/**********************************************************************
* fill_filtered_fragment_list
*
* Filter the fragment list so that the filtered_choices only contain
* fragments that are in the correct position. choices is the list
* that we are going to filter. fragment_pos is the position in the
* fragment that we are looking for and num_frag_parts is the the
* total number of pieces. The result will be appended to
* filtered_choices.
**********************************************************************/
void Wordrec::fill_filtered_fragment_list(BLOB_CHOICE_LIST *choices,
int fragment_pos,
int num_frag_parts,
BLOB_CHOICE_LIST *filtered_choices) {
BLOB_CHOICE_IT filtered_choices_it(filtered_choices);
BLOB_CHOICE_IT choices_it(choices);
for (choices_it.mark_cycle_pt(); !choices_it.cycled_list();
choices_it.forward()) {
UNICHAR_ID choice_unichar_id = choices_it.data()->unichar_id();
const CHAR_FRAGMENT *frag = unicharset.get_fragment(choice_unichar_id);
if (frag != NULL && frag->get_pos() == fragment_pos &&
frag->get_total() == num_frag_parts) {
// Recover the unichar_id of the unichar that this fragment is
// a part of
BLOB_CHOICE *b = new BLOB_CHOICE(*choices_it.data());
int original_unichar = unicharset.unichar_to_id(frag->get_unichar());
b->set_unichar_id(original_unichar);
filtered_choices_it.add_to_end(b);
}
}
filtered_choices->sort(SortByUnicharID<BLOB_CHOICE>);
}
// Returns true if *this and other agree on the baseline and x-height
// to within some tolerance based on a given estimate of the x-height.
bool BLOB_CHOICE::PosAndSizeAgree(const BLOB_CHOICE& other, float x_height,
bool debug) const {
double baseline_diff = fabs(yshift() - other.yshift());
if (baseline_diff > kMaxBaselineDrift * x_height) {
if (debug) {
tprintf("Baseline diff %g for %d v %d\n",
baseline_diff, unichar_id_, other.unichar_id_);
}
return false;
}
double this_range = max_xheight() - min_xheight();
double other_range = other.max_xheight() - other.min_xheight();
double denominator = ClipToRange(MIN(this_range, other_range),
1.0, kMaxOverlapDenominator * x_height);
double overlap = MIN(max_xheight(), other.max_xheight()) -
MAX(min_xheight(), other.min_xheight());
overlap /= denominator;
if (debug) {
tprintf("PosAndSize for %d v %d: bl diff = %g, ranges %g, %g / %g ->%g\n",
unichar_id_, other.unichar_id_, baseline_diff,
this_range, other_range, denominator, overlap);
}
return overlap >= kMinXHeightMatch;
}
/**********************************************************************
* BLOB_CHOICE::BLOB_CHOICE
*
* Constructor to build a BLOB_CHOICE from another BLOB_CHOICE.
**********************************************************************/
BLOB_CHOICE::BLOB_CHOICE(const BLOB_CHOICE &other) {
unichar_id_ = other.unichar_id();
rating_ = other.rating();
certainty_ = other.certainty();
config_ = other.config();
script_id_ = other.script_id();
}
// Returns a bigger MATRIX with a new column and row in the matrix in order
// to split the blob at the given (ind,ind) diagonal location.
// Entries are relocated to the new MATRIX using the transformation defined
// by MATRIX_COORD::MapForSplit.
// Transfers the pointer data to the new MATRIX and deletes *this.
MATRIX* MATRIX::ConsumeAndMakeBigger(int ind) {
int dim = dimension();
int band_width = bandwidth();
// Check to see if bandwidth needs expanding.
for (int col = ind; col >= 0 && col > ind - band_width; --col) {
if (array_[col * band_width + band_width - 1] != empty_) {
++band_width;
break;
}
}
MATRIX* result = new MATRIX(dim + 1, band_width);
for (int col = 0; col < dim; ++col) {
for (int row = col; row < dim && row < col + bandwidth(); ++row) {
MATRIX_COORD coord(col, row);
coord.MapForSplit(ind);
BLOB_CHOICE_LIST* choices = get(col, row);
if (choices != NULL) {
// Correct matrix location on each choice.
