unimore/bda-6-steffo
unimore/bda-6-steffo
1
Fork 0
mirror of https://github.com/Steffo99/unimore-bda-6.git synced 2024-11-21 07:24:19 +00:00
Code Activity

Refactor things to work better

Browse source
This commit is contained in:
Steffo 2023-02-02 17:24:11 +01:00
parent 4c3f892038
commit 965cea692a
Signed by: steffo
GPG key ID: 2A24051445686895
12 changed files with 239 additions and 234 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
.editorconfig Normal file
View file

@ -0,0 +1,12 @@
# EditorConfig is awesome: https://EditorConfig.org
# top-most EditorConfig file
root = true
[*]
indent_style = space
indent_size = 4
end_of_line = lf
charset = utf-8
trim_trailing_whitespace = true
insert_final_newline = true

2
.vscode/launch.json vendored
View file

@ -12,7 +12,7 @@
"justMyCode": true,
"env": {
"NLTK_DATA": "./data/nltk",
"DATA_SET_SIZE": "100",
"DATA_SET_SIZE": "250",
},
"cwd": "${workspaceFolder}",
}

98
unimore_bda_6/__main__.py
View file

@ -1,36 +1,98 @@
import logging
from .config import config, DATA_SET_SIZE
from .database import mongo_reviews_collection_from_config, get_reviews_dataset_polar, get_reviews_dataset_uniform
from .analysis.vanilla import VanillaReviewSA, polar_categorizer, stars_categorizer
from .analysis.potts import PottsReviewSA
from .database import Review, mongo_reviews_collection_from_config, dataset_polar, dataset_varied
from .analysis.vanilla import VanillaSA
from .tokenization import all_tokenizers
from .log import install_log_handler
log = logging.getLogger(__name__)
def review_vanilla_extractor(review: Review) -> tuple[str, float]:
"""
Extract review text and rating from a `Review`.
"""
return review["reviewText"], review["overall"]
def polar_categorizer(rating: float) -> str:
"""
Return the polar label corresponding to the given rating.
Possible categories are:
* negative (1.0, 2.0)
* positive (3.0, 4.0, 5.0)
* unknown (everything else)
"""
match rating:
case 1.0 | 2.0:
return "negative"
case 3.0 | 4.0 | 5.0:
return "positive"
case _:
return "unknown"
def varied_categorizer(rating: float) -> str:
"""
Return the "stars" label corresponding to the given rating.
Possible categories are:
* terrible (1.0)
* negative (2.0)
* mixed (3.0)
* positive (4.0)
* great (5.0)
* unknown (everything else)
"""
match rating:
case 1.0:
return "terrible"
case 2.0:
return "negative"
case 3.0:
return "mixed"
case 4.0:
return "positive"
case 5.0:
return "great"
case _:
return "unknown"
def main():
with mongo_reviews_collection_from_config() as reviews:
reviews_polar_training = get_reviews_dataset_polar(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_polar_evaluation = get_reviews_dataset_polar(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_uniform_training = get_reviews_dataset_uniform(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_uniform_evaluation = get_reviews_dataset_uniform(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_polar_training = dataset_polar(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_polar_evaluation = dataset_polar(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
vanilla_polar = VanillaReviewSA(categorizer=polar_categorizer)
vanilla_polar.train(reviews_polar_training)
log.info("Vanilla polar evaluation results: %s", vanilla_polar.evaluate(reviews_polar_evaluation))
for tokenizer in all_tokenizers:
log.info("Training polar model with %s tokenizer", tokenizer)
model = VanillaSA(extractor=review_vanilla_extractor, tokenizer=tokenizer, categorizer=polar_categorizer)
model.train(reviews_polar_training)
log.info("Evaluating polar model with %s tokenizer", tokenizer)
evaluation = model.evaluate(reviews_polar_evaluation)
log.info("Polar model with %s results: %s", tokenizer, evaluation)
potts_polar = PottsReviewSA()
potts_polar.train(reviews_polar_training)
log.info("Potts polar evaluation results: %s", potts_polar.evaluate(reviews_polar_evaluation))
del reviews_polar_training
del reviews_polar_evaluation
vanilla_uniform = VanillaReviewSA(categorizer=stars_categorizer)
vanilla_uniform.train(reviews_uniform_training)
log.info("Vanilla uniform evaluation results: %s", vanilla_polar.evaluate(reviews_polar_evaluation))
with mongo_reviews_collection_from_config() as reviews:
reviews_varied_training = dataset_varied(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
reviews_varied_evaluation = dataset_varied(collection=reviews, amount=DATA_SET_SIZE.__wrapped__)
while True:
print(vanilla_uniform.use(input("> ")))
for tokenizer in all_tokenizers:
log.info("Training varied model with %s tokenizer", tokenizer)
model = VanillaSA(extractor=review_vanilla_extractor, tokenizer=tokenizer, categorizer=varied_categorizer)
model.train(reviews_varied_training)
log.info("Evaluating varied model with %s tokenizer", tokenizer)
evaluation = model.evaluate(reviews_varied_evaluation)
log.info("Varied model with %s results: %s", tokenizer, evaluation)
del reviews_varied_training
del reviews_varied_evaluation
if __name__ == "__main__":

