/* -*-C-*-
********************************************************************************
*
* File: permdawg.c (Formerly permdawg.c)
* Description: Scale word choices by a dictionary
* Author: Mark Seaman, OCR Technology
* Created: Fri Oct 16 14:37:00 1987
* Modified: Tue Jul 9 15:43:18 1991 (Mark Seaman) marks@hpgrlt
* Language: C
* Package: N/A
* Status: Reusable Software Component
*
* (c) Copyright 1987, Hewlett-Packard Company.
** Licensed under the Apache License, Version 2.0 (the "License");
** you may not use this file except in compliance with the License.
** You may obtain a copy of the License at
** http://www.apache.org/licenses/LICENSE-2.0
** Unless required by applicable law or agreed to in writing, software
** distributed under the License is distributed on an "AS IS" BASIS,
** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
** See the License for the specific language governing permissions and
** limitations under the License.
*
*********************************************************************************/
/*----------------------------------------------------------------------
I n c l u d e s
----------------------------------------------------------------------*/
#include "context.h"
#include "conversion.h"
#include "cutil.h"
#include "dawg.h"
#include "freelist.h"
#include "globals.h"
#include "ndminx.h"
#include "permdawg.h"
#include "permute.h"
#include "stopper.h"
#include "tordvars.h"
#include "tprintf.h"
#include "varable.h"
#include <ctype.h>
#include "dict.h"
#include "image.h"
/*----------------------------------------------------------------------
V a r i a b l e s
----------------------------------------------------------------------*/
BOOL_VAR(segment_dawg_debug, 0, "Debug mode for word segmentation");
double_VAR(segment_penalty_dict_case_bad, OK_WERD,
"Default score multiplier for word matches, which may have "
"case issues (lower is better).");
double_VAR(segment_penalty_dict_case_ok, GOOD_WERD,
"Score multiplier for word matches that have good case "
"(lower is better).");
double_VAR(segment_penalty_dict_frequent_word, FREQ_WERD,
"Score multiplier for word matches which have good case and are "
"frequent in the given language (lower is better).");
/*----------------------------------------------------------------------
F u n c t i o n s
----------------------------------------------------------------------*/
namespace tesseract {
static const float kPermDawgRatingPad = 5.0;
/**********************************************************************
* adjust_word
*
* Assign an adjusted value to a string that is a word. The value
* that this word choice has is based on case and punctuation rules.
**********************************************************************/
void Dict::adjust_word(WERD_CHOICE *word,
float *certainty_array) {
float adjust_factor;
float new_rating = word->rating();
if (segment_dawg_debug) {
tprintf ("Word: %s %4.2f ",
word->debug_string(getUnicharset()).string(), word->rating());
}
new_rating += RATING_PAD;
if (Context::case_ok(*word, getUnicharset())) {
if (freq_dawg_ != NULL && freq_dawg_->word_in_dawg(*word)) {
word->set_permuter(FREQ_DAWG_PERM);
new_rating *= segment_penalty_dict_frequent_word;
adjust_factor = segment_penalty_dict_frequent_word;
if (segment_dawg_debug)
tprintf(", F, %4.2f ", (double)segment_penalty_dict_frequent_word);
} else {
new_rating *= segment_penalty_dict_case_ok;
adjust_factor = segment_penalty_dict_case_ok;
if (segment_dawg_debug)
tprintf(", %4.2f ", (double)segment_penalty_dict_case_ok);
}
} else {
new_rating *= segment_penalty_dict_case_bad;
adjust_factor = segment_penalty_dict_case_bad;
if (segment_dawg_debug) {
tprintf(", C %4.2f ", (double)segment_penalty_dict_case_bad);
}
}
new_rating -= RATING_PAD;
word->set_rating(new_rating);
LogNewChoice(*word, adjust_factor, certainty_array, false);
if (segment_dawg_debug)
tprintf(" --> %4.2f\n", new_rating);
}
/**********************************************************************
* go_deeper_dawg_fxn
*
* If the choice being composed so far could be a dictionary word
* keep exploring choices.
**********************************************************************/
void Dict::go_deeper_dawg_fxn(
const char *debug, const BLOB_CHOICE_LIST_VECTOR &char_choices,
int char_choice_index,
const CHAR_FRAGMENT_INFO *prev_char_frag_info,
bool word_ending, WERD_CHOICE *word, float certainties[],
float *limit, WERD_CHOICE *best_choice, void *void_more_args) {
DawgArgs *more_args = reinterpret_cast<DawgArgs*>(void_more_args);
int word_index = word->length() - 1;
// There are two modes for deciding whether to go deeper: regular dawg
// permuter mode and the special ambigs mode. If *limit is <= 0.0 the
// function switches to the ambigs mode (this is the case when
// dawg_permute_and_select() function is called from NoDangerousAmbigs()) and
// only searches for the first choice that has a rating better than *limit
// (in this case ratings are fake, since the real ratings can not be < 0).
