使用字符级别特征的RNN网络进行姓氏分类

优质
小牛编辑
137浏览
2023-12-01

译者:hhxx2015

作者: Sean Robertson

我们将构建和训练字符级RNN来对单词进行分类。 字符级RNN将单词作为一系列字符读取,在每一步输出预测和“隐藏状态”,将其先前的隐藏状态输入至下一时刻。 我们将最终时刻输出作为预测结果,即表示该词属于哪个类。

具体来说,我们将在18种语言构成的几千个姓氏的数据集上训练模型,根据一个姓氏的拼写预测它是哪种语言的姓氏:

$ python predict.py Hinton
(-0.47) Scottish
(-1.52) English
(-3.57) Irish

$ python predict.py Schmidhuber
(-0.19) German
(-2.48) Czech
(-2.68) Dutch

阅读建议:

我默认你已经安装好了PyTorch,熟悉Python语言,理解“张量”的概念:

  • https://pytorch.org/ PyTorch安装指南
  • Deep Learning with PyTorch: A 60 Minute Blitz PyTorch入门
  • Learning PyTorch with Examples 一些PyTorch的例子
  • PyTorch for Former Torch Users Lua Torch 用户参考

事先学习并了解RNN的工作原理对理解这个例子十分有帮助:

  • The Unreasonable Effectiveness of Recurrent Neural Networks shows a bunch of real life examples
  • Understanding LSTM Networks is about LSTMs specifically but also informative about RNNs in general

准备数据

点击这里下载数据 并将其解压到当前文件夹。

在"data/names"文件夹下是名称为"[language].txt"的18个文本文件。每个文件的每一行都有一个姓氏,它们几乎都是罗马化的文本(但是我们仍需要将其从Unicode转换为ASCII编码)

我们最终会得到一个语言对应姓氏列表的字典,{language: [names ...]}

通用变量“category”和“line”(例子中的语言和姓氏单词)用于以后的可扩展性。

from __future__ import unicode_literals, print_function, division
from io import open
import glob
import os

def findFiles(path): return glob.glob(path)

print(findFiles('data/names/*.txt'))

import unicodedata
import string

all_letters = string.ascii_letters + " .,;'"
n_letters = len(all_letters)

# Turn a Unicode string to plain ASCII, thanks to https://stackoverflow.com/a/518232/2809427
def unicodeToAscii(s):
    return ''.join(
        c for c in unicodedata.normalize('NFD', s)
        if unicodedata.category(c) != 'Mn'
        and c in all_letters
    )

print(unicodeToAscii('Ślusàrski'))

# Build the category_lines dictionary, a list of names per language
category_lines = {}
all_categories = []

# Read a file and split into lines
def readLines(filename):
    lines = open(filename, encoding='utf-8').read().strip().split('\n')
    return [unicodeToAscii(line) for line in lines]

for filename in findFiles('data/names/*.txt'):
    category = os.path.splitext(os.path.basename(filename))[0]
    all_categories.append(category)
    lines = readLines(filename)
    category_lines[category] = lines

n_categories = len(all_categories)

输出:

['data/names/Italian.txt', 'data/names/German.txt', 'data/names/Portuguese.txt', 'data/names/Chinese.txt', 'data/names/Greek.txt', 'data/names/Polish.txt', 'data/names/French.txt', 'data/names/English.txt', 'data/names/Spanish.txt', 'data/names/Arabic.txt', 'data/names/Czech.txt', 'data/names/Russian.txt', 'data/names/Irish.txt', 'data/names/Dutch.txt', 'data/names/Scottish.txt', 'data/names/Vietnamese.txt', 'data/names/Korean.txt', 'data/names/Japanese.txt']
Slusarski

现在我们有了category_lines,一个字典变量存储每一种语言及其对应的每一行文本(姓氏)列表的映射关系。

变量all_categories是全部语言种类的列表,

变量n_categories 是语言种类的数量,后续会使用

print(category_lines['Italian'][:5])

输出:

['Abandonato', 'Abatangelo', 'Abatantuono', 'Abate', 'Abategiovanni']

