random incomplete metric stuff, including string subsequence kernel impl

6 files changed, 1013 insertions, 0 deletions

diff --git a/mteval/ns.cc b/mteval/ns.cc
new file mode 100644
index 00000000..1045a51f
--- /dev/null
+++ b/mteval/ns.cc
@@ -0,0 +1,241 @@
+#include "ns.h"
+#include "ns_ter.h"
+
+#include <cassert>
+#include <cmath>
+#include <cstdlib>
+#include <iostream>
+#include <sstream>
+
+using namespace std;
+using boost::shared_ptr;
+
+map<string, EvaluationMetric*> EvaluationMetric::instances_;
+
+SegmentEvaluator::~SegmentEvaluator() {}
+EvaluationMetric::~EvaluationMetric() {}
+
+struct DefaultSegmentEvaluator : public SegmentEvaluator {
+  DefaultSegmentEvaluator(const vector<vector<WordID> >& refs, const EvaluationMetric* em) : refs_(refs), em_(em) {}
+  void Evaluate(const vector<WordID>& hyp, SufficientStats* out) const {
+    em_->ComputeSufficientStatistics(hyp, refs_, out);
+  }
+  const vector<vector<WordID> > refs_;
+  const EvaluationMetric* em_;
+};
+
+shared_ptr<SegmentEvaluator> EvaluationMetric::CreateSegmentEvaluator(const vector<vector<WordID> >& refs) const {
+  return shared_ptr<SegmentEvaluator>(new DefaultSegmentEvaluator(refs, this));
+}
+
+void EvaluationMetric::ComputeSufficientStatistics(const vector<WordID>&,
+                                                   const vector<vector<WordID> >&,
+                                                   SufficientStats*) const {
+  cerr << "Base class ComputeSufficientStatistics should not be called.\n";
+  abort();
+}
+
+enum BleuType { IBM, Koehn, NIST };
+template <unsigned int N = 4u, BleuType BrevityType = IBM>
+struct BleuSegmentEvaluator : public SegmentEvaluator {
+  BleuSegmentEvaluator(const vector<vector<WordID> >& refs, const EvaluationMetric* em) : evaluation_metric(em) {
+    assert(refs.size() > 0);
+    float tot = 0;
+    int smallest = 9999999;
+    for (vector<vector<WordID> >::const_iterator ci = refs.begin();
+         ci != refs.end(); ++ci) {
+      lengths_.push_back(ci->size());
+      tot += lengths_.back();
+      if (lengths_.back() < smallest) smallest = lengths_.back();
+      CountRef(*ci);
+    }
+    if (BrevityType == Koehn)
+      lengths_[0] = tot / refs.size();
+    if (BrevityType == NIST)
+      lengths_[0] = smallest;
+  }
+
+  void Evaluate(const vector<WordID>& hyp, SufficientStats* out) const {
+    out->fields.resize(N + N + 2);
+    out->evaluation_metric = evaluation_metric;
+    for (unsigned i = 0; i < N+N+2; ++i) out->fields[i] = 0;
+
+    ComputeNgramStats(hyp, &out->fields[0], &out->fields[N], true);
+    float& hyp_len = out->fields[2*N];
+    float& ref_len = out->fields[2*N + 1];
+    hyp_len = hyp.size();
+    ref_len = lengths_[0];
+    if (lengths_.size() > 1 && BrevityType == IBM) {
+      float bestd = 2000000;
+      float hl = hyp.size();
+      float bl = -1;
+      for (vector<float>::const_iterator ci = lengths_.begin(); ci != lengths_.end(); ++ci) {
+        if (fabs(*ci - hl) < bestd) {
+          bestd = fabs(*ci - hl);
+          bl = *ci;
+        }
+      }
+      ref_len = bl;
+    }
+  }
+
+  struct NGramCompare {
+    int operator() (const vector<WordID>& a, const vector<WordID>& b) {
+      const size_t as = a.size();
+      const size_t bs = b.size();
+      const size_t s = (as < bs ? as : bs);
+      for (size_t i = 0; i < s; ++i) {
+         int d = a[i] - b[i];
+         if (d < 0) return true;
+         if (d > 0) return false;
+      }
+      return as < bs;
+    }
+  };
+  typedef map<vector<WordID>, pair<int,int>, NGramCompare> NGramCountMap;
+
+  void CountRef(const vector<WordID>& ref) {
+    NGramCountMap tc;
+    vector<WordID> ngram(N);
+    int s = ref.size();
+    for (int j=0; j<s; ++j) {
+      int remaining = s-j;
+      int k = (N < remaining ? N : remaining);
+      ngram.clear();
+      for (int i=1; i<=k; ++i) {
+        ngram.push_back(ref[j + i - 1]);
+        tc[ngram].first++;
+      }
+    }
+    for (typename NGramCountMap::iterator i = tc.begin(); i != tc.end(); ++i) {
+      pair<int,int>& p = ngrams_[i->first];
+      if (p.first < i->second.first)
+        p = i->second;
+    }
+  }
+
+  void ComputeNgramStats(const vector<WordID>& sent,
+                         float* correct,  // N elements reserved
+                         float* hyp,      // N elements reserved
+                         bool clip_counts = true) const {
+    vector<WordID> ngram(N);
+    *correct *= 0;
+    *hyp *= 0;
+    int s = sent.size();
+    for (int j=0; j<s; ++j) {
+      int remaining = s-j;
+      int k = (N < remaining ? N : remaining);
+      ngram.clear();
+      for (int i=1; i<=k; ++i) {
