word2vec源碼解析之word2vec.c
最近研究了一下google的開源項目word2vector，http://code.google.com/p/word2vec/。
其實這玩意算是神經網絡在文本挖掘的一項成功應用。
//下面是我對word2vec.c的註釋
//詳細算法可以參考論文，或者看這篇博客 http://www.cnblogs.com/downtjs/p/3784440.html
// Copyright 2013 Google Inc. 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.
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <pthread.h>
#define MAX_STRING 100
#define EXP_TABLE_SIZE 1000
#define MAX_EXP 6
#define MAX_SENTENCE_LENGTH 1000
#define MAX_CODE_LENGTH 40
const int vocab_hash_size = 30000000; // Maximum 30 * 0.7 = 21M words in the vocabulary
typedef float real; // Precision of float numbers
struct vocab_word
{
long long cn;//詞頻
int *point;//huffman編碼對應內節點的路徑
char *word, *code, codelen;//huffman編碼
};
char train_file[MAX_STRING], output_file[MAX_STRING];
char save_vocab_file[MAX_STRING], read_vocab_file[MAX_STRING];
struct vocab_word *vocab;
int binary = 0, cbow = 0, debug_mode = 2, window = 5, min_count = 5, num_threads = 1, min_reduce = 1;
int *vocab_hash;
long long vocab_max_size = 1000, vocab_size = 0, layer1_size = 100;
long long train_words = 0, word_count_actual = 0, file_size = 0, classes = 0;
real alpha = 0.025, starting_alpha, sample = 0;
real *syn0, *syn1, *syn1neg, *expTable;
clock_t start;
int hs = 1, negative = 0;
const int table_size = 1e8;
int *table;
//每個單詞的能量分佈表，table在負樣本抽樣中用到
void InitUnigramTable()
{
int a, i;
long long train_words_pow = 0;
real d1, power = 0.75;
table = (int *)malloc(table_size * sizeof(int));
for (a = 0; a < vocab_size; a++) //遍歷詞彙表，統計詞的能量總值train_words_pow，指數power應該是縮小值的吧。
train_words_pow += pow(vocab[a].cn, power);
i = 0;
d1 = pow(vocab[i].cn, power) / (real)train_words_pow;//表示已遍歷的詞的能量值佔總能力值的比例
for (a = 0; a < table_size; a++)//遍歷table。a表示table的位置，i表示詞彙表的位置
{
table[a] = i;//單詞i佔用table的a位置
//table反映的是一個單詞能量的分佈，一個單詞能量越大，所佔用的table的位置越多
if (a / (real)table_size > d1)
{
i++;//移動到下一個詞
d1 += pow(vocab[i].cn, power) / (real)train_words_pow;
}
if (i >= vocab_siInitNetze) i = vocab_size - 1;
}
}
// Reads a single word from a file, assuming space + tab + EOL to be word boundaries
//從文件中讀取一個詞
void ReadWord(char *word, FILE *fin) {
int a = 0, ch;
while (!feof(fin)) {
ch = fgetc(fin);
if (ch == 13) continue;
if ((ch == ' ') || (ch == '\t') || (ch == '\n')) {
if (a > 0) {
if (ch == '\n') ungetc(ch, fin);
break;
}
if (ch == '\n') {
strcpy(word, (char *)"</s>");
return;
} else continue;
}
word[a] = ch;
a++;
if (a >= MAX_STRING - 1) a--; // Truncate too long words
}
word[a] = 0;
}
// Returns hash value of a word返回一個詞的hash值，一個詞跟hash值一一對應（可能衝突）
int GetWordHash(char *word)
{
unsigned long long a, hash = 0;
for (a = 0; a < strlen(word); a++)
hash = hash * 257 + word[a];//採取257進制
hash = hash % vocab_hash_size;
return hash;
}
// Returns position of a word in the vocabulary; if the word is not found, returns -1
// 返回一個詞在詞彙表中的位置，如果不存在則返回-1
int SearchVocab(char *word)
{
unsigned int hash = GetWordHash(word);
while (1)
{
if (vocab_hash[hash] == -1) return -1;
if (!strcmp(word, vocab[vocab_hash[hash]].word))
return vocab_hash[hash];
hash = (hash + 1) % vocab_hash_size;