单词转化为张量

现在我们已经加载了所有的姓氏,我们需要将它们转换为张量来使用它们。

我们使用大小为<1 x n_letters>的“one-hot 向量”表示一个字母。

一个one-hot向量所有位置都填充为0,并在其表示的字母的位置表示为1,例如"b" = <0 1 0 0 0 ...>.(字母b的编号是2,第二个位置是1,其他位置是0)

我们使用一个<line_length x 1 x n_letters>的2D矩阵表示一个单词

额外的1维是batch的维度,PyTorch默认所有的数据都是成batch处理的。我们这里只设置了batch的大小为1。

import torch

# 从所有的字母中得到某个letter的索引编号, 例如 "a" = 0
def letterToIndex(letter):
    return all_letters.find(letter)

# Just for demonstration, turn a letter into a <1 x n_letters> Tensor
def letterToTensor(letter):
    tensor = torch.zeros(1, n_letters)
    tensor[0][letterToIndex(letter)] = 1
    return tensor

# Turn a line into a <line_length x 1 x n_letters>,
# or an array of one-hot letter vectors
def lineToTensor(line):
    tensor = torch.zeros(len(line), 1, n_letters)
    for li, letter in enumerate(line):
        tensor[li][0][letterToIndex(letter)] = 1
    return tensor

print(letterToTensor('J'))

print(lineToTensor('Jones').size())

输出:

tensor([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
         0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.,
         0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
         0., 0., 0.]])
torch.Size([5, 1, 57])

构造神经网络

在autograd之前,要在Torch中构建一个可以复制之前时刻层参数的循环神经网络。

layer的隐藏状态和梯度将交给计算图自己处理。

这意味着你可以像实现正规的 feed-forward 层一样,以很常纯粹的方式实现RNN。

这个RNN组件 (几乎是从这里复制的 the PyTorch for Torch users tutorial) 仅使用两层 linear 层对输入和隐藏层做处理,

在最后添加一层 LogSoftmax 层预测最终输出。

import torch.nn as nn

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()

        self.hidden_size = hidden_size

        self.i2h = nn.Linear(input_size + hidden_size, hidden_size)
        self.i2o = nn.Linear(input_size + hidden_size, output_size)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, input, hidden):
        combined = torch.cat((input, hidden), 1)
        hidden = self.i2h(combined)
        output = self.i2o(combined)
        output = self.softmax(output)
        return output, hidden

    def initHidden(self):
        return torch.zeros(1, self.hidden_size)

n_hidden = 128
rnn = RNN(n_letters, n_hidden, n_categories)

要运行此网络的一个步骤,我们需要传递一个输入(在我们的例子中,是当前字母的Tensor)和一个先前隐藏的状态(我们首先将其初始化为零)。

我们将返回输出(每种语言的概率)和下一个隐藏状态(为我们下一步保留使用)。

input = letterToTensor('A')
hidden =torch.zeros(1, n_hidden)

output, next_hidden = rnn(input, hidden)

为了提高效率,我们不希望为每一步都创建一个新的Tensor,因此我们将使用lineToTensor函数而不是letterToTensor函数,并使用切片方法。

这一步可以通过预先计算批量的张量进一步优化。

input = lineToTensor('Albert')
hidden = torch.zeros(1, n_hidden)

output, next_hidden = rnn(input[0], hidden)
print(output)

输出:

tensor([[-2.8857, -2.9005, -2.8386, -2.9397, -2.8594, -2.8785, -2.9361, -2.8270,
         -2.9602, -2.8583, -2.9244, -2.9112, -2.8545, -2.8715, -2.8328, -2.8233,
         -2.9685, -2.9780]], grad_fn=<LogSoftmaxBackward>)

可以看到输出是一个<1 x n_categories>的张量,其中每一条代表这个单词属于某一类的可能性(越高可能性越大)

训练

训练前的准备

进行训练步骤之前我们需要构建一些帮助函数。

第一个是当我们知道输出结果对应每种类别的可能性时,解析神经网络的输出。

我们可以使用 Tensor.topk函数得到最大值在结果中的位置索引

def categoryFromOutput(output):
    top_n, top_i = output.topk(1)
    category_i = top_i[0].item()
    return all_categories[category_i], category_i

print(categoryFromOutput(output))

输出:

('Vietnamese', 15)