+        ngram.push_back(sent[j + i - 1]);
+        pair<int,int>& p = ngrams_[ngram];
+        if(clip_counts){
+          if (p.second < p.first) {
+            ++p.second;
+            correct[i-1]++;
+          }
+        } else {
+          ++p.second;
+          correct[i-1]++;
+        }
+        // if the 1 gram isn't found, don't try to match don't need to match any 2- 3- .. grams:
+        if (!p.first) {
+          for (; i<=k; ++i)
+            hyp[i-1]++;
+        } else {
+          hyp[i-1]++;
+        }
+      }
+    }
+  }
+
+  const EvaluationMetric* evaluation_metric;
+  vector<float> lengths_;
+  mutable NGramCountMap ngrams_;
+};
+
+template <unsigned int N = 4u, BleuType BrevityType = IBM>
+struct BleuMetric : public EvaluationMetric {
+  BleuMetric() : EvaluationMetric("IBM_BLEU") {}
+  float ComputeScore(const SufficientStats& stats) const {
+    float log_bleu = 0;
+    int count = 0;
+    for (int i = 0; i < N; ++i) {
+      if (stats.fields[i+N] > 0) {
+        float cor_count = stats.fields[i];  // correct_ngram_hit_counts[i];
+        // smooth bleu
+        if (!cor_count) { cor_count = 0.01; }
+        float lprec = log(cor_count) - log(stats.fields[i+N]); // log(hyp_ngram_counts[i]);
+        // if (precs) precs->push_back(exp(lprec));
+        log_bleu += lprec;
+        ++count;
+      }
+    }
+    log_bleu /= count;
+    float lbp = 0.0;
+    const float& hyp_len = stats.fields[2*N];
+    const float& ref_len = stats.fields[2*N + 1];
+    if (hyp_len < ref_len)
+      lbp = (hyp_len - ref_len) / hyp_len;
+    log_bleu += lbp;
+    //if (bp) *bp = exp(lbp);
+    return exp(log_bleu);
+  }
+  shared_ptr<SegmentEvaluator> CreateSegmentEvaluator(const vector<vector<WordID> >& refs) const {
+    return shared_ptr<SegmentEvaluator>(new BleuSegmentEvaluator<N,BrevityType>(refs, this));
+  }
+};
+
+EvaluationMetric* EvaluationMetric::Instance(const string& metric_id) {
+  static bool is_first = true;
+  if (is_first) {
+    instances_["NULL"] = NULL;
+    is_first = false;
+  }
+
+  map<string, EvaluationMetric*>::iterator it = instances_.find(metric_id);
+  if (it == instances_.end()) {
+    EvaluationMetric* m = NULL; 
+    if (metric_id == "IBM_BLEU") {
+      m = new BleuMetric<4, IBM>;
+    } else if (metric_id == "NIST_BLEU") {
+      m = new BleuMetric<4, NIST>;
+    } else if (metric_id == "Koehn_BLEU") {
+      m = new BleuMetric<4, Koehn>;
+    } else if (metric_id == "TER") {
+      m = new TERMetric;
+    } else {
+      cerr << "Implement please: " << metric_id << endl;
+      abort();
+    }
+    if (m->MetricId() != metric_id) {
+      cerr << "Registry error: " << metric_id << " vs. " << m->MetricId() << endl;
+      abort();
+    }
+    return instances_[metric_id] = m;
+  } else {
+    return it->second;
+  }
+}
+
+SufficientStats::SufficientStats(const string& encoded) {
+  istringstream is(encoded);
+  string type;
+  is >> type;
+  evaluation_metric = EvaluationMetric::Instance(type);
+  float val;
+  while(is >> val)
+    fields.push_back(val);
+}
+
+void SufficientStats::Encode(string* out) const {
+  ostringstream os;
+  if (evaluation_metric)
+    os << evaluation_metric->MetricId();
+  else
+    os << "NULL";
+  for (unsigned i = 0; i < fields.size(); ++i)
+    os << ' ' << fields[i];
+  *out = os.str();
+}
+
diff --git a/mteval/ns.h b/mteval/ns.h
new file mode 100644
index 00000000..f19b7509
--- /dev/null
+++ b/mteval/ns.h
@@ -0,0 +1,106 @@
+#ifndef _NS_H_
+#define _NS_H_
+
+#include <string>
+#include <vector>
+#include <map>
+#include <boost/shared_ptr.hpp>
+#include "wordid.h"
+
+class EvaluationMetric;
+
+class SufficientStats {
+ public:
+  SufficientStats() : evaluation_metric() {}
+  explicit SufficientStats(const std::string& encoded);
+  explicit SufficientStats(const EvaluationMetric* s) : evaluation_metric(s) {}
+  SufficientStats(const EvaluationMetric* s, const std::vector<float>& f) :
+    evaluation_metric(s), fields(f) {}
+
+  SufficientStats& operator+=(const SufficientStats& delta) {
+    if (delta.evaluation_metric) evaluation_metric = delta.evaluation_metric;
+    if (fields.size() != delta.fields.size())
+      fields.resize(std::max(fields.size(), delta.fields.size()));
+    for (unsigned i = 0; i < delta.fields.size(); ++i)
+      fields[i] += delta.fields[i];
+    return *this;
+  }
+  SufficientStats& operator-=(const SufficientStats& delta) {
+    if (delta.evaluation_metric) evaluation_metric = delta.evaluation_metric;
+    if (fields.size() != delta.fields.size())
+      fields.resize(std::max(fields.size(), delta.fields.size()));
+    for (unsigned i = 0; i < delta.fields.size(); ++i)
+      fields[i] -= delta.fields[i];