BLOB_CHOICE_IT bc_it(choices);
for (bc_it.mark_cycle_pt(); !bc_it.cycled_list(); bc_it.forward()) {
BLOB_CHOICE* choice = bc_it.data();
choice->set_matrix_cell(coord.col, coord.row);
}
ASSERT_HOST(coord.Valid(*result));
result->put(coord.col, coord.row, choices);
}
}
}
delete this;
return result;
}
// Returns true if there are any real classification results.
bool MATRIX::Classified(int col, int row, int wildcard_id) const {
if (get(col, row) == NOT_CLASSIFIED) return false;
BLOB_CHOICE_IT b_it(get(col, row));
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
BLOB_CHOICE* choice = b_it.data();
if (choice->IsClassified())
return true;
}
return false;
}
// Helper to find the BLOB_CHOICE in the bc_list that matches the given
// unichar_id, or NULL if there is no match.
BLOB_CHOICE* FindMatchingChoice(UNICHAR_ID char_id,
BLOB_CHOICE_LIST* bc_list) {
// Find the corresponding best BLOB_CHOICE.
BLOB_CHOICE_IT choice_it(bc_list);
for (choice_it.mark_cycle_pt(); !choice_it.cycled_list();
choice_it.forward()) {
BLOB_CHOICE* choice = choice_it.data();
if (choice->unichar_id() == char_id) {
return choice;
}
}
return NULL;
}
/**
* append_choices
*
* Checks to see whether or not the next choice is worth appending to
* the word being generated. If so then keeps going deeper into the word.
*
* This function assumes that Dict::go_deeper_fxn_ is set.
*/
void Dict::append_choices(
const char *debug,
const BLOB_CHOICE_LIST_VECTOR &char_choices,
const BLOB_CHOICE &blob_choice,
int char_choice_index,
const CHAR_FRAGMENT_INFO *prev_char_frag_info,
WERD_CHOICE *word,
float certainties[],
float *limit,
WERD_CHOICE *best_choice,
int *attempts_left,
void *more_args) {
int word_ending =
(char_choice_index == char_choices.length() - 1) ? true : false;
// Deal with fragments.
CHAR_FRAGMENT_INFO char_frag_info;
if (!fragment_state_okay(blob_choice.unichar_id(), blob_choice.rating(),
blob_choice.certainty(), prev_char_frag_info, debug,
word_ending, &char_frag_info)) {
return; // blob_choice must be an invalid fragment
}
// Search the next letter if this character is a fragment.
if (char_frag_info.unichar_id == INVALID_UNICHAR_ID) {
permute_choices(debug, char_choices, char_choice_index + 1,
&char_frag_info, word, certainties, limit,
best_choice, attempts_left, more_args);
return;
}
// Add the next unichar.
float old_rating = word->rating();
float old_certainty = word->certainty();
uint8_t old_permuter = word->permuter();
certainties[word->length()] = char_frag_info.certainty;
word->append_unichar_id_space_allocated(
char_frag_info.unichar_id, char_frag_info.num_fragments,
char_frag_info.rating, char_frag_info.certainty);
// Explore the next unichar.
(this->*go_deeper_fxn_)(debug, char_choices, char_choice_index,
&char_frag_info, word_ending, word, certainties,
limit, best_choice, attempts_left, more_args);
// Remove the unichar we added to explore other choices in it's place.
word->remove_last_unichar_id();
word->set_rating(old_rating);
word->set_certainty(old_certainty);
word->set_permuter(old_permuter);
}
/**
* BLOB_CHOICE::BLOB_CHOICE
*
* Constructor to build a BLOB_CHOICE from another BLOB_CHOICE.
*/
BLOB_CHOICE::BLOB_CHOICE(const BLOB_CHOICE &other) {
unichar_id_ = other.unichar_id();
rating_ = other.rating();
certainty_ = other.certainty();
config_ = other.config();
config2_ = other.config2();
script_id_ = other.script_id();
language_model_state_ = NULL;
}
/**
* BLOB_CHOICE::BLOB_CHOICE
*
* Constructor to build a BLOB_CHOICE from another BLOB_CHOICE.
