RSVP for your your local TensorFlow Everywhere event today!

如何使用 TF-Hub 解决 Kaggle 上的问题

在 TensorFlow.org 上查看 在 Google Colab 中运行 在 GitHub 中查看源代码 下载笔记本

TF-Hub 是一个平台，用于共享打包为可重用资源的机器学习专业知识，尤其是经过预训练的模块。在本教程中，我们将使用 TF-Hub 文本嵌入向量模块来训练具有合理基线准确率的简单情感分类器。之后，我们会将预测结果提交给 Kaggle。

有关使用 TF-Hub 进行文本分类的更详细教程，以及提高准确率的后续步骤，请查看使用 TF-Hub 进行分本分类。

设置

pip install -q kaggle

import tensorflow as tf
import tensorflow_hub as hub
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
import zipfile

from sklearn import model_selection

由于本教程将使用 Kaggle 中的数据集，因此需要为您的 Kaggle 帐号创建 API 令牌，并将其上传到 Colab 环境。

import os
import pathlib

# Upload the API token.
def get_kaggle():
  try:
    import kaggle
    return kaggle
  except OSError:
    pass

  token_file = pathlib.Path("~/.kaggle/kaggle.json").expanduser()
  token_file.parent.mkdir(exist_ok=True, parents=True)

  try:
    from google.colab import files
  except ImportError:
    raise ValueError("Could not find kaggle token.")

  uploaded = files.upload()
  token_content = uploaded.get('kaggle.json', None)
  if token_content:
    token_file.write_bytes(token_content)
    token_file.chmod(0o600)
  else:
    raise ValueError('Need a file named "kaggle.json"')

  import kaggle
  return kaggle


kaggle = get_kaggle()

开始

数据

我们将尝试完成 Kaggle 的 Sentiment Analysis on Movie Reviews 任务。数据集由 Rotten Tomatoes 电影评论中符合句法的子短语组成。任务是采用五分制将短语标记为负面或正面。

在使用此 API 下载数据之前，您必须接受竞赛规则。

SENTIMENT_LABELS = [
    "negative", "somewhat negative", "neutral", "somewhat positive", "positive"
]

# Add a column with readable values representing the sentiment.
def add_readable_labels_column(df, sentiment_value_column):
  df["SentimentLabel"] = df[sentiment_value_column].replace(
      range(5), SENTIMENT_LABELS)

# Download data from Kaggle and create a DataFrame.
def load_data_from_zip(path):
  with zipfile.ZipFile(path, "r") as zip_ref:
    name = zip_ref.namelist()[0]
    with zip_ref.open(name) as zf:
      return pd.read_csv(zf, sep="\t", index_col=0)


# The data does not come with a validation set so we'll create one from the
# training set.
def get_data(competition, train_file, test_file, validation_set_ratio=0.1):
  data_path = pathlib.Path("data")
  kaggle.api.competition_download_files(competition, data_path)
  competition_path = (data_path/competition)
  competition_path.mkdir(exist_ok=True, parents=True)
  competition_zip_path = competition_path.with_suffix(".zip")

  with zipfile.ZipFile(competition_zip_path, "r") as zip_ref:
    zip_ref.extractall(competition_path)

  train_df = load_data_from_zip(competition_path/train_file)
  test_df = load_data_from_zip(competition_path/test_file)

  # Add a human readable label.
  add_readable_labels_column(train_df, "Sentiment")

  # We split by sentence ids, because we don't want to have phrases belonging
  # to the same sentence in both training and validation set.
  train_indices, validation_indices = model_selection.train_test_split(
      np.unique(train_df["SentenceId"]),
      test_size=validation_set_ratio,
      random_state=0)

  validation_df = train_df[train_df["SentenceId"].isin(validation_indices)]
  train_df = train_df[train_df["SentenceId"].isin(train_indices)]
  print("Split the training data into %d training and %d validation examples." %
        (len(train_df), len(validation_df)))

  return train_df, validation_df, test_df


train_df, validation_df, test_df = get_data(
    "sentiment-analysis-on-movie-reviews",
    "train.tsv.zip", "test.tsv.zip")

Split the training data into 140315 training and 15745 validation examples.

注：本竞赛的任务不是对整个评论进行评分，而是对评论中的各个短语进行评分。这是一项更加艰巨的任务。

train_df.head(20)

训练模型

注：我们也可以将此任务建模为回归模型，请参阅使用 TF-Hub 进行文本分类。

class MyModel(tf.keras.Model):
  def __init__(self, hub_url):
    super().__init__()
    self.hub_url = hub_url
    self.embed = hub.load(self.hub_url).signatures['default']
    self.sequential = tf.keras.Sequential([
      tf.keras.layers.Dense(500),
      tf.keras.layers.Dense(100),
      tf.keras.layers.Dense(5),
    ])

  def call(self, inputs):
    phrases = inputs['Phrase'][:,0]
    embedding = 5*self.embed(phrases)['default']
    return self.sequential(embedding)

  def get_config(self):
    return {"hub_url":self.hub_url}

model = MyModel("https://tfhub.dev/google/nnlm-en-dim128/1")
model.compile(
    loss = tf.losses.SparseCategoricalCrossentropy(from_logits=True),
    optimizer=tf.optimizers.Adam(), 
    metrics = [tf.keras.metrics.SparseCategoricalAccuracy(name="accuracy")])

history = model.fit(x=dict(train_df), y=train_df['Sentiment'],
          validation_data=(dict(validation_df), validation_df['Sentiment']),
          epochs = 25)