+    return *this;
+  }
+  SufficientStats& operator*=(const double& scalar) {
+    for (unsigned i = 0; i < fields.size(); ++i)
+      fields[i] *= scalar;
+    return *this;
+  }
+  SufficientStats& operator/=(const double& scalar) {
+    for (unsigned i = 0; i < fields.size(); ++i)
+      fields[i] /= scalar;
+    return *this;
+  }
+  bool operator==(const SufficientStats& other) const {
+    return other.fields == fields;
+  }
+  size_t size() const { return fields.size(); }
+  float operator[](size_t i) const {
+    if (i < fields.size()) return fields[i];
+    return 0;
+  }
+  void Encode(std::string* out) const;
+
+  const EvaluationMetric* evaluation_metric;
+  std::vector<float> fields;
+};
+
+inline const SufficientStats& operator+(const SufficientStats& a, const SufficientStats& b) {
+  SufficientStats res(a);
+  return res += b;
+}
+
+inline const SufficientStats& operator-(const SufficientStats& a, const SufficientStats& b) {
+  SufficientStats res(a);
+  return res -= b;
+}
+
+struct SegmentEvaluator {
+  virtual ~SegmentEvaluator();
+  virtual void Evaluate(const std::vector<WordID>& hyp, SufficientStats* out) const = 0;
+};
+
+// Instructions for implementing a new metric
+//   Override MetricId() and give the metric a unique string name (no spaces)
+//   To Instance(), add something that creates the metric
+//   Implement ONE of the following:
+//      1) void ComputeSufficientStatistics(const std::vector<std::vector<WordID> >& refs, SufficientStats* out) const;
+//      2) a new SegmentEvaluator class AND CreateSegmentEvaluator(const std::vector<std::vector<WordID> >& refs) const;
+//   The later (#2) is only used when it is necessary to precompute per-segment data from a set of refs
+//   Implement ComputeScore(const SufficientStats& stats) const;
+class EvaluationMetric {
+ public:
+  static EvaluationMetric* Instance(const std::string& metric_id = "IBM_BLEU");
+
+ protected:
+  EvaluationMetric(const std::string& id) : name_(id) {}
+  virtual ~EvaluationMetric();
+
+ public:
+  const std::string& MetricId() const { return name_; }
+
+  virtual float ComputeScore(const SufficientStats& stats) const = 0;
+  virtual boost::shared_ptr<SegmentEvaluator> CreateSegmentEvaluator(const std::vector<std::vector<WordID> >& refs) const;
+  virtual void ComputeSufficientStatistics(const std::vector<WordID>& hyp,
+                                           const std::vector<std::vector<WordID> >& refs,
+                                           SufficientStats* out) const;
+
+ private:
+  static std::map<std::string, EvaluationMetric*> instances_;
+  const std::string name_;
+};
+
+#endif
+
diff --git a/mteval/ns_ter.cc b/mteval/ns_ter.cc
new file mode 100644
index 00000000..14dc6e49
--- /dev/null
+++ b/mteval/ns_ter.cc
@@ -0,0 +1,551 @@
+#include "ns_ter.h"
+
+#include <cstdio>
+#include <cassert>
+#include <iostream>
+#include <limits>
+#include <sstream>
+#include <tr1/unordered_map>
+#include <set>
+#include <valarray>
+#include <boost/functional/hash.hpp>
+#include <stdexcept>
+#include "tdict.h"
+
+static const bool ter_use_average_ref_len = true;
+static const int ter_short_circuit_long_sentences = -1;
+
+static const unsigned kINSERTIONS = 0;
+static const unsigned kDELETIONS = 1;
+static const unsigned kSUBSTITUTIONS = 2;
+static const unsigned kSHIFTS = 3;
+static const unsigned kREF_WORDCOUNT = 4;
+static const unsigned kDUMMY_LAST_ENTRY = 5;
+
+using namespace std;
+using namespace std::tr1;
+
+#if 0
+
+struct COSTS {
+  static const float substitution;
+  static const float deletion;
+  static const float insertion;
+  static const float shift;
+};
+const float COSTS::substitution = 1.0f;
+const float COSTS::deletion = 1.0f;
+const float COSTS::insertion = 1.0f;
+const float COSTS::shift = 1.0f;
+
+static const int MAX_SHIFT_SIZE = 10;
+static const int MAX_SHIFT_DIST = 50;
+
+struct Shift {
+  unsigned int d_;
+  Shift() : d_() {}
+  Shift(int b, int e, int m) : d_() {
+    begin(b);
+    end(e);
+    moveto(m);
+  }
+  inline int begin() const {
+    return d_ & 0x3ff;
+  }
+  inline int end() const {
+    return (d_ >> 10) & 0x3ff;
+  }
+  inline int moveto() const {
+    int m = (d_ >> 20) & 0x7ff;
+    if (m > 1024) { m -= 1024; m *= -1; }
+    return m;
+  }
+  inline void begin(int b) {
+    d_ &= 0xfffffc00u;
+    d_ |= (b & 0x3ff);
+  }
+  inline void end(int e) {
+    d_ &= 0xfff003ffu;
+    d_ |= (e & 0x3ff) << 10;
+  }
+  inline void moveto(int m) {
+    bool neg = (m < 0);
+    if (neg) { m *= -1; m += 1024; }
+    d_ &= 0xfffff;
+    d_ |= (m & 0x7ff) << 20;
+  }
+};
+
+class TERScorerImpl {
+
+ public:
+  enum TransType { MATCH, SUBSTITUTION, INSERTION, DELETION };
+
+  explicit TERScorerImpl(const vector<WordID>& ref) : ref_(ref) {
+    for (int i = 0; i < ref.size(); ++i)
+      rwexists_.insert(ref[i]);
+  }
+
+  float Calculate(const vector<WordID>& hyp, int* subs, int* ins, int* dels, int* shifts) const {
+    return CalculateAllShifts(hyp, subs, ins, dels, shifts);
+  }
+
+  inline int GetRefLength() const {
+    return ref_.size();
+  }
+
+ private:
+  vector<WordID> ref_;
+  set<WordID> rwexists_;
+
+  typedef unordered_map<vector<WordID>, set<int>, boost::hash<vector<WordID> > > NgramToIntsMap;
+  mutable NgramToIntsMap nmap_;
+
+  static float MinimumEditDistance(
+      const vector<WordID>& hyp,
+      const vector<WordID>& ref,
+      vector<TransType>* path) {
+    vector<vector<TransType> > bmat(hyp.size() + 1, vector<TransType>(ref.size() + 1, MATCH));
+    vector<vector<float> > cmat(hyp.size() + 1, vector<float>(ref.size() + 1, 0));
+    for (int i = 0; i <= hyp.size(); ++i)
+      cmat[i][0] = i;
+    for (int j = 0; j <= ref.size(); ++j)
+      cmat[0][j] = j;
+    for (int i = 1; i <= hyp.size(); ++i) {
+      const WordID& hw = hyp[i-1];
+      for (int j = 1; j <= ref.size(); ++j) {
+        const WordID& rw = ref[j-1];
+	float& cur_c = cmat[i][j];
+	TransType& cur_b = bmat[i][j];
+
+        if (rw == hw) {
+          cur_c = cmat[i-1][j-1];
+          cur_b = MATCH;
+        } else {
+          cur_c = cmat[i-1][j-1] + COSTS::substitution;
+          cur_b = SUBSTITUTION;
+        }
+	float cwoi = cmat[i-1][j];
+        if (cur_c > cwoi + COSTS::insertion) {
+          cur_c = cwoi + COSTS::insertion;
+          cur_b = INSERTION;
+        }
+        float cwod = cmat[i][j-1];
+        if (cur_c > cwod + COSTS::deletion) {
+          cur_c = cwod + COSTS::deletion;
+          cur_b = DELETION;
+        }
+      }
+    }
+
+    // trace back along the best path and record the transition types
+    path->clear();
+    int i = hyp.size();
+    int j = ref.size();
+    while (i > 0 || j > 0) {
+      if (j == 0) {
+        --i;
+        path->push_back(INSERTION);
+      } else if (i == 0) {
+        --j;
+        path->push_back(DELETION);
+      } else {
+        TransType t = bmat[i][j];
+        path->push_back(t);
+        switch (t) {
+          case SUBSTITUTION:
+          case MATCH:
+            --i; --j; break;
+          case INSERTION:
+            --i; break;
+          case DELETION:
+            --j; break;
+        }
+      }
+    }
+    reverse(path->begin(), path->end());
+    return cmat[hyp.size()][ref.size()];
+  }
+
+  void BuildWordMatches(const vector<WordID>& hyp, NgramToIntsMap* nmap) const {
+    nmap->clear();
+    set<WordID> exists_both;
+    for (int i = 0; i < hyp.size(); ++i)
+      if (rwexists_.find(hyp[i]) != rwexists_.end())
+        exists_both.insert(hyp[i]);
+    for (int start=0; start<ref_.size(); ++start) {
+      if (exists_both.find(ref_[start]) == exists_both.end()) continue;
+      vector<WordID> cp;
+      int mlen = min(MAX_SHIFT_SIZE, static_cast<int>(ref_.size() - start));
+      for (int len=0; len<mlen; ++len) {
+        if (len && exists_both.find(ref_[start + len]) == exists_both.end()) break;
+        cp.push_back(ref_[start + len]);
+	(*nmap)[cp].insert(start);
+      }
+    }
+  }
+
+  static void PerformShift(const vector<WordID>& in,
+    int start, int end, int moveto, vector<WordID>* out) {
+    // cerr << "ps: " << start << " " << end << " " << moveto << endl;
+    out->clear();
+    if (moveto == -1) {
+      for (int i = start; i <= end; ++i)
+       out->push_back(in[i]);
+      for (int i = 0; i < start; ++i)
+       out->push_back(in[i]);
+      for (int i = end+1; i < in.size(); ++i)
+       out->push_back(in[i]);
+    } else if (moveto < start) {
+      for (int i = 0; i <= moveto; ++i)
+       out->push_back(in[i]);
+      for (int i = start; i <= end; ++i)
+       out->push_back(in[i]);
+      for (int i = moveto+1; i < start; ++i)
+       out->push_back(in[i]);
+      for (int i = end+1; i < in.size(); ++i)
+       out->push_back(in[i]);
+    } else if (moveto > end) {
+      for (int i = 0; i < start; ++i)
+       out->push_back(in[i]);
+      for (int i = end+1; i <= moveto; ++i)
+       out->push_back(in[i]);
+      for (int i = start; i <= end; ++i)
+       out->push_back(in[i]);
+      for (int i = moveto+1; i < in.size(); ++i)
+       out->push_back(in[i]);
+    } else {
+      for (int i = 0; i < start; ++i)
+       out->push_back(in[i]);
+      for (int i = end+1; (i < in.size()) && (i <= end + (moveto - start)); ++i)
+       out->push_back(in[i]);
+      for (int i = start; i <= end; ++i)
