// Method to reduce, if possible the spans
void ReorderingStack::Reduce(Range current)
{
bool cont_loop = true;
while (cont_loop && m_stack.size() > 0) {
Range previous = m_stack.back();
if (current.GetStartPos() - previous.GetEndPos() == 1) { //mono&merge
m_stack.pop_back();
Range t(previous.GetStartPos(), current.GetEndPos());
current = t;
}
else if (previous.GetStartPos() - current.GetEndPos() == 1) { //swap&merge
m_stack.pop_back();
Range t(current.GetStartPos(), previous.GetEndPos());
current = t;
}
else { // discontinuous, no more merging
cont_loop = false;
}
} // finished reducing, exit
// add to stack
m_stack.push_back(current);
}
// Method to push (shift element into the stack and reduce if reqd)
int ReorderingStack::ShiftReduce(Range input_span)
{
int distance; // value to return: the initial distance between this and previous span
// stack is empty
if(m_stack.empty()) {
m_stack.push_back(input_span);
return input_span.GetStartPos() + 1; // - (-1)
}
// stack is non-empty
Range prev_span = m_stack.back(); //access last element added
//calculate the distance we are returning
if(input_span.GetStartPos() > prev_span.GetStartPos()) {
distance = input_span.GetStartPos() - prev_span.GetEndPos();
} else {
distance = input_span.GetEndPos() - prev_span.GetStartPos();
}
if(distance == 1) { //monotone
m_stack.pop_back();
Range new_span(prev_span.GetStartPos(), input_span.GetEndPos());
Reduce(new_span);
} else if(distance == -1) { //swap
m_stack.pop_back();
Range new_span(input_span.GetStartPos(), prev_span.GetEndPos());
Reduce(new_span);
} else { // discontinuous
m_stack.push_back(input_span);
}
return distance;
}
void ChartParser::Create(const Range &range, ChartParserCallback &to)
{
assert(m_decodeGraphList.size() == m_ruleLookupManagers.size());
std::vector <DecodeGraph*>::const_iterator iterDecodeGraph;
std::vector <ChartRuleLookupManager*>::const_iterator iterRuleLookupManagers = m_ruleLookupManagers.begin();
for (iterDecodeGraph = m_decodeGraphList.begin(); iterDecodeGraph != m_decodeGraphList.end(); ++iterDecodeGraph, ++iterRuleLookupManagers) {
const DecodeGraph &decodeGraph = **iterDecodeGraph;
assert(decodeGraph.GetSize() == 1);
ChartRuleLookupManager &ruleLookupManager = **iterRuleLookupManagers;
size_t maxSpan = decodeGraph.GetMaxChartSpan();
size_t last = m_source.GetSize()-1;
if (maxSpan != 0) {
last = min(last, range.GetStartPos()+maxSpan);
}
if (maxSpan == 0 || range.GetNumWordsCovered() <= maxSpan) {
const InputPath &inputPath = GetInputPath(range);
ruleLookupManager.GetChartRuleCollection(inputPath, last, to);
}
}
if (range.GetNumWordsCovered() == 1 && range.GetStartPos() != 0 && range.GetStartPos() != m_source.GetSize()-1) {
bool alwaysCreateDirectTranslationOption = StaticData::Instance().IsAlwaysCreateDirectTranslationOption();
if (to.Empty() || alwaysCreateDirectTranslationOption) {
// create unknown words for 1 word coverage where we don't have any trans options
const Word &sourceWord = m_source.GetWord(range.GetStartPos());
m_unknown.Process(sourceWord, range, to);
}
}
}
LRModel::ReorderingType
LRModel::
GetOrientation(Range const& prev, Range const& cur) const
{
UTIL_THROW_IF2(m_modelType == None, "No reordering model type specified");
return ((m_modelType == LeftRight)
? prev.GetEndPos() <= cur.GetStartPos() ? R : L
: (cur.GetStartPos() == prev.GetEndPos() + 1) ? M
: (m_modelType == Monotonic) ? NM
: (prev.GetStartPos() == cur.GetEndPos() + 1) ? S