}
return -1;
}
// Reads a word and returns its index in the vocabulary
// 從文件流中讀取一個詞，並返回這個詞在詞彙表中的位置
int ReadWordIndex(FILE *fin)
{
char word[MAX_STRING];
ReadWord(word, fin);
if (feof(fin)) return -1;
return SearchVocab(word);
}
// Adds a word to the vocabulary 將一個詞添加到一個詞彙中
int AddWordToVocab(char *word)
{
unsigned int hash, length = strlen(word) + 1;
if (length > MAX_STRING)
length = MAX_STRING;
vocab[vocab_size].word = (char *)calloc(length, sizeof(char));
strcpy(vocab[vocab_size].word, word);
vocab[vocab_size].cn = 0;
vocab_size++;
// Reallocate memory if needed
if (vocab_size + 2 >= vocab_max_size)
{
vocab_max_size += 1000;
vocab = (struct vocab_word *)realloc(vocab, vocab_max_size * sizeof(struct vocab_word));
}
hash = GetWordHash(word);
while (vocab_hash[hash] != -1)//如果hash值衝突了
hash = (hash + 1) % vocab_hash_size;//使用開放地址法解決衝突
vocab_hash[hash] = vocab_size - 1;//由詞的hash值找到她所在詞彙表的排序位置
return vocab_size - 1;
}
// Used later for sorting by word counts
int VocabCompare(const void *a, const void *b)
{
return ((struct vocab_word *)b)->cn - ((struct vocab_word *)a)->cn;
}
// Sorts the vocabulary by frequency using word counts
// 根據詞頻排序
void SortVocab()
{
int a, size;
unsigned int hash;
// Sort the vocabulary and keep </s> at the first position
qsort(&vocab[1], vocab_size - 1, sizeof(struct vocab_word), VocabCompare);
for (a = 0; a < vocab_hash_size; a++)
vocab_hash[a] = -1;
size = vocab_size;
train_words = 0;
for (a = 0; a < size; a++)
{
// Words occuring less than min_count times will be discarded from the vocab
//出現太少的詞直接丟棄
if (vocab[a].cn < min_count)
{
vocab_size--;
free(vocab[vocab_size].word);
}
else
{
// Hash will be re-computed, as after the sorting it is not actual
// 重新計算hash查找。vocab_hash是由hash值找到該詞所在位置
hash=GetWordHash(vocab[a].word);
while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size;
vocab_hash[hash] = a;
train_words += vocab[a].cn;
}
}
vocab = (struct vocab_word *)realloc(vocab, (vocab_size + 1) * sizeof(struct vocab_word));
// Allocate memory for the binary tree construction
for (a = 0; a < vocab_size; a++)
{
vocab[a].code = (char *)calloc(MAX_CODE_LENGTH, sizeof(char));
vocab[a].point = (int *)calloc(MAX_CODE_LENGTH, sizeof(int));
}
}
// Reduces the vocabulary by removing infrequent tokens
// 再次移除詞頻過小的詞，縮減詞彙表
void ReduceVocab()
{
int a, b = 0;
unsigned int hash;
for (a = 0; a < vocab_size; a++)//我草，這很容易看錯啊
if (vocab[a].cn > min_reduce)
{
vocab[b].cn = vocab[a].cn;
vocab[b].word = vocab[a].word;
b++;
}
else free(vocab[a].word);
vocab_size = b;
for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
for (a = 0; a < vocab_size; a++) {
// Hash will be re-computed, as it is not actual
hash = GetWordHash(vocab[a].word);
while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size;
vocab_hash[hash] = a;
}
fflush(stdout);
min_reduce++;
}
// Create binary Huffman tree using the word counts根據詞頻創建huffman樹
// Frequent words will have short uniqe binary codes詞頻越大的單詞有越短的huffman編碼
void CreateBinaryTree() {
long long a, b, i, min1i, min2i, pos1, pos2, point[MAX_CODE_LENGTH];
