diff --git a/machine_translation/.gitignore b/machine_translation/.gitignore new file mode 100644 index 00000000..3efbccb6 --- /dev/null +++ b/machine_translation/.gitignore @@ -0,0 +1,8 @@ +data/wmt14 +data/pre-wmt14 +pretrained/wmt14_model +gen.log +gen_result +train.log +dataprovider_copy_1.py +*.pyc diff --git a/machine_translation/README.md b/machine_translation/README.md index 46b5a71c..84edb20c 100644 --- a/machine_translation/README.md +++ b/machine_translation/README.md @@ -1 +1,580 @@ -TODO: Write about https://github.com/PaddlePaddle/Paddle/tree/develop/demo/seqToseq +# 机器翻译 + +## 背景介绍 + +机器翻译（machine translation, MT）就是用计算机来实现不同语言之间的翻译。被翻译的语言通常称为源语言（source language），翻译成的结果语言称为目标语言（target language）。机器翻译就是实现从源语言到目标语言转换的过程，是自然语言处理的重要研究领域之一。 + +传统机器翻译方法是让语言学专家编写两种语言之间的转换规则，再将这些规则录入给计算机。但这个方法对语言学专家的要求非常高，并且，我们不可能知道一门语言究竟有多少隐含的使用规则，更何况两种甚至更多的语言。因此，我们往往无法得到一个完备的规则集合，这也是传统方法面临的主要挑战。 + +统计机器翻译（Statistical Machine Translation, SMT），其转化规则是由机器自动从大规模的语料中学习得到的，而非我们人主动提供规则。因此，该方法克服了传统方法面临的知识获取瓶颈问题，但仍然存在许多挑战：1）需要人为设计许多特征，但永远无法覆盖所有的语言规则；2）难以利用全局的特征；3）依赖于许多预处理环节，如词语对齐、中文分词、规则抽取、句法分析等，而每个环节的错误会逐步累积，对翻译的影响也越来也大。 + +面对上述挑战，一个较好的解决方法是利用神经网络。利用神经网络的机器翻译系统大致分为两类：1）仍以统计机器翻译系统为框架，只是利用神经网络来改进其中的关键模块（见图1的左半部分），如翻译模型、语言模型、调序模型等。2）不再以统计机器翻译系统为框架，而是直接用神经网络将源语言映射到目标语言（见图1的右半部分），即端到端的神经网络机器翻译（Neural Machine Translation, NMT）。 +

+
+图. 基于神经网络的机器翻译系统 +

+ +本教程主要介绍NMT模型，以及如何用PaddlePaddle来训练一个法英翻译（从法语翻译到英语）的模型。 + +## 效果展示 + +当法英翻译的模型训练完毕后，以下面的两句法语为例： +```text +Les avionneurs se querellent au sujet de la largeur des sièges alors que de grosses commandes sont en jeu Jet makers feud over seat width with big orders at stake +La dispute fait rage entre les grands constructeurs aéronautiques à propos de la largeur des sièges de la classe touriste sur les vols long-courriers , ouvrant la voie à une confrontation amère lors du salon aéronautique de Dubaï qui a lieu de mois-ci . A row has flared up between leading plane makers over the width of tourist-class seats on long-distance flights , setting the tone for a bitter confrontation at this month 's Dubai Airshow . + +``` +如果设定显示的条数（即[集束搜索算法](### 集束搜索算法)的宽度）为3，生成的英语句子如下： +```text +0 +0 -19.0179 The will be rotated about the width of the seats , while large orders are at stake . +1 -19.1114 The will be rotated about the width of the seats , while large commands are at stake . +2 -19.5112 The will be rotated about the width of the seats , while large commands are at play . + +1 +0 -28.1139 The dispute is between the large aviation manufacturers about the width of the tourist seats on the flights , paving the way for a confrontation at the Dubai aviation fair , which is a month . +1 -28.7138 The dispute is between the large aviation manufacturers about the width of the tourist seats on the flights , paving the way for a confrontation at the Dubai aviation fair , which takes place months . +2 -29.3381 The dispute is between the large aviation manufacturers about the width of the tourist seats on the flights , paving the way for a confrontation at the Dubai aviation fair , which takes place in month . +``` +- 第一行的“0”和第6行的“1”表示翻译第几条法语句子。 +- 其他六行列出了翻译结果，其中右起第一列是生成的英语句子，右起第二列是该条英语句子的得分（从大到小），右起第三列是序号。 +- 另外有两个特殊表示：``表示句子的结尾，`unk`表示无法识别（即不在训练字典中）的单词。 + +## 模型概览 + +本节依次介绍GRU模型，NMT中典型的“编码器-解码器”（Encoder-Decoder）框架和“注意力”（Attention）机制，以及集束搜索（Beam Search）算法。 + +### GRU模型 + +GRU（Gated Recurrent Unit，门控循环单元）是Cho等人在LSTM上提出的简化版本，如下图所示。GRU单元只有两个门，分别是重置门（Reset Gate）和更新门（Update Gate），用于控制记忆内容是否能继续保存下去。 +

+
+图. GRU模型 +

+ +Chung等人通过多组实验表明，GRU虽然参数更少，但是在多个任务上都和LSTM有相近的表现。 + +### 编码器-解码器框架 + +编码器-解码器（Encoder-Decoder）框架用于解决一个任意长度的源序列到另一个任意长度的目标序列的变换问题。即编码阶段将整个源序列压缩成一个固定维度的向量，解码阶段以概率的方式从中解码出整个目标序列。其中编码和解码的过程通常使用RNN实现，但由于普通的RNN存在梯度消失和梯度爆炸问题，因此通常引入LSTM或GRU。 +

+
+图. 编码器-解码器框架 +