我们还需要一种快速获取训练示例(得到一个姓氏及其所属的语言类别)的方法:

import random

def randomChoice(l):
    return l[random.randint(0, len(l) - 1)]

def randomTrainingExample():
    category = randomChoice(all_categories)
    line = randomChoice(category_lines[category])
    category_tensor = torch.tensor([all_categories.index(category)], dtype=torch.long)
    line_tensor = lineToTensor(line)
    return category, line, category_tensor, line_tensor

for i in range(10):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    print('category =', category, '/ line =', line)

输出:

category = Russian / line = Minkin
category = French / line = Masson
category = German / line = Hasek
category = Dutch / line = Kloeten
category = Scottish / line = Allan
category = Italian / line = Agostini
category = Japanese / line = Fumihiko
category = Polish / line = Gajos
category = Scottish / line = Duncan
category = Arabic / line = Gerges

训练神经网络

现在,训练过程只需要向神经网络输入大量的数据,让它做出预测,并将对错反馈给它。

nn.LogSoftmax作为最后一层layer时,nn.NLLLoss作为损失函数是合适的。

criterion = nn.NLLLoss()

训练过程的每次循环将会发生:

  • 构建输入和目标张量
  • 构建0初始化的隐藏状态
  • 读入每一个字母
    • 将当前隐藏状态传递给下一字母
  • 比较最终结果和目标
  • 反向传播
  • 返回结果和损失
learning_rate = 0.005 # If you set this too high, it might explode. If too low, it might not learn

def train(category_tensor, line_tensor):
    hidden = rnn.initHidden()

    rnn.zero_grad()

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    loss = criterion(output, category_tensor)
    loss.backward()

    # Add parameters' gradients to their values, multiplied by learning rate
    for p in rnn.parameters():
        p.data.add_(-learning_rate, p.grad.data)

    return output, loss.item()

现在我们只需要准备一些例子来运行程序。

由于train函数同时返回输出和损失,我们可以打印其输出结果并跟踪其损失画图。

由于有1000个示例,我们每print_every次打印样例,并求平均损失。

import time
import math

n_iters = 100000
print_every = 5000
plot_every = 1000

# Keep track of losses for plotting
current_loss = 0
all_losses = []

def timeSince(since):
    now = time.time()
    s = now - since
    m = math.floor(s / 60)
    s -= m * 60
    return '%dm %ds' % (m, s)

start = time.time()

for iter in range(1, n_iters + 1):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output, loss = train(category_tensor, line_tensor)
    current_loss += loss

    # Print iter number, loss, name and guess
    if iter % print_every == 0:
        guess, guess_i = categoryFromOutput(output)
        correct = '✓' if guess == category else '✗ (%s)' % category
        print('%d  %d%% (%s) %.4f  %s / %s  %s' % (iter, iter / n_iters * 100, timeSince(start), loss, line, guess, correct))

    # Add current loss avg to list of losses
    if iter % plot_every == 0:
        all_losses.append(current_loss / plot_every)
        current_loss = 0

输出:

5000 5% (0m 11s) 2.0318 Jaeger / German ✓
10000 10% (0m 18s) 2.1296 Sokolofsky / Russian ✗ (Polish)
15000 15% (0m 26s) 1.2620 Jo / Korean ✓
20000 20% (0m 34s) 1.9295 Livson / Scottish ✗ (Russian)
25000 25% (0m 41s) 1.2325 Fortier / French ✓
30000 30% (0m 49s) 2.5714 Purdes / Dutch ✗ (Czech)
35000 35% (0m 56s) 2.3312 Bayer / Arabic ✗ (German)
40000 40% (1m 4s) 2.3792 Mitchell / Dutch ✗ (Scottish)
45000 45% (1m 12s) 1.3536 Maes / Dutch ✓
50000 50% (1m 20s) 2.6095 Sai / Chinese ✗ (Vietnamese)
55000 55% (1m 28s) 0.5883 Cheung / Chinese ✓
60000 60% (1m 35s) 1.5788 William / Irish ✓
65000 65% (1m 43s) 2.5809 Mulder / Scottish ✗ (Dutch)
70000 70% (1m 51s) 1.3440 Bruce / German ✗ (Scottish)
75000 75% (1m 58s) 1.1839 Romero / Italian ✗ (Spanish)
80000 80% (2m 6s) 2.6453 Reyes / Portuguese ✗ (Spanish)
85000 85% (2m 14s) 0.0290 Mcmillan / Scottish ✓
90000 90% (2m 22s) 0.7337 Riagan / Irish ✓
95000 95% (2m 30s) 2.6208 Maneates / Dutch ✗ (Greek)
100000 100% (2m 37s) 0.5170 Szwarc / Polish ✓