*/
BLOB_CHOICE::BLOB_CHOICE(const BLOB_CHOICE &other) {
unichar_id_ = other.unichar_id();
rating_ = other.rating();
certainty_ = other.certainty();
fontinfo_id_ = other.fontinfo_id();
fontinfo_id2_ = other.fontinfo_id2();
script_id_ = other.script_id();
matrix_cell_ = other.matrix_cell_;
min_xheight_ = other.min_xheight_;
max_xheight_ = other.max_xheight_;
yshift_ = other.yshift();
classifier_ = other.classifier_;
}
/**
* BLOB_CHOICE::BLOB_CHOICE
*
* Constructor to build a BLOB_CHOICE from another BLOB_CHOICE.
*/
BLOB_CHOICE::BLOB_CHOICE(const BLOB_CHOICE &other) {
unichar_id_ = other.unichar_id();
rating_ = other.rating();
certainty_ = other.certainty();
fontinfo_id_ = other.fontinfo_id();
fontinfo_id2_ = other.fontinfo_id2();
script_id_ = other.script_id();
language_model_state_ = NULL;
min_xheight_ = other.min_xheight_;
max_xheight_ = other.max_xheight_;
adapted_ = other.adapted_;
}
// Creates a fake blob choice from the combination of the given fragments.
// unichar is the class to be made from the combination,
// expanded_fragment_lengths[choice_index] is the number of fragments to use.
// old_choices[choice_index] has the classifier output for each fragment.
// choice index initially indexes the last fragment and should be decremented
// expanded_fragment_lengths[choice_index] times to get the earlier fragments.
// Guarantees to return something non-null, or abort!
BLOB_CHOICE* Wordrec::rebuild_fragments(
const char* unichar,
const char* expanded_fragment_lengths,
int choice_index,
BLOB_CHOICE_LIST_VECTOR *old_choices) {
float rating = 0.0f;
float certainty = 0.0f;
inT16 min_xheight = -MAX_INT16;
inT16 max_xheight = MAX_INT16;
for (int fragment_pieces = expanded_fragment_lengths[choice_index] - 1;
fragment_pieces >= 0; --fragment_pieces, --choice_index) {
// Get a pointer to the classifier results from the old_choices.
BLOB_CHOICE_LIST *current_choices = old_choices->get(choice_index);
// Populate fragment with updated values and look for the
// fragment with the same values in current_choices.
// Update rating and certainty of the character being composed.
CHAR_FRAGMENT fragment;
fragment.set_all(unichar, fragment_pieces,
expanded_fragment_lengths[choice_index], false);
BLOB_CHOICE_IT choice_it(current_choices);
for (choice_it.mark_cycle_pt(); !choice_it.cycled_list();
choice_it.forward()) {
BLOB_CHOICE* choice = choice_it.data();
const CHAR_FRAGMENT *current_fragment =
getDict().getUnicharset().get_fragment(choice->unichar_id());
if (current_fragment && fragment.equals(current_fragment)) {
rating += choice->rating();
if (choice->certainty() < certainty) {
certainty = choice->certainty();
}
IntersectRange(choice->min_xheight(), choice->max_xheight(),
&min_xheight, &max_xheight);
break;
}
}
if (choice_it.cycled_list()) {
print_ratings_list("Failure", current_choices, unicharset);
tprintf("Failed to find fragment %s at index=%d\n",
fragment.to_string().string(), choice_index);
}
ASSERT_HOST(!choice_it.cycled_list()); // Be sure we found the fragment.