+       out->push_back(in[i]);
+      for (int i = (end + (moveto - start))+1; i < in.size(); ++i)
+       out->push_back(in[i]);
+    }
+    if (out->size() != in.size()) {
+      cerr << "ps: " << start << " " << end << " " << moveto << endl;
+      cerr << "in=" << TD::GetString(in) << endl;
+      cerr << "out=" << TD::GetString(*out) << endl;
+    }
+    assert(out->size() == in.size());
+    // cerr << "ps: " << TD::GetString(*out) << endl;
+  }
+
+  void GetAllPossibleShifts(const vector<WordID>& hyp,
+      const vector<int>& ralign,
+      const vector<bool>& herr,
+      const vector<bool>& rerr,
+      const int min_size,
+      vector<vector<Shift> >* shifts) const {
+    for (int start = 0; start < hyp.size(); ++start) {
+      vector<WordID> cp(1, hyp[start]);
+      NgramToIntsMap::iterator niter = nmap_.find(cp);
+      if (niter == nmap_.end()) continue;
+      bool ok = false;
+      int moveto;
+      for (set<int>::iterator i = niter->second.begin(); i != niter->second.end(); ++i) {
+        moveto = *i;
+        int rm = ralign[moveto];
+        ok = (start != rm &&
+              (rm - start) < MAX_SHIFT_DIST &&
+              (start - rm - 1) < MAX_SHIFT_DIST);
+        if (ok) break;
+      }
+      if (!ok) continue;
+      cp.clear();
+      for (int end = start + min_size - 1;
+           ok && end < hyp.size() && end < (start + MAX_SHIFT_SIZE); ++end) {
+        cp.push_back(hyp[end]);
+	vector<Shift>& sshifts = (*shifts)[end - start];
+        ok = false;
+        NgramToIntsMap::iterator niter = nmap_.find(cp);
+        if (niter == nmap_.end()) break;
+        bool any_herr = false;
+        for (int i = start; i <= end && !any_herr; ++i)
+          any_herr = herr[i];
+        if (!any_herr) {
+          ok = true;
+          continue;
+        }
+        for (set<int>::iterator mi = niter->second.begin();
+             mi != niter->second.end(); ++mi) {
+          int moveto = *mi;
+	  int rm = ralign[moveto];
+	  if (! ((rm != start) &&
+	        ((rm < start) || (rm > end)) &&
+		(rm - start <= MAX_SHIFT_DIST) &&
+		((start - rm - 1) <= MAX_SHIFT_DIST))) continue;
+          ok = true;
+	  bool any_rerr = false;
+	  for (int i = 0; (i <= end - start) && (!any_rerr); ++i)
+            any_rerr = rerr[moveto+i];
+	  if (!any_rerr) continue;
+	  for (int roff = 0; roff <= (end - start); ++roff) {
+	    int rmr = ralign[moveto+roff];
+	    if ((start != rmr) && ((roff == 0) || (rmr != ralign[moveto])))
+	      sshifts.push_back(Shift(start, end, moveto + roff));
+	  }
+        }
+      }
+    }
+  }
+
+  bool CalculateBestShift(const vector<WordID>& cur,
+                          const vector<WordID>& hyp,
+                          float curerr,
+                          const vector<TransType>& path,
+                          vector<WordID>* new_hyp,
+                          float* newerr,
+                          vector<TransType>* new_path) const {
+    vector<bool> herr, rerr;
+    vector<int> ralign;
+    int hpos = -1;
+    for (int i = 0; i < path.size(); ++i) {
+      switch (path[i]) {
+        case MATCH:
+	  ++hpos;
+	  herr.push_back(false);
+	  rerr.push_back(false);
+	  ralign.push_back(hpos);
+          break;
+        case SUBSTITUTION:
+	  ++hpos;
+	  herr.push_back(true);
+	  rerr.push_back(true);
+	  ralign.push_back(hpos);
+          break;
+        case INSERTION:
+	  ++hpos;
+	  herr.push_back(true);
+          break;
+	case DELETION:
+	  rerr.push_back(true);
+	  ralign.push_back(hpos);
+          break;
+      }
+    }
+#if 0
+    cerr << "RALIGN: ";
+    for (int i = 0; i < rerr.size(); ++i)
+      cerr << ralign[i] << " ";
+    cerr << endl;
+    cerr << "RERR: ";
+    for (int i = 0; i < rerr.size(); ++i)
+      cerr << (bool)rerr[i] << " ";
+    cerr << endl;
+    cerr << "HERR: ";
+    for (int i = 0; i < herr.size(); ++i)
+      cerr << (bool)herr[i] << " ";
+    cerr << endl;
+#endif
+
+    vector<vector<Shift> > shifts(MAX_SHIFT_SIZE + 1);
+    GetAllPossibleShifts(cur, ralign, herr, rerr, 1, &shifts);
+    float cur_best_shift_cost = 0;
+    *newerr = curerr;
+    vector<TransType> cur_best_path;
+    vector<WordID> cur_best_hyp;
+
+    bool res = false;
+    for (int i = shifts.size() - 1; i >=0; --i) {
+      float curfix = curerr - (cur_best_shift_cost + *newerr);
+      float maxfix = 2.0f * (1 + i) - COSTS::shift;
+      if ((curfix > maxfix) || ((cur_best_shift_cost == 0) && (curfix == maxfix))) break;
+      for (int j = 0; j < shifts[i].size(); ++j) {
+        const Shift& s = shifts[i][j];
+	curfix = curerr - (cur_best_shift_cost + *newerr);
+	maxfix = 2.0f * (1 + i) - COSTS::shift;  // TODO remove?