// Modification of the hyphen state is turned off in the ambigs mode.
// When in the regular dawg permuter mode, the function explores all the
// possible words and chooses the one with the best rating. The letters with
// ratings that are far worse than the ones seen so far are pruned out.
bool ambigs_mode = (*limit <= 0.0);
if (ambigs_mode) {
if (best_choice->rating() < *limit) return;
} else {
// Prune bad subwords
if (more_args->rating_array[word_index] == NO_RATING) {
more_args->rating_array[word_index] = word->rating();
} else {
float permdawg_limit = more_args->rating_array[word_index] *
more_args->rating_margin + kPermDawgRatingPad;
if (permdawg_limit < word->rating()) {
if (segment_dawg_debug) {
tprintf("early pruned word rating=%4.2f,"
" permdawg_limit=%4.2f, word=%s\n", word->rating(),
permdawg_limit, word->debug_string(getUnicharset()).string());
}
return;
}
}
}
// Deal with hyphens
if (word_ending && has_hyphen_end(*word) && !ambigs_mode) {
if (segment_dawg_debug)
tprintf("new hyphen choice = %s\n",
word->debug_string(getUnicharset()).string());
word->set_permuter(more_args->permuter);
adjust_word(word, certainties);
set_hyphen_word(*word, *(more_args->active_dawgs),
*(more_args->constraints));
update_best_choice(*word, best_choice);
} else { // Look up char in DAWG
// TODO(daria): update the rest of the code that specifies alternative
// letter_is_okay_ functions (e.g. TessCharNgram class) to work with
// multi-byte unichars and/or unichar ids.
// If the current unichar is an ngram first try calling
// letter_is_okay() for each unigram it contains separately.
UNICHAR_ID orig_uch_id = word->unichar_id(word_index);
bool checked_unigrams = false;
if (getUnicharset().get_isngram(orig_uch_id)) {
if (segment_dawg_debug) {
tprintf("checking unigrams in an ngram %s\n",
getUnicharset().debug_str(orig_uch_id).string());
}
int orig_num_fragments = word->fragment_length(word_index);
int num_unigrams = 0;
word->remove_last_unichar_id();
const char *ngram_str = getUnicharset().id_to_unichar(orig_uch_id);
const char *ngram_str_end = ngram_str + strlen(ngram_str);
const char *ngram_ptr = ngram_str;
bool unigrams_ok = true;
// Construct DawgArgs that reflect the current state.
DawgInfoVector unigram_active_dawgs = *(more_args->active_dawgs);
DawgInfoVector unigram_constraints = *(more_args->constraints);
DawgInfoVector unigram_updated_active_dawgs;
DawgInfoVector unigram_updated_constraints;
DawgArgs unigram_dawg_args(&unigram_active_dawgs, &unigram_constraints,
&unigram_updated_active_dawgs,
&unigram_updated_constraints, 0.0);
unigram_dawg_args.permuter = more_args->permuter;
// Check unigrams in the ngram with letter_is_okay().
while (unigrams_ok && ngram_ptr < ngram_str_end) {
int step = getUnicharset().step(ngram_ptr);
UNICHAR_ID uch_id = (step <= 0) ? INVALID_UNICHAR_ID :
getUnicharset().unichar_to_id(ngram_ptr, step);
ngram_ptr += step;
++num_unigrams;
word->append_unichar_id(uch_id, 1, 0.0, 0.0);
unigrams_ok = unigrams_ok && (this->*letter_is_okay_)(
&unigram_dawg_args, word_index+num_unigrams-1, word,
word_ending && (ngram_ptr == ngram_str_end));
(*unigram_dawg_args.active_dawgs) =
*(unigram_dawg_args.updated_active_dawgs);
(*unigram_dawg_args.constraints) =
*(unigram_dawg_args.updated_constraints);
if (segment_dawg_debug) {
tprintf("unigram %s is %s\n",
getUnicharset().debug_str(uch_id).string(),
unigrams_ok ? "OK" : "not OK");
}
}
// Restore the word and copy the updated dawg state if needed.
while (num_unigrams-- > 0) word->remove_last_unichar_id();
word->append_unichar_id_space_allocated(
orig_uch_id, orig_num_fragments, 0.0, 0.0);
if (unigrams_ok) {
checked_unigrams = true;
more_args->permuter = unigram_dawg_args.permuter;
*(more_args->updated_active_dawgs) =
*(unigram_dawg_args.updated_active_dawgs);
*(more_args->updated_constraints) =
*(unigram_dawg_args.updated_constraints);
}
}
// Check which dawgs from dawgs_ vector contain the word
// up to and including the current unichar.
