File size: 34,557 Bytes
0ed25b0 f422723 0ed25b0 f422723 0ed25b0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 |
import fitz
import random
import logging
import zipfile
import re, string
import unicodedata
import numpy as np
import pandas as pd
import seaborn as sns
from tqdm import tqdm
from scipy import stats
from pathlib import Path
import matplotlib.pyplot as plt
from collections import Counter
import nltk
from nltk.tokenize import word_tokenize, sent_tokenize
import warnings
warnings.simplefilter("ignore", DeprecationWarning)
import pickle
import pyLDAvis
import pyLDAvis.lda_model as lda
from bertopic import BERTopic
from wordcloud import WordCloud
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.decomposition import LatentDirichletAllocation as LDA
nltk.download('stopwords')
nltk.download('punkt')
nltk.download('averaged_perceptron_tagger')
nltk.download('universal_tagset')
tqdm.pandas()
plt.rcParams["font.family"] = "Tahoma"
sns.set_theme(style="whitegrid", font="Tahoma")
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
HOME_DIR = Path("..")
RAW_DATA_DIR = HOME_DIR / "raw"
PROCESSED_DATA_DIR = HOME_DIR / "processed"
GLOVE_EMBEDDINGS_FILE = PROCESSED_DATA_DIR / "glove.6B.100d.txt"
DATA_ANALYSIS_PATH = HOME_DIR / "data_analysis"
FIGURES_DIR = DATA_ANALYSIS_PATH / "plots"
FIGURES_DIR.mkdir(parents=True, exist_ok=True)
PROCESSED_DATA_DIR.mkdir(parents=True, exist_ok=True)
POST_TAGS = ['ADJ','ADP','ADV','CONJ','DET','NOUN','NUM','PRT','PRON','VERB','.','X']
class FileManager:
"""Handles file operations, including zip and unzipping folders and saving text to files."""
@staticmethod
def unzip_data(zip_path, extract_to):
"""
Unzips a ZIP file to a specified directory.
Parameters:
- zip_path (str or Path): Path to the ZIP file.
- extract_to (str or Path): Target directory to extract files to.
Raises:
- FileNotFoundError: If the ZIP file does not exist.
- RuntimeError: If the file is not a valid ZIP archive.
"""
zip_file = Path(zip_path)
extract_to = Path(extract_to)
if not zip_file.exists():
raise FileNotFoundError(f"ZIP file not found: {zip_file}")
target_dir = extract_to / zip_file.stem
if target_dir.exists():
logging.info(f"Directory already exists: {target_dir}")
return
try:
with zipfile.ZipFile(zip_file, 'r') as zip_ref:
zip_ref.extractall(target_dir)
logging.info(f"Extracted {zip_file} to {target_dir}")
except zipfile.BadZipFile as e:
raise RuntimeError(f"Invalid ZIP file: {zip_file}") from e
@staticmethod
def save_text(text, file_path):
"""
Saves text to a file.
Parameters:
- text (str): Text to save.
- file_path (str or Path): Target file path.
Raises:
- IOError: If writing to the file fails.
"""
file_path = Path(file_path)
file_path.parent.mkdir(parents=True, exist_ok=True)
try:
with open(file_path, 'w', encoding='utf-8') as file:
file.write(text)
logging.info(f"Saved text to {file_path}")
except IOError as e:
logging.error(f"Failed to save text to {file_path}: {e}")
raise
class PDFExtractor:
"""Extracts and cleans text from PDF documents."""
@staticmethod
def extract_text(pdf_path):
"""
Extracts and processes text from a PDF file.
Parameters:
- pdf_path (str or Path): Path to the PDF file.
Returns:
- str: Cleaned and processed text.
Raises:
- FileNotFoundError: If the PDF file does not exist.
- RuntimeError: If the PDF cannot be opened.
"""
pdf_path = Path(pdf_path)
if not pdf_path.exists():
logging.error(f"PDF file not found: {pdf_path}")
raise FileNotFoundError(f"PDF file not found: {pdf_path}")
try:
doc = fitz.open(pdf_path)
text_lines = [
PDFExtractor._clean_line(page.get_text("text"))
for page in doc
]
doc.close()
return '\n'.join(PDFExtractor._combine_paragraphs(text_lines))
except Exception as e:
logging.error(f"Error extracting text from {pdf_path}: {e}")
raise RuntimeError(f"Error extracting text from {pdf_path}: {e}")
@staticmethod
def _clean_line(text):
"""
Cleans a line of text by removing unwanted content.