6
unimore_bda_6/analysis/__init__.py
View file

@ -0,0 +1,6 @@
from .vanilla import BaseSA
__all__ = (
"BaseSA",
)

26
unimore_bda_6/analysis/base.py
View file

@ -1,4 +1,9 @@
import abc
import typing as t
Input = t.TypeVar("Input")
Category = t.TypeVar("Category")
class BaseSA(metaclass=abc.ABCMeta):
@ -6,30 +11,15 @@ class BaseSA(metaclass=abc.ABCMeta):
Abstract base class for sentiment analyzers implemented in this project.
"""
def __init__(self) -> None:
"""
Create the empty shell of the sentiment analyzer.
"""
self.trained = False
"If :meth:`train` has been called at least once, and the analyzer is ready or not to be evaluated or used."
@abc.abstractmethod
def train(self, training_set) -> None:
def train(self, training_set: list[tuple[Input, Category]]) -> None:
"""
Train the analyzer with the given training set.
"""
raise NotImplementedError()
@abc.abstractmethod
def evaluate(self, test_set) -> None:
"""
Evaluate the analyzer with the given test set.
"""
raise NotImplementedError()
@abc.abstractmethod
def use(self, text: str) -> str:
def use(self, text: Input) -> Category:
"""
Use the sentiment analyzer.
"""
@ -48,6 +38,8 @@ class NotTrainedError(Exception):
__all__ = (
"Input",
"Category",
"BaseSA",
"AlreadyTrainedError",
"NotTrainedError",

30
unimore_bda_6/analysis/potts.py
View file

@ -1,30 +0,0 @@
from ..vendor.potts import Tokenizer
from .vanilla import VanillaSA, VanillaReviewSA
class PottsSA(VanillaSA):
"""
A sentiment analyzer using Potts' tokenizer.
"""
def __init__(self) -> None:
super().__init__()
def _tokenize_text(self, text: str) -> list[str]:
"""
Convert a text string into a list of tokens, using the language of the model.
"""
tokenizer: Tokenizer = Tokenizer(preserve_case=False)
return list(tokenizer.tokenize(text))
class PottsReviewSA(VanillaReviewSA, PottsSA):
"""
A `PottsSA` to be used with `Review`s.
"""
__all__ = (
"PottsSA",
"PottsReviewSA",
)

