int main(int argc, char * argv[]){
FILE * input = stdin;
setlocale(LC_ALL, "");
GError * error = NULL;
GOptionContext * context;
context = g_option_context_new("- generate n-gram");
g_option_context_add_main_entries(context, entries, NULL);
if (!g_option_context_parse(context, &argc, &argv, &error)) {
g_print("option parsing failed:%s\n", error->message);
exit(EINVAL);
}
SystemTableInfo system_table_info;
bool retval = system_table_info.load(SYSTEM_TABLE_INFO);
if (!retval) {
fprintf(stderr, "load table.conf failed.\n");
exit(ENOENT);
}
PhraseLargeTable2 phrase_table;
/* init phrase table */
MemoryChunk * chunk = new MemoryChunk;
chunk->load(SYSTEM_PHRASE_INDEX);
phrase_table.load(chunk);
FacadePhraseIndex phrase_index;
const pinyin_table_info_t * phrase_files =
system_table_info.get_table_info();
if (!load_phrase_index(phrase_files, &phrase_index))
exit(ENOENT);
Bigram bigram;
bigram.attach(bigram_filename, ATTACH_CREATE|ATTACH_READWRITE);
char* linebuf = NULL; size_t size = 0;
phrase_token_t last_token, cur_token = last_token = 0;
while( getline(&linebuf, &size, input) ){
if ( feof(input) )
break;
if ( '\n' == linebuf[strlen(linebuf) - 1] ) {
linebuf[strlen(linebuf) - 1] = '\0';
}
TAGLIB_PARSE_SEGMENTED_LINE(&phrase_index, token, linebuf);
last_token = cur_token;
cur_token = token;
/* skip null_token in second word. */
if ( null_token == cur_token )
continue;
/* training uni-gram */
phrase_index.add_unigram_frequency(cur_token, 1);
/* skip pi-gram training. */
if ( null_token == last_token ){
if ( !train_pi_gram )
continue;
last_token = sentence_start;
}
/* train bi-gram */
SingleGram * single_gram = NULL;
bigram.load(last_token, single_gram);
if ( NULL == single_gram ){
single_gram = new SingleGram;
}
guint32 freq, total_freq;
/* increase freq */
if (single_gram->get_freq(cur_token, freq))
assert(single_gram->set_freq(cur_token, freq + 1));
else
assert(single_gram->insert_freq(cur_token, 1));
/* increase total freq */
single_gram->get_total_freq(total_freq);
single_gram->set_total_freq(total_freq + 1);
bigram.store(last_token, single_gram);
delete single_gram;
}
free(linebuf);
if (!save_phrase_index(phrase_files, &phrase_index))
exit(ENOENT);
return 0;
}
bool read_document(PhraseLargeTable2 * phrase_table,
FacadePhraseIndex * phrase_index,
FILE * document,
HashofDocument hash_of_document,
HashofUnigram hash_of_unigram){
char * linebuf = NULL;size_t size = 0;
phrase_token_t last_token, cur_token = last_token = 0;
while ( getline(&linebuf, &size, document) ){
if ( feof(document) )
break;
if ( '\n' == linebuf[strlen(linebuf) - 1] ) {
linebuf[strlen(linebuf) - 1] = '\0';
}
TAGLIB_PARSE_SEGMENTED_LINE(phrase_index, token, linebuf);
last_token = cur_token;
cur_token = token;
/* skip null_token in second word. */
if ( null_token == cur_token )
continue;
gpointer value = NULL;
gboolean lookup_result = g_hash_table_lookup_extended
(hash_of_unigram, GUINT_TO_POINTER(cur_token),
NULL, &value);
if ( !lookup_result ){
g_hash_table_insert(hash_of_unigram, GUINT_TO_POINTER(cur_token),
GUINT_TO_POINTER(1));
} else {
guint32 freq = GPOINTER_TO_UINT(value);
freq ++;
g_hash_table_insert(hash_of_unigram, GUINT_TO_POINTER(cur_token),
GUINT_TO_POINTER(freq));
}
/* skip pi-gram training. */
if ( null_token == last_token ){
if ( !g_train_pi_gram )
continue;
last_token = sentence_start;
}
/* remember the (last_token, cur_token) word pair. */
HashofSecondWord hash_of_second_word = NULL;
lookup_result = g_hash_table_lookup_extended
(hash_of_document, GUINT_TO_POINTER(last_token),
NULL, &value);
if ( !lookup_result ){
hash_of_second_word = g_hash_table_new
(g_direct_hash, g_direct_equal);
} else {
hash_of_second_word = (HashofSecondWord) value;
}
value = NULL;
lookup_result = g_hash_table_lookup_extended
(hash_of_second_word, GUINT_TO_POINTER(cur_token),
NULL, &value);
guint32 count = 0;
if ( lookup_result ) {
count = GPOINTER_TO_UINT(value);
}
count ++;
g_hash_table_insert(hash_of_second_word,
GUINT_TO_POINTER(cur_token),
GUINT_TO_POINTER(count));
g_hash_table_insert(hash_of_document,
GUINT_TO_POINTER(last_token),
hash_of_second_word);
}
free(linebuf);
return true;
}
int main(int argc, char * argv[]){
int i = 1;
bool train_pi_gram = true;
const char * bigram_filename = "deleted_bigram.db";
setlocale(LC_ALL, "");
while ( i < argc ){
if ( strcmp("--help", argv[i]) == 0){
print_help();
exit(0);
} else if ( strcmp("--skip-pi-gram-training", argv[i]) == 0 ){
train_pi_gram = false;
} else if ( strcmp("--deleted-bigram-file", argv[i]) == 0){
if ( ++i >= argc ) {
print_help();
exit(EINVAL);
}
bigram_filename = argv[i];
} else {
print_help();
exit(EINVAL);
}
++i;
}
/* load phrase table. */
PhraseLargeTable2 phrase_table;
MemoryChunk * new_chunk = new MemoryChunk;
new_chunk->load("phrase_index.bin");
phrase_table.load(new_chunk);
FacadePhraseIndex phrase_index;
if (!load_phrase_index(&phrase_index))
exit(ENODATA);
Bigram bigram;
bigram.attach(bigram_filename, ATTACH_CREATE|ATTACH_READWRITE);
char* linebuf = NULL; size_t size = 0;
phrase_token_t last_token, cur_token = last_token = 0;
while( getline(&linebuf, &size, stdin) ){
if ( feof(stdin) )
break;
if ( '\n' == linebuf[strlen(linebuf) - 1] ) {
linebuf[strlen(linebuf) - 1] = '\0';
}
TAGLIB_PARSE_SEGMENTED_LINE(&phrase_index, token, linebuf);
last_token = cur_token;
cur_token = token;
/* skip null_token in second word. */
if ( null_token == cur_token )
continue;
/* skip pi-gram training. */
if ( null_token == last_token ){
if ( !train_pi_gram )
continue;
last_token = sentence_start;
}
/* train bi-gram */
SingleGram * single_gram = NULL;
bigram.load(last_token, single_gram);
if ( NULL == single_gram ){
single_gram = new SingleGram;
}
guint32 freq, total_freq;
//increase freq
if (single_gram->get_freq(cur_token, freq))
assert(single_gram->set_freq(cur_token, freq + 1));
else
assert(single_gram->insert_freq(cur_token, 1));
//increase total freq
single_gram->get_total_freq(total_freq);
single_gram->set_total_freq(total_freq + 1);
bigram.store(last_token, single_gram);
delete single_gram;
}
free(linebuf);
return 0;
}