Parameters:
- text (str): The text to clean.
Returns:
- list: List of cleaned sentences.
"""
paragraphs = [line.strip() for line in sent_tokenize(text)]
return [p for p in paragraphs if not PDFExtractor._is_numeric_string(p)]
@staticmethod
def _combine_paragraphs(lines):
"""
Combines lines into paragraphs based on paragraph markers.
Parameters:
- lines (list of str): List of text lines.
Returns:
- list: Combined paragraphs.
"""
combined = []
for line in lines:
if PDFExtractor._is_paragraph_marker(line):
if combined:
combined[-1] += f' {line}'
else:
combined.append(line)
else:
combined.append(line)
return combined
@staticmethod
def _is_numeric_string(string):
"""
Checks if a string is numeric and less than 1000.
Parameters:
- string (str): The string to check.
Returns:
- bool: True if numeric and less than 1000, otherwise False.
"""
return string.isdigit() and int(string) < 1000
@staticmethod
def _is_paragraph_marker(line):
"""
Determines if a line is a paragraph marker.
Parameters:
- line (str): The line to check.
Returns:
- bool: True if it matches paragraph marker criteria, otherwise False.
"""
return line.startswith("[") and line.endswith("]") and line[1:-1].isdigit()
class DataLoader:
"""Loads and processes TSV data files into DataFrames."""
def __init__(self, base_dir=PROCESSED_DATA_DIR, file_extension="tsv"):
"""
Initialize the DataLoader.
Parameters:
- base_dir (Path): Base directory containing the processed data.
- file_extension (str): Extension of data files to read (default: 'tsv').
"""
self.base_dir = Path(base_dir)
self.file_extension = file_extension
def load_data(self, data_type, column_name=None):
"""
Load data based on the specified type.
Parameters:
- data_type (str): One of ['with_summaries', 'without_summaries', 'all'].
Returns:
- pd.DataFrame: Concatenated DataFrame with a 'split' column.
"""
paths = {
'with_summaries': [self.base_dir / "with_summaries" / f"{split}.{self.file_extension}" for split in ['train', 'dev', 'test']],
'without_summaries': [self.base_dir / "without_summaries" / f"all_data.{self.file_extension}"],
'all': [self.base_dir / "without_summaries" / f"all_data.{self.file_extension}"] +
[self.base_dir / "with_summaries" / f"{split}.{self.file_extension}" for split in ['train', 'dev', 'test']]
}
if data_type not in paths:
raise ValueError(f"Invalid data type specified: {data_type}. Expected one of {list(paths.keys())}.")
valid_paths = [path for path in paths[data_type] if path.exists()]
missing_paths = [path for path in paths[data_type] if not path.exists()]
if missing_paths:
logging.warning(f"Missing files: {missing_paths}")
if not valid_paths:
raise FileNotFoundError("No valid data files found to load.")
if column_name:
return self._read_files(valid_paths)[column_name]
return self._read_files(valid_paths)
@staticmethod
def _read_files(paths):
"""
Read and concatenate data files into a single DataFrame.
Parameters:
- paths (list of Path): Paths to the files to read.
Returns:
- pd.DataFrame: Combined DataFrame with a 'split' column.
"""
df_list = []
for path in paths:
logging.info(f"Loading file: {path}")
try:
df = pd.read_csv(path, sep='\t')
df['split'] = path.stem
df_list.append(df)
except Exception as e:
logging.error(f"Failed to read {path}: {e}")
return pd.concat(df_list, ignore_index=True) if df_list else pd.DataFrame()
class GloveVectorizer:
"""
Maps words to GloVe embeddings and computes sentence embeddings
by averaging word vectors.
"""
def __init__(self, embedding_file):
"""
Initializes the vectorizer with GloVe embeddings.
Args:
embedding_file (str): Path to the GloVe embedding file.