140
unimore_bda_6/analysis/vanilla.py
View file

@ -1,4 +1,3 @@
import abc
import nltk
import nltk.classify
import nltk.sentiment
@ -6,31 +5,30 @@ import nltk.sentiment.util
import logging
import typing as t
from ..database import Review
from .base import BaseSA, AlreadyTrainedError, NotTrainedError
from .base import Input, Category, BaseSA, AlreadyTrainedError, NotTrainedError
TokenBag = list[str]
IntermediateValue = t.TypeVar("IntermediateValue")
log = logging.getLogger(__name__)
class VanillaSA(BaseSA, metaclass=abc.ABCMeta):
class VanillaSA(BaseSA):
"""
A sentiment analyzer resembling the one implemented in structure the one implemented in the classroom, using the basic sentiment analyzer of NLTK.
"""
def __init__(self) -> None:
def __init__(self, *, extractor: t.Callable[[Input], tuple[str, Category]], tokenizer: t.Callable[[str], TokenBag], categorizer: t.Callable[[Input], Category]) -> None:
super().__init__()
self.model: nltk.sentiment.SentimentAnalyzer = nltk.sentiment.SentimentAnalyzer()
self.trained: bool = False
def _tokenize_text(self, text: str) -> list[str]:
"""
Convert a text string into a list of tokens.
"""
tokens = nltk.word_tokenize(text)
nltk.sentiment.util.mark_negation(tokens, shallow=True)
return tokens
self.extractor: t.Callable[[Input], tuple[str, IntermediateValue]] = extractor
self.tokenizer: t.Callable[[str], TokenBag] = tokenizer
self.categorizer: t.Callable[[IntermediateValue], Category] = categorizer
def __add_feature_unigrams(self, training_set: list[tuple[list[str], str]]) -> None:
def __add_feature_unigrams(self, training_set: list[tuple[TokenBag, Category]]) -> None:
"""
Add the `nltk.sentiment.util.extract_unigram_feats` feature to the model.
"""
@ -38,116 +36,64 @@ class VanillaSA(BaseSA, metaclass=abc.ABCMeta):
unigrams = self.model.unigram_word_feats(words=all_words, min_freq=4)
self.model.add_feat_extractor(nltk.sentiment.util.extract_unigram_feats, unigrams=unigrams)
def _featurize_documents(self, documents: list[tuple[list[str], str]]):
def _add_features(self, training_set: list[tuple[TokenBag, Category]]):
"""
Apply features to a document.
Add new features to the sentiment analyzer.
"""
return self.model.apply_features(documents, labeled=True)
self.__add_feature_unigrams(training_set)
def _train_with_set(self, training_set: list[tuple[list[str], str]]) -> None:
def _train_from_dataset(self, dataset: list[tuple[TokenBag, Category]]) -> None:
"""
Train the model with the given **pre-classified but not pre-tokenized** training set.
Train the model with the given training set.
"""
if self.trained:
raise AlreadyTrainedError()
self.__add_feature_unigrams(training_set)
training_set_with_features = self._featurize_documents(training_set)
self.__add_feature_unigrams(dataset)
training_set_with_features = self.model.apply_features(dataset, labeled=True)
self.model.train(trainer=nltk.classify.NaiveBayesClassifier.train, training_set=training_set_with_features)
self.trained = True
def _evaluate_with_set(self, test_set: list[tuple[list[str], str]]) -> dict:
def _evaluate_from_dataset(self, dataset: list[tuple[TokenBag, Category]]) -> dict:
"""
Perform a model evaluation with the given test set.
"""
if not self.trained:
raise NotTrainedError()
test_set_with_features = self._featurize_documents(test_set)
test_set_with_features = self.model.apply_features(dataset, labeled=True)
return self.model.evaluate(test_set_with_features)
def _use_with_tokens(self, tokens: list[str]) -> str:
def _use_from_tokenbag(self, tokens: TokenBag) -> Category:
"""
Categorize the given token bag.
"""
if not self.trained:
raise NotTrainedError()
return self.model.classify(instance=tokens)
def _extract_data(self, inp: Input) -> tuple[TokenBag, Category]:
text, value = self.extractor(inp)
return self.tokenizer(text), self.categorizer(value)
class VanillaReviewSA(VanillaSA):
"""
A `VanillaSA` to be used with `Review`s.
"""
def _extract_dataset(self, inp: list[Input]) -> list[tuple[TokenBag, Category]]:
return list(map(self._extract_data, inp))
def __init__(self, categorizer: t.Callable[[Review], str]) -> None:
super().__init__()
self.categorizer: t.Callable[[Review], str] = categorizer
def train(self, training_set: list[Input]) -> None:
dataset = self._extract_dataset(training_set)
self._train_from_dataset(dataset)
def _review_to_data_set(self, review: Review) -> tuple[list[str], str]:
"""
Convert a review to a NLTK-compatible dataset.
"""
return self._tokenize_text(text=review["reviewText"]), self.categorizer(rating=review["overall"])
def train(self, reviews: t.Iterable[Review]) -> None:
data_set = list(map(self._review_to_data_set, reviews))
self._train_with_set(data_set)
def evaluate(self, test_set: list[tuple[Input, Category]]) -> None:
dataset = self._extract_dataset(test_set)
return self._evaluate_from_dataset(dataset)
def evaluate(self, reviews: t.Iterable[Review]):
data_set = list(map(self._review_to_data_set, reviews))
return self._evaluate_with_set(data_set)
def use(self, text: str) -> str:
return self._use_with_tokens(self._tokenize_text(text))
def polar_categorizer(rating: float) -> str:
"""
Return the polar label corresponding to the given rating.
Possible categories are:
* negative (1.0, 2.0)
* positive (3.0, 4.0, 5.0)
* unknown (everything else)
"""
match rating:
case 1.0 | 2.0:
return "negative"
case 3.0 | 4.0 | 5.0:
return "positive"
case _:
return "unknown"
def stars_categorizer(rating: float) -> str:
"""
Return the "stars" label corresponding to the given rating.
Possible categories are:
* terrible (1.0)
* negative (2.0)
* mixed (3.0)
* positive (4.0)
* great (5.0)
* unknown (everything else)
"""
match rating:
case 1.0:
return "terrible"
case 2.0:
return "negative"
case 3.0:
return "mixed"
case 4.0:
return "positive"
case 5.0:
return "great"
case _:
return "unknown"
def use(self, text: Input) -> Category:
tokens = self.tokenizer(text)
return self._use_from_tokenbag(tokens)
__all__ = (
"VanillaSA",
"VanillaReviewSA",
"polar_categorizer",
"stars_categorizer",
)