+ +### 编码器 + +编码阶段分为三步： +- 独热编码（one-hot coding，1-of-K coding）表示：将源语言句子$x=\left \{ x_1,x_2,...,x_T \right \}$的每个词表示成一个向量$w_i\epsilon R^{\left | V \right |,i=1,2,...,T}$ （T暂不包括句子结束标记``）。这个向量的维度与词汇表大小$\left | V \right |$ 相同，并且只有一个维度上有值1（该位置对应该词在词汇表中的位置），其余全是0。 +- 映射到低维语义空间的词向量：独热编码表示存在两个问题，1）生成的向量维度往往很大，容易造成维数灾难；2）无法刻画词与词之间的关系（如语义相似性，也就是无法很好地表达语义）。因此需再映射到低维的语义空间，由一个固定维度的稠密向量表示，称为词向量。记映射矩阵为$C\epsilon R^{K\times \left | V \right |}$，用$s_i=Cw_i$表示第$i$个词的词向量，向量维度$K$通常取100-500之间。 +- 用RNN压缩源语言词序列：这一过程的计算公式为$h_i=\varnothing _\theta \left ( h_{i-1}, s_i \right )$，其中$h_0$是一个全零的向量，最后得到的$h_T$就是整个源语言句子的压缩表示。 + +### 解码器 + +解码阶段的总体思路是：根据当前单词$u_i$，当前RNN隐层状态$z_i$，计算出下一个隐层状态$z_{i+1}$。然后用softmax归一化得到概率分布$p_{i+1}$，以此采样出单词$u_{i+1}$。不断重复上述操作，直到获得句子结束标记``或超过RNN循环次数为止。 + +$z_i$的计算公式如下： +$$z_{i+1}=\phi _{\theta '}\left ( c,u_i,z_i \right )$$ +其中$c=h_T$是源语言句子的压缩表示，$z_0$是一个全零的向量，$u_0$是源语言句子的结束标记``。 + +概率分布公式如下： +$$p\left ( u_i|u_{ +
+图. 基于注意力机制的编码器 +

+ +### 解码器 + +与简单的解码器不同，这里$z_i$的计算公式如下： +$$z_{i+1}=\phi _{\theta '}\left ( c_i,u_i,z_i \right )$$ +即源语言句子的压缩表示由原来的$c$变成了$c_i$，同时$c_i$可以通过各个$h_i$的加权平均得到：$c_i=\sum _{j=1}^{T}\alpha _{ij}h_j$，$\alpha _{ij}$为权重。 + +### 集束搜索算法 + +集束搜索（Beam Search）是一种启发式图搜索算法，通常用在解空间较大的系统中，如机器翻译、语音识别等。原因是这些系统的数据集通常非常大，而且也没有唯一正确解，因此用最快的方法找到最接近正确的解才是系统的目标。 + +集束搜索算法使用广度优先策略建立搜索树，在树的每一层，按照启发代价（Heuristic Cost）对节点进行排序，然后仅留下预先确定的个数（Beam Width，Beam Size，集束宽度）的节点。只有这些节点会在下一层继续扩展，其他节点就被剪掉了，也就是说保留了质量较高的节点，剪掉了质量较差的节点。因此，搜索所占用的空间和时间大幅减少，但缺点是可能存在最优解被丢弃。 + +## 数据准备 + +### 下载与解压缩 + +本教程使用[WMT-14](http://www-lium.univ-lemans.fr/~schwenk/cslm_joint_paper/)数据集中的[bitexts(after selection)](http://www-lium.univ-lemans.fr/~schwenk/cslm_joint_paper/data/bitexts.tgz)作为训练集，[dev+test data](dev 与 test 数据)作为测试集和生成集。 + +在Linux下，只需简单地运行以下命令： +```bash +cd data +./wmt14_data.sh +``` +得到的数据集`data/wmt14`包含如下三个文件夹： +

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
文件夹名法英平行语料文件文件数文件大小
trainccb2_pc30.src, ccb2_pc30.trg, etc123.55G
testntst1213.src, ntst1213.trg21636k
genntst14.src, ntst14.trg2864k
+

+ +- `XXX.src`是源法语文件，`XXX.trg`是目标英语文件，文件中的每行存放一个句子 +- `XXX.src`和`XXX.trg`的行数一致，且两者任意第$i$行的句子之间都有着一一对应的关系。 + +### 用户自定义数据集（可选） + +如果您想使用自己的数据集，只需按照如下方式组织，并将它们放在`data`目录下： +```text +user_dataset +├── train +│   ├── train_file1.src +│   ├── train_file1.trg +│   └── ... +├── test +│   ├── test_file1.src +│   ├── test_file1.trg +│   └── ... +├── gen +│   ├── gen_file1.src +│   ├── gen_file1.trg +│   └── ... +``` + +- 一级目录`user_dataset`：用户自定义的数据集名字。 +- 二级目录`train`、`test`和`gen`：必须使用这三个文件夹名字。 +- 三级目录：存放源语言到目标语言的平行语料库文件，后缀名必须使用`.src`和`.trg`。 + +### 数据预处理 + +我们的预处理流程包括两步： +- 将每个源语言到目标语言的平行语料库文件合并为一个文件： + - 合并每个`XXX.src`和`XXX.trg`文件为`XXX`。 + - `XXX`中的第$i$行 = `XXX.src`中的第$i$行 + '\t' + `XXX.trg`中的第$i$行。 +- 创建训练数据的“源字典”和“目标字典”。每个字典都有**DICTSIZE**个单词，包括：词频最高的（DICTSIZE - 3）个单词，和3个特殊符号``（序列的开始）、``（序列的结束）和``（未包含在字典中的单词）。 + +预处理可以使用`preprocess.py`： +```python +python preprocess.py -i INPUT [-d DICTSIZE] [-m] +``` +- `-i INPUT`：输入的原始数据集路径 +- `-d DICTSIZE`：指定的字典单词数，如果没有设置，字典会包含输入数据集中的所有单词 +- `-m --mergeDict`：合并 “源字典”和“目标字典”，使得两个字典有相同的上下文 + +本教程的具体命令如下： +```python +python preprocess.py -i data/wmt14 -d 30000 +``` +请耐心等待几分钟的时间，您会在屏幕上看到： +```text +concat parallel corpora for dataset +build source dictionary for train data +build target dictionary for train data +dictionary size is 30000 +``` +预处理好的数据集存放在`data/pre-wmt14`目录下： +```text +pre-wmt14 +├── train +│   └── train +├── test +│   └── test +├── gen +│   └── gen +├── train.list +├── test.list +├── gen.list +├── src.dict +└── trg.dict +``` +- `train`、`test`和`gen`：分别包含了法英平行语料库的训练、测试和生成数据。其每个文件的每一行均分为两部分，前半部分是法语序列，后半部分是对应的英语序列。 +- `train.list`、`test.list`和`gen.list`：分别记录了`train`、`test`和`gen`文件夹中的文件路径。 +- `src.dict`和`trg.dict`：源（法语）和目标（英语）字典。每个字典都含有30000个单词，包括29997个最高频单词和3个特殊符号。 + +### 提供数据给PaddlePaddle + +我们通过`dataprovider.py`将数据提供给PaddlePaddle。具体步骤如下： + +1. 首先，引入PaddlePaddle的PyDataProvider2包，并定义三个特殊符号。 +```python +from paddle.trainer.PyDataProvider2 import * +UNK_IDX = 2 #未包含在字典中的单词 +START = "" #序列的开始 +END = "" #序列的结束 +``` +2. 