コード例 #1
0
ファイル: SWE_Test_SynSel.cpp プロジェクト: jiangfeng1124/SWE
int main (int argc, char *argv[])
{		
	if (argc < 4)
	{
		printf("SemWE_Test_SynSel synonym_question word_embed synonym_result candidate_num\n");
		printf("Primary Designed for the TOEFL Synonym Selection Task\n");
		exit(1);
	}
	
	strcpy(word_pair, argv[1]);
	strcpy(word_embed, argv[2]);
	strcpy(sim_result, argv[3]);
	candidate_num = atoi(argv[4]);
				
	char tmp_word[WORD_LEN];
	float tmp_value = 0.0;
	
	printf(">> Synonym Selection on Word Embedding Model\n");
	
	FILE *fTEST = fopen(word_pair, "r");
	if (fTEST == NULL)
	{
		printf(">> Error, can not open file %s\n", word_pair);
		exit(1);
	}
	FILE *fRES = fopen(sim_result, "w");
	if (fRES == NULL)
	{
		printf(">> Error, can not open file %s\n", sim_result);
		exit(1);
	}

	printf("--- Load Word Embedding from: %s\n", word_embed);
	FILE *fEMB = fopen(word_embed, "r");
	fscanf(fEMB, "%d%d", &word_num, &vect_dim);
	printf("--- word num: %d\n--- vec dimen: %d\n", word_num, vect_dim);
	wordEmbed = (real*)malloc(sizeof(real)*word_num*vect_dim);
	
	for (int i = 0; i < word_num; i++)
	{
		fscanf(fEMB, "%s", tmp_word);		
		for (int j = 0; j < vect_dim; j++)
		{
			fscanf(fEMB, "%f", &tmp_value);
			wordEmbed[i*vect_dim+j] = tmp_value;
			//tmp_vector.push_back(tmp_value);
		}
		wordMapper.insert(make_pair(tmp_word, i));
		wordSet.insert(make_pair(tmp_word, i));
	}
	/*if (wordMapper.size() != word_num)
	{
	}*/
	fclose(fEMB);
	printf("--- Load finish\n");
				
	//printf(">> Synonym Selection\n");
	printf("--- Synonym question : %s\n", word_pair);
	printf("--- Candidate num: %d\n", candidate_num);
	printf("--- Word Embed: %s\n", word_embed);	
	printf("--- Selection result: %s\n", sim_result);
	clock_t start = clock();
	WordSim tmp_sim;
	
	char word_A[1024];
	char word_B[1024];
	
	int all_num = 0;
	int use_num = 0;
	int cand_id = 0;
	while (fscanf(fTEST, "%s", word_A) != EOF)
	{
		//printf("--- %s %s: ", word_A, word_B);
		all_num++;
		
		fprintf(fRES, "%s\t", word_A);

		float max_sim = -10000.0;
		string sel_synword = "";
		// candidate
		for (cand_id = 0; cand_id < candidate_num; cand_id++)
		{
			fscanf(fTEST, "%s", word_B);
			if (cand_id == 0){
				sel_synword = word_B;
			}
			float sim_AB = -100.0;

			if (wordSet.find(word_A) == wordSet.end() || wordSet.find(word_B) == wordSet.end())
			{
				;
			}
			else{
				int index_A = wordSet[word_A];
				int index_B = wordSet[word_B];
				
				
				sim_AB = SemWE_CalcCosine(&wordEmbed[index_A*vect_dim], &wordEmbed[index_B*vect_dim], vect_dim);
				if (sim_AB > max_sim)
				{
					max_sim = sim_AB;
					sel_synword = word_B;
				}				
			}	
			fprintf(fRES, "%s(%.6f) ", word_B, sim_AB);
		}
		fprintf(fRES, "=>best: %s\n", sel_synword.c_str());
		//fprintf(fRES, "%s\t%s", sel_synword.c_str());
	}
	fclose(fRES);
	fclose(fTEST);
	//////////////////////////////////////////////////////////////////////////
	double timeCost = (clock()-start)/CLOCKS_PER_SEC;		
	//printf("--- calculate nums: %d (/%d)\n", use_num, all_num);
	printf("--- elapsed time: %f\n", timeCost);		
	//printf(">> Finish.\n");
	