}
return new BLOB_CHOICE(getDict().getUnicharset().unichar_to_id(unichar),
rating, certainty, -1, -1, 0,
min_xheight, max_xheight, false);
}
LIST call_matcher( //call a matcher
TBLOB *ptblob, //previous
TBLOB *tessblob, //blob to match
TBLOB *ntblob, //next
void *, //unused parameter
TEXTROW * //always null anyway
) {
PBLOB *pblob; //converted blob
PBLOB *blob; //converted blob
PBLOB *nblob; //converted blob
LIST result; //tess output
BLOB_CHOICE *choice; //current choice
BLOB_CHOICE_LIST ratings; //matcher result
BLOB_CHOICE_IT it; //iterator
char choice_lengths[2] = {0, 0};
blob = make_ed_blob (tessblob);//convert blob
if (blob == NULL) {
// Since it is actually possible to get a NULL blob here, due to invalid
// segmentations, fake a really bad classification.
choice_lengths[0] = strlen(unicharset.id_to_unichar(1));
return append_choice(NULL, unicharset.id_to_unichar(1), choice_lengths,
static_cast<float>(MAX_NUM_INT_FEATURES),
static_cast<float>(kReallyBadCertainty), 0);
}
pblob = ptblob != NULL ? make_ed_blob (ptblob) : NULL;
nblob = ntblob != NULL ? make_ed_blob (ntblob) : NULL;
(*tess_matcher) (pblob, blob, nblob, tess_word, tess_denorm, ratings);
//match it
delete blob; //don't need that now
if (pblob != NULL)
delete pblob;
if (nblob != NULL)
delete nblob;
it.set_to_list (&ratings); //get list
result = NULL;
for (it.mark_cycle_pt (); !it.cycled_list (); it.forward ()) {
choice = it.data ();
choice_lengths[0] = strlen(choice->unichar ());
result = append_choice (result, choice->unichar (),
choice_lengths, choice->rating (),
choice->certainty (), choice->config ());
}
return result; //converted list
}
/**********************************************************************
* merge_and_put_fragment_lists
*
* Merge the fragment lists in choice_lists and append it to the
* ratings matrix.
**********************************************************************/
void Wordrec::merge_and_put_fragment_lists(inT16 row, inT16 column,
inT16 num_frag_parts,
BLOB_CHOICE_LIST *choice_lists,
MATRIX *ratings) {
BLOB_CHOICE_IT *choice_lists_it = new BLOB_CHOICE_IT[num_frag_parts];
for (int i = 0; i < num_frag_parts; i++) {
choice_lists_it[i].set_to_list(&choice_lists[i]);
choice_lists_it[i].mark_cycle_pt();
}
BLOB_CHOICE_LIST *merged_choice = ratings->get(row, column);
if (merged_choice == NULL)
merged_choice = new BLOB_CHOICE_LIST;
bool end_of_list = false;
BLOB_CHOICE_IT merged_choice_it(merged_choice);
while (!end_of_list) {
// Find the maximum unichar_id of the current entry the iterators
// are pointing at
UNICHAR_ID max_unichar_id = choice_lists_it[0].data()->unichar_id();
for (int i = 0; i < num_frag_parts; i++) {
UNICHAR_ID unichar_id = choice_lists_it[i].data()->unichar_id();
if (max_unichar_id < unichar_id) {
max_unichar_id = unichar_id;
}
}
// Move the each iterators until it gets to an entry that has a
// value greater than or equal to max_unichar_id
for (int i = 0; i < num_frag_parts; i++) {
UNICHAR_ID unichar_id = choice_lists_it[i].data()->unichar_id();
while (!choice_lists_it[i].cycled_list() &&
unichar_id < max_unichar_id) {
choice_lists_it[i].forward();
unichar_id = choice_lists_it[i].data()->unichar_id();
}
if (choice_lists_it[i].cycled_list()) {
end_of_list = true;
break;
}
}
if (end_of_list)
break;
// Checks if the fragments are parts of the same character
UNICHAR_ID first_unichar_id = choice_lists_it[0].data()->unichar_id();
bool same_unichar = true;
for (int i = 1; i < num_frag_parts; i++) {
UNICHAR_ID unichar_id = choice_lists_it[i].data()->unichar_id();
if (unichar_id != first_unichar_id) {
same_unichar = false;
break;
}
}
if (same_unichar) {
// Add the merged character to the result
UNICHAR_ID merged_unichar_id = first_unichar_id;
GenericVector<ScoredFont> merged_fonts =
choice_lists_it[0].data()->fonts();
float merged_min_xheight = choice_lists_it[0].data()->min_xheight();
float merged_max_xheight = choice_lists_it[0].data()->max_xheight();
float positive_yshift = 0, negative_yshift = 0;
int merged_script_id = choice_lists_it[0].data()->script_id();
BlobChoiceClassifier classifier = choice_lists_it[0].data()->classifier();
float merged_rating = 0, merged_certainty = 0;
for (int i = 0; i < num_frag_parts; i++) {
float rating = choice_lists_it[i].data()->rating();
float certainty = choice_lists_it[i].data()->certainty();
if (i == 0 || certainty < merged_certainty)
merged_certainty = certainty;
merged_rating += rating;
choice_lists_it[i].forward();
if (choice_lists_it[i].cycled_list())
end_of_list = true;
IntersectRange(choice_lists_it[i].data()->min_xheight(),
choice_lists_it[i].data()->max_xheight(),
&merged_min_xheight, &merged_max_xheight);
float yshift = choice_lists_it[i].data()->yshift();
if (yshift > positive_yshift) positive_yshift = yshift;
if (yshift < negative_yshift) negative_yshift = yshift;
// Use the min font rating over the parts.