char code[MAX_CODE_LENGTH];
long long *count = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
long long *binary = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
long long *parent_node = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long));
for (a = 0; a < vocab_size; a++) count[a] = vocab[a].cn;
for (a = vocab_size; a < vocab_size * 2; a++) count[a] = 1e15;
pos1 = vocab_size - 1;
pos2 = vocab_size;
// Following algorithm constructs the Huffman tree by adding one node at a time
for (a = 0; a < vocab_size - 1; a++)
{
// First, find two smallest nodes 'min1, min2' 找出目前權值最小的兩個節點
if (pos1 >= 0)//第一個權值最小的節點
{
if (count[pos1] < count[pos2])
{
min1i = pos1;
pos1--;
}
else
{
min1i = pos2;
pos2++;
}
}
else
{
min1i = pos2;
pos2++;
}
if (pos1 >= 0)//第二個權值最小的節點
{
if (count[pos1] < count[pos2])
{
min2i = pos1;
pos1--;
}
else
{
min2i = pos2;
pos2++;
}
}
else
{
min2i = pos2;
pos2++;
}
count[vocab_size + a] = count[min1i] + count[min2i];
parent_node[min1i] = vocab_size + a;
parent_node[min2i] = vocab_size + a;
binary[min2i] = 1;//節點編碼爲1，之前默認是0。
}
// Now assign binary code to each vocabulary word
for (a = 0; a < vocab_size; a++)
{
b = a;
i = 0;
while (1)
{
code[i] = binary[b];
point[i] = b;
i++;
b = parent_node[b];
if (b == vocab_size * 2 - 2) break;
}
vocab[a].codelen = i;
vocab[a].point[0] = vocab_size - 2;
for (b = 0; b < i; b++)
{
vocab[a].code[i - b - 1] = code[b];
vocab[a].point[i - b] = point[b] - vocab_size;
}
}
free(count);
free(binary);
free(parent_node);
}
//從分詞文件中統計每個單詞的詞頻
void LearnVocabFromTrainFile()
{
char word[MAX_STRING];
FILE *fin;
long long a, i;
for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1;
fin = fopen(train_file, "rb");
if (fin == NULL)
{
printf("ERROR: training data file not found!\n");
exit(1);
}
vocab_size = 0;
AddWordToVocab((char *)"</s>");
while (1)
{
ReadWord(word, fin);
if (feof(fin)) break;
train_words++;
if ((debug_mode > 1) && (train_words % 100000 == 0))
{
printf("%lldK%c", train_words / 1000, 13);
fflush(stdout);
}
i = SearchVocab(word);//返回該詞在詞彙表中的位置
if (i == -1)//該詞之前不存在
{
a = AddWordToVocab(word);//把該詞添加到詞彙表中
vocab[a].cn = 1;
}
else vocab[i].cn++;//更新詞頻
if (vocab_size > vocab_hash_size * 0.7)//如果詞彙表太龐大，就縮減
ReduceVocab();
}
SortVocab();//根據詞頻排序詞彙表
if (debug_mode > 0)
{
printf("Vocab size: %lld\n", vocab_size);
printf("Words in train file: %lld\n", train_words);
}
file_size = ftell(fin);
fclose(fin);
}
void SaveVocab() {
long long i;
FILE *fo = fopen(save_vocab_file, "wb");
for (i = 0; i < vocab_size; i++) fprintf(fo, "%s %lld\n", vocab[i].word, vocab[i].cn);
fclose(fo);
}
//從文件讀取詞彙，該文件已經統計好了每個詞彙的詞頻
void ReadVocab()
{
long long a, i = 0;
char c;
char word[MAX_STRING];
FILE *fin = fopen(read_vocab_file, "rb");//打開詞彙文件
if (fin == NULL)
{
printf("Vocabulary file not found\n");
exit(1);
}
for (a = 0; a < vocab_hash_size; a++)
vocab_hash[a] = -1;
vocab_size = 0;
while (1)
{
ReadWord(word, fin);//從fin進入一個詞到word中
if (feof(fin)) break;
a = AddWordToVocab(word);//把該詞添加到詞彙中，並返回該詞的位置
fscanf(fin, "%lld%c", &vocab[a].cn, &c);//讀取詞頻？c是幹啥的嗎，讀取空格嗎