Epoch 1/25
4385/4385 [==============================] - 16s 4ms/step - loss: 1.0250 - accuracy: 0.5853 - val_loss: 1.0063 - val_accuracy: 0.5815
Epoch 2/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9993 - accuracy: 0.5950 - val_loss: 0.9866 - val_accuracy: 0.5983
Epoch 3/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9961 - accuracy: 0.5960 - val_loss: 0.9880 - val_accuracy: 0.5883
Epoch 4/25
4385/4385 [==============================] - 17s 4ms/step - loss: 0.9927 - accuracy: 0.5973 - val_loss: 0.9828 - val_accuracy: 0.5897
Epoch 5/25
4385/4385 [==============================] - 17s 4ms/step - loss: 0.9910 - accuracy: 0.5976 - val_loss: 0.9844 - val_accuracy: 0.5923
Epoch 6/25
4385/4385 [==============================] - 17s 4ms/step - loss: 0.9903 - accuracy: 0.5987 - val_loss: 0.9831 - val_accuracy: 0.5909
Epoch 7/25
4385/4385 [==============================] - 17s 4ms/step - loss: 0.9901 - accuracy: 0.5985 - val_loss: 0.9816 - val_accuracy: 0.5947
Epoch 8/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9891 - accuracy: 0.5981 - val_loss: 0.9779 - val_accuracy: 0.5950
Epoch 9/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9890 - accuracy: 0.5993 - val_loss: 0.9842 - val_accuracy: 0.5976
Epoch 10/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9882 - accuracy: 0.5992 - val_loss: 0.9814 - val_accuracy: 0.6008
Epoch 11/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9882 - accuracy: 0.5991 - val_loss: 0.9884 - val_accuracy: 0.5891
Epoch 12/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9880 - accuracy: 0.5990 - val_loss: 0.9893 - val_accuracy: 0.5978
Epoch 13/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9877 - accuracy: 0.5993 - val_loss: 0.9794 - val_accuracy: 0.5957
Epoch 14/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9876 - accuracy: 0.5996 - val_loss: 0.9785 - val_accuracy: 0.5964
Epoch 15/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9870 - accuracy: 0.5995 - val_loss: 0.9767 - val_accuracy: 0.5940
Epoch 16/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9869 - accuracy: 0.5998 - val_loss: 0.9769 - val_accuracy: 0.5978
Epoch 17/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9866 - accuracy: 0.6005 - val_loss: 0.9823 - val_accuracy: 0.5938
Epoch 18/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9867 - accuracy: 0.5996 - val_loss: 0.9747 - val_accuracy: 0.6002
Epoch 19/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9865 - accuracy: 0.5997 - val_loss: 0.9762 - val_accuracy: 0.6018
Epoch 20/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9865 - accuracy: 0.5995 - val_loss: 0.9775 - val_accuracy: 0.5961
Epoch 21/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9862 - accuracy: 0.5997 - val_loss: 0.9789 - val_accuracy: 0.5940
Epoch 22/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9864 - accuracy: 0.5997 - val_loss: 0.9776 - val_accuracy: 0.5943
Epoch 23/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9864 - accuracy: 0.5991 - val_loss: 0.9820 - val_accuracy: 0.5903
Epoch 24/25
4385/4385 [==============================] - 16s 4ms/step - loss: 0.9856 - accuracy: 0.6005 - val_loss: 0.9808 - val_accuracy: 0.5957
Epoch 25/25
4385/4385 [==============================] - 17s 4ms/step - loss: 0.9863 - accuracy: 0.5998 - val_loss: 0.9768 - val_accuracy: 0.5952

预测

为验证集和训练集运行预测。

plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])

[<matplotlib.lines.Line2D at 0x7f4cec67c8d0>]

png

train_eval_result = model.evaluate(dict(train_df), train_df['Sentiment'])
validation_eval_result = model.evaluate(dict(validation_df), validation_df['Sentiment'])

print(f"Training set accuracy: {train_eval_result[1]}")
print(f"Validation set accuracy: {validation_eval_result[1]}")

4385/4385 [==============================] - 15s 4ms/step - loss: 0.9849 - accuracy: 0.6005
493/493 [==============================] - 1s 2ms/step - loss: 0.9768 - accuracy: 0.5952
Training set accuracy: 0.6004988551139832
Validation set accuracy: 0.595236599445343

混淆矩阵

另一个非常有趣的统计数据（尤其对于多类问题而言）是混淆矩阵。混淆矩阵允许可视化显示正确和错误标记的样本的比例。我们可以很容易看出分类器出现了多大偏差，以及标签分布是否有意义。理想情况下，预测中的最大数值应沿对角线分布。

predictions = model.predict(dict(validation_df))
predictions = tf.argmax(predictions, axis=-1)
predictions

<tf.Tensor: shape=(15745,), dtype=int64, numpy=array([1, 1, 2, ..., 2, 2, 2])>

cm = tf.math.confusion_matrix(validation_df['Sentiment'], predictions)
cm = cm/cm.numpy().sum(axis=1)[:, tf.newaxis]

sns.heatmap(
    cm, annot=True,
    xticklabels=SENTIMENT_LABELS,
    yticklabels=SENTIMENT_LABELS)
plt.xlabel("Predicted")
plt.ylabel("True")

Text(32.99999999999999, 0.5, 'True')

png

我们可以将以下代码粘贴到代码单元，然后执行该代码，从而轻松将预测值提交回 Kaggle：

test_predictions = model.predict(dict(test_df))
test_predictions = np.argmax(test_predictions, axis=-1)

result_df = test_df.copy()

result_df["Predictions"] = test_predictions

result_df.to_csv(
    "predictions.csv",
    columns=["Predictions"],
    header=["Sentiment"])
kaggle.api.competition_submit("predictions.csv", "Submitted from Colab",
                              "sentiment-analysis-on-movie-reviews")

提交后，查看排行榜了解您的表现。