// TODO(rays) font lists are unsorted. Need to be faster?
const GenericVector<ScoredFont>& frag_fonts =
choice_lists_it[i].data()->fonts();
for (int f = 0; f < frag_fonts.size(); ++f) {
int merged_f = 0;
for (merged_f = 0; merged_f < merged_fonts.size() &&
merged_fonts[merged_f].fontinfo_id != frag_fonts[f].fontinfo_id;
++merged_f) {}
if (merged_f == merged_fonts.size()) {
merged_fonts.push_back(frag_fonts[f]);
} else if (merged_fonts[merged_f].score > frag_fonts[f].score) {
merged_fonts[merged_f].score = frag_fonts[f].score;
}
}
}
float merged_yshift = positive_yshift != 0
? (negative_yshift != 0 ? 0 : positive_yshift)
: negative_yshift;
BLOB_CHOICE* choice = new BLOB_CHOICE(merged_unichar_id,
merged_rating,
merged_certainty,
merged_script_id,
merged_min_xheight,
merged_max_xheight,
merged_yshift,
classifier);
choice->set_fonts(merged_fonts);
merged_choice_it.add_to_end(choice);
}
}
if (classify_debug_level)
print_ratings_list("Merged Fragments", merged_choice,
unicharset);
if (merged_choice->empty())
delete merged_choice;
else
ratings->put(row, column, merged_choice);
delete [] choice_lists_it;
}
void Dict::ReplaceAmbig(int wrong_ngram_begin_index, int wrong_ngram_size,
UNICHAR_ID correct_ngram_id, WERD_CHOICE *werd_choice,
MATRIX *ratings) {
int num_blobs_to_replace = 0;
int begin_blob_index = 0;
int i;
// Rating and certainty for the new BLOB_CHOICE are derived from the
// replaced choices.
float new_rating = 0.0f;
float new_certainty = 0.0f;
BLOB_CHOICE* old_choice = nullptr;
for (i = 0; i < wrong_ngram_begin_index + wrong_ngram_size; ++i) {
if (i >= wrong_ngram_begin_index) {
int num_blobs = werd_choice->state(i);
int col = begin_blob_index + num_blobs_to_replace;
int row = col + num_blobs - 1;
BLOB_CHOICE_LIST* choices = ratings->get(col, row);
ASSERT_HOST(choices != nullptr);
old_choice = FindMatchingChoice(werd_choice->unichar_id(i), choices);
ASSERT_HOST(old_choice != nullptr);
new_rating += old_choice->rating();
new_certainty += old_choice->certainty();
num_blobs_to_replace += num_blobs;
} else {
begin_blob_index += werd_choice->state(i);
}
}
new_certainty /= wrong_ngram_size;
// If there is no entry in the ratings matrix, add it.