其次，使用初始化函数`hook`，分别定义了训练模式和应用（这里为生成）模式下的数据输入格式（`input_types`）。其中`src_dict_path`（源语言字典路径）、`trg_dict_path`（目标语言字典路径）和`is_generating`（训练或生成模式）是从模型配置中传入的对象，调用方式请见[训练模型配置说明](### 训练模型配置说明)。 +```python +def hook(settings, src_dict_path, trg_dict_path, is_generating, file_list, + **kwargs): + # job_mode = 1: 训练模式；0: 生成模式 + settings.job_mode = not is_generating + + def fun(dict_path): # 根据字典路径加载字典 + out_dict = dict() + with open(dict_path, "r") as fin: + out_dict = { + line.strip(): line_count + for line_count, line in enumerate(fin) + } + return out_dict + + settings.src_dict = fun(src_dict_path) + settings.trg_dict = fun(trg_dict_path) + + if settings.job_mode: #训练模式 + settings.input_types = { + 'source_language_word': #源语言序列 + integer_value_sequence(len(settings.src_dict)), + 'target_language_word': #目标语言序列 + integer_value_sequence(len(settings.trg_dict)), + 'target_language_next_word': #目标语言序列的下一个词 + integer_value_sequence(len(settings.trg_dict)) + } + else: #生成模式 + settings.input_types = { + 'source_language_word': #源语言序列 + integer_value_sequence(len(settings.src_dict)), + 'sent_id': #源语言序列编号 + integer_value_sequence(len(open(file_list[0], "r").readlines())) + } +``` +3. 最后，使用`process`函数打开文本文件`file_name`，读取每一行，将行中的数据转换成与`input_types`一致的格式，再用`yield`关键字返回给PaddlePaddle进程。注意：由于本示例中的训练数据有3.55G，对于内存较小的机器，不能一次性加载进内存，所以推荐使用`pool_size`变量来设置内存中暂存的数据条数。 + +```python +def _get_ids(s, dictionary): # 获得源语言序列中的每个单词在字典中的位置 + words = s.strip().split() + return [dictionary[START]] + \ + [dictionary.get(w, UNK_IDX) for w in words] + \ + [dictionary[END]] + +@provider(init_hook=hook, pool_size=50000) +def process(settings, file_name): + with open(file_name, 'r') as f: + for line_count, line in enumerate(f): + line_split = line.strip().split('\t') + if settings.job_mode and len(line_split) != 2: + continue + src_seq = line_split[0] + src_ids = _get_ids(src_seq, settings.src_dict) + + if settings.job_mode: + trg_seq = line_split[1] + trg_words = trg_seq.split() + trg_ids = [settings.trg_dict.get(w, UNK_IDX) for w in trg_words] + + # 如果任意一个序列长度超过80个单词，在训练模式下会移除这条样本，以防止RNN过深。 + if len(src_ids) > 80 or len(trg_ids) > 80: + continue + trg_ids_next = trg_ids + [settings.trg_dict[END]] + trg_ids = [settings.trg_dict[START]] + trg_ids + yield { + 'source_language_word': src_ids, + 'target_language_word': trg_ids, + 'target_language_next_word': trg_ids_next + } + else: + yield {'source_language_word': src_ids, 'sent_id': [line_count]} +``` + +## 模型配置说明 + +### 数据定义 + +1. 首先，定义数据集路径和源/目标语言字典路径，并用`is_generating`变量定义当前配置是训练模式（默认）还是应用模式。该变量接受从命令行传入的参数，使用方法见[应用命令与结果](### 应用命令与结果)。 + +```python +import os +from paddle.trainer_config_helpers import * + +data_dir = "./data/pre-wmt14" # 数据集路径 +src_lang_dict = os.path.join(data_dir, 'src.dict') # 源语言字典路径 +trg_lang_dict = os.path.join(data_dir, 'trg.dict') # 目标语言字典路径 +is_generating = get_config_arg("is_generating", bool, False) # 配置模式 +``` +2. 其次，通过`define_py_data_sources2`函数从`dataprovider.py`中读取数据，并用`args`变量传入源/目标语言的字典路径以及配置模式。 + +```python +if not is_generating: + train_list = os.path.join(data_dir, 'train.list') + test_list = os.path.join(data_dir, 'test.list') +else: + train_list = None + test_list = os.path.join(data_dir, 'gen.list') + +define_py_data_sources2( + train_list, + test_list, + module="dataprovider", + obj="process", + args={ + "src_dict_path": src_lang_dict, + "trg_dict_path": trg_lang_dict, + "is_generating": is_generating + }) +``` + +### 算法配置 + +```python +settings( + learning_method = AdamOptimizer(), + batch_size = 50 is not is_generating else 1, + learning_rate = 5e-4 is not is_generating else 0) +``` + +### 模型结构 +```python +source_dict_dim = len(open(src_lang_dict, "r").readlines()) +target_dict_dim = len(open(trg_lang_dict, "r").readlines()) +word_vector_dim = 512 # dimension of word vector +decoder_size = 512 # dimension of hidden unit in GRU Decoder network +encoder_size = 512 # dimension