+        if ((curfix > maxfix) || ((cur_best_shift_cost == 0) && (curfix == maxfix))) continue;
+	vector<WordID> shifted(cur.size());
+	PerformShift(cur, s.begin(), s.end(), ralign[s.moveto()], &shifted);
+	vector<TransType> try_path;
+	float try_cost = MinimumEditDistance(shifted, ref_, &try_path);
+	float gain = (*newerr + cur_best_shift_cost) - (try_cost + COSTS::shift);
+	if (gain > 0.0f || ((cur_best_shift_cost == 0.0f) && (gain == 0.0f))) {
+	  *newerr = try_cost;
+	  cur_best_shift_cost = COSTS::shift;
+	  new_path->swap(try_path);
+	  new_hyp->swap(shifted);
+	  res = true;
+	  // cerr << "Found better shift " << s.begin() << "..." << s.end() << " moveto " << s.moveto() << endl;
+	}
+      }
+    }
+
+    return res;
+  }
+
+  static void GetPathStats(const vector<TransType>& path, int* subs, int* ins, int* dels) {
+    *subs = *ins = *dels = 0;
+    for (int i = 0; i < path.size(); ++i) {
+      switch (path[i]) {
+        case SUBSTITUTION:
+	  ++(*subs);
+        case MATCH:
+          break;
+        case INSERTION:
+          ++(*ins); break;
+	case DELETION:
+          ++(*dels); break;
+      }
+    }
+  }
+
+  float CalculateAllShifts(const vector<WordID>& hyp,
+      int* subs, int* ins, int* dels, int* shifts) const {
+    BuildWordMatches(hyp, &nmap_);
+    vector<TransType> path;
+    float med_cost = MinimumEditDistance(hyp, ref_, &path);
+    float edits = 0;
+    vector<WordID> cur = hyp;
+    *shifts = 0;
+    if (ter_short_circuit_long_sentences < 0 ||
+        ref_.size() < ter_short_circuit_long_sentences) {
+      while (true) {
+        vector<WordID> new_hyp;
+        vector<TransType> new_path;
+        float new_med_cost;
+        if (!CalculateBestShift(cur, hyp, med_cost, path, &new_hyp, &new_med_cost, &new_path))
+          break;
+        edits += COSTS::shift;
+        ++(*shifts);
+        med_cost = new_med_cost;
+        path.swap(new_path);
+        cur.swap(new_hyp);
+      }
+    }
+    GetPathStats(path, subs, ins, dels);
+    return med_cost + edits;
+  }
+};
+
+class TERScore : public ScoreBase<TERScore> {
+  friend class TERScorer;
+
+ public:
+
+ TERScore() : stats(0,kDUMMY_LAST_ENTRY) {}
+  float ComputePartialScore() const { return 0.0;}
+  float ComputeScore() const {
+    float edits = static_cast<float>(stats[kINSERTIONS] + stats[kDELETIONS] + stats[kSUBSTITUTIONS] + stats[kSHIFTS]);
+    return edits / static_cast<float>(stats[kREF_WORDCOUNT]);
+  }
+  void ScoreDetails(string* details) const;
+  void PlusPartialEquals(const Score& rhs, int oracle_e_cover, int oracle_f_cover, int src_len){}
+  void PlusEquals(const Score& delta, const float scale) {
+    if (scale==1)
+      stats += static_cast<const TERScore&>(delta).stats;
+    if (scale==-1)
+      stats -= static_cast<const TERScore&>(delta).stats;
+    throw std::runtime_error("TERScore::PlusEquals with scale != +-1");
+ }
+  void PlusEquals(const Score& delta) {
+    stats += static_cast<const TERScore&>(delta).stats;
+  }
+
+  ScoreP GetZero() const {
+    return ScoreP(new TERScore);
+  }
+  ScoreP GetOne() const {
+    return ScoreP(new TERScore);
+  }
+  void Subtract(const Score& rhs, Score* res) const {
+    static_cast<TERScore*>(res)->stats = stats - static_cast<const TERScore&>(rhs).stats;
+  }
+  void Encode(std::string* out) const {
+    ostringstream os;
+    os << stats[kINSERTIONS] << ' '
+       << stats[kDELETIONS] << ' '
+       << stats[kSUBSTITUTIONS] << ' '
+       << stats[kSHIFTS] << ' '
+       << stats[kREF_WORDCOUNT];
+    *out = os.str();
+  }
+  bool IsAdditiveIdentity() const {
+    for (int i = 0; i < kDUMMY_LAST_ENTRY; ++i)
+      if (stats[i] != 0) return false;
+    return true;
+  }
+ private:
+  valarray<int> stats;