MATRIX_COORD coord(begin_blob_index,
begin_blob_index + num_blobs_to_replace - 1);
if (!coord.Valid(*ratings)) {
ratings->IncreaseBandSize(coord.row - coord.col + 1);
}
if (ratings->get(coord.col, coord.row) == nullptr)
ratings->put(coord.col, coord.row, new BLOB_CHOICE_LIST);
BLOB_CHOICE_LIST* new_choices = ratings->get(coord.col, coord.row);
BLOB_CHOICE* choice = FindMatchingChoice(correct_ngram_id, new_choices);
if (choice != nullptr) {
// Already there. Upgrade if new rating better.
if (new_rating < choice->rating())
choice->set_rating(new_rating);
if (new_certainty < choice->certainty())
choice->set_certainty(new_certainty);
// DO NOT SORT!! It will mess up the iterator in LanguageModel::UpdateState.
} else {
// Need a new choice with the correct_ngram_id.
choice = new BLOB_CHOICE(*old_choice);
choice->set_unichar_id(correct_ngram_id);
choice->set_rating(new_rating);
choice->set_certainty(new_certainty);
choice->set_classifier(BCC_AMBIG);
choice->set_matrix_cell(coord.col, coord.row);
BLOB_CHOICE_IT it (new_choices);
it.add_to_end(choice);
}
// Remove current unichar from werd_choice. On the last iteration
// set the correct replacement unichar instead of removing a unichar.
for (int replaced_count = 0; replaced_count < wrong_ngram_size;
++replaced_count) {
if (replaced_count + 1 == wrong_ngram_size) {
werd_choice->set_blob_choice(wrong_ngram_begin_index,
num_blobs_to_replace, choice);
} else {
werd_choice->remove_unichar_id(wrong_ngram_begin_index + 1);
}
}
if (stopper_debug_level >= 1) {
werd_choice->print("ReplaceAmbig() ");
tprintf("Modified blob_choices: ");
print_ratings_list("\n", new_choices, getUnicharset());
}
}
void LMPainPoints::GenerateFromPath(float rating_cert_scale,
ViterbiStateEntry *vse,
WERD_RES *word_res) {
ViterbiStateEntry *curr_vse = vse;
BLOB_CHOICE *curr_b = vse->curr_b;
// The following pain point generation and priority calculation approaches
// prioritize exploring paths with low average rating of the known part of
// the path, while not relying on the ratings of the pieces to be combined.
//
// A pain point to combine the neighbors is generated for each pair of
// neighboring blobs on the path (the path is represented by vse argument
// given to GenerateFromPath()). The priority of each pain point is set to
// the average rating (per outline length) of the path, not including the
// ratings of the blobs to be combined.
// The ratings of the blobs to be combined are not used to calculate the
// priority, since it is not possible to determine from their magnitude
// whether it will be beneficial to combine the blobs. The reason is that
// chopped junk blobs (/ | - ') can have very good (low) ratings, however
// combining them will be beneficial. Blobs with high ratings might be
// over-joined pieces of characters, but also could be blobs from an unseen
// font or chopped pieces of complex characters.
while (curr_vse->parent_vse != NULL) {
ViterbiStateEntry* parent_vse = curr_vse->parent_vse;
const MATRIX_COORD& curr_cell = curr_b->matrix_cell();
const MATRIX_COORD& parent_cell = parent_vse->curr_b->matrix_cell();
MATRIX_COORD pain_coord(parent_cell.col, curr_cell.row);
if (!pain_coord.Valid(*word_res->ratings) ||
!word_res->ratings->Classified(parent_cell.col, curr_cell.row,
dict_->WildcardID())) {
// rat_subtr contains ratings sum of the two adjacent blobs to be merged.
// rat_subtr will be subtracted from the ratings sum of the path, since
// the blobs will be joined into a new blob, whose rating is yet unknown.
float rat_subtr = curr_b->rating() + parent_vse->curr_b->rating();
// ol_subtr contains the outline length of the blobs that will be joined.
float ol_subtr =
AssociateUtils::ComputeOutlineLength(rating_cert_scale, *curr_b) +
AssociateUtils::ComputeOutlineLength(rating_cert_scale,
*(parent_vse->curr_b));
// ol_dif is the outline of the path without the two blobs to be joined.
float ol_dif = vse->outline_length - ol_subtr;
// priority is set to the average rating of the path per unit of outline,
// not counting the ratings of the pieces to be joined.