画出结果

all_losses得到历史损失记录,反映了神经网络的学习情况:

import matplotlib.pyplot as plt
import matplotlib.ticker as ticker

plt.figure()
plt.plot(all_losses)

https://pytorch.org/tutorials/_images/sphx_glr_char_rnn_classification_tutorial_001.png

评价结果

为了了解网络在不同类别上的表现,我们将创建一个混淆矩阵,显示每种语言(行)和神经网络将其预测为哪种语言(列)。

为了计算混淆矩阵,使用evaluate()函数处理了一批数据,evaluate()函数与去掉反向传播的train()函数大体相同。

# Keep track of correct guesses in a confusion matrix
confusion = torch.zeros(n_categories, n_categories)
n_confusion = 10000

# Just return an output given a line
def evaluate(line_tensor):
    hidden = rnn.initHidden()

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    return output

# Go through a bunch of examples and record which are correctly guessed
for i in range(n_confusion):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output = evaluate(line_tensor)
    guess, guess_i = categoryFromOutput(output)
    category_i = all_categories.index(category)
    confusion[category_i][guess_i] += 1

# Normalize by dividing every row by its sum
for i in range(n_categories):
    confusion[i] = confusion[i] / confusion[i].sum()

# Set up plot
fig = plt.figure()
ax = fig.add_subplot(111)
cax = ax.matshow(confusion.numpy())
fig.colorbar(cax)

# Set up axes
ax.set_xticklabels([''] + all_categories, rotation=90)
ax.set_yticklabels([''] + all_categories)

# Force label at every tick
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(1))

# sphinx_gallery_thumbnail_number = 2
plt.show()

https://pytorch.org/tutorials/_images/sphx_glr_char_rnn_classification_tutorial_002.png

你可以从主轴线以外挑出亮的点,显示模型预测错了哪些语言,例如汉语预测为了韩语,西班牙预测为了意大利。

看上去在希腊语上效果很好,在英语上表现欠佳。(可能是因为英语与其他语言的重叠较多)。

处理用户输入

def predict(input_line, n_predictions=3):
    print('\n> %s' % input_line)
    with torch.no_grad():
        output = evaluate(lineToTensor(input_line))

        # Get top N categories
        topv, topi = output.topk(n_predictions, 1, True)
        predictions = []

        for i in range(n_predictions):
            value = topv[0][i].item()
            category_index = topi[0][i].item()
            print('(%.2f) %s' % (value, all_categories[category_index]))
            predictions.append([value, all_categories[category_index]])

predict('Dovesky')
predict('Jackson')
predict('Satoshi')

输出:

> Dovesky
(-0.74) Russian
(-0.77) Czech
(-3.31) English

> Jackson
(-0.80) Scottish
(-1.69) English
(-1.84) Russian

> Satoshi
(-1.16) Japanese
(-1.89) Arabic
(-1.90) Polish

最终版的脚本 in the Practical PyTorch repo 将上述代码拆分为几个文件:

  • data.py (读取文件)
  • model.py (构造RNN网络)
  • train.py (运行训练过程)
  • predict.py (在命令行中和参数一起运行predict()函数)
  • server.py (使用bottle.py构建JSON API的预测服务)

运行 train.py 来训练和保存网络

predict.py和一个姓氏的单词一起运行查看预测结果 :

$ python predict.py Hazaki
(-0.42) Japanese
(-1.39) Polish
(-3.51) Czech

运行 server.py 并访问http://localhost:5533/Yourname 得到JSON格式的预测输出

练习

  • 尝试其它 (类别->行) 格式的数据集,比如:
    • 任何单词 -> 语言
    • 姓名 -> 性别
    • 角色姓名 -> 作者
    • 页面标题 -> blog 或 subreddit
  • 通过更大和更复杂的网络获得更好的结果
    • 增加更多linear层
    • 尝试 nn.LSTMnn.GRU
    • 组合这些 RNN构造更复杂的神经网络