+};
+
+ScoreP TERScorer::ScoreFromString(const std::string& data) {
+  istringstream is(data);
+  TERScore* r = new TERScore;
+  is >> r->stats[TERScore::kINSERTIONS]
+     >> r->stats[TERScore::kDELETIONS]
+     >> r->stats[TERScore::kSUBSTITUTIONS]
+     >> r->stats[TERScore::kSHIFTS]
+     >> r->stats[TERScore::kREF_WORDCOUNT];
+  return ScoreP(r);
+}
+
+void TERScore::ScoreDetails(std::string* details) const {
+  char buf[200];
+  sprintf(buf, "TER = %.2f, %3d|%3d|%3d|%3d (len=%d)",
+     ComputeScore() * 100.0f,
+     stats[kINSERTIONS],
+     stats[kDELETIONS],
+     stats[kSUBSTITUTIONS],
+     stats[kSHIFTS],
+     stats[kREF_WORDCOUNT]);
+  *details = buf;
+}
+
+TERScorer::~TERScorer() {
+  for (vector<TERScorerImpl*>::iterator i = impl_.begin(); i != impl_.end(); ++i)
+    delete *i;
+}
+
+TERScorer::TERScorer(const vector<vector<WordID> >& refs) : impl_(refs.size()) {
+  for (int i = 0; i < refs.size(); ++i)
+    impl_[i] = new TERScorerImpl(refs[i]);
+}
+
+ScoreP TERScorer::ScoreCCandidate(const vector<WordID>& hyp) const {
+  return ScoreP();
+}
+
+ScoreP TERScorer::ScoreCandidate(const std::vector<WordID>& hyp) const {
+  float best_score = numeric_limits<float>::max();
+  TERScore* res = new TERScore;
+  int avg_len = 0;
+  for (int i = 0; i < impl_.size(); ++i)
+    avg_len += impl_[i]->GetRefLength();
+  avg_len /= impl_.size();
+  for (int i = 0; i < impl_.size(); ++i) {
+    int subs, ins, dels, shifts;
+    float score = impl_[i]->Calculate(hyp, &subs, &ins, &dels, &shifts);
+    // cerr << "Component TER cost: " << score << endl;
+    if (score < best_score) {
+      res->stats[TERScore::kINSERTIONS] = ins;
+      res->stats[TERScore::kDELETIONS] = dels;
+      res->stats[TERScore::kSUBSTITUTIONS] = subs;
+      res->stats[TERScore::kSHIFTS] = shifts;
+      if (ter_use_average_ref_len) {
+        res->stats[TERScore::kREF_WORDCOUNT] = avg_len;
+      } else {
+        res->stats[TERScore::kREF_WORDCOUNT] = impl_[i]->GetRefLength();
+      }
+
+      best_score = score;
+    }
+  }
+  return ScoreP(res);
+}
+#endif
+
+void TERMetric::ComputeSufficientStatistics(const vector<WordID>& hyp,
+                                            const vector<vector<WordID> >& refs,
+                                            SufficientStats* out) const {
+  out->fields.resize(kDUMMY_LAST_ENTRY);
+}
+
+float TERMetric::ComputeScore(const SufficientStats& stats) const {
+  float edits = static_cast<float>(stats[kINSERTIONS] + stats[kDELETIONS] + stats[kSUBSTITUTIONS] + stats[kSHIFTS]);
+  return edits / static_cast<float>(stats[kREF_WORDCOUNT]);
+}
+
diff --git a/mteval/ns_ter.h b/mteval/ns_ter.h
new file mode 100644
index 00000000..bb90f95e
--- /dev/null
+++ b/mteval/ns_ter.h
@@ -0,0 +1,18 @@
+#ifndef _NS_TER_H_
+#define _NS_TER_H_
+
+#include "ns.h"
+
+class TERMetric : public EvaluationMetric {
+  friend class EvaluationMetric;
+ protected:
+  TERMetric() : EvaluationMetric("TER") {}
+
+ public:
+  virtual void ComputeSufficientStatistics(const std::vector<WordID>& hyp,
+                                           const std::vector<std::vector<WordID> >& refs,
+                                           SufficientStats* out) const;
+  virtual float ComputeScore(const SufficientStats& stats) const;
+};
+
+#endif
diff --git a/mteval/scorer_test.cc b/mteval/scorer_test.cc
index a07a8c4b..09da250c 100644
--- a/mteval/scorer_test.cc
+++ b/mteval/scorer_test.cc
@@ -3,9 +3,11 @@
 #include <valarray>
 #include <gtest/gtest.h>
 
+#include "ns.h"
 #include "tdict.h"
 #include "scorer.h"
 #include "aer_scorer.h"
+#include "kernel_string_subseq.h"
 
 using namespace std;
 
@@ -175,6 +177,50 @@ TEST_F(ScorerTest, AERTest) {
   EXPECT_EQ(d2, details);