i++;
}
SortVocab();//根據詞頻排序
if (debug_mode > 0)
{
printf("Vocab size: %lld\n", vocab_size);
printf("Words in train file: %lld\n", train_words);
}
//讀取訓練數據
fin = fopen(train_file, "rb");
if (fin == NULL)
{
printf("ERROR: training data file not found!\n");
exit(1);
}
fseek(fin, 0, SEEK_END);
file_size = ftell(fin);
fclose(fin);
}
void InitNet()
{
long long a, b;
a = posix_memalign((void **)&syn0, 128, (long long)vocab_size * layer1_size * sizeof(real));
//先知道這個也是申請動態數組，對齊還有128這個參數以後再瞭解
if (syn0 == NULL)
{
printf("Memory allocation failed\n"); exit(1);
}
if (hs)//採用softmax
{
a = posix_memalign((void **)&syn1, 128, (long long)vocab_size * layer1_size * sizeof(real));
if (syn1 == NULL)
{
printf("Memory allocation failed\n"); exit(1);
}
for (b = 0; b < layer1_size; b++)
for (a = 0; a < vocab_size; a++)
syn1[a * layer1_size + b] = 0;
}
if (negative>0)//還有負樣本
{
a = posix_memalign((void **)&syn1neg, 128, (long long)vocab_size * layer1_size * sizeof(real));
if (syn1neg == NULL)
{
printf("Memory allocation failed\n"); exit(1);
}
for (b = 0; b < layer1_size; b++)
for (a = 0; a < vocab_size; a++)
syn1neg[a * layer1_size + b] = 0;
}
for (b = 0; b < layer1_size; b++)
for (a = 0; a < vocab_size; a++)
syn0[a * layer1_size + b] = (rand() / (real)RAND_MAX - 0.5) / layer1_size;
CreateBinaryTree();//建立huffman樹，對每個單詞進行編碼
}
//這個線程函數執行之前，已經做好了一些工作：根據詞頻排序的詞彙表，每個單詞的huffman編碼
void *TrainModelThread(void *id)
{
long long a, b, d, word, last_word, sentence_length = 0, sentence_position = 0;
long long word_count = 0, last_word_count = 0, sen[MAX_SENTENCE_LENGTH + 1];
long long l1, l2, c, target, label;
unsigned long long next_random = (long long)id;
real f, g;
clock_t now;
real *neu1 = (real *)calloc(layer1_size, sizeof(real));
real *neu1e = (real *)calloc(layer1_size, sizeof(real));
FILE *fi = fopen(train_file, "rb");
//每個線程對應一段文本。根據線程id找到自己負責的文本的初始位置
fseek(fi, file_size / (long long)num_threads * (long long)id, SEEK_SET);
while (1)
{
if (word_count - last_word_count > 10000)
{
word_count_actual += word_count - last_word_count;
last_word_count = word_count;
if ((debug_mode > 1))
{
now=clock();
printf("%cAlpha: %f Progress: %.2f%% Words/thread/sec: %.2fk ", 13, alpha,
word_count_actual / (real)(train_words + 1) * 100,
word_count_actual / ((real)(now - start + 1) / (real)CLOCKS_PER_SEC * 1000));
fflush(stdout);
}
alpha = starting_alpha * (1 - word_count_actual / (real)(train_words + 1));
if (alpha < starting_alpha * 0.0001) alpha = starting_alpha * 0.0001;
}
if (sentence_length == 0)
{
while (1)
{
word = ReadWordIndex(fi);//從文件流中讀取一個詞，並返回這個詞在詞彙表中的位置
if (feof(fi)) break;
if (word == -1) continue;
word_count++;
if (word == 0) break;
// The subsampling randomly discards frequent words while keeping the ranking same
if (sample > 0)//對高頻詞進行下采樣，不過要保持排序不變。
{
real ran = (sqrt(vocab[word].cn / (sample * train_words)) + 1) * (sample * train_words) / vocab[word].cn;
next_random = next_random * (unsigned long long)25214903917 + 11;
if (ran < (next_random & 0xFFFF) / (real)65536) continue;
}
sen[sentence_length] = word;