"""
self.word2vec = {}
self.embedding = []
self.idx2word = []
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(message)s")
try:
logging.info("Loading word vectors...")
with open(embedding_file, encoding='utf-8') as f:
for line in f:
values = line.split()
word = values[0]
vec = np.asarray(values[1:], dtype='float32')
self.word2vec[word] = vec
self.embedding.append(vec)
self.idx2word.append(word)
self.embedding = np.array(self.embedding)
self.word2idx = {word: idx for idx, word in enumerate(self.idx2word)}
self.V, self.D = self.embedding.shape
logging.info(f"Found {len(self.word2vec)} word vectors.")
except FileNotFoundError:
logging.error(f"Embedding file '{embedding_file}' not found.")
raise FileNotFoundError(f"Embedding file '{embedding_file}' not found.")
except Exception as e:
logging.error(f"Error loading embeddings: {e}")
raise RuntimeError(f"Error loading embeddings: {e}")
def fit(self, data):
"""Placeholder for potential future implementation."""
pass
def get_vocabulary(self):
"""
Returns the vocabulary of the embeddings.
Returns:
list: A list of all words in the GloVe vocabulary.
"""
return list(self.word2vec.keys())
def transform(self, data, return_unknowns=False):
"""
Transforms a list of sentences into mean GloVe embeddings.
Args:
data (list of str): Sentences to transform.
return_unknowns (bool): If True, also return unknown words.
Returns:
np.ndarray: Mean GloVe embeddings for each sentence.
list: (Optional) List of unknown words for each sentence.
"""
X = np.zeros((len(data), self.D))
unknown_words = []
emptycount = 0
for n, sentence in enumerate(data):
tokens = sentence.lower().split()
vecs = []
unknowns = []
for word in tokens:
if word in self.word2vec:
vecs.append(self.word2vec[word])
else:
unknowns.append(word)
if vecs:
vecs = np.array(vecs)
X[n] = vecs.mean(axis=0)
else:
emptycount += 1
if return_unknowns:
unknown_words.append(unknowns)
if emptycount > 0:
print(f"Warning: {emptycount} sentences had no known words.")
return (X, unknown_words) if return_unknowns else X
def fit_transform(self, data, return_unknowns=False):
"""
Fits and transforms the data.
Args:
data (list of str): Sentences to transform.
return_unknowns (bool): If True, also return unknown words.
Returns:
np.ndarray: Mean GloVe embeddings for each sentence.
list: (Optional) List of unknown words for each sentence.
"""
self.fit(data)
return self.transform(data, return_unknowns)
class TextProcessor:
"""Processes text data for analysis and visualization."""
def __init__(self, data_loader):
self.data_loader = data_loader
@staticmethod
def tokenize_stats(df, col_name, tokenize_type):
tokenizer = sent_tokenize if tokenize_type == 'sent' else word_tokenize
stats = df[col_name].dropna().apply(lambda x: len(tokenizer(x)))
return stats
@staticmethod
def get_punctuation():
return string.punctuation
@staticmethod
def get_stopwords(language='english'):
return set(nltk.corpus.stopwords.words(language))
@staticmethod
def unicode_to_ascii(s):
return ''.join(c for c in unicodedata.normalize('NFD', s)
if unicodedata.category(c) != 'Mn')
@staticmethod
def count_stopwords(text, stopwords):
word_tokens = word_tokenize(text)
stopwords_x = [w for w in word_tokens if w in stopwords]
return len(stopwords_x)
@staticmethod
def replace_punctuation(text, punctuation):
table = str.maketrans(punctuation, ' ' * len(punctuation))
return text.translate(table)
@staticmethod
def get_unknown_words(text, vocab):
tokens = word_tokenize(text)
unknown = [t for t in tokens if t not in vocab.word2vec]
return unknown
@staticmethod
def get_pos_tags(sentences, columns, data_type, tagset='universal'):
''' Extract the part-of-speech taggings of the sentence
Input:
- sentence: string, sentence to tag
- tagset: string, tagset or the set of tags to search for
'''
tags = []
columns = [f'{data_type}_{c}' for c in columns]
for sent in tqdm(sentences):
pos_tags = Counter([j for _,j in nltk.pos_tag(word_tokenize(sent), tagset=tagset)])
pos_tags = {f'{data_type}_{k}':v for k,v in dict(pos_tags).items()}
tags.append(pos_tags)
return pd.DataFrame(tags, columns=columns).fillna(0)
