
||||
|
||||
Arguments:
|
||||
into: The dictionary to modify
|
||||
outof: The dictionary to merge into into.
|
||||
|
||||
Returns:
|
||||
into (after modification)
|
||||
|
||||
"""
|
||||
def stats_merge_function(i, o):
|
||||
out = [0, 0]
|
||||
out[0] = i[0] + o[0]
|
||||
out[1] = 0
|
||||
return out
|
||||
return merge_dict(into, outof, stats_merge_function)
|
||||
|
||||
|
||||
def weighted_choice(random_state: Random, values: Sequence, weights: Sequence):
|
||||
"""
|
||||
Choose a random value in a weighted manner.
|
||||
|
||||
Arguments:
|
||||
random_state: A random.Random instance
|
||||
values: A list of values to choose from
|
||||
weights: The weights that corrospond to each value
|
||||
|
||||
Returns:
|
||||
The selected value
|
||||
"""
|
||||
total: float = 0
|
||||
cum_weights: List[float] = []
|
||||
for w in weights:
|
||||
total += w
|
||||
cum_weights.append(total)
|
||||
x = random_state.random() * total
|
||||
i = bisect(cum_weights, x)
|
||||
return values[i]
|
||||
|
||||
|
||||
def weighted_stat_choice(random_state: Random, stats: Dict):
|
||||
"""
|
||||
Perform a weighted choice on a stat dictionary as used in chainers.
|
||||
|
||||
Arguments:
|
||||
random_state: A random.Random instance
|
||||
stats: A stats dictionary from a chainer
|
||||
"""
|
||||
values = tuple(stats.keys())
|
||||
weights = tuple(stats[x][1] for x in values)
|
||||
return weighted_choice(random_state, values, weights)
|
||||
|
||||
|
||||
def make_word(seq: Sequence[str]):
|
||||
return "".join(seq)
|
||||
|
||||
|
||||
def make_sent(seq: Sequence[str]) -> str:
|
||||
output = ""
|
||||
|
||||
for item in seq:
|
||||
if item in string.punctuation:
|
||||
output += item
|
||||
else:
|
||||
output += (" " + item) if output else (item)
|
||||
|
||||
return output
|
|
@ -0,0 +1,5 @@
|
|||
#!/bin/sh
|
||||
|
||||
rm -rv build/
|
||||
rm -rv carkov.egg-info/
|
||||
rm -rv dist/
|
|
@ -0,0 +1,3 @@
|
|||
[build-system]
|
||||
requires = ["setuptools", "wheel"]
|
||||
build-backend = "setuptools.build_meta"
|
|
@ -0,0 +1,41 @@
|
|||
[metadata]
|
||||
name = carkov
|
||||
version = 0.1.2
|
||||
description = A markov chainer library
|
||||
author = Aldercone Studio
|
||||
author_email = alderconestudio@gmail.com
|
||||
keywords = text, markov, ebooks, chainer, generator, generative
|
||||
long_description = file: README.md
|
||||
long_description_content_type = text/markdown
|
||||
license_file = LICENSE
|
||||
url=https://github.com/aldercone/carkov
|
||||
license = BSD
|
||||
platform = any
|
||||
classifiers =
|
||||
Development Status :: 3 - Alpha
|
||||
Intended Audience :: Developers
|
||||
License :: OSI Approved :: BSD License
|
||||
Operating System :: OS Independent
|
||||
Programming Language :: Python
|
||||
Programming Language :: Python :: 3.6
|
||||
Programming Language :: Python :: 3.7
|
||||
Programming Language :: Python :: 3.8
|
||||
Programming Language :: Python :: 3.9
|
||||
Topic :: Artistic Software
|
||||
Topic :: Text Processing
|
||||
Topic :: Text Processing :: Linguistic
|
||||
|
||||
[options]
|
||||
packages =
|
||||
carkov
|
||||
carkov.analyze
|
||||
|
||||
zip_safe = true
|
||||
install_requires =
|
||||
unidecode
|
||||
nltk
|
||||
msgpack
|
||||
|
||||
[options.entry_points]
|
||||
console_scripts =
|
||||
carkov = carkov.__main__:main
|
|
@ -0,0 +1,17 @@
|
|||
[tox]
|
||||
envlist = py36, py37, py38, py39
|
||||
|
||||
[testenv]
|
||||
deps =
|
||||
flake8
|
||||
mypy
|
||||
commands =
|
||||
flake8
|
||||
mypy carkov typehints
|
||||
|
||||
[flake8]
|
||||
max-line-length = 120
|
||||
max-complexity = 15
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,4 @@
|
|||
from typing import Any, Callable, Dict
|
||||
|
||||
def packb(o: Any, use_bin_type: bool, default: Callable) -> bytes: ...
|
||||
def unpackb(data: bytes, raw: bool, object_hook: Callable) -> Dict: ...
|
|
@ -0,0 +1,4 @@
|
|||
from typing import Iterator
|
||||
|
||||
def sent_tokenize(sent: str) -> Iterator[str]: ...
|
||||
def word_tokenize(word: str) -> Iterator[str]: ...
|
Loading…
Reference in New Issue