of hidden unit in GRU Encoder network + +if is_generating: + beam_size=3 # expand width in beam search + max_length=250 # a stop condition of sequence generation + gen_trans_file = get_config_arg("gen_trans_file", str, None) + +#### Encoder +src_word_id = data_layer(name='source_language_word', size=source_dict_dim) +src_embedding = embedding_layer( + input=src_word_id, + size=word_vector_dim, + param_attr=ParamAttr(name='_source_language_embedding')) +src_forward = simple_gru(input=src_embedding, size=encoder_size) +src_backward = simple_gru( + input=src_embedding, size=encoder_size, reverse=True) +encoded_vector = concat_layer(input=[src_forward, src_backward]) + +with mixed_layer(size=decoder_size) as encoded_proj: + encoded_proj += full_matrix_projection(input=encoded_vector) + +backward_first = first_seq(input=src_backward) +with mixed_layer( + size=decoder_size, + act=TanhActivation(), ) as decoder_boot: + decoder_boot += full_matrix_projection(input=backward_first) + +#### Decoder +def gru_decoder_with_attention(enc_vec, enc_proj, current_word): + decoder_mem = memory( + name='gru_decoder', size=decoder_size, boot_layer=decoder_boot) + + context = simple_attention( + encoded_sequence=enc_vec, + encoded_proj=enc_proj, + decoder_state=decoder_mem, ) + + with mixed_layer(size=decoder_size * 3) as decoder_inputs: + decoder_inputs += full_matrix_projection(input=context) + decoder_inputs += full_matrix_projection(input=current_word) + + gru_step = gru_step_layer( + name='gru_decoder', + input=decoder_inputs, + output_mem=decoder_mem, + size=decoder_size) + + with mixed_layer( + size=target_dict_dim, bias_attr=True, + act=SoftmaxActivation()) as out: + out += full_matrix_projection(input=gru_step) + return out + +decoder_group_name = "decoder_group" +group_inputs = [ + StaticInput( + input=encoded_vector, is_seq=True), StaticInput( + input=encoded_proj, is_seq=True) +] + +if not is_generating: + trg_embedding = embedding_layer( + input=data_layer( + name='target_language_word', size=target_dict_dim), + size=word_vector_dim, + param_attr=ParamAttr(name='_target_language_embedding')) + group_inputs.append(trg_embedding) + + # For decoder equipped with attention mechanism, in training, + # target embeding (the groudtruth) is the data input, + # while encoded source sequence is accessed to as an unbounded memory. + # Here, the StaticInput defines a read-only memory + # for the recurrent_group. + decoder = recurrent_group( + name=decoder_group_name, + step=gru_decoder_with_attention, + input=group_inputs) + + lbl = data_layer(name='target_language_next_word', size=target_dict_dim) + cost = classification_cost(input=decoder, label=lbl) + outputs(cost) +else: + # In generation, the decoder predicts a next target word based on + # the encoded source sequence and the last generated target word. + + # The encoded source sequence (encoder's output) must be specified by + # StaticInput, which is a read-only memory. + # Embedding of the last generated word is automatically gotten by + # GeneratedInputs, which is initialized by a start mark, such as , + # and must be included in generation. + + + trg_embedding = GeneratedInput( + size=target_dict_dim, + embedding_name='_target_language_embedding', + embedding_size=word_vector_dim) + group_inputs.append(trg_embedding) + + beam_gen = beam_search( + name=decoder_group_name, + step=gru_decoder_with_attention, + input=group_inputs, + bos_id=0, + eos_id=1, + beam_size=beam_size, + max_length=max_length) + + seqtext_printer_evaluator( + input=beam_gen, + id_input=data_layer( + name="sent_id", size=1), + dict_file=trg_lang_dict, + result_file=gen_trans_file) + outputs(beam_gen) +``` + +## 训练模型 + +可以通过以下命令来训练模型： + +```bash +./train.sh +``` +其中`train.sh` 的内容为： + +```bash +paddle train \ +--config='seqToseq_net.py' \ +--save_dir='model' \ +--use_gpu=false \ +--num_passes=16 \ +--show_parameter_stats_period=100 \ +--trainer_count=4 \ +--log_period=10 \ +--dot_period=5 \ +2>&1 | tee 'train.log' +``` +- config: 设置神经网络的配置文件。 +- save_dir: 设置保存模型的输出路径。 +- use_gpu: 是否使用GPU训练，这里使用CPU。 +- num_passes: 设置passes的数量。PaddlePaddle中的一个pass表示对数据集中所有样本的一次完整训练。 +- show_parameter_stats_period: 这里每隔100个batch显示一次参数统计信息。 +- trainer_count: 设置CPU线程数或者GPU设备数。 +- log_period: 这里每隔10个batch打印一次日志。 +- dot_period: 这里每个5个batch打印一个点"."。 + +训练的损失函数每隔10个batch打印一次，您将会看到如下消息： +```text +I0719 19:16:45.952062 15563 TrainerInternal.cpp:160] Batch=10 samples=500 AvgCost=198.475 CurrentCost=198.475 Eval: classification_error_evaluator=0.737155 CurrentEval: classification_error_evaluator=0.737155 +I0719 19:17:56.707319 15563 TrainerInternal.cpp:160] Batch=20 samples=1000 AvgCost=157.479 CurrentCost=116.483 Eval: classification_error_evaluator=0.698392 CurrentEval: classification_error_evaluator=0.659065 +..... +``` +- AvgCost：从第0个batch到当前batch的平均损失值。 +- CurrentCost：当前batch的损失值。 +- classification\_error\_evaluator(Eval)：从第0个评估到当前评估中，每个单词的预测错误率。 +- classification\_error\_evaluator(CurrentEval)：当前评估中，每个单词的预测错误率。 + +当classification\_error\_evaluator的值低于0.35时，模型就训练成功了。 + +## 应用模型 + +### 下载预训练的模型 + +由于机器翻译模型的训练非常耗时，我们在50个物理节点（每节点含有2颗6核CPU）的集群中，花了5天时间训练了16个pass，其中每个pass耗时7个小时。因此，我们提供了一个预训练好的模型（pass-00012）供大家直接下载使用。该模型大小为205MB，在所有16个模型中有最高的[BLEU评估](### BLEU评估)值27.77。下载并解压模型的命令如下： +```bash +cd pretrained +./wmt14_model.sh +``` + +### 应用命令与结果 + +可以通过以下命令来进行法英翻译： + +```bash +./gen.sh +``` +其中`gen.sh` 的内容为： + +```bash +paddle train \ +--job=test \ +--config='seqToseq_net.py' \ +--save_dir='pretrained/wmt14_model' \ +--use_gpu=true \ +--num_passes=13 \ +--test_pass=12 \ +--trainer_count=1 \ +--config_args=is_generating=1,gen_trans_file="gen_result" \ +2>&1 | tee 'translation/gen.log' +``` +与训练命令不同的参数如下： +- job：设置任务的模式为测试。 +- save_dir：设置存放预训练模型的路径。 +- num_passes和test_pass：加载第$i\epsilon \left [ test\_pass,num\_passes-1 \right ]$轮的模型参数，这里只加载 `data/wmt14_model/pass-00012`。 +- config_args：将命令行中的自定义参数传递给模型配置。 + +翻译结果请见[效果展示](## 效果展示)。 + +### BLEU评估 + +BLEU(Bilingual Evaluation understudy)是一种广泛使用的机器翻译自动评测指标，由IBM的watson研究中心于2002年提出，基本出发点是：机器译文越接近专业翻译人员的翻译结果，翻译系统的性能越好。其中，机器译文与人工参考译文之间的接近程度，采用句子精确度（precision）的计算方法，即比较两者的n元词组相匹配的个数，匹配的个数越多，BLEU得分越好。 + +[Moses](http://www.statmt.org/moses/) 是一个统计学的开源机器翻译系统，我们使用其中的 [multi-bleu.perl](https://github.com/moses-smt/mosesdecoder/blob/master/scripts/generic/multi-bleu.perl) 来做BLEU评估。下载脚本的命令如下： +```bash +./moses_bleu.sh +``` + +## 总结 + +## 参考文献 + +1. Bahdanau D, Cho K, Bengio Y. [Neural machine translation by jointly learning to align and translate](https://arxiv.org/abs/1409.0473)[J]. arXiv preprint arXiv:1409.0473, 2014. +1. Cho K, Van Merriënboer B, Gulcehre C, et al. [Learning phrase representations using RNN encoder-decoder for statistical machine translation](https://arxiv.org/abs/1406.1078)[J]. arXiv preprint arXiv:1406.1078, 2014. +1. Chung J, Gulcehre C, Cho K H, et al. [Empirical evaluation of gated recurrent neural networks on sequence modeling](https://arxiv.org/abs/1412.3555)[J]. arXiv preprint arXiv:1412.3555, 2014. +1. http://en.wikipedia.org/wiki/Beam_search +1. Papineni K, Roukos S, Ward T, et al. [BLEU: a method for automatic evaluation of machine translation](http://dl.acm.org/citation.cfm?id=1073135)[C]//Proceedings of the 40th annual meeting on association for computational linguistics. Association for Computational Linguistics, 2002: 311-318. +1. Clear.的博客. http://blog.csdn.net/u011414416 + diff --git a/machine_translation/data/wmt14_data.sh b/machine_translation/data/wmt14_data.sh new file mode 100755 index 00000000..43f67168 --- /dev/null +++ b/machine_translation/data/wmt14_data.sh @@ -0,0 +1,53 @@ +#!