def remove_stopwords(self, df, target_columns=None):
''' Apply some basic techniques for cleaning a text for an analysis of words
Input:
- text: text to be cleaned
Output:
- result: cleaned text
'''
def clean_text(text, stopwords):
text = text.lower()
pattern = r'[^a-zA-Z\s]'
text = re.sub(pattern, '', text)
tokens = nltk.word_tokenize(text)
tokens = [token.strip() for token in tokens]
text = ' '.join([token for token in tokens if token not in stopwords])
return text
if target_columns:
logging.info(f"Removing stopwords for columns: {target_columns}")
stopwords = self.get_stopwords()
cleaned_text = []
for col in target_columns:
cleaned_text.append(df[col].progress_apply(lambda x: clean_text(x, stopwords)).tolist())
return cleaned_text
def prepare_text(self, df, target_columns=None, drop_duplicates=True, drop_na=True):
if target_columns and len(target_columns) == 2:
logging.info(f"Preparing text data for columns: {target_columns}")
try:
df = df[target_columns]
except KeyError as e:
logging.error(f"Invalid columns specified: {e}")
raise ValueError(f"Invalid columns specified: {e}")
if drop_duplicates:
df.drop_duplicates(subset=target_columns[0], inplace=True)
logging.info(f"Dropped duplicates, new shape: {df.shape}")
if drop_na:
df.dropna(inplace=True)
logging.info(f"Dropped NA values, new shape: {df.shape}")
df.reset_index(drop=True, inplace=True)
df.columns = ['text', 'summary']
logging.info(f"Renamed columns to 'text' and 'summary'")
logging.info("Cleaning unicode characters and extra spaces...")
df['text'] = df['text'].apply(lambda x: self.unicode_to_ascii(x.strip()))
df['summary'] = df['summary'].apply(lambda x: self.unicode_to_ascii(x.strip()))
logging.info(f"Data prepared, new shape: {df.shape}")
return df
else:
logging.error("Invalid columns or number of target columns specified.")
raise ValueError('No target columns specified, or invalid number of columns.')
def get_vectorizer_features(self, texts, max_df=0.9, min_df=25, max_features=5000):
tf_vectorizer = CountVectorizer(max_df=max_df, min_df=min_df, max_features=max_features)
tf = tf_vectorizer.fit_transform(texts)
tf_feature_names = tf_vectorizer.get_feature_names_out()
return tf, tf_feature_names
def get_all_stats(self, df):
"""
Generate and add statistical metrics for text and summary columns in a DataFrame.
Parameters:
df (pd.DataFrame): Input DataFrame containing 'text' and 'summary' columns.
Returns:
pd.DataFrame: DataFrame with added statistical columns.
"""
punc = self.get_punctuation()
stopwords = self.get_stopwords()
vocab = GloveVectorizer(GLOVE_EMBEDDINGS_FILE)
def add_stat_column(column_name, compute_func, *args, **kwargs):
if column_name not in df.columns:
logging.info(f"Calculating {column_name}...")
df[column_name] = compute_func(*args, **kwargs)
else:
logging.info(f"{column_name} already present in stats, skipping...")
logging.info("Calculating text statistics (sentences, tokens, characters, etc.)...")
add_stat_column('text_sent_count', self.tokenize_stats, df, 'text', 'sent')
add_stat_column('text_word_count', self.tokenize_stats, df, 'text', 'word')
add_stat_column('text_char_count', lambda x: x['text'].progress_apply(lambda t: len(t.replace(" ", ""))), df)
add_stat_column('text_sent_density', lambda x: x['text_sent_count'] / (x['text_word_count'] + 1), df)
add_stat_column('text_word_density', lambda x: x['text_word_count'] / (x['text_char_count'] + 1), df)
add_stat_column('text_punc_count', lambda x: x['text'].progress_apply(lambda t: sum(1 for char in t if char in punc)), df)
add_stat_column('text_stopw_count', lambda x: x['text'].progress_apply(lambda t: self.count_stopwords(t, stopwords)), df)
add_stat_column('text_unknown_words', lambda x: x['text'].progress_apply(lambda t: self.get_unknown_words(self.replace_punctuation(t.lower(), string.punctuation), vocab)), df)
add_stat_column('text_unknown_count', lambda x: x['text_unknown_words'].progress_apply(lambda t: len(t) if isinstance(t, list) else 0), df)
logging.info("Calculating summary statistics (sentences, tokens, characters, etc.)...")
