Publish

6 files changed, 374 insertions, 0 deletions

diff --git a/fsa/queryproc/.gitignore b/fsa/queryproc/.gitignore
new file mode 100644
index 00000000000..a073ef1dd72
--- /dev/null
+++ b/fsa/queryproc/.gitignore
@@ -0,0 +1,9 @@
+.deps
+.libs
+Makefile
+Makefile.in
+count_plain_grams
+count_sorted_grams
+p2s_ratio
+permute_query
+sort_grams
diff --git a/fsa/queryproc/count_plain_grams.cpp b/fsa/queryproc/count_plain_grams.cpp
new file mode 100644
index 00000000000..197c958149b
--- /dev/null
+++ b/fsa/queryproc/count_plain_grams.cpp
@@ -0,0 +1,89 @@
+// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+#include <iostream>
+#include <iomanip>
+#include <map>
+#include <string>
+
+#include "fsa.h"
+#include "permuter.h"
+#include "selector.h"
+#include "ngram.h"
+#include "base64.h"
+#include "wordchartokenizer.h"
+
+using namespace fsa;
+
+unsigned int gram_count(unsigned int mg, unsigned int q)
+{
+  unsigned int i,j,c1,c2,ct=0;
+
+  for(i=2;i<=mg;i++){
+    c1=1;c2=1;
+    for(j=(i>q-i)?(i+1):(q-i+1);j<=q;j++){
+      c1*=j;
+      c2*=(q-j)+1;
+    }
+    ct+=c1/c2;
+  }
+  return ct;
+}
+
+int main(int argc, char **argv)
+{
+  const unsigned int MAXQUERY = 10;
+  const unsigned int MAXGRAM  = 6;
+
+  Permuter p;
+  NGram freq_s,query;
+  WordCharTokenizer tokenizer(WordCharTokenizer::PUNCTUATION_WHITESPACEONLY);
+  unsigned int freq;
+  Selector s;
+  std::string qstr;
+  unsigned int qlen,glen;
+
+  if(argc!=2){
+    std::cerr << "usage: " << argv[0] << " fsa_file" << std::endl;
+    exit(1);
+  }
+
+  FSA fsa(argv[1]);
+  FSA::State state(fsa);
+  std::map<std::string,unsigned int> grams,gq;
+  std::map<std::string,unsigned int>::iterator grams_it,gq_it;
+  std::string gram_str;
+
+  while(!std::cin.eof()){
+    getline(std::cin,qstr);
+    query.set(qstr,tokenizer,1,-1);
+    qlen = query.length();
+    if(2<=qlen && qlen<=MAXQUERY){
+      freq_s.set(qstr,tokenizer,0,1);
+      freq=atoi(freq_s[0].c_str());
+      gq.clear();
+      for(unsigned int i=0;i<qlen-1;i++){
+        for(unsigned int j=2;j<=MAXGRAM&&i+j<=qlen;j++){
+          state.startWord(query[i]);
+          for(unsigned int k=1;state.isValid()&&k<j;k++){
+            state.deltaWord(query[i+k]);
+          }
+          if(state.isFinal()){
+            gram_str = query.join(" ",i,j);
+            gq[gram_str]=freq;
+          }
+        }
+      }
+      for(gq_it=gq.begin();gq_it!=gq.end();++gq_it){
+        grams_it=grams.find(gq_it->first);
+        if(grams_it!=grams.end())
+          grams[gq_it->first]=grams_it->second+gq_it->second;
+        else
+          grams[gq_it->first]=gq_it->second;
+      }
+    }
+  }
+
+  for(grams_it=grams.begin();grams_it!=grams.end();++grams_it)
+    std::cout << grams_it->first << '\t' << grams_it->second << std::endl;
+
+  return 0;
+}
diff --git a/fsa/queryproc/count_sorted_grams.cpp b/fsa/queryproc/count_sorted_grams.cpp
new file mode 100644
index 00000000000..58be4548fac
--- /dev/null
+++ b/fsa/queryproc/count_sorted_grams.cpp
@@ -0,0 +1,78 @@
+// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+#include <iostream>
+#include <iomanip>
+#include <map>
+#include <string>
+
+#include "fsa.h"
+#include "permuter.h"
+#include "selector.h"
+#include "ngram.h"
+#include "base64.h"
+#include "wordchartokenizer.h"
+
+using namespace fsa;
+
+int main(int argc, char **argv)
+{
+  const unsigned int MAXQUERY = 10;
+  const unsigned int MAXGRAM  = 6;
+
+  Permuter p;
+  NGram freq_s,query,gram;
+  WordCharTokenizer tokenizer(WordCharTokenizer::PUNCTUATION_WHITESPACEONLY);
+  unsigned int freq;