sentence_length++;
//1000個單詞視作一個句子？
if (sentence_length >= MAX_SENTENCE_LENGTH) break;
}
sentence_position = 0;
}
if (feof(fi)) break;
if (word_count > train_words / num_threads) break;//如果當前線程已處理的單詞超過了 閾值，則退出。
word = sen[sentence_position];
if (word == -1) continue;
for (c = 0; c < layer1_size; c++) neu1[c] = 0;
for (c = 0; c < layer1_size; c++) neu1e[c] = 0;
next_random = next_random * (unsigned long long)25214903917 + 11;
b = next_random % window;
if (cbow)
{ //train the cbow architecture
// in -> hidden
for (a = b; a < window * 2 + 1 - b; a++) if (a != window)//掃描目標單詞的左右幾個單詞
{
c = sentence_position - window + a;
if (c < 0) continue;
if (c >= sentence_length) continue;
last_word = sen[c];
if (last_word == -1) continue;
for (c = 0; c < layer1_size; c++)//layer1_size詞向量的維度，默認值是100
neu1[c] += syn0[c + last_word * layer1_size];//傳說中的向量和？
}
if (hs) for (d = 0; d < vocab[word].codelen; d++)//開始遍歷huffman樹，每次一個節點
{
f = 0;
l2 = vocab[word].point[d] * layer1_size;//point應該記錄的是huffman的路徑。找到當前節點，並算出偏移
// Propagate hidden -> output
for (c = 0; c < layer1_size; c++) f += neu1[c] * syn1[c + l2];//計算內積
if (f <= -MAX_EXP) continue;//內積不在範圍內直接丟棄
else if (f >= MAX_EXP) continue;
else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))];//內積之後sigmoid函數
// 'g' is the gradient multiplied by the learning rate
g = (1 - vocab[word].code[d] - f) * alpha;//偏導數的一部分
//layer1_size是向量的維度
// Propagate errors output -> hidden 反向傳播誤差，從huffman樹傳到隱藏層。下面就是把當前內節點的誤差傳播給隱藏層，syn1[c + l2]是偏導數的一部分。
for (c = 0; c < layer1_size; c++) neu1e[c] += g * syn1[c + l2];
// Learn weights hidden -> output 更新當前內節點的向量，後面的neu1[c]其實是偏導數的一部分
for (c = 0; c < layer1_size; c++) syn1[c + l2] += g * neu1[c];
}
// NEGATIVE SAMPLING
if (negative > 0)
for (d = 0; d < negative + 1; d++)
{
if (d == 0)
{
target = word;//目標單詞
label = 1;//正樣本
}
else
{
next_random = next_random * (unsigned long long)25214903917 + 11;
target = table[(next_random >> 16) % table_size];
if (target == 0) target = next_random % (vocab_size - 1) + 1;
if (target == word) continue;
label = 0;//負樣本
}
l2 = target * layer1_size;
f = 0;
for (c = 0; c < layer1_size; c++)
f += neu1[c] * syn1neg[c + l2];//內積
if (f > MAX_EXP)
g = (label - 1) * alpha;
else if (f < -MAX_EXP)
g = (label - 0) * alpha;
else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha;
for (c = 0; c < layer1_size; c++)
neu1e[c] += g * syn1neg[c + l2];//隱藏層的誤差
for (c = 0; c < layer1_size; c++)
syn1neg[c + l2] += g * neu1[c];//更新負樣本向量
}
// hidden -> in
for (a = b; a < window * 2 + 1 - b; a++)
if (a != window)//cbow模型 更新的不是中間詞語的向量，而是周圍幾個詞語的向量。
{
c = sentence_position - window + a;
if (c < 0) continue;
if (c >= sentence_length) continue;
last_word = sen[c];
if (last_word == -1) continue;
for (c = 0; c < layer1_size; c++)
syn0[c + last_word * layer1_size] += neu1e[c];//更新詞向量
}
}
else
{ //train skip-gram
for (a = b; a < window * 2 + 1 - b; a++)
if (a != window)//掃描周圍幾個詞語
{
c = sentence_position - window + a;
if (c < 0) continue;
if (c >= sentence_length) continue;
last_word = sen[c];
if (last_word == -1) continue;
