implemented pruning of subsumed ngrams

proycon · proycon · commit 33957eddbeae · 2020-04-10T20:12:46.000+02:00
diff --git a/colibricore_classes.in.pxd b/colibricore_classes.in.pxd
@@ -331,6 +331,7 @@ cdef extern from "patternmodel.h":
         bool DOREMOVEFLEXGRAMS
         bool DORESET
         int PRUNENONSUBSUMED
+        int PRUNESUBSUMED
 
     cdef cppclass IndexedDataHandler:
         unsigned int count(IndexedData &)
diff --git a/colibricore_wrapper.in.pyx b/colibricore_wrapper.in.pyx
@@ -899,7 +899,8 @@ cdef class PatternModelOptions:
     * DOREMOVESKIPGRAMS - Remove skipgrams from the model
     * DOREMOVEFLEXGRAMS - Remove flexgrams from the model
     * DORESET - Reset all counts before training
-    * PRUNENONSUBSUMED - Prune all n-grams up to this length that are not subsumed by higher-order ngrams
+    * PRUNENONSUBSUMED - Prune all n-grams up to this length that are *NOT* subsumed by higher-order ngrams
+    * PRUNESUBSUMED - Prune all n-grams up to this length that are subsumed by higher-order ngrams
     * DEBUG
     * QUIET (default: False)
 
@@ -950,6 +951,8 @@ cdef class PatternModelOptions:
             self.coptions.QUIET = value
         elif key == 'PRUNENONSUBSUMED':
             self.coptions.PRUNENONSUBSUMED = value
+        elif key == 'PRUNESUBSUMED':
+            self.coptions.PRUNESUBSUMED = value
         else:
             raise KeyError
 
@@ -990,6 +993,8 @@ cdef class PatternModelOptions:
             return self.coptions.QUIET
         elif key == 'PRUNENONSUBSUMED':
             return self.coptions.PRUNENONSUBSUMED
+        elif key == 'PRUNESUBSUMED':
+            return self.coptions.PRUNESUBSUMED
         else:
             raise KeyError
 
diff --git a/configure.ac b/configure.ac
@@ -4,7 +4,7 @@
 # $URL: https://ilk.uvt.nl/svn/sources/ucto/trunk/configure.ac $
 
 AC_PREREQ([2.67])
-AC_INIT([colibri-core],[2.5.3],[proycon@anaproy.nl])
+AC_INIT([colibri-core],[2.5.4],[proycon@anaproy.nl])
 AC_CONFIG_SRCDIR([configure.ac])
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_HEADER([config.h])
diff --git a/include/patternmodel.h b/include/patternmodel.h
@@ -141,7 +141,8 @@ class PatternModelOptions {
         bool DOREVERSEINDEX; ///< Obsolete now, only here for backward-compatibility with v1
         bool DOPATTERNPERLINE; ///< Assume each line contains one integral pattern, rather than actively extracting all subpatterns on a line (default: false)
 
-        int PRUNENONSUBSUMED; //< Prune all n-grams that are not subsumed by higher-order ngrams
+        int PRUNENONSUBSUMED; //< Prune all n-grams that are **NOT** subsumed by higher-order ngrams
+        int PRUNESUBSUMED; //< Prune all n-grams that are subsumed by higher-order ngrams
 
         bool DOREMOVEINDEX; ///< Do not load index information (for indexed models), loads just the patterns without any counts
         bool DOREMOVENGRAMS; ///< Remove n-grams from the model upon loading it
@@ -178,7 +179,8 @@ class PatternModelOptions {
             DOREMOVESKIPGRAMS = false;
             DOREMOVEFLEXGRAMS = false;
 
-            PRUNENONSUBSUMED = false;
+            PRUNENONSUBSUMED = 0;
+            PRUNESUBSUMED = 0;
 
             DEBUG = false;
             QUIET = false;
@@ -210,6 +212,7 @@ class PatternModelOptions {
             DOREMOVEFLEXGRAMS = ref.DOREMOVEFLEXGRAMS;
 
             PRUNENONSUBSUMED = ref.PRUNENONSUBSUMED;
+            PRUNESUBSUMED = ref.PRUNESUBSUMED;
 
