将word2vec bin文件转换为文本

Question

我可以从word2vec网站下载GoogleNews-vectors-
negative300.bin.gz。.bin文件（大约3.4GB）是对我无用的二进制格式。Tomas
Mikolov向我们保证：“将二进制格式转换为文本格式应该很简单（尽管这将占用更多的磁盘空间）。检查距离工具中的代码，读取二进制文件相当简单。”
不幸的是，我对C的了解不足，无法理解http://word2vec.googlecode.com/svn/trunk/distance.c。

据说gensim能做到这一点还，但我已经找到了教程似乎是有关转换 的 文本，而不是相反。

有人可以建议修改C代码或gensim发出文本的指令吗？

Answer 1

在word2vec-toolkit邮件列表上，Thomas
Mensink以小型C程序的形式提供了一个答案，该程序会将.bin文件转换为文本。这是distance.c文件的修改。我用下面的Thomas代码替换了原来的distance.c，并重建了word2vec（make
clean; make），并将编译后的距离重命名为readbin。然后./readbin vector.bin将创建vector.bin的文本版本。

//  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 <string.h>
#include <math.h>
#include <malloc.h>

const long long max_size = 2000;         // max length of strings
const long long N = 40;                  // number of closest words that will be shown
const long long max_w = 50;              // max length of vocabulary entries

int main(int argc, char **argv) {
  FILE *f;
  char file_name[max_size];
  float len;
  long long words, size, a, b;
  char ch;
  float *M;
  char *vocab;
  if (argc < 2) {
    printf("Usage: ./distance <FILE>\nwhere FILE contains word projections in the BINARY FORMAT\n");
    return 0;
  }
  strcpy(file_name, argv[1]);
  f = fopen(file_name, "rb");
  if (f == NULL) {
    printf("Input file not found\n");
    return -1;
  }
  fscanf(f, "%lld", &words);
  fscanf(f, "%lld", &size);
  vocab = (char *)malloc((long long)words * max_w * sizeof(char));
  M = (float *)malloc((long long)words * (long long)size * sizeof(float));
  if (M == NULL) {
    printf("Cannot allocate memory: %lld MB    %lld  %lld\n", (long long)words * size * sizeof(float) / 1048576, words, size);
    return -1;
  }
  for (b = 0; b < words; b++) {
    fscanf(f, "%s%c", &vocab[b * max_w], &ch);
    for (a = 0; a < size; a++) fread(&M[a + b * size], sizeof(float), 1, f);
    len = 0;
    for (a = 0; a < size; a++) len += M[a + b * size] * M[a + b * size];
    len = sqrt(len);
    for (a = 0; a < size; a++) M[a + b * size] /= len;
  }
  fclose(f);
  //Code added by Thomas Mensink
  //output the vectors of the binary format in text
  printf("%lld %lld #File: %s\n",words,size,file_name);
  for (a = 0; a < words; a++){
    printf("%s ",&vocab[a * max_w]);
    for (b = 0; b< size; b++){ printf("%f ",M[a*size + b]); }
    printf("\b\b\n");
  }

  return 0;
}

我从中删除了“ \ b \ b” printf。

顺便说一句，生成的文本文件仍然包含文本单词和一些不必要的空格，我不希望这些空格用于某些数值计算。我使用bash命令从每一行中删除了初始文本列和尾随空白。

cut --complement -d ' ' -f 1 GoogleNews-vectors-negative300.txt > GoogleNews-vectors-negative300_tuples-only.txt
sed 's/ $//' GoogleNews-vectors-negative300_tuples-only.txt