if (checked_unigrams ||
(this->*letter_is_okay_)(more_args, word_index, word, word_ending)) {
// Add a new word choice
if (word_ending) {
if (segment_dawg_debug) {
tprintf("found word = %s\n",
word->debug_string(getUnicharset()).string());
}
WERD_CHOICE *adjusted_word = word;
WERD_CHOICE hyphen_tail_word;
if (!ambigs_mode && hyphen_base_size() > 0) {
hyphen_tail_word = *word;
remove_hyphen_head(&hyphen_tail_word);
adjusted_word = &hyphen_tail_word;
}
adjusted_word->set_permuter(more_args->permuter);
if (!ambigs_mode) {
adjust_word(adjusted_word, &certainties[hyphen_base_size()]);
}
update_best_choice(*adjusted_word, best_choice);
} else { // search the next letter
// Make updated_* point to the next entries in the DawgInfoVector
// arrays (that were originally created in dawg_permute_and_select)
++(more_args->updated_active_dawgs);
++(more_args->updated_constraints);
// Make active_dawgs and constraints point to the updated ones.
++(more_args->active_dawgs);
++(more_args->constraints);
permute_choices(debug, char_choices, char_choice_index + 1,
prev_char_frag_info, word, certainties, limit,
best_choice, more_args);
// Restore previous state to explore another letter in this position.
--(more_args->updated_active_dawgs);
--(more_args->updated_constraints);
--(more_args->active_dawgs);
--(more_args->constraints);
}
} else {
if (segment_dawg_debug) {
tprintf("last unichar not OK at index %d in %s\n",
word_index, word->debug_string(getUnicharset()).string());
}
}
}
}
/**********************************************************************
* dawg_permute_and_select
*
* Recursively explore all the possible character combinations in
* the given char_choices. Use go_deeper_dawg_fxn() to search all the
* dawgs in the dawgs_ vector in parallel and discard invalid words.
*
* Allocate and return a WERD_CHOICE with the best valid word found.
* **********************************************************************/
WERD_CHOICE *Dict::dawg_permute_and_select(
const BLOB_CHOICE_LIST_VECTOR &char_choices, float rating_limit) {
WERD_CHOICE *best_choice = new WERD_CHOICE();
best_choice->make_bad();
best_choice->set_rating(rating_limit);
if (char_choices.length() == 0) return best_choice;
DawgInfoVector *active_dawgs = new DawgInfoVector[char_choices.length() + 1];
DawgInfoVector *constraints = new DawgInfoVector[char_choices.length() + 1];
init_active_dawgs(&(active_dawgs[0]));
init_constraints(&(constraints[0]));
DawgArgs dawg_args(&(active_dawgs[0]), &(constraints[0]),
&(active_dawgs[1]), &(constraints[1]),
(segment_penalty_dict_case_bad /
segment_penalty_dict_case_ok));
WERD_CHOICE word(MAX_WERD_LENGTH);
copy_hyphen_info(&word);
// Discard rating and certainty of the hyphen base (if any).
word.set_rating(0.0);
word.set_certainty(0.0);
if (word.length() + char_choices.length() > MAX_WERD_LENGTH) {
return best_choice; // the word is too long to permute
}
float certainties[MAX_WERD_LENGTH];
this->go_deeper_fxn_ = &tesseract::Dict::go_deeper_dawg_fxn;
permute_choices(segment_dawg_debug ? "segment_dawg_debug" : NULL,
char_choices, 0, NULL, &word, certainties,
&rating_limit, best_choice, &dawg_args);
delete[] active_dawgs;
delete[] constraints;
return best_choice;
}
// Fill the given active_dawgs vector with dawgs that could contain the
// beginning of the word. If hyphenated() returns true, copy the entries
// from hyphen_active_dawgs_ instead.
void Dict::init_active_dawgs(DawgInfoVector *active_dawgs) {
int i;
if (hyphenated()) {
*active_dawgs = hyphen_active_dawgs_;
if (dawg_debug_level >= 3) {
for (i = 0; i < hyphen_active_dawgs_.size(); ++i) {
tprintf("Adding hyphen beginning dawg [%d, " REFFORMAT "]\n",
hyphen_active_dawgs_[i].dawg_index,
hyphen_active_dawgs_[i].ref);
}
}
} else {
for (i = 0; i < dawgs_.length(); ++i) {
if (kBeginningDawgsType[(dawgs_[i])->type()]) {
*active_dawgs += DawgInfo(i, NO_EDGE);
if (dawg_debug_level >= 3) {
tprintf("Adding beginning dawg [%d, " REFFORMAT "]\n", i, NO_EDGE);
}
}
}
}
}
// If hyphenated() returns true, copy the entries from hyphen_constraints_
// into the given constraints vector.
void Dict::init_constraints(DawgInfoVector *constraints) {
if (hyphenated()) {
*constraints = hyphen_constraints_;
if (dawg_debug_level >= 3) {
for (int i = 0; i < hyphen_constraints_.size(); ++i) {
tprintf("Adding hyphen constraint [%d, " REFFORMAT "]\n",
hyphen_constraints_[i].dawg_index,
hyphen_constraints_[i].ref);
}
}
}
}
} // namespace tesseract