l1 = last_word * layer1_size;
for (c = 0; c < layer1_size; c++)
neu1e[c] = 0;
// HIERARCHICAL SOFTMAX
if (hs)
for (d = 0; d < vocab[word].codelen; d++)//遍歷葉子節點
{
f = 0;
l2 = vocab[word].point[d] * layer1_size;//point記錄的是huffman的路徑
// Propagate hidden -> output 感覺源代碼這個英語註釋有點誤導人，這裏的隱藏層就是輸入層，就是詞向量。
for (c = 0; c < layer1_size; c++)
f += syn0[c + l1] * syn1[c + l2];//計算兩個詞向量的內積
if (f <= -MAX_EXP) continue;
else if (f >= MAX_EXP) continue;
else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))];
// 'g' is the gradient multiplied by the learning rate
g = (1 - vocab[word].code[d] - f) * alpha;//偏導數的一部分
// Propagate errors output -> hidden
for (c = 0; c < layer1_size; c++)
neu1e[c] += g * syn1[c + l2];//隱藏層的誤差
// Learn weights hidden -> output
for (c = 0; c < layer1_size; c++)
syn1[c + l2] += g * syn0[c + l1];//更新葉子節點向量
}
// NEGATIVE SAMPLING
if (negative > 0)//這個同cobow差不多
for (d = 0; d < negative + 1; d++)
{
if (d == 0)
{
target = word;
label = 1;
}
else
{
next_random = next_random * (unsigned long long)25214903917 + 11;
target = table[(next_random >> 16) % table_size];
if (target == 0) target = next_random % (vocab_size - 1) + 1;
if (target == word) continue;
label = 0;
}
l2 = target * layer1_size;
f = 0;
for (c = 0; c < layer1_size; c++)
f += syn0[c + l1] * syn1neg[c + l2];
if (f > MAX_EXP) g = (label - 1) * alpha;
else if (f < -MAX_EXP)
g = (label - 0) * alpha;
else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha;
for (c = 0; c < layer1_size; c++)
neu1e[c] += g * syn1neg[c + l2];
for (c = 0; c < layer1_size; c++)
syn1neg[c + l2] += g * syn0[c + l1];
}
// Learn weights input -> hidden
for (c = 0; c < layer1_size; c++)
syn0[c + l1] += neu1e[c];//更新周圍幾個詞語的向量
}
}
sentence_position++;
if (sentence_position >= sentence_length)
{
sentence_length = 0;
continue;
}
}
fclose(fi);
free(neu1);
free(neu1e);
pthread_exit(NULL);
}
void TrainModel()
{
long a, b, c, d;
FILE *fo;
pthread_t *pt = (pthread_t *)malloc(num_threads * sizeof(pthread_t));
printf("Starting training using file %s\n", train_file);
starting_alpha = alpha;
if (read_vocab_file[0] != 0)
ReadVocab();//從文件讀入詞彙
else
LearnVocabFromTrainFile();//從訓練文件學習詞彙
if (save_vocab_file[0] != 0)
SaveVocab();//保存詞彙
if (output_file[0] == 0)
return;
InitNet();
if (negative > 0) InitUnigramTable();
start = clock();
for (a = 0; a < num_threads; a++) pthread_create(&pt[a], NULL, TrainModelThread, (void *)a);
for (a = 0; a < num_threads; a++) pthread_join(pt[a], NULL);
fo = fopen(output_file, "wb");
if (classes == 0) //不需要聚類，只需要輸出詞向量
{
// Save the word vectors
fprintf(fo, "%lld %lld\n", vocab_size, layer1_size);
for (a = 0; a < vocab_size; a++)
{
fprintf(fo, "%s ", vocab[a].word);
if (binary)
for (b = 0; b < layer1_size; b++)
fwrite(&syn0[a * layer1_size + b], sizeof(real), 1, fo);
else
for (b = 0; b < layer1_size; b++)
fprintf(fo, "%lf ", syn0[a * layer1_size + b]);
fprintf(fo, "\n");
}
}
else //使用k-means進行聚類
{
// Run K-means on the word vectors
int clcn = classes, iter = 10, closeid;
int *centcn = (int *)malloc(classes * sizeof(int));//該類別的數量