+  Selector s;
+  std::string qstr;
+  unsigned int qlen,glen;
+
+  if(argc!=2){
+    std::cerr << "usage: " << argv[0] << " sorted_fsa_file" << std::endl;
+    exit(1);
+  }
+
+  FSA fsa(argv[1]);
+  FSA::State state(fsa);
+  std::map<std::string,unsigned int> grams;
+  std::map<std::string,unsigned int>::iterator grams_it;
+  std::string gram_str;
+
+  while(!std::cin.eof()){
+    getline(std::cin,qstr);
+    query.set(qstr,tokenizer,1,-1);
+    qlen = query.length();
+    if(2<=qlen && qlen<=MAXQUERY){
+      freq_s.set(qstr,tokenizer,0,1);
+      freq=atoi(freq_s[0].c_str());
+      query.sort();
+      qlen = query.uniq();
+      unsigned int glen=qlen<MAXGRAM?qlen:MAXGRAM;
+      for(unsigned int n=2;n<=glen;n++){
+        unsigned int c=Permuter::firstComb(n,qlen);
+        while(c>0){
+          s.clear();
+          s.set(c);
+          gram.set(query,s);
+          state.startWord(gram[0]);
+          for(unsigned int i=1;state.isValid()&&i<gram.size();i++){
+            state.deltaWord(gram[i]);
+          }
+          if(state.isFinal()){
+            gram_str = gram.join(" ");
+            grams_it=grams.find(gram_str);
+            if(grams_it!=grams.end())
+              grams[gram_str]=grams_it->second+freq;
+            else
+              grams[gram_str]=freq;
+          }
+          c=Permuter::nextComb(c,qlen);
+        }
+      }
+    }
+  }
+
+  for(grams_it=grams.begin();grams_it!=grams.end();++grams_it)
+    std::cout << grams_it->first << '\t' << grams_it->second << std::endl;
+
+  return 0;
+}
diff --git a/fsa/queryproc/p2s_ratio.cpp b/fsa/queryproc/p2s_ratio.cpp
new file mode 100644
index 00000000000..cbc61c45d53
--- /dev/null
+++ b/fsa/queryproc/p2s_ratio.cpp
@@ -0,0 +1,59 @@
+// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+#include <iostream>
+#include <iomanip>
+#include <map>
+#include <string>
+
+#include "fsa.h"
+#include "permuter.h"
+#include "ngram.h"
+#include "base64.h"
+
+using namespace fsa;
+
+int main(int argc, char **argv)
+{
+  const unsigned int MAXQUERY = 10;
+  const unsigned int MAXGRAM  = 6;
+
+  Permuter p;
+  NGram freq_s,gram,sorted_gram;
+  unsigned int freq;
+  Selector s(10);
+  std::string gstr;
+
+  if(argc!=3){
+    std::cerr << "usage: " << argv[0] << " plain_count_fsa_file sorted_count_fsa_file" << std::endl;
+    exit(1);
+  }
+
+  FSA plain_fsa(argv[1]);
+  FSA sorted_fsa(argv[2]);
+  FSA::State state1(plain_fsa),state2(sorted_fsa);
+
+  while(!std::cin.eof()){
+    getline(std::cin,gstr);
+    gram.set(gstr);
+    if(gram.length()>1){
+      sorted_gram.set(gram);
+      sorted_gram.sort();
+      sorted_gram.uniq();
+      state1.startWord(gram[0]);
+      for(unsigned int i=1;state1.isValid()&&i<gram.length();i++){
+        state1.deltaWord(gram[i]);
+      }
+      state2.startWord(sorted_gram[0]);
+      for(unsigned int i=1;state2.isValid()&&i<sorted_gram.length();i++){
+        state2.deltaWord(sorted_gram[i]);
+      }
+      if(state1.isFinal() && state2.isFinal()){
+        unsigned int c1,c2;
+        c1=*((unsigned int*)state1.data());
+        c2=*((unsigned int*)state2.data());
+        std::cout << gram << "\t" << c1 << "," << c2 << "," << (double)c1/(double)c2 << std::endl;
+      }
+    }
+  }
+
+  return 0;
+}
diff --git a/fsa/queryproc/permute_query.cpp b/fsa/queryproc/permute_query.cpp
new file mode 100644
index 00000000000..7645e864a44
--- /dev/null
+++ b/fsa/queryproc/permute_query.cpp
@@ -0,0 +1,110 @@
+// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+#include <iostream>
+#include <iomanip>
+