/bin/bash +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +set -e +set -x +mkdir wmt14 +cd wmt14 + +# download the dataset +wget http://www-lium.univ-lemans.fr/~schwenk/cslm_joint_paper/data/bitexts.tgz +wget http://www-lium.univ-lemans.fr/~schwenk/cslm_joint_paper/data/dev+test.tgz + +# untar the dataset +tar -zxvf bitexts.tgz +tar -zxvf dev+test.tgz +gunzip bitexts.selected/* +mv bitexts.selected train +rm bitexts.tgz +rm dev+test.tgz + +# separate the dev and test dataset +mkdir test gen +mv dev/ntst1213.* test +mv dev/ntst14.* gen +rm -rf dev + +set +x +# rename the suffix, .fr->.src, .en->.trg +for dir in train test gen +do + filelist=`ls $dir` + cd $dir + for file in $filelist + do + if [ ${file##*.} = "fr" ]; then + mv $file ${file/%fr/src} + elif [ ${file##*.} = 'en' ]; then + mv $file ${file/%en/trg} + fi + done + cd .. +done diff --git a/machine_translation/dataprovider.py b/machine_translation/dataprovider.py new file mode 100755 index 00000000..c2b49804 --- /dev/null +++ b/machine_translation/dataprovider.py @@ -0,0 +1,94 @@ +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from paddle.trainer.PyDataProvider2 import * + +UNK_IDX = 2 +START = "" +END = "" + + +def hook(settings, src_dict_path, trg_dict_path, is_generating, file_list, + **kwargs): + # job_mode = 1: training mode + # job_mode = 0: generating mode + settings.job_mode = not is_generating + + def fun(dict_path): + out_dict = dict() + with open(dict_path, "r") as fin: + out_dict = { + line.strip(): line_count + for line_count, line in enumerate(fin) + } + return out_dict + + settings.src_dict = fun(src_dict_path) + settings.trg_dict = fun(trg_dict_path) + + settings.logger.info("src dict len : %d" % (len(settings.src_dict))) + + if settings.job_mode: + settings.slots = { + 'source_language_word': + integer_value_sequence(len(settings.src_dict)), + 'target_language_word': + integer_value_sequence(len(settings.trg_dict)), + 'target_language_next_word': + integer_value_sequence(len(settings.trg_dict)) + } + settings.logger.info("trg dict len : %d" % (len(settings.trg_dict))) + else: + settings.slots = { + 'source_language_word': + integer_value_sequence(len(settings.src_dict)), + 'sent_id': + integer_value_sequence(len(open(file_list[0], "r").readlines())) + } + + +def _get_ids(s, dictionary): + words = s.strip().split() + return [dictionary[START]] + \ + [dictionary.get(w, UNK_IDX) for w in words] + \ + [dictionary[END]] + + +@provider(init_hook=hook, pool_size=50000) +def process(settings, file_name): + with open(file_name, 'r') as f: + for line_count, line in enumerate(f): + line_split = line.strip().split('\t') + if settings.job_mode and len(line_split) != 2: + continue + src_seq = line_split[0] # one source sequence + src_ids = _get_ids(src_seq, settings.src_dict) + + if settings.job_mode: + trg_seq = line_split[1] # one target sequence + trg_words = trg_seq.split() + trg_ids = [settings.trg_dict.get(w, UNK_IDX) for w in trg_words] + + # remove sequence whose length > 80 in training mode + if len(src_ids) > 80 or len(trg_ids) > 80: + continue + trg_ids_next = trg_ids + [settings.trg_dict[END]] + trg_ids = [settings.trg_dict[START]] + trg_ids + yield { + 'source_language_word': src_ids, + 'target_language_word': trg_ids, + 'target_language_next_word': trg_ids_next + } + else: + yield {'source_language_word': src_ids, 'sent_id': [line_count]} diff --git a/machine_translation/gen.sh b/machine_translation/gen.sh new file mode 100755 index 00000000..140d5626 --- /dev/null +++ b/machine_translation/gen.sh @@ -0,0 +1,26 @@ +#!