int *cl = (int *)calloc(vocab_size, sizeof(int));//詞到類別的映射
real closev, x;
real *cent = (real *)calloc(classes * layer1_size, sizeof(real));//質心數組
for (a = 0; a < vocab_size; a++)
cl[a] = a % clcn;//任意分類？
for (a = 0; a < iter; a++)
{
for (b = 0; b < clcn * layer1_size; b++)
cent[b] = 0;//質心清零
for (b = 0; b < clcn; b++)
centcn[b] = 1;
for (c = 0; c < vocab_size; c++)
{
for (d = 0; d < layer1_size; d++)
cent[layer1_size * cl[c] + d] += syn0[c * layer1_size + d];//求和放到質心數組中
centcn[cl[c]]++;//類別數量加1
}
for (b = 0; b < clcn; b++)//遍歷所有類別
{
closev = 0;
for (c = 0; c < layer1_size; c++)
{
cent[layer1_size * b + c] /= centcn[b];//均值，就是求新的質心
closev += cent[layer1_size * b + c] * cent[layer1_size * b + c];
}
closev = sqrt(closev);
for (c = 0; c < layer1_size; c++)
cent[layer1_size * b + c] /= closev;//對質心進行歸一化？
}
for (c = 0; c < vocab_size; c++)//對所有詞語重新分類
{
closev = -10;
closeid = 0;
for (d = 0; d < clcn; d++)
{
x = 0;
for (b = 0; b < layer1_size; b++)
x += cent[layer1_size * d + b] * syn0[c * layer1_size + b];//內積
if (x > closev)
{
closev = x;
closeid = d;
}
}
cl[c] = closeid;
}
}
// Save the K-means classes
for (a = 0; a < vocab_size; a++)
fprintf(fo, "%s %d\n", vocab[a].word, cl[a]);
free(centcn);
free(cent);
free(cl);
}
fclose(fo);
}
int ArgPos(char *str, int argc, char **argv)
{
int a;
for (a = 1; a < argc; a++) if (!strcmp(str, argv[a]))
{
if (a == argc - 1)
{
printf("Argument missing for %s\n", str);
exit(1);
}
return a;
}
return -1;
}
int main(int argc, char **argv) {
int i;
if (argc == 1) {
printf("WORD VECTOR estimation toolkit v 0.1b\n\n");
printf("Options:\n");
printf("Parameters for training:\n");
//輸入文件：已分詞的語料
printf("\t-train <file>\n");
printf("\t\tUse text data from <file> to train the model\n");
//輸出文件：詞向量或者詞聚類
printf("\t-output <file>\n");
printf("\t\tUse <file> to save the resulting word vectors / word clusters\n");
//詞向量的維度，默認值是100
printf("\t-size <int>\n");
printf("\t\tSet size of word vectors; default is 100\n");
//窗口大小，默認是5
printf("\t-window <int>\n");
printf("\t\tSet max skip length between words; default is 5\n");
//設定詞出現頻率的閾值，對於常出現的詞會被隨機下采樣
printf("\t-sample <float>\n");
printf("\t\tSet threshold for occurrence of words. Those that appear with higher frequency");
printf(" in the training data will be randomly down-sampled; default is 0 (off), useful value is 1e-5\n");
//是否採用softmax體系
printf("\t-hs <int>\n");
printf("\t\tUse Hierarchical Softmax; default is 1 (0 = not used)\n");
//負樣本的數量，默認是0，通常使用5-10。0表示不使用。
printf("\t-negative <int>\n");
printf("\t\tNumber of negative examples; default is 0, common values are 5 - 10 (0 = not used)\n");
//開啓的線程數量
printf("\t-threads <int>\n");
printf("\t\tUse <int> threads (default 1)\n");
//最小閾值。對於出現次數少於該值的詞，會被拋棄掉。
printf("\t-min-count <int>\n");
printf("\t\tThis will discard words that appear less than <int> times; default is 5\n");
//學習速率初始值，默認是0.025
printf("\t-alpha <float>\n");
printf("\t\tSet the starting learning rate; default is 0.025\n");
//輸出詞類別，而不是詞向量
printf("\t-classes <int>\n");
printf("\t\tOutput word classes rather than word vectors; default number of classes is 0 (vectors are written)\n");