add_stat_column('sum_sent_count', self.tokenize_stats, df, 'summary', 'sent')
add_stat_column('sum_word_count', self.tokenize_stats, df, 'summary', 'word')
add_stat_column('sum_char_count', lambda x: x['summary'].progress_apply(lambda t: len(t.replace(" ", ""))), df)
add_stat_column('sum_sent_density', lambda x: x['sum_sent_count'] / (x['sum_word_count'] + 1), df)
add_stat_column('sum_word_density', lambda x: x['sum_word_count'] / (x['sum_char_count'] + 1), df)
add_stat_column('sum_punc_count', lambda x: x['summary'].progress_apply(lambda t: sum(1 for char in t if char in punc)), df)
add_stat_column('sum_stopw_count', lambda x: x['summary'].progress_apply(lambda t: self.count_stopwords(t, stopwords)), df)
add_stat_column('sum_unknown_words', lambda x: x['summary'].progress_apply(lambda t: self.get_unknown_words(self.replace_punctuation(t.lower(), string.punctuation), vocab)), df)
add_stat_column('sum_unknown_count', lambda x: x['sum_unknown_words'].progress_apply(lambda t: len(t) if isinstance(t, list) else 0), df)
logging.info("Adding POS tags for text and summary...")
text_columns = [f'text_{p}' for p in POST_TAGS]
if not all(col in df.columns for col in text_columns):
df = pd.concat([df, self.get_pos_tags(df['text'], POST_TAGS, 'text')], axis=1)
else:
logging.info("Text POS tags already present in stats, skipping...")
sum_columns = [f'sum_{p}' for p in POST_TAGS]
if not all(col in df.columns for col in sum_columns):
df = pd.concat([df, self.get_pos_tags(df['summary'], POST_TAGS, 'sum')], axis=1)
else:
logging.info("Summary POS tags already present in stats, skipping...")
logging.info("All statistics have been calculated successfully.")
return df
class SCAPlotter:
"""Generates plots for data visualization."""
def __init__(self):
self.labels_dict = {
'sum_word_count': 'Word Count of Summaries', 'text_word_count': 'Word Count of Judgments',
'sum_char_count': 'Chararacter Count of Summaries', 'text_char_count': 'Chararacter Count of Judgments',
'sum_word_density': 'Word Density of Summaries', 'text_word_density': 'Word Density of Judgments',
'sum_punc_count': 'Punctuation Count of Summaries', 'text_punc_count': 'Punctuation Count of Judgments',
'text_sent_count': 'Sentence Count of Judgments', 'sum_sent_count': 'Sentence Count of Summaries',
'text_sent_density': 'Sentence Density of Judgments', 'sum_sent_density': 'Sentence Density of Summaries',
'text_stopw_count': 'Stopwords Count of Judgments', 'sum_stopw_count': 'Stopwords Count of Summaries',
'ADJ': 'adjective','ADP': 'adposition', 'ADV': 'adverb','CONJ': 'conjunction',
'DET': 'determiner','NOUN': 'noun', 'text_unknown_count': 'Unknown words in Judgments',
'sum_unknown_count': 'Unknown words in Summaries'
}
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(message)s")
def plot_case_distribution(self, df):
plt.figure(figsize=(7.5, 6))
sns.countplot(data=df, x='type', hue='type', palette='muted', width=0.5)
plt.ylabel('Number of Cases')
plt.xlabel('Case Type')
plt.xticks(rotation=0)
plt.savefig(FIGURES_DIR / 'number_of_cases_by_type.png')
plt.close()
def plot_summary_vs_judgment_length(self, df):
slope, intercept, _, _, _ = stats.linregress(df['text_word_count'], df['sum_word_count'])
plt.figure(figsize=(7.5, 6))
sns.scatterplot(x='text_word_count', y='sum_word_count', data=df, s=10, label='Data', color="dodgerblue")
plt.xlabel('Judgment Length')
plt.ylabel('Summary Length')
plt.plot(df['text_word_count'], intercept + slope * df['text_word_count'], 'b', label=f'Best Fit: y = {slope:.2f}x + {intercept:.2f}')
self._add_capacity_shading(df['text_word_count'], df['sum_word_count'])
plt.legend()
plt.savefig(FIGURES_DIR / 'data_summary_lengths.png')
plt.close()
def plot_length_distribution(self, df, columns, plot_histogram=True, plot_boxplots=True, file_name='stats'):
if plot_histogram or plot_boxplots:
if plot_histogram:
self._plot_histograms(
df,
np.array([columns]),
self.labels_dict,
show_kde=False,
output_path=FIGURES_DIR / f'{file_name}_histograms.png'
)
if plot_boxplots:
self._plot_boxplots(
df,
np.array([columns]),
self.labels_dict,
output_path=FIGURES_DIR / f'{file_name}_boxplots.png'
)
else:
raise ValueError('No plots selected to be generated.')