+#include "permuter.h"
+#include "selector.h"
+#include "ngram.h"
+#include "base64.h"
+
+#if (__GNUG__ <3 || (__GNUG__ == 3 && __GNUC_MINOR__ < 1))
+namespace std {
+const char *fixed = "";
+}
+#endif
+
+using namespace fsa;
+
+unsigned int gram_count(unsigned int mg, unsigned int q)
+{
+  unsigned int i,j,c1,c2,ct=0;
+
+  for(i=2;i<=mg;i++){
+    c1=1;c2=1;
+    for(j=(i>q-i)?(i+1):(q-i+1);j<=q;j++){
+      c1*=j;
+      c2*=(q-j)+1;
+    }
+    ct+=c1/c2;
+  }
+  return ct;
+}
+
+int main(int argc, char **argv)
+{
+  const unsigned int MAXQUERY = 10;
+  const unsigned int MAXGRAM  = 6;
+
+  Permuter p;
+  NGram query,gram;
+  Selector s;
+  std::string qstr;
+  unsigned int qlen,glen;
+  bool verbose=true;
+  unsigned int i;
+  double total,ctotal;
+  int stats[MAXQUERY+1];
+
+  for(i=0;i<=MAXQUERY;i++)
+    stats[i]=0;
+  while(!std::cin.eof()){
+    getline(std::cin,qstr);
+    query.set(qstr,1);
+    qlen = query.length();
+    if(2<=qlen && qlen<=MAXQUERY){
+      stats[qlen]++;
+      std::cout << "QUERY: " << query << std::endl;
+      query.sort();
+      qlen = query.uniq();
+      unsigned int glen=qlen<MAXGRAM?qlen:MAXGRAM;
+      for(unsigned int n=2;n<=glen;n++){
+        unsigned int c=Permuter::firstComb(n,qlen);
+        while(c>0){
+          s.clear();
+          s.set(c);
+          gram.set(query,s);
+          std::cout << "   " << gram << std::endl;
+          c=Permuter::nextComb(c,qlen);
+        }
+      }
+    }
+    else{
+      if(qlen<2)
+        stats[0]++;
+      else
+        stats[1]++;
+    }
+  }
+
+
+
+  if(verbose){
+    total=0.0;ctotal=0.0;
+    for(i=0;i<=MAXQUERY;i++)
+      total+=stats[i];
+    std::cerr << std::fixed << std::setprecision(4) << std::endl;
+    std::cerr << "Statistics:" << std::endl;
+    std::cerr << std::endl;
+    std::cerr << "  Empty or single term:  " <<
+      std::setw(12) << stats[0] << "   " <<
+      std::setw(7) << double(stats[0])*100.0/total << "%" << std::endl;
+    std::cerr << "  Too long:              " <<
+      std::setw(12) << stats[1] << "   " <<
+      std::setw(7) << double(stats[1])*100.0/total << "%" << std::endl;
+    for(i=2;i<=MAXQUERY;i++){
+      std::cerr << "  Length " << std::setw(2) << i << " (grams " << std::setw(3) <<
+        gram_count(i<MAXGRAM?i:MAXGRAM,i) << "): " <<
+        std::setw(12) << stats[i] << "   " <<
+        std::setw(7) << double(stats[i])*100.0/total << "%" << std::endl;
+      ctotal+=stats[i]*gram_count(i<MAXGRAM?i:MAXGRAM,i);
+    }
+    std::cerr << "  Total:                 " <<
+      std::setw(12) << std::setprecision(0) << total << std::endl;
+    std::cerr << std::endl;
+    std::cerr << "Average number of grams per query: " <<
+      std::setprecision(2) << ctotal/total << std::endl;
+    std::cerr << std::endl;
+  }
+
+  return 0;
+}
diff --git a/fsa/queryproc/sort_grams.cpp b/fsa/queryproc/sort_grams.cpp
new file mode 100644
index 00000000000..427dba129ff
--- /dev/null
+++ b/fsa/queryproc/sort_grams.cpp
@@ -0,0 +1,29 @@
+// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+#include <iostream>
+#include <iomanip>
+
+#include "permuter.h"
+#include "ngram.h"
+#include "base64.h"
+#include "wordchartokenizer.h"
+
+using namespace fsa;
+
+int main(int argc, char **argv)
+{
+
+  NGram query;
+  WordCharTokenizer tokenizer(WordCharTokenizer::PUNCTUATION_WHITESPACEONLY);
+  std::string qstr;
+
+  while(!std::cin.eof()){
+    getline(std::cin,qstr);
+    query.set(qstr,tokenizer,0,-1);
+    query.sort();
+    query.uniq();
+    std::cout << query << std::endl;
+  }
+
+
+  return 0;
+}