//debug模式，默認是2，表示在訓練過程中會輸出更多信息
printf("\t-debug <int>\n");
printf("\t\tSet the debug mode (default = 2 = more info during training)\n");
//是否用binary模式保存數據，默認是0，表示否。
printf("\t-binary <int>\n");
printf("\t\tSave the resulting vectors in binary moded; default is 0 (off)\n");
//保存詞彙到這個文件
printf("\t-save-vocab <file>\n");
printf("\t\tThe vocabulary will be saved to <file>\n");
//詞彙從該文件讀取，而不是由訓練數據重組
printf("\t-read-vocab <file>\n");
printf("\t\tThe vocabulary will be read from <file>, not constructed from the training data\n");
//是否採用continuous bag of words算法。默認是0，表示採用另一個叫skip-gram的算法。
printf("\t-cbow <int>\n");
printf("\t\tUse the continuous bag of words model; default is 0 (skip-gram model)\n");
//工具使用樣例
printf("\nExamples:\n");
printf("./word2vec -train data.txt -output vec.txt -debug 2 -size 200 -window 5 -sample 1e-4 -negative 5 -hs 0 -binary 0 -cbow 1\n\n");
return 0;
}
output_file[0] = 0;
save_vocab_file[0] = 0;
read_vocab_file[0] = 0;
if ((i = ArgPos((char *)"-size", argc, argv)) > 0) layer1_size = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-train", argc, argv)) > 0) strcpy(train_file, argv[i + 1]);
if ((i = ArgPos((char *)"-save-vocab", argc, argv)) > 0) strcpy(save_vocab_file, argv[i + 1]);
if ((i = ArgPos((char *)"-read-vocab", argc, argv)) > 0) strcpy(read_vocab_file, argv[i + 1]);
if ((i = ArgPos((char *)"-debug", argc, argv)) > 0) debug_mode = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-binary", argc, argv)) > 0) binary = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-cbow", argc, argv)) > 0) cbow = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-alpha", argc, argv)) > 0) alpha = atof(argv[i + 1]);
if ((i = ArgPos((char *)"-output", argc, argv)) > 0) strcpy(output_file, argv[i + 1]);
if ((i = ArgPos((char *)"-window", argc, argv)) > 0) window = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-sample", argc, argv)) > 0) sample = atof(argv[i + 1]);
if ((i = ArgPos((char *)"-hs", argc, argv)) > 0) hs = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-negative", argc, argv)) > 0) negative = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-threads", argc, argv)) > 0) num_threads = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-min-count", argc, argv)) > 0) min_count = atoi(argv[i + 1]);
if ((i = ArgPos((char *)"-classes", argc, argv)) > 0) classes = atoi(argv[i + 1]);
vocab = (struct vocab_word *)calloc(vocab_max_size, sizeof(struct vocab_word));
vocab_hash = (int *)calloc(vocab_hash_size, sizeof(int));
expTable = (real *)malloc((EXP_TABLE_SIZE + 1) * sizeof(real));
for (i = 0; i < EXP_TABLE_SIZE; i++)
{
//expTable[i] = exp((i -500)/ 500 * 6) 即 e^-6 ~ e^6
expTable[i] = exp((i / (real)EXP_TABLE_SIZE * 2 - 1) * MAX_EXP); // Precompute the exp() table
//expTable[i] = 1/(1+e^6) ~ 1/(1+e^-6)即 0.01 ~ 1 的樣子
expTable[i] = expTable[i] / (expTable[i] + 1); // Precompute f(x) = x / (x + 1)
}
TrainModel();
return 0;
}
本文作者：linger
本文鏈接：http://blog.csdn.net/lingerlanlan/article/details/38232755