/bin/bash +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +set -e + +paddle train \ + --job=test \ + --config='seqToseq_net.py' \ + --save_dir='pretrained/wmt14_model' \ + --use_gpu=false \ + --num_passes=13 \ + --test_pass=12 \ + --trainer_count=1 \ + --config_args=is_generating=1,gen_trans_file="gen_result" \ + 2>&1 | tee 'gen.log' diff --git a/machine_translation/image/encoder_attention.PNG b/machine_translation/image/encoder_attention.PNG new file mode 100644 index 00000000..dde2fc60 Binary files /dev/null and b/machine_translation/image/encoder_attention.PNG differ diff --git a/machine_translation/image/encoder_attention.png b/machine_translation/image/encoder_attention.png new file mode 100644 index 00000000..dde2fc60 Binary files /dev/null and b/machine_translation/image/encoder_attention.png differ diff --git a/machine_translation/image/encoder_decoder.png b/machine_translation/image/encoder_decoder.png new file mode 100644 index 00000000..2bfc5d35 Binary files /dev/null and b/machine_translation/image/encoder_decoder.png differ diff --git a/machine_translation/image/gru.PNG b/machine_translation/image/gru.PNG new file mode 100644 index 00000000..d0a3e21c Binary files /dev/null and b/machine_translation/image/gru.PNG differ diff --git a/machine_translation/image/gru.png b/machine_translation/image/gru.png new file mode 100644 index 00000000..d0a3e21c Binary files /dev/null and b/machine_translation/image/gru.png differ diff --git a/machine_translation/image/nmt.png b/machine_translation/image/nmt.png new file mode 100644 index 00000000..0c8f6853 Binary files /dev/null and b/machine_translation/image/nmt.png differ diff --git a/machine_translation/moses_bleu.sh b/machine_translation/moses_bleu.sh new file mode 100755 index 00000000..2f230d7f --- /dev/null +++ b/machine_translation/moses_bleu.sh @@ -0,0 +1,18 @@ +#!/bin/bash +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +set -e +set -x +echo "Downloading multi-bleu.perl" +wget https://raw.githubusercontent.com/moses-smt/mosesdecoder/master/scripts/generic/multi-bleu.perl --no-check-certificate diff --git a/machine_translation/pretrained/wmt14_model.sh b/machine_translation/pretrained/wmt14_model.sh new file mode 100755 index 00000000..c4b55b90 --- /dev/null +++ b/machine_translation/pretrained/wmt14_model.sh @@ -0,0 +1,23 @@ +#!/bin/bash +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +set -e +set -x + +# download the pretrained model +wget http://paddlepaddle.bj.bcebos.com/model_zoo/wmt14_model.tar.gz + +# untar the model +tar -zxvf wmt14_model.tar.gz +rm wmt14_model.tar.gz diff --git a/machine_translation/seqToseq_net.py b/machine_translation/seqToseq_net.py new file mode 100644 index 00000000..ae2e0fbd --- /dev/null +++ b/machine_translation/seqToseq_net.py @@ -0,0 +1,168 @@ +# edit-mode: -*- python -*- + +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from paddle.trainer_config_helpers import * + +### Data Definiation +data_dir = "./data/pre-wmt14" +src_lang_dict = os.path.join(data_dir, 'src.dict') +trg_lang_dict = os.path.join(data_dir, 'trg.dict') +is_generating = get_config_arg("is_generating", bool, False) + +if not is_generating: + train_list = os.path.join(data_dir, 'train.list') + test_list = os.path.join(data_dir, 'test.list') +else: + train_list = None + test_list = os.path.join(data_dir, 'gen.list') + +define_py_data_sources2( + train_list, + test_list, + module="dataprovider", + obj="process", + args={ + "src_dict_path": src_lang_dict, + "trg_dict_path": trg_lang_dict, + "is_generating": is_generating + }) + +### Algorithm Configuration +settings( + learning_method = AdamOptimizer(), + batch_size = 50 if not is_generating else 1, + learning_rate = 5e-4 if not is_generating else 0) + +### Network Architecture +source_dict_dim = len(open(src_lang_dict, "r").readlines()) +target_dict_dim = len(open(trg_lang_dict, "r").readlines()) +word_vector_dim = 512 # dimension of word vector +decoder_size = 512 # dimension of hidden unit in GRU Decoder network +encoder_size = 512 # dimension of hidden unit in GRU Encoder network + +if is_generating: + beam_size=3 # expand width in beam search + max_length=250 # a stop condition of sequence generation + gen_trans_file = get_config_arg("gen_trans_file", str, None) + +#### Encoder +src_word_id = data_layer(name='source_language_word', size=source_dict_dim) +src_embedding = embedding_layer( + input=src_word_id, + size=word_vector_dim, + param_attr=ParamAttr(name='_source_language_embedding')) +src_forward = simple_gru(input=src_embedding, size=encoder_size) +src_backward = simple_gru( + input=src_embedding, size=encoder_size, reverse=True) +encoded_vector = concat_layer(input=[src_forward, src_backward]) + +with mixed_layer(size=decoder_size) as encoded_proj: + encoded_proj += full_matrix_projection(input=encoded_vector) + +backward_first = first_seq(input=src_backward) +with mixed_layer( + size=decoder_size, + act=TanhActivation(), ) as decoder_boot: + decoder_boot += full_matrix_projection(input=backward_first) + +#### Decoder +def gru_decoder_with_attention(enc_vec, enc_proj, current_word): + decoder_mem = memory( + name='gru_decoder', size=decoder_size, boot_layer=decoder_boot) + + context = simple_attention( + encoded_sequence=enc_vec, + encoded_proj=enc_proj, + decoder_state=decoder_mem, ) + + with mixed_layer(size=decoder_size * 3) as decoder_inputs: + decoder_inputs += full_matrix_projection(input=context) + decoder_inputs += full_matrix_projection(input=current_word) + + gru_step = gru_step_layer( + name='gru_decoder', + input=decoder_inputs, + output_mem=decoder_mem, + size=decoder_size) + + with mixed_layer( + size=target_dict_dim, bias_attr=True, + act=SoftmaxActivation()) as out: + out += full_matrix_projection(input=gru_step) + return out + +decoder_group_name = "decoder_group" +group_inputs = [ + StaticInput( + input=encoded_vector, is_seq=True), StaticInput( + input=encoded_proj, is_seq=True) +] + +if not is_generating: + trg_embedding = embedding_layer( + input=data_layer( + name='target_language_word', size=target_dict_dim), + size=word_vector_dim, + param_attr=ParamAttr(name='_target_language_embedding')) + group_inputs.append(trg_embedding) + + # For decoder equipped with attention mechanism, in training, + # target embeding (the groudtruth) is the data input, + # while encoded source sequence is accessed to as an unbounded memory. + # Here, the StaticInput defines a read-only memory + # for the recurrent_group. + decoder = recurrent_group( + name=decoder_group_name, + step=gru_decoder_with_attention, + input=group_inputs) + + lbl = data_layer(name='target_language_next_word', size=target_dict_dim) + cost = classification_cost(input=decoder, label=lbl) + outputs(cost) +else: + # In generation, the decoder predicts a next target word based on + # the encoded source sequence and the last generated target word. + + # The encoded source sequence (encoder's output) must be specified by + # StaticInput, which is a read-only memory. + # Embedding of the last generated word is automatically gotten by + # GeneratedInputs, which is initialized by a start mark, such as , + # and must be included in generation. + + + trg_embedding = GeneratedInput( + size=target_dict_dim, + embedding_name='_target_language_embedding', + embedding_size=word_vector_dim) + group_inputs.append(trg_embedding) + + beam_gen = beam_search( + name=decoder_group_name, + step=gru_decoder_with_attention, + input=group_inputs, + bos_id=0, + eos_id=1, + beam_size=beam_size, + max_length=max_length) + + seqtext_printer_evaluator( + input=beam_gen, + id_input=data_layer( + name="sent_id", size=1), + dict_file=trg_lang_dict, + result_file=gen_trans_file) + outputs(beam_gen) diff --git a/machine_translation/train.sh b/machine_translation/train.sh new file mode 100755 index 00000000..724f61fa --- /dev/null +++ b/machine_translation/train.sh @@ -0,0 +1,26 @@ +#!/bin/bash +# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +set -e + +paddle train \ +--config='seqToseq_net.py' \ +--save_dir='model' \ +--use_gpu=false \ +--num_passes=16 \ +--show_parameter_stats_period=100 \ +--trainer_count=4 \ +--log_period=10 \ +--dot_period=5 \ +2>&1 | tee 'train.log'