Added word frequency information

6 files changed, 77 insertions, 15 deletions

@@ -28,17 +28,22 @@ namespace verbly {
     {
       // Serialize the form first.
       {
-        db.insertIntoTable(
-          "forms",
-          {
-            { "form_id", arg.getId() },
-            { "form", arg.getText() },
-            { "complexity", arg.getComplexity() },
-            { "proper", arg.isProper() },
-            { "length", arg.getLength() },
-            { "anagram_set_id", arg.getAnagramSetId() },
-            { "reverse_form_id", arg.getReverseId() }
-          });
+        std::list<hatkirby::column> fields = {
+          { "form_id", arg.getId() },
+          { "form", arg.getText() },
+          { "complexity", arg.getComplexity() },
+          { "proper", arg.isProper() },
+          { "length", arg.getLength() },
+          { "anagram_set_id", arg.getAnagramSetId() },
+          { "reverse_form_id", arg.getReverseId() }
+        };
+
+        if (arg.getFrequency() > 0)
+        {
+          fields.emplace_back("frequency", arg.getFrequency());
+        }
+
+        db.insertIntoTable("forms", std::move(fields));
       }
 
       // Then, serialize the form/pronunciation relationship.
@@ -63,6 +63,16 @@ namespace verbly {
         return reverse_id_;
       }
 
+      void setFrequency(int freq)
+      {
+        frequency_ = freq;
+      }
+
+      int getFrequency() const
+      {
+        return frequency_;
+      }
+
       std::set<const pronunciation*> getPronunciations() const
       {
         return pronunciations_;
@@ -79,6 +89,7 @@ namespace verbly {
       const int length_;
       const int anagram_set_id_;
       int reverse_id_ = -1;
+      int frequency_ = 0;
 
       std::set<const pronunciation*> pronunciations_;
 
@@ -1,4 +1,5 @@
 #include "generator.h"
+#include <cstdlib>
 #include <stdexcept>
 #include <iostream>
 #include <regex>
@@ -23,6 +24,7 @@ namespace verbly {
       std::string wordNetPath,
       std::string cmudictPath,
       std::string imageNetPath,
+      std::string wordfreqPath,
       std::string outputPath,
       std::string imageNetOutput) :
         verbNetPath_(verbNetPath),
@@ -30,6 +32,7 @@ namespace verbly {
         wordNetPath_(wordNetPath),
         cmudictPath_(cmudictPath),
         imageNetPath_(imageNetPath),
+        wordfreqPath_(wordfreqPath),
         db_(outputPath, hatkirby::dbmode::create),
         imageNetOutput_(imageNetOutput)
     {
@@ -117,6 +120,10 @@ namespace verbly {
       // and then only generate pronunciations for already-exisiting forms.
       readCmudictPronunciations();
 
+      // Reads word (really form) frequency information from a corpus, formatted
+      // as a CSV file with the form in one column and the frequency in another.
+      readWordFrequency();
+
       // Writes the database schema
       writeSchema();
 
@@ -624,6 +631,39 @@ namespace verbly {
       }
     }
 
+    void generator::readWordFrequency()
+    {
+      std::list<std::string> lines(readFile(wordfreqPath_));
+
+      hatkirby::progress ppgs(
+        "Reading word frequencies...",
+        lines.size());
+
+      for (std::string line : lines)
+      {
+        ppgs.update();
+
+        std::regex freqline("([a-z]+),([0-9]+)");
+        std::smatch freqline_data;
+        if (std::regex_search(line, freqline_data, freqline))
+        {
+          std::string text = freqline_data[1];
+
+          if (!formByText_.count(text))
+          {
+            continue;
+          }
+
+          std::string freqnumstr = freqline_data[2];
+          long long freqnumnum = std::atoll(freqnumstr.c_str());
+          formByText_[text]->setFrequency(
+            freqnumnum > std::numeric_limits<int>::max()
+              ? std::numeric_limits<int>::max()
+              : freqnumnum);
+        }
+      }
+    }
+
     void generator::writeSchema()
     {
       std::ifstream file("schema.sql");
@@ -35,6 +35,7 @@ namespace verbly {
         std::string wordNetPath,
         std::string cmudictPath,
         std::string imageNetPath,
+        std::string wordfreqPath,
         std::string outputPath,
         std::string imageNetOutput);
 
@@ -62,6 +63,8 @@ namespace verbly {
 
       void readCmudictPronunciations();
 
+      void readWordFrequency();
+
       void writeSchema();
 
       void writeVersion();
@@ -125,6 +128,7 @@ namespace verbly {
       std::string wordNetPath_;
       std::string cmudictPath_;
       std::string imageNetPath_;
+      std::string wordfreqPath_;
 
       // Output
 
@@ -4,23 +4,24 @@
 
 void printUsage()
 {
-  std::cout << "usage: generator verbnet agid wordnet cmudict imagenet output ino" << std::endl;
+  std::cout << "usage: generator verbnet agid wordnet cmudict imagenet wordfreq output ino" << std::endl;
   std::cout << "verbnet  :: path to a VerbNet data directory" << std::endl;
   std::cout << "agid     :: path to an AGID infl.txt file" << std::endl;
   std::cout << "wordnet  :: path to a WordNet prolog data directory" << std::endl;
   std::cout << "cmudict  :: path to a CMUDICT pronunciation file" << std::endl;
   std::cout << "imagenet :: path to an ImageNet urls.txt file" << std::endl;
+  std::cout << "wordfreq :: path to a word frequency CSV file" << std::endl;
   std::cout << "output   :: datafile output path" << std::endl;
   std::cout << "ino      :: imagenet directory output path" << std::endl;
 }
 
 int main(int argc, char** argv)
 {
-  if (argc == 8)
+  if (argc == 9)
   {
     try
     {
-      verbly::generator::generator app(argv[1], argv[2], argv[3], argv[4], argv[5], argv[6], argv[7]);
+      verbly::generator::generator app(argv[1], argv[2], argv[3], argv[4], argv[5], argv[6], argv[7], argv[8]);
 
       try
       {
@@ -162,7 +162,8 @@ CREATE TABLE `forms` (
   `proper` SMALLINT NOT NULL,
   `length` SMALLINT NOT NULL,
   `anagram_set_id` INTEGER NOT NULL,
-  `reverse_form_id` INTEGER NOT NULL
+  `reverse_form_id` INTEGER NOT NULL,
+  `frequency` INTEGER
 );
 
 CREATE UNIQUE INDEX `form_by_string` ON `forms`(`form`);