	free(wordEmbed); wordEmbed = NULL;	
	return 0;
}
コード例 #2
0
ファイル: SWE_Test_WordSim.cpp プロジェクト: MorLong/SWE
int main (int argc, char *argv[])
{		
	if (argc < 4)
	{
		printf("SemWE_WordSim.exe word_pair word_embed sim_result distance_ID\n");
		printf("distance_ID: cosine; euclidean\n");
		exit(1);
	}
	
	strcpy(word_pair, argv[1]);
	strcpy(word_embed, argv[2]);
	strcpy(sim_result, argv[3]);
	distance_ID = argv[4];
				
	char tmp_word[WORD_LEN];
	float tmp_value = 0.0;

	printf(">> Word Similarity Calculation on Word Embedding Model\n");
	printf("--- distance flag: %s\n", distance_ID.c_str());

	FILE *fTEST = fopen(word_pair, "r");
	if (fTEST == NULL)
	{
		printf(">> Error, can not open file %s\n", word_pair);
		exit(1);
	}
	FILE *fRES = fopen(sim_result, "w");
	if (fRES == NULL)
	{
		printf(">> Error, can not open file %s\n", sim_result);
		exit(1);
	}

	printf(">> Load Word Embedding from: %s\n", word_embed);
	FILE *fEMB = fopen(word_embed, "r");
	fscanf(fEMB, "%d%d", &word_num, &vect_dim);
	printf("--- word num: %d\n--- vec dimen: %d\n", word_num, vect_dim);
	wordEmbed = (real*)malloc(sizeof(real)*word_num*vect_dim);

	for (int i = 0; i < word_num; i++)
	{
		fscanf(fEMB, "%s", tmp_word);		
		for (int j = 0; j < vect_dim; j++)
		{
			fscanf(fEMB, "%f", &tmp_value);
			wordEmbed[i*vect_dim+j] = tmp_value;
			//tmp_vector.push_back(tmp_value);
		}
		wordMapper.insert(make_pair(tmp_word, i));
		wordSet.insert(make_pair(tmp_word, i));
	}
	/*if (wordMapper.size() != word_num)
	{
	}*/
	fclose(fEMB);
	printf("--- Load finish\n");
				
	printf(">> Calculate Word Similarity: %s\n", distance_ID.c_str());
	printf("--- word pair : %s\n", word_pair);
	printf("--- word embed: %s\n", word_embed);	
	printf("--- calc result: %s\n", sim_result);
	clock_t start = clock();
	WordSim tmp_sim;
	
	char word_A[1024];
	char word_B[1024];
	
	int all_num = 0;
	int use_num = 0;
	while (fscanf(fTEST, "%s%s", word_A, word_B) != EOF)
	{
		//printf("--- %s %s: ", word_A, word_B);
		all_num++;
		float sim_AB = 0.0;
		if (wordSet.find(word_A) == wordSet.end() || wordSet.find(word_B) == wordSet.end())
		{
			;
		}
		else{
			use_num++;
			int index_A = wordSet[word_A];
			int index_B = wordSet[word_B];
			if (distance_ID == "cosine")
			{
				sim_AB = SemWE_CalcCosine(&wordEmbed[index_A*vect_dim], &wordEmbed[index_B*vect_dim], vect_dim);
			}	
			if (distance_ID == "euclidean")
			{
				sim_AB = SemWE_CalcEuclidean(&wordEmbed[index_A*vect_dim], &wordEmbed[index_B*vect_dim], vect_dim);				
			}
			fprintf(fRES, "%s %s %.6f\n", word_A, word_B, sim_AB);
		}		
	}
	fclose(fRES);
	fclose(fTEST);
	//////////////////////////////////////////////////////////////////////////
	double timeCost = (clock()-start)/CLOCKS_PER_SEC;		
	printf("--- calculate nums: %d (/%d)\n", use_num, all_num);
	printf("--- elapsed time: %f\n", timeCost);		
	//printf(">> Finish.\n");
	
	free(wordEmbed); wordEmbed = NULL;	
	return 0;
}