def plot_most_common_words(self, count_data, words, figsize=(15, 7), no_words=20, file_name=None, show_plot=False):
"""
Draw a barplot showing the most common words in the data.
Parameters:
- count_data (sparse matrix): Document-term matrix containing word occurrences.
- count_vectorizer (CountVectorizer): Fitted CountVectorizer object.
- figsize (tuple): Figure size for the plot.
- no_words (int): Number of most common words to display.
- output_path (str): Path to save the plot.
"""
total_counts = np.zeros(len(words))
for t in count_data:
total_counts += t.toarray()[0]
count_dict = sorted(zip(words, total_counts), key=lambda x: x[1], reverse=True)[:no_words]
words = [w[0] for w in count_dict]
counts = [w[1] for w in count_dict]
x_pos = np.arange(len(words))
plt.figure(figsize=figsize)
plt.subplot(title=f'{no_words} most common words')
sns.set_context("notebook", font_scale=1.25, rc={"lines.linewidth": 2.5})
sns.barplot(x=x_pos, y=counts, palette='husl')
plt.xticks(x_pos, words, rotation=45)
plt.ylabel('Frequency')
plt.tight_layout()
if file_name:
plt.savefig(FIGURES_DIR / f'{file_name}.png')
if show_plot:
plt.show()
plt.close()
def plot_bertopic_visualizations(self, model, output_path):
"""
Generate and save BERTopic visualizations.
"""
fig = model.visualize_barchart(top_n_topics=12)
fig.write_html(output_path / "topic_barchart.html")
hierarchical_fig = model.visualize_hierarchy()
hierarchical_fig.write_html(output_path / "topic_hierarchy.html")
heatmap_fig = model.visualize_heatmap()
heatmap_fig.write_html(output_path / "topic_heatmap.html")
word_cloud_fig = model.visualize_topics()
word_cloud_fig.write_html(output_path / "topic_wordcloud.html")
def plot_overlap_heatmap(self, overlap_matrix, file_name=None):
"""
Plot a heatmap for the overlap matrix.
Parameters:
overlap_matrix (np.array): Overlap matrix between judgment and summary topics.
output_path (str): Path to save the heatmap.
"""
plt.figure(figsize=(12, 8))
sns.heatmap(overlap_matrix, annot=False, cmap="coolwarm", cbar=True)
plt.title("Topic Overlap Between Judgments and Summaries")
plt.xlabel("Summary Topics")
plt.ylabel("Judgment Topics")
plt.savefig(FIGURES_DIR / f'{file_name}.png')
plt.close()
def plot_wordcloud(self, texts, background_color="white", max_words=1000, contour_width=3, contour_color='steelblue', file_name='wordcloud'):
long_string = ','.join(texts)
wordcloud = WordCloud(background_color=background_color, max_words=max_words, contour_width=contour_width, contour_color=contour_color)
wordcloud.generate(long_string)
wordcloud.to_image()
wordcloud.to_file(FIGURES_DIR / f'{file_name}.png')
def plot_lda_results(self, lda_model, tf, tf_vectorizer, file_name='lda_topics'):
LDAvis_prepared = lda.prepare(lda_model, tf, tf_vectorizer)
with open(FIGURES_DIR / f'{file_name}.pkl', 'wb') as f:
pickle.dump(LDAvis_prepared, f)
with open(FIGURES_DIR / f'{file_name}.pkl', 'rb') as f:
LDAvis_prepared = pickle.load(f)
pyLDAvis.save_html(LDAvis_prepared, FIGURES_DIR / f'{file_name}.html')
@staticmethod
def _plot_boxplots(data, plot_vars, labels, figsize=(15, 5), output_path=None, show_plot=False):
"""
Plot boxplots for the specified variables with appropriate labels.