22
unimore_bda_6/database.py
View file

@ -80,11 +80,11 @@ def sample_reviews_by_rating(reviews: pymongo.collection.Collection, rating: flo
])
def get_reviews_dataset_polar(collection: pymongo.collection.Collection, amount: int) -> list[Review]:
def dataset_polar(collection: pymongo.collection.Collection, amount: int) -> list[Review]:
"""
Get a list of shuffled 1-star and 5-star reviews.
Get a list of the same amount of 1-star and 5-star reviews.
"""
log.info("Building dataset with %d polar reviews...", amount * 2)
log.info("Building polar dataset with %d reviews...", amount * 2)
# Sample the required reviews
positive = sample_reviews_by_rating(collection, rating=5.0, amount=amount)
@ -93,18 +93,14 @@ def get_reviews_dataset_polar(collection: pymongo.collection.Collection, amount:
# Randomness here does not matter, so just merge the lists
both = [*positive, *negative]
# Shuffle the dataset, just in case it affects the performance
# TODO: does it actually?
random.shuffle(both)
return both
def get_reviews_dataset_uniform(collection: pymongo.collection.Collection, amount: int) -> list[Review]:
def dataset_varied(collection: pymongo.collection.Collection, amount: int) -> list[Review]:
"""
Get a list of shuffled reviews of any rating.
Get a list of the same amount of reviews for each rating.
"""
log.info("Building dataset with %d uniform reviews...", amount * 5)
log.info("Building varied dataset with %d reviews...", amount * 5)
# Sample the required reviews
terrible = sample_reviews_by_rating(collection, rating=1.0, amount=amount)
@ -116,10 +112,6 @@ def get_reviews_dataset_uniform(collection: pymongo.collection.Collection, amoun
# Randomness here does not matter, so just merge the lists
full = [*terrible, *negative, *mixed, *positive, *great]
# Shuffle the dataset, just in case it affects the performance
# TODO: does it actually?
random.shuffle(full)
return full
@ -129,5 +121,5 @@ __all__ = (
"mongo_reviews_collection_from_config",
"sample_reviews",
"sample_reviews_by_rating",
"get_reviews_dataset_polar",
"dataset_polar",
)

15
unimore_bda_6/tokenization/__init__.py Normal file
View file

@ -0,0 +1,15 @@
from . import nltk_based
from . import potts_based
all_tokenizers = [
nltk_based.tokenizer,
potts_based.tokenizer,
]
__all__ = (
"nltk_based",
"potts_based",
"all_tokenizers",
)

16
unimore_bda_6/tokenization/nltk_based.py Normal file
View file

@ -0,0 +1,16 @@
import nltk
import nltk.sentiment.util
def tokenizer(text: str) -> list[str]:
"""
Convert a text string into a list of tokens.
"""
tokens = nltk.word_tokenize(text)
nltk.sentiment.util.mark_negation(tokens, shallow=True)
return tokens
__all__ = (
"tokenizer",
)

97
unimore_bda_6/vendor/potts.py → unimore_bda_6/tokenization/potts_based.py
View file

