// Function to constituency parse all sentence in the given input file
// and write parsed sentences to the given output file
extern "C" void parse_file(const char *sInputFile, const char *sOutputFile, bool tokenize)
{
std::cerr << "Processing file " << sInputFile << std::endl;
// initialize the input reader
CSentenceReader input_reader(sInputFile);
// open the output file
FILE *outfp = NULL;
outfp = fopen(sOutputFile, "w");
// initialize the temporary sentence variables
CStringVector tokenized_sent[1];
CTwoStringVector tagged_sent[1];
english::CCFGTree parsed_sent[1];
// get the tagger and the parser that were stored earlier
CTagger *tagger = (CTagger *)zpm->tagger;
CConParser *conparser = (CConParser *)zpm->conparser;
// read in and tokenize the given input file if asked
bool readSomething;
if (tokenize) {
readSomething = input_reader.readSegmentedSentenceAndTokenize(tokenized_sent);
}
else {
readSomething = input_reader.readSegmentedSentence(tokenized_sent);
}
while ( readSomething )
{
if ( tokenized_sent->back() == "\n" )
{
tokenized_sent->pop_back();
}
std::string parse = "";
if(tokenized_sent->size() < MAX_SENTENCE_SIZE){
tagger->tag(tokenized_sent, tagged_sent);
conparser->parse(*tagged_sent, parsed_sent);
parse = parsed_sent->str_unbinarized();
} else {
std::cerr << "Sentence too long. Writing empty string. Sentence: " << tokenized_sent << std::endl;
}
fprintf(outfp, "%s\n", parse.c_str());
if (tokenize) {
readSomething = input_reader.readSegmentedSentenceAndTokenize(tokenized_sent);
}
else {
readSomething = input_reader.readSegmentedSentence(tokenized_sent);
}
}
// close the output file
std::cerr << "Wrote output to " << sOutputFile << std::endl;
fclose(outfp);
}
void process(const std::string &sInputFile, const std::string &sOutputFile){
std::cerr<<"Converting start ..." << std::endl;
std::ifstream is(sInputFile.c_str());
ASSERT(is.is_open(), "The training file is unaccessible.");
std::ofstream os(sOutputFile.c_str());
static CSentenceParsed sent;
int nCount=0;
CConParser parser;
CCoNLLOutput o_conll;
is >> sent;
while( ! sent.empty() ) {
std::cerr << "Sentence " << nCount << " ... ";
nCount ++;
parser.convert( sent, &o_conll );
os<<o_conll;
os.flush();
std::cerr << "done." << std::endl;
is >> sent;
}
is.close();
os.close();
std::cerr << "Done. " << std::endl;
}
// Function to constituency parse a sentence
extern "C" char* parse_sentence(void* vzps, const char *input_sentence, bool tokenize)
{
zparSession_t* zps = static_cast<zparSession_t *>(vzps);
try {
// create a temporary string stream from the input char *
CSentenceReader input_reader(std::string(input_sentence), false);
// tokenize the sentence
CStringVector tokenized_sent[1];
if (tokenize) {
input_reader.readSegmentedSentenceAndTokenize(tokenized_sent);
}
else {
input_reader.readSegmentedSentence(tokenized_sent);
}
if (zps->output_buffer != NULL) {
delete zps->output_buffer;
zps->output_buffer = NULL;
}
if(tokenized_sent->size() >= MAX_SENTENCE_SIZE){
// The ZPar code asserts that length < MAX_SENTENCE_SIZE...
std::cerr << "Sentence too long. Returning empty string. Sentence: " << input_sentence << std::endl;
zps->output_buffer = new char[1];
strcpy(zps->output_buffer, "");
} else {
// initialize the variables that will hold the tagged and parsed sentences
CTwoStringVector tagged_sent[1];
english::CCFGTree parsed_sent[1];
// get the tagger and parser that were stored earlier
CTagger *tagger = zps->tagger;
CConParser *conparser = zps->conparser;
// tag and parse the sentence
tagger->tag(tokenized_sent, tagged_sent);
conparser->parse(*tagged_sent, parsed_sent);
// now put the parsed sentence into a string stream
std::string parse = parsed_sent->str_unbinarized();
int parselen = parse.length();
zps->output_buffer = new char[parselen + 1];
strcpy(zps->output_buffer, parse.c_str());
}
} catch (const std::string &e) {
std::cerr << e << std::endl;
zps->output_buffer = new char[1];
strcpy(zps->output_buffer, "");
}
return zps->output_buffer;
}
extern "C" void parse_tagged_file(void* vzps, const char *sInputFile, const char *sOutputFile, const char seperator='/')
{
zparSession_t* zps = static_cast<zparSession_t *>(vzps);
std::cerr << "Processing file " << sInputFile << std::endl;
// initialize the input reader
CSentenceReader input_reader(sInputFile);
// open the output file
FILE *outfp = NULL;
outfp = fopen(sOutputFile, "w");
// initialize the temporary sentence variables
CTwoStringVector tagged_sent[1];
english::CCFGTree parsed_sent[1];
// get the parser that was stored earlier
CConParser *conparser = zps->conparser;
// read in and tokenize the given input file if asked
bool readSomething;
readSomething = input_reader.readTaggedSentence(tagged_sent, false, seperator);
while ( readSomething )
{
std::string parse = "";
if(tagged_sent->size() < MAX_SENTENCE_SIZE){
conparser->parse(*tagged_sent, parsed_sent);
parse = parsed_sent->str_unbinarized();
} else {
std::cerr << "Sentence too long. Writing empty string. Sentence: " << tagged_sent << std::endl;
}
fprintf(outfp, "%s\n", parse.c_str());
readSomething = input_reader.readTaggedSentence(tagged_sent, false, seperator);
}
// close the output file
std::cerr << "Wrote output to " << sOutputFile << std::endl;
fclose(outfp);
}