Parameters:
- data (pd.DataFrame): The data points to plot.
- plot_vars (array-like): A (1, x) or (n, m) array containing column names to plot.
- labels (dict): A dictionary mapping column names to their respective labels.
- figsize (tuple): The size of the figure (default: (15, 5)).
- output_path (str, optional): File path to save the plot.
- show_plot (bool, optional): Whether to display the plot.
Returns:
- None
"""
plot_vars = np.atleast_2d(plot_vars)
nrows, ncols = plot_vars.shape
fig, axes = plt.subplots(nrows, ncols, figsize=figsize, squeeze=False)
for i in range(nrows):
for j in range(ncols):
var = plot_vars[i, j]
ax = axes[i, j]
if var is not None:
ax.set_title(labels.get(var, var))
ax.grid(True)
ax.tick_params(
axis='x',
which='both',
bottom=False,
top=False,
labelbottom=False
)
if var in data.columns:
ax.boxplot(data[var])
else:
ax.set_visible(False)
else:
ax.set_visible(False)
fig.tight_layout()
if output_path:
plt.savefig(output_path)
if show_plot:
plt.show()
plt.close()
@staticmethod
def _plot_histograms(data, plot_vars, labels, figsize=(15,5), show_kde=False, output_path=None, show_plot=False):
''' Function to plot the histograms of the variables in plot_vars
Input:
- data: a dataframe, containing the data points to plot
- plot_vars: a (1,x) array, containing the columns to plot
- xlim: a list, defines the max x value for every column to plot
- labels: a dictionary, to map the column names to its label
- figsize: a tuple, indicating the size of the figure
- show_kde: a boolean, indicating if the kde should be shown
- output_path: a string, indicating the path to save the file
'''
fig, axes = plt.subplots(1, plot_vars.shape[1], figsize=figsize, sharey=False, dpi=100)
if plot_vars.shape[1] == 1:
axes = [axes]
for i in range(plot_vars.shape[1]):
color = (random.uniform(0, 1), random.uniform(0, 1), random.uniform(0, 1))
sns.histplot(
data[plot_vars[0, i]],
color=color,
ax=axes[i],
bins=50,
kde=show_kde,
)
x_label = plot_vars[0, i].replace('sent', 'sentence')
axes[i].set_xlabel(' '.join([l.capitalize() for l in x_label.split('_')[1:]]))
axes[i].set_ylabel('Frequency')
axes[i].set_title(labels[plot_vars[0, i]])
fig.tight_layout()
if output_path:
plt.savefig(output_path)
if show_plot:
plt.show()
plt.close()
@staticmethod
def _add_capacity_shading(input_stats, output_stats):
model_input_length, model_output_length = 16384, 1024
plt.gca().add_patch(
plt.Rectangle((0, 0), model_input_length, max(output_stats) + 50,
color='red', alpha=0.3, linestyle='--', linewidth=1.5,
label=f"Judgments accommodated: {len([x for x in input_stats if x < model_input_length]):,}")
)
plt.gca().add_patch(
plt.Rectangle((0, 0), max(input_stats) + 400, model_output_length,
color='green', alpha=0.3, linestyle='-', linewidth=1.5,
label=f"Summaries accommodated: {len([y for y in output_stats if y < model_output_length]):,}")
)
class TopicModeling:
"""
Class to perform topic modeling using LDA, UMAP, and HDBSCAN.
"""
def __init__(self):
self.plotter = SCAPlotter()
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(message)s")
def perform_lda_analysis(self, texts, tf_vectorizer, no_top_words=8, n_components=10, max_iter=500, random_state=0, learning_method='online', file_name='lda_topics'):
"""
Perform LDA topic modeling and save top words per topic.
Parameters:
texts (list of str): Input texts for LDA.
tf_vectorizer (TfidfVectorizer or CountVectorizer): Vectorizer for text processing.
no_top_words (int): Number of top words to display per topic.
n_components (int): Number of topics.
max_iter (int): Maximum number of iterations.
random_state (int): Random state for reproducibility.
learning_method (str): Learning method for LDA ('batch' or 'online').
file_name (str): Name of the file to save topics.