float priority = ol_dif > 0 ? (vse->ratings_sum-rat_subtr)/ol_dif : 0.0;
GeneratePainPoint(pain_coord.col, pain_coord.row, LM_PPTYPE_PATH,
priority, true, max_char_wh_ratio_, word_res);
} else if (debug_level_ > 3) {
tprintf("NO pain point (Classified) for col=%d row=%d type=%s\n",
pain_coord.col, pain_coord.row,
LMPainPointsTypeName[LM_PPTYPE_PATH]);
BLOB_CHOICE_IT b_it(word_res->ratings->get(pain_coord.col,
pain_coord.row));
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
BLOB_CHOICE* choice = b_it.data();
choice->print_full();
}
}
curr_vse = parent_vse;
curr_b = curr_vse->curr_b;
}
}
/**
* Return whether this is believable superscript or subscript text.
*
* We insist that:
* + there are no punctuation marks.
* + there are no italics.
* + no normal-sized character is smaller than superscript_scaledown_ratio
* of what it ought to be, and
* + each character is at least as certain as certainty_threshold.
*
* @param[in] debug If true, spew debug output
* @param[in] word The word whose best_choice we're evaluating
* @param[in] certainty_threshold If any of the characters have less
* certainty than this, reject.
* @param[out] left_ok How many left-side characters were ok?
* @param[out] right_ok How many right-side characters were ok?
* @return Whether the complete best choice is believable as a superscript.
*/
bool Tesseract::BelievableSuperscript(bool debug,
const WERD_RES &word,
float certainty_threshold,
int *left_ok,
int *right_ok) const {
int initial_ok_run_count = 0;
int ok_run_count = 0;
float worst_certainty = 0.0f;
const WERD_CHOICE &wc = *word.best_choice;
const UnicityTable<FontInfo>& fontinfo_table = get_fontinfo_table();
for (int i = 0; i < wc.length(); i++) {
TBLOB *blob = word.rebuild_word->blobs[i];
UNICHAR_ID unichar_id = wc.unichar_id(i);
float char_certainty = wc.certainty(i);
bool bad_certainty = char_certainty < certainty_threshold;
bool is_punc = wc.unicharset()->get_ispunctuation(unichar_id);
bool is_italic = word.fontinfo && word.fontinfo->is_italic();
BLOB_CHOICE *choice = word.GetBlobChoice(i);
if (choice && fontinfo_table.size() > 0) {
// Get better information from the specific choice, if available.
int font_id1 = choice->fontinfo_id();
bool font1_is_italic = font_id1 >= 0
? fontinfo_table.get(font_id1).is_italic() : false;
int font_id2 = choice->fontinfo_id2();
is_italic = font1_is_italic &&
(font_id2 < 0 || fontinfo_table.get(font_id2).is_italic());
}
float height_fraction = 1.0f;
float char_height = blob->bounding_box().height();
float normal_height = char_height;
if (wc.unicharset()->top_bottom_useful()) {
int min_bot, max_bot, min_top, max_top;
wc.unicharset()->get_top_bottom(unichar_id,
&min_bot, &max_bot,
&min_top, &max_top);
float hi_height = max_top - max_bot;
float lo_height = min_top - min_bot;
normal_height = (hi_height + lo_height) / 2;
if (normal_height >= kBlnXHeight) {
// Only ding characters that we have decent information for because
// they're supposed to be normal sized, not tiny specks or dashes.
height_fraction = char_height / normal_height;
}
}
bool bad_height = height_fraction < superscript_scaledown_ratio;
if (debug) {
if (is_italic) {
tprintf(" Rejecting: superscript is italic.\n");
}
if (is_punc) {
tprintf(" Rejecting: punctuation present.\n");
}
const char *char_str = wc.unicharset()->id_to_unichar(unichar_id);
if (bad_certainty) {
tprintf(" Rejecting: don't believe character %s with certainty %.2f "
"which is less than threshold %.2f\n", char_str,
char_certainty, certainty_threshold);
}
if (bad_height) {
tprintf(" Rejecting: character %s seems too small @ %.2f versus "
"expected %.2f\n", char_str, char_height, normal_height);
}
}
if (bad_certainty || bad_height || is_punc || is_italic) {
if (ok_run_count == i) {
initial_ok_run_count = ok_run_count;
}
ok_run_count = 0;
} else {
ok_run_count++;
}
if (char_certainty < worst_certainty) {
worst_certainty = char_certainty;
}
}
bool all_ok = ok_run_count == wc.length();
if (all_ok && debug) {
tprintf(" Accept: worst revised certainty is %.2f\n", worst_certainty);
}
if (!all_ok) {
if (left_ok) *left_ok = initial_ok_run_count;
if (right_ok) *right_ok = ok_run_count;
}
return all_ok;
}
/// Recursive helper to find a match to the target_text (from text_index
/// position) in the choices (from choices_pos position).
