|
| 1 | +// Copyright 2013 Google Inc. All Rights Reserved. |
| 2 | +// |
| 3 | +// Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | +// you may not use this file except in compliance with the License. |
| 5 | +// You may obtain a copy of the License at |
| 6 | +// |
| 7 | +// http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | +// |
| 9 | +// Unless required by applicable law or agreed to in writing, software |
| 10 | +// distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | +// See the License for the specific language governing permissions and |
| 13 | +// limitations under the License. |
| 14 | + |
| 15 | +#include <stdio.h> |
| 16 | +#include <string.h> |
| 17 | +#include <math.h> |
| 18 | +#include <malloc.h> |
| 19 | + |
| 20 | +const long long max_size = 2000; // max length of strings |
| 21 | +const long long N = 40; // number of closest words that will be shown |
| 22 | +const long long max_w = 50; // max length of vocabulary entries |
| 23 | + |
| 24 | +int main(int argc, char **argv) { |
| 25 | + FILE *f; |
| 26 | + char st1[max_size]; |
| 27 | + char bestw[N][max_size]; |
| 28 | + char file_name[max_size], st[100][max_size]; |
| 29 | + float dist, len, bestd[N], vec[max_size]; |
| 30 | + long long words, size, a, b, c, d, cn, bi[100]; |
| 31 | + char ch; |
| 32 | + float *M; |
| 33 | + char *vocab; |
| 34 | + if (argc < 2) { |
| 35 | + printf("Usage: ./word-analogy <FILE>\nwhere FILE contains word projections in the BINARY FORMAT\n"); |
| 36 | + return 0; |
| 37 | + } |
| 38 | + strcpy(file_name, argv[1]); |
| 39 | + f = fopen(file_name, "rb"); |
| 40 | + if (f == NULL) { |
| 41 | + printf("Input file not found\n"); |
| 42 | + return -1; |
| 43 | + } |
| 44 | + fscanf(f, "%lld", &words); |
| 45 | + fscanf(f, "%lld", &size); |
| 46 | + vocab = (char *)malloc((long long)words * max_w * sizeof(char)); |
| 47 | + M = (float *)malloc((long long)words * (long long)size * sizeof(float)); |
| 48 | + if (M == NULL) { |
| 49 | + printf("Cannot allocate memory: %lld MB %lld %lld\n", (long long)words * size * sizeof(float) / 1048576, words, size); |
| 50 | + return -1; |
| 51 | + } |
| 52 | + for (b = 0; b < words; b++) { |
| 53 | + fscanf(f, "%s%c", &vocab[b * max_w], &ch); |
| 54 | + for (a = 0; a < size; a++) fread(&M[a + b * size], sizeof(float), 1, f); |
| 55 | + len = 0; |
| 56 | + for (a = 0; a < size; a++) len += M[a + b * size] * M[a + b * size]; |
| 57 | + len = sqrt(len); |
| 58 | + for (a = 0; a < size; a++) M[a + b * size] /= len; |
| 59 | + } |
| 60 | + fclose(f); |
| 61 | + while (1) { |
| 62 | + for (a = 0; a < N; a++) bestd[a] = 0; |
| 63 | + for (a = 0; a < N; a++) bestw[a][0] = 0; |
| 64 | + printf("Enter three words (EXIT to break): "); |
| 65 | + a = 0; |
| 66 | + while (1) { |
| 67 | + st1[a] = fgetc(stdin); |
| 68 | + if ((st1[a] == '\n') || (a >= max_size - 1)) { |
| 69 | + st1[a] = 0; |
| 70 | + break; |
| 71 | + } |
| 72 | + a++; |
| 73 | + } |
| 74 | + if (!strcmp(st1, "EXIT")) break; |
| 75 | + cn = 0; |
| 76 | + b = 0; |
| 77 | + c = 0; |
| 78 | + while (1) { |
| 79 | + st[cn][b] = st1[c]; |
| 80 | + b++; |
| 81 | + c++; |
| 82 | + st[cn][b] = 0; |
| 83 | + if (st1[c] == 0) break; |
| 84 | + if (st1[c] == ' ') { |
| 85 | + cn++; |
| 86 | + b = 0; |
| 87 | + c++; |
| 88 | + } |
| 89 | + } |
| 90 | + cn++; |
| 91 | + if (cn < 3) { |
| 92 | + printf("Only %lld words were entered.. three words are needed at the input to perform the calculation\n", cn); |
| 93 | + continue; |
| 94 | + } |
| 95 | + for (a = 0; a < cn; a++) { |
| 96 | + for (b = 0; b < words; b++) if (!strcmp(&vocab[b * max_w], st[a])) break; |
| 97 | + if (b == words) b = 0; |
| 98 | + bi[a] = b; |
| 99 | + printf("\nWord: %s Position in vocabulary: %lld\n", st[a], bi[a]); |
| 100 | + if (b == 0) { |
| 101 | + printf("Out of dictionary word!\n"); |
| 102 | + break; |
| 103 | + } |
| 104 | + } |
| 105 | + if (b == 0) continue; |
| 106 | + printf("\n Word Distance\n------------------------------------------------------------------------\n"); |
| 107 | + for (a = 0; a < size; a++) vec[a] = M[a + bi[1] * size] - M[a + bi[0] * size] + M[a + bi[2] * size]; |
| 108 | + len = 0; |
| 109 | + for (a = 0; a < size; a++) len += vec[a] * vec[a]; |
| 110 | + len = sqrt(len); |
| 111 | + for (a = 0; a < size; a++) vec[a] /= len; |
| 112 | + for (a = 0; a < N; a++) bestd[a] = 0; |
| 113 | + for (a = 0; a < N; a++) bestw[a][0] = 0; |
| 114 | + for (c = 0; c < words; c++) { |
| 115 | + if (c == bi[0]) continue; |
| 116 | + if (c == bi[1]) continue; |
| 117 | + if (c == bi[2]) continue; |
| 118 | + a = 0; |
| 119 | + for (b = 0; b < cn; b++) if (bi[b] == c) a = 1; |
| 120 | + if (a == 1) continue; |
| 121 | + dist = 0; |
| 122 | + for (a = 0; a < size; a++) dist += vec[a] * M[a + c * size]; |
| 123 | + for (a = 0; a < N; a++) { |
| 124 | + if (dist > bestd[a]) { |
| 125 | + for (d = N - 1; d > a; d--) { |
| 126 | + bestd[d] = bestd[d - 1]; |
| 127 | + strcpy(bestw[d], bestw[d - 1]); |
| 128 | + } |
| 129 | + bestd[a] = dist; |
| 130 | + strcpy(bestw[a], &vocab[c * max_w]); |
| 131 | + break; |
| 132 | + } |
| 133 | + } |
| 134 | + } |
| 135 | + for (a = 0; a < N; a++) printf("%50s\t\t%f\n", bestw[a], bestd[a]); |
| 136 | + } |
| 137 | + return 0; |
| 138 | +} |
0 commit comments