Returns:
lda_model (LDA): Fitted LDA model.
"""
logging.info("Vectorizing text data...")
tf = tf_vectorizer.fit_transform(texts)
logging.info("Fitting LDA model...")
lda_model = LDA(
n_components=n_components,
learning_method=learning_method,
max_iter=max_iter,
random_state=random_state
).fit(tf)
words = tf_vectorizer.get_feature_names_out()
with open(FIGURES_DIR / f'{file_name}.txt', 'w') as f:
for topic_idx, topic in enumerate(lda_model.components_):
f.write(f"\nTopic #{topic_idx}:\n")
f.write(" ".join([words[i] for i in topic.argsort()[:-no_top_words - 1:-1]]) + "\n")
self.plotter.plot_lda_results(lda_model, tf, tf_vectorizer, file_name)
return lda_model
def perform_bertopic_analysis(self, cleaned_text=None, cleaned_summary=None, output_path='bertopic', save_topic_info=True):
"""
Perform BERTopic modeling and generate plots.
Parameters:
cleaned_text (list of str): List of cleaned text strings.
cleaned_summary (list of str): List of cleaned summary strings.
output_path (str): Directory path to save results.
save_topic_info (bool): Save topic information as a CSV file.
Returns:
model (BERTopic): Trained BERTopic model.
topic_info (pd.DataFrame): DataFrame containing topic information.
"""
if cleaned_text is None and cleaned_summary is None:
logging.error("No cleaned text or summary data provided.")
raise ValueError("Please provide cleaned text and/or summary data.")
if cleaned_text and cleaned_summary:
logging.info('merging text and summary data...')
elif cleaned_text:
logging.info('using only text data...')
elif cleaned_summary:
logging.info('using only summary data...')
combined_texts = cleaned_text or [] + cleaned_summary or []
logging.info("Initializing and fitting BERTopic model...")
model = BERTopic()
model.fit_transform(combined_texts)
topic_info = None
topic_info_path = FIGURES_DIR / output_path
topic_info_path.mkdir(parents=True, exist_ok=True)
if save_topic_info:
logging.info("Saving topic information to CSV file...")
topic_info = model.get_topic_info()
topic_info.to_csv(topic_info_path / "topic_info.csv", index=False)
logging.info("Generating BERTopic visualizations...")
self.plotter.plot_bertopic_visualizations(model, topic_info_path)
return model, topic_info
def calculate_overlap_matrix(self, judgment_model, summary_model, top_n=12):
"""
Calculate the overlap matrix between judgment and summary topics.
Args:
judgment_model: The model containing judgment topics.
summary_model: The model containing summary topics.
top_n (int): The number of top topics to consider.
Returns:
np.ndarray: Overlap matrix between judgment and summary topics.
"""
logging.info("Getting topic information from judgment and summary models.")
# Get topic information
judgment_topics = judgment_model.get_topic_info()['Topic'][:top_n].values
summary_topics = summary_model.get_topic_info()['Topic'][:top_n].values
logging.info("Initializing overlap matrix.")
# Initialize overlap matrix
overlap_matrix = np.zeros((top_n, top_n))
for i, j_topic_id in enumerate(judgment_topics):
if j_topic_id == -1: # Skip outliers
logging.info(f"Skipping outlier topic in judgment model at index {i}.")
continue
logging.info(f"Processing judgment topic {j_topic_id} at index {i}.")
j_terms = {term for term, _ in judgment_model.get_topic(j_topic_id)}
for j, s_topic_id in enumerate(summary_topics):
if s_topic_id == -1: # Skip outliers
logging.info(f"Skipping outlier topic in summary model at index {j}.")
continue
logging.info(f"Processing summary topic {s_topic_id} at index {j}.")
s_terms = {term for term, _ in summary_model.get_topic(s_topic_id)}
# Calculate Jaccard similarity
overlap_matrix[i, j] = len(j_terms & s_terms) / len(j_terms | s_terms)
logging.info(f"Calculated Jaccard similarity for judgment topic {j_topic_id} and summary topic {s_topic_id}: {overlap_matrix[i, j]}")
logging.info("Overlap matrix calculation complete.")
return overlap_matrix |