@ -74,7 +74,7 @@ emoticon_string = r"""
[<>]?
[:;=8] # eyes
[\-o\*\']? # optional nose
[\)\]\(\[dDpP/\:\}\{@\|\\] # mouth
[\)\]\(\[dDpP/\:\}\{@\|\\] # mouth
|
[\)\]\(\[dDpP/\:\}\{@\|\\] # mouth
[\-o\*\']? # optional nose
@ -90,20 +90,20 @@ regex_strings = (
(?: # (international)
\+?[01]
[\-\s.]*
)?
)?
(?: # (area code)
[\(]?
\d{3}
[\-\s.\)]*
)?
)?
\d{3} # exchange
[\-\s.]*
[\-\s.]*
\d{4} # base
)"""
,
# Emoticons:
emoticon_string
,
,
# HTML tags:
r"""<[^>]+>"""
,
@ -121,7 +121,7 @@ regex_strings = (
|
(?:[\w_]+) # Words without apostrophes or dashes.
|
(?:\.(?:\s*\.){1,}) # Ellipsis dots.
(?:\.(?:\s*\.){1,}) # Ellipsis dots.
|
(?:\S) # Everything else that isn't whitespace.
"""
@ -129,7 +129,7 @@ regex_strings = (
######################################################################
# This is the core tokenizing regex:
word_re = re.compile(r"""(%s)""" % "|".join(regex_strings), re.VERBOSE | re.I | re.UNICODE)
# The emoticon string gets its own regex so that we can preserve case for them as needed:
@ -142,47 +142,50 @@ amp = "&amp;"
######################################################################
class Tokenizer:
def __init__(self, preserve_case=False):
self.preserve_case = preserve_case
def tokenize(self, s: str) -> t.Iterable[str]:
"""
Argument: s -- any string object
Value: a tokenize list of strings; conatenating this list returns the original string if preserve_case=False
"""
# Fix HTML character entitites:
s = self.__html2string(s)
# Tokenize:
words = word_re.findall(s)
# Possible alter the case, but avoid changing emoticons like :D into :d:
if not self.preserve_case:
words = map((lambda x : x if emoticon_re.search(x) else x.lower()), words)
return words
def tokenizer(text: str) -> t.Iterable[str]:
"""
Argument: s -- any string object
Value: a tokenize list of strings; conatenating this list returns the original string if preserve_case=False
"""
# Fix HTML character entitites:
s = __html2string(text)
# Tokenize:
words = word_re.findall(s)
# Possible alter the case, but avoid changing emoticons like :D into :d:
words = map((lambda x : x if emoticon_re.search(x) else x.lower()), words)
# Return the results
return words
def __html2string(self, s: str) -> str:
"""
Internal metod that seeks to replace all the HTML entities in
s with their corresponding unicode characters.
"""
# First the digits:
ents = set(html_entity_digit_re.findall(s))
if len(ents) > 0:
for ent in ents:
entnum = ent[2:-1]
try:
entnum = int(entnum)
s = s.replace(ent, chr(entnum))
except:
pass
# Now the alpha versions:
ents = set(html_entity_alpha_re.findall(s))
ents = filter((lambda x : x != amp), ents)
def __html2string(html: str) -> str:
"""
Internal metod that seeks to replace all the HTML entities in
s with their corresponding unicode characters.
"""
# First the digits:
ents = set(html_entity_digit_re.findall(html))
if len(ents) > 0:
for ent in ents:
entname = ent[1:-1]
try:
s = s.replace(ent, chr(html.entities.name2codepoint[entname]))
entnum = ent[2:-1]
try:
entnum = int(entnum)
html = html.replace(ent, chr(entnum))
except:
pass
s = s.replace(amp, " and ")
return s
pass
# Now the alpha versions:
ents = set(html_entity_alpha_re.findall(html))
ents = filter((lambda x : x != amp), ents)
for ent in ents:
entname = ent[1:-1]
try:
html = html.replace(ent, chr(html.entities.name2codepoint[entname]))
except:
pass
html = html.replace(amp, " and ")
return html
__all__ = (
"tokenizer",
)

9
unimore_bda_6/vendor/__init__.py vendored
View file

@ -1,9 +0,0 @@
"""
This module contains modules downloaded from the Internet and adapted for the project.
Edits to the respective modules are released under the same license as the modules themselves.
Currently:
* the adaptation of :mod:`potts` is released under the CC BY-NC-SA 3.0 license.
"""