// file: $isip/class/stat/NGramModel/ngrm_05.cc
// version: $Id: ngrm_05.cc 8326 2002-07-10 16:23:24Z alphonso $
//

// isip include files
//
#include "NGramModel.h"

// method: load
//
// arguments:
//  Sof& sof: (input) input ngram source file
//  const int32& tag: (input) sof object instance name
//  const Vector<String> symbol_table: (input) input symbol table mapping index
//  const String& name: (input) sof object instance name
//
// return: a bool8 indicating status
//
// this method reads in the ngram from a file
//
bool8 NGramModel::load(Sof& sof_a, const int32& tag_a,
			 const Vector<String>& symbol_table_a,
			 const String& name_a) {

  // check the order of the language model
  //
  if ((int32)order_d < 1) {
    Console::put(L"*** order of the language model is zero ***");
    return false;
  }
  
  // store the symbol table for future use
  //
  symbol_table_d.assign(symbol_table_a);
  
  // parse the ngram file
  //
  NGramParser parser(order_d);
  if (!parser.load(sof_a, gram_hash_d, symbol_table_a, tag_a, name_a)) {
    return Error::handle(name(), L"load", Error::ARG, __FILE__, __LINE__);
  }

  // exit gracefully
  //
  return true;
}

// method: store
//
// arguments:
//  Sof& sof: (output) output ngram source file
//  const int32& tag: (input) sof object instance name
//  const String& name: (input) sof object instance name
//
// return: a bool8 indicating status
//
// this method store ngrams into a source ngram file
//
bool8 NGramModel::store(Sof& sof_a, 
			  const int32& tag_a, const String& name_a) const{

  // parse the ngram file
  //
  NGramParser parser(order_d);
  if (!parser.store(sof_a, gram_hash_d, symbol_table_d, tag_a, name_a)) {
    return Error::handle(name(), L"store", Error::ARG, __FILE__, __LINE__);
  }

  // exit gracefully
  //
  return true;
}
// method: getScore
//
// arguments:
//  const VectorLong& index: (input) SearchSymbol index sequence
//  
// return: a float_4 score represented by the ngram
//
// this method returns the score represented by the ngram give a
// SearchSymbol index sequence
//
float32 NGramModel::getScore(const VectorLong& index_a) {
  
  // define temporary variables used in function
  //
  int32 skipped = 0;
  float32 score = DEF_LM_SCORE;
  
  NGramNode* ngnode = (NGramNode*)NULL;
  NGramNode* pre_node = (NGramNode*)NULL;

  HashTable<Long, NGramNode>* hash_table = &gram_hash_d;
  int32 num_indices = index_a.length();

  // get the correct ngram node
  //
  for (int32 i = 0; i < num_indices; i++) {

    // store the prefix node for reading backoff score
    //
    pre_node = ngnode;
    
    // find the successive nodes
    //
    ngnode = hash_table->get(index_a(i));
    
    // if any of these ngram word sequences do not exist
    //
    if (ngnode == (NGramNode*)NULL) {
      skipped++;
      if (pre_node == (NGramNode*)NULL) {
	continue;
      }
      break;
    }

    // get the hash talbe of the next gram
    //
    hash_table = ngnode->getNextGram();
    
  } // end for loop

  // get lm score if ngram exists
  //
  if (ngnode != (NGramNode*)NULL) {
    score = ngnode->getLmScore();
  }
  
  // else, use back-off to get ngram score
  //  
  else if (pre_node != (NGramNode*)NULL) {

    // get the backoff score
    //
    score = pre_node->getBackoff();
    
    // add the lm score of the current word for a shorter history list
    //
    VectorLong shorter_index;
    shorter_index.shift(index_a, -skipped);
    shorter_index.setLength(num_indices - skipped);
    score += getScore(shorter_index);
  }

  // otherwise this word sequence is invalid
  //
  else {
    index_a.debug(L"index_a");
    Console::put(L"*** symbol sequence not found in language model ***");
  }
  
  // return score and exit gracefully
  //
  return score;
}