 }
 
+TEST_F(ScorerTest, Kernel) {
+  for (int i = 1; i < 10; ++i) {
+    const float l = (i / 10.0);
+    float f = ssk<4>(refs0[0], hyp1, l) +
+              ssk<4>(refs0[1], hyp1, l) +
+              ssk<4>(refs0[2], hyp1, l) +
+              ssk<4>(refs0[3], hyp1, l);
+    float f2= ssk<4>(refs1[0], hyp2, l) +
+              ssk<4>(refs1[1], hyp2, l) +
+              ssk<4>(refs1[2], hyp2, l) +
+              ssk<4>(refs1[3], hyp2, l);
+    f /= 4;
+    f2 /= 4;
+    float f3= ssk<4>(refs0[0], hyp2, l) +
+              ssk<4>(refs0[1], hyp2, l) +
+              ssk<4>(refs0[2], hyp2, l) +
+              ssk<4>(refs0[3], hyp2, l);
+    float f4= ssk<4>(refs1[0], hyp1, l) +
+              ssk<4>(refs1[1], hyp1, l) +
+              ssk<4>(refs1[2], hyp1, l) +
+              ssk<4>(refs1[3], hyp1, l);
+    f3 += f4;
+    f3 /= 8;
+    cerr << "LAMBDA=" << l << "\t" << f << " " << f2 << "\tf=" << ((f + f2)/2 - f3) << " (bad=" << f3 << ")" << endl;
+  }
+}
+
+TEST_F(ScorerTest, NewScoreAPI) {
+  EvaluationMetric* metric = EvaluationMetric::Instance("IBM_BLEU");
+  boost::shared_ptr<SegmentEvaluator> e1 = metric->CreateSegmentEvaluator(refs0);
+  boost::shared_ptr<SegmentEvaluator> e2 = metric->CreateSegmentEvaluator(refs1);
+  SufficientStats stats1;
+  e1->Evaluate(hyp2, &stats1);
+  SufficientStats stats2;
+  e2->Evaluate(hyp1, &stats2);
+  stats1 += stats2;
+  string ss;
+  stats1.Encode(&ss);
+  cerr << "SS: " << ss << endl;
+  cerr << metric->ComputeScore(stats1) << endl;
+  SufficientStats statse("IBM_BLEU 53 32 18 11 65 63 61 59 65 72");
+  cerr << metric->ComputeScore(statse) << endl;
+}
+
 int main(int argc, char **argv) {
   testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
diff --git a/utils/kernel_string_subseq.h b/utils/kernel_string_subseq.h
new file mode 100644
index 00000000..516e8b89
--- /dev/null
+++ b/utils/kernel_string_subseq.h
@@ -0,0 +1,51 @@
+#ifndef _KERNEL_STRING_SUBSEQ_H_
+#define _KERNEL_STRING_SUBSEQ_H_
+
+#include <vector>
+#include <cmath>
+#include <boost/multi_array.hpp>
+
+template <unsigned N, typename T>
+float ssk(const T* s, const size_t s_size, const T* t, const size_t t_size, const float lambda) {
+  assert(N > 0);
+  boost::multi_array<float, 3> kp(boost::extents[N + 1][s_size + 1][t_size + 1]);
+  const float l2 = lambda * lambda;
+  for (unsigned j = 0; j < s_size; ++j)
+    for (unsigned k = 0; k < t_size; ++k)
+      kp[0][j][k] = 1.0f;
+  for (unsigned i = 0; i < N; ++i) {
+    for (unsigned j = 0; j < s_size; ++j) {
+      float kpp = 0.0f;
+      for (unsigned k = 0; k < t_size; ++k) {
+        kpp = lambda * (kpp + lambda * (s[j]==t[k]) * kp[i][j][k]);
+        kp[i + 1][j + 1][k + 1] = lambda * kp[i + 1][j][k + 1] + kpp;
+      }
+    }
+  }
+  float kn = 0.0f;
+  for (int i = 0; i < N; ++i)
+    for (int j = 0; j < s_size; ++j)
+      for (int k = 0; k < t_size; ++k)
+        kn += l2 * (s[j] == t[k]) * kp[i][j][k];
+  return kn;
+}
+
+template <unsigned N, typename T>
+float ssk(const std::vector<T>& s, const std::vector<T>& t, const float lambda) {
+  float kst = ssk<N, T>(&s[0], s.size(), &t[0], t.size(), lambda);
+  if (!kst) return 0.0f;
+  float kss = ssk<N, T>(&s[0], s.size(), &s[0], s.size(), lambda);
+  float ktt = ssk<N, T>(&t[0], t.size(), &t[0], t.size(), lambda);
+  return kst / std::sqrt(kss * ktt);
+}
+
+template <unsigned N>
+float ssk(const std::string& s, const std::string& t, const float lambda) {
+  float kst = ssk<N, char>(&s[0], s.size(), &t[0], t.size(), lambda);
+  if (!kst) return 0.0f;
+  float kss = ssk<N, char>(&s[0], s.size(), &s[0], s.size(), lambda);
+  float ktt = ssk<N, char>(&t[0], t.size(), &t[0], t.size(), lambda);
+  return kst / std::sqrt(kss * ktt);
+}
+
+#endif