: (m_modelType == MSD) ? D
: (cur.GetStartPos() > prev.GetEndPos()) ? DR : DL);
}
LRModel::ReorderingType
LRModel::
GetOrientation(Range const& prev, Range const& cur,
Bitmap const& cov) const
{
return ((m_modelType == LeftRight)
? cur.GetStartPos() > prev.GetEndPos() ? R : L
: IsMonotonicStep(prev,cur,cov) ? M
: (m_modelType == Monotonic) ? NM
: IsSwap(prev,cur,cov) ? S
: (m_modelType == MSD) ? D
: cur.GetStartPos() > prev.GetEndPos() ? DR : DL);
}
bool
IsSwap(Range const& prev, Range const& cur, Bitmap const& cov)
{
size_t s = prev.GetStartPos();
size_t e = cur.GetEndPos();
return (e+1 == s || (e < s && !cov.GetValue(s-1)));
}
float
DistortionScoreProducer::
CalculateDistortionScore(const Hypothesis& hypo,
const Range &prev, const Range &curr, const int FirstGap)
{
// if(!StaticData::Instance().UseEarlyDistortionCost()) {
if(!hypo.GetManager().options()->reordering.use_early_distortion_cost) {
return - (float) hypo.GetInput().ComputeDistortionDistance(prev, curr);
} // else {
/* Pay distortion score as soon as possible, from Moore and Quirk MT Summit 2007
Definitions:
S : current source range
S' : last translated source phrase range
S'' : longest fully-translated initial segment
*/
int prefixEndPos = (int)FirstGap-1;
if((int)FirstGap==-1)
prefixEndPos = -1;
// case1: S is adjacent to S'' => return 0
if ((int) curr.GetStartPos() == prefixEndPos+1) {
IFVERBOSE(4) std::cerr<< "MQ07disto:case1" << std::endl;
return 0;
}
// case2: S is to the left of S' => return 2(length(S))
if ((int) curr.GetEndPos() < (int) prev.GetEndPos()) {
IFVERBOSE(4) std::cerr<< "MQ07disto:case2" << std::endl;
return (float) -2*(int)curr.GetNumWordsCovered();
}
// case3: S' is a subsequence of S'' => return 2(nbWordBetween(S,S'')+length(S))
if ((int) prev.GetEndPos() <= prefixEndPos) {
IFVERBOSE(4) std::cerr<< "MQ07disto:case3" << std::endl;
int z = (int)curr.GetStartPos()-prefixEndPos - 1;
return (float) -2*(z + (int)curr.GetNumWordsCovered());
}
// case4: otherwise => return 2(nbWordBetween(S,S')+length(S))
IFVERBOSE(4) std::cerr<< "MQ07disto:case4" << std::endl;
return (float) -2*((int)curr.GetNumWordsBetween(prev) + (int)curr.GetNumWordsCovered());
}
bool
IsMonotonicStep(Range const& prev, // words range of last source phrase
Range const& cur, // words range of current source phrase
Bitmap const& cov) // coverage bitmap
{
size_t e = prev.GetEndPos() + 1;
size_t s = cur.GetStartPos();
return (s == e || (s >= e && !cov.GetValue(e)));
}
/// return orientation for the first phrase
LRModel::ReorderingType
LRModel::
GetOrientation(Range const& cur) const
{
UTIL_THROW_IF2(m_modelType == None, "Reordering Model Type is None");
return ((m_modelType == LeftRight) ? R :
(cur.GetStartPos() == 0) ? M :
(m_modelType == MSD) ? D :
(m_modelType == MSLR) ? DR : NM);
}
Phrase Phrase::GetSubString(const Range &range) const
{
Phrase retPhrase(range.GetNumWordsCovered());
for (size_t currPos = range.GetStartPos() ; currPos <= range.GetEndPos() ; currPos++) {
Word &word = retPhrase.AddWord();
word = GetWord(currPos);
}
return retPhrase;
}
Phrase Phrase::GetSubString(const Range &range, FactorType factorType) const
{
Phrase retPhrase(range.GetNumWordsCovered());
for (size_t currPos = range.GetStartPos() ; currPos <= range.GetEndPos() ; currPos++) {
const Factor* f = GetFactor(currPos, factorType);
Word &word = retPhrase.AddWord();
word.SetFactor(factorType, f);
}
return retPhrase;
}