             DEBUG = ref.DEBUG;
             QUIET = ref.QUIET;
@@ -1209,6 +1212,28 @@ class PatternModel: public MapType, public PatternModelInterface {
                     if (!options.QUIET) std::cerr << " pruned " << prunednonsubsumed << " non-subsumed " << (n-1) << "-grams"  << std::endl;
                 }
             }
+            if (options.PRUNESUBSUMED) {
+                if (!options.QUIET) std::cerr << "Pruning subsumed n-grams"  << std::endl;
+                int end_n = options.PRUNESUBSUMED;
+                if ((end_n > options.MAXLENGTH)) end_n = options.MAXLENGTH;
+                for (int n = 2; n <= end_n; n++) {
+                    std::unordered_set<Pattern> subsumed;
+                    unsigned int prunedsubsumed = 0;
+                    PatternModel::iterator iter = this->begin();
+                    while (iter != this->end()) {
+                        const unsigned int pattern_n = iter->first.n();
+                        if (pattern_n == (unsigned int) n) {
+                            subngrams.clear();
+                            iter->first.ngrams(subngrams, n-1);
+                            for (std::vector<PatternPointer>::iterator iter2 = subngrams.begin(); iter2 != subngrams.end(); iter2++) subsumed.insert(Pattern(*iter2));
+                        }
+                        iter++;
+                    };
+                    prunedsubsumed += this->pruneinset(subsumed, n-1);
+                    if (!options.QUIET) std::cerr << " pruned " << prunedsubsumed << " subsumed " << (n-1) << "-grams"  << std::endl;
+                }
+
+            }
             if ((options.MINLENGTH > 1) && (options.DOSKIPGRAMS || options.DOSKIPGRAMS_EXHAUSTIVE))  {
                 unsigned int pruned = this->prunebylength(options.MINLENGTH-1);
                 if (!options.QUIET) std::cerr << " pruned " << pruned << " patterns below minimum length (" << options.MINLENGTH << ")"  << std::endl;
@@ -2081,6 +2106,34 @@ class PatternModel: public MapType, public PatternModelInterface {
             return pruned;
         }
 
+        /**
+         * Prune all patterns that are in the specified set.
+         * @param s The set containing the patterns not to prune
+         * @param _n The size constraint, limit to patterns of this size only (set to 0 for no constraint, default)
+         * @return the number of distinct patterns pruned
+         */
+        unsigned int pruneinset(const std::unordered_set<Pattern> & s, int _n) {
+            unsigned int pruned = 0;
+            if (s.empty()) {
+                return pruned;
+            }
+            PatternModel::iterator iter = this->begin();
+            while (iter != this->end()) {
+                const PatternType pattern = iter->first;
+                if ( (_n == 0) || (pattern.n() == (unsigned int) _n) ) {
+                    if (s.find(pattern) != s.end()) {
+                        //found in set
+                        iter = this->erase(iter);
+                        pruned++;
+                        continue;
+                    }
+                }
+                iter++;
+            };
+
+            return pruned;
+        }
+
         /**
          * Prune all patterns that are not in the second model
          * @return the number of distinct patterns pruned
diff --git a/setup.py b/setup.py
@@ -183,7 +183,7 @@ def read(fname):
     license = "GPL",
     keywords = "nlp computational_linguistics frequency ngram skipgram pmi cooccurrence linguistics",
     long_description=read('README.rst'),
-    version = '2.5.3',
+    version = '2.5.4',
     ext_modules = extensions,
     cmdclass = {'build_ext': build_ext},
     classifiers=[
diff --git a/src/patternmodeller.cpp b/src/patternmodeller.cpp
@@ -47,6 +47,7 @@ void usage() {
     cerr << "\t-W|--wordthreshold  <number>  Word occurrence threshold (secondary threshold): only count patterns in which the words/unigrams occur at least this many times, only effective when the primary " << endl;
     cerr << "\t                              occurrence threshold (-t) is lower than this threshold (default: disabled)" << endl;
     cerr << "\t-p|--prune <number>           Prune all lower-order n-grams below the specified order that are *NOT* subsumed by higher order n-grams (default: 0, disabled). Only effective when used with -l, usually set to equal values" << endl;
+    cerr << "\t   --prunesubsumed <number>           Prune all lower-order n-grams below the specified order that are subsumed by higher order n-grams (default: 0, disabled). Only effective when used with -l, usually set to equal values" << endl;
     cerr << "\t-s|--skipgrams                Compute skipgrams (costs extra memory and time)" << endl;
     cerr << "\t-y|--skipthreshold <number>   Occurrence threshold for skipgrams (overrides -t for skipgrams, defaults to -t). Skipgrams occurring less than this will be pruned. Value must be equal to or higher than -t." << endl;
     cerr << "\t-T|--skiptypes <number>       Skip type threshold (for use with -s): only skipgrams with at least x possible types for the skip will be considered, otherwise the skipgram " << endl;