/// @param choices is an array of GenericVectors, of length choices_length,
/// with each element representing a starting position in the word, and the
/// #GenericVector holding classification results for a sequence of consecutive
/// blobs, with index 0 being a single blob, index 1 being 2 blobs etc.
/// @param choices_pos
/// @param choices_length
/// @param target_text
/// @param text_index
/// @param rating
/// @param segmentation
/// @param best_rating
/// @param best_segmentation
void Tesseract::SearchForText(const GenericVector<BLOB_CHOICE_LIST*>* choices,
int choices_pos, int choices_length,
const GenericVector<UNICHAR_ID>& target_text,
int text_index,
float rating, GenericVector<int>* segmentation,
float* best_rating,
GenericVector<int>* best_segmentation) {
const UnicharAmbigsVector& table = getDict().getUnicharAmbigs().dang_ambigs();
for (int length = 1; length <= choices[choices_pos].size(); ++length) {
// Rating of matching choice or worst choice if no match.
float choice_rating = 0.0f;
// Find the corresponding best BLOB_CHOICE.
BLOB_CHOICE_IT choice_it(choices[choices_pos][length - 1]);
for (choice_it.mark_cycle_pt(); !choice_it.cycled_list();
choice_it.forward()) {
BLOB_CHOICE* choice = choice_it.data();
choice_rating = choice->rating();
UNICHAR_ID class_id = choice->unichar_id();
if (class_id == target_text[text_index]) {
break;
}
// Search ambigs table.
if (class_id < table.size() && table[class_id] != NULL) {
AmbigSpec_IT spec_it(table[class_id]);
for (spec_it.mark_cycle_pt(); !spec_it.cycled_list();
spec_it.forward()) {
const AmbigSpec *ambig_spec = spec_it.data();
// We'll only do 1-1.
if (ambig_spec->wrong_ngram[1] == INVALID_UNICHAR_ID &&
ambig_spec->correct_ngram_id == target_text[text_index])
break;
}
if (!spec_it.cycled_list())
break; // Found an ambig.
}
}
if (choice_it.cycled_list())
continue; // No match.
segmentation->push_back(length);
if (choices_pos + length == choices_length &&
text_index + 1 == target_text.size()) {
// This is a complete match. If the rating is good record a new best.
if (applybox_debug > 2) {
tprintf("Complete match, rating = %g, best=%g, seglength=%d, best=%d\n",
rating + choice_rating, *best_rating, segmentation->size(),
best_segmentation->size());
}
if (best_segmentation->empty() || rating + choice_rating < *best_rating) {
*best_segmentation = *segmentation;
*best_rating = rating + choice_rating;
}
} else if (choices_pos + length < choices_length &&
text_index + 1 < target_text.size()) {
if (applybox_debug > 3) {
tprintf("Match found for %d=%s:%s, at %d+%d, recursing...\n",
target_text[text_index],
unicharset.id_to_unichar(target_text[text_index]),
choice_it.data()->unichar_id() == target_text[text_index]
? "Match" : "Ambig",
choices_pos, length);
}
SearchForText(choices, choices_pos + length, choices_length, target_text,
text_index + 1, rating + choice_rating, segmentation,
best_rating, best_segmentation);
if (applybox_debug > 3) {
tprintf("End recursion for %d=%s\n", target_text[text_index],
unicharset.id_to_unichar(target_text[text_index]));
}
}
segmentation->truncate(segmentation->size() - 1);
}
}