tommy/libpostal
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[similarity] adding a match count in Soft-TFIDF to allow answering questions about subsets i.e. the set of tokens in "Park Pl" contain the set of tokens in "Park". Setting Jaro-Winkler minimum length of 4 chars on, more specific option name for possible abbeviation detection

Browse Source
This commit is contained in:
Al
2018-01-06 03:50:03 -05:00
parent 0cb488ecea
commit 4356174630
2 changed files with 36 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
src/soft_tfidf.c
View File

@@ -7,9 +7,10 @@
static soft_tfidf_options_t DEFAULT_SOFT_TFIDF_OPTIONS = {
.jaro_winkler_min = 0.9,
.jaro_winkler_min_length = 4,
.damerau_levenshtein_max = 1,
.damerau_levenshtein_min_length = 4,
.use_abbreviations = true
.possible_affine_gap_abbreviations = true
};
@@ -53,7 +54,7 @@ bool compare_canonical(address_expansion_t e1, char **tokens1, phrase_t match1,
return false;
}
}
return true;
return canonical_index == canonical_len;
}
}
@@ -112,7 +113,7 @@ bool phrases_have_same_canonical(size_t num_tokens1, char **tokens1, size_t num_
}
double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char **tokens1, double *token_scores1, phrase_array *phrases1, size_t num_tokens2, char **tokens2, double *token_scores2, phrase_array *phrases2, phrase_array *acronym_alignments, soft_tfidf_options_t options) {
double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char **tokens1, double *token_scores1, phrase_array *phrases1, size_t num_tokens2, char **tokens2, double *token_scores2, phrase_array *phrases2, phrase_array *acronym_alignments, soft_tfidf_options_t options, size_t *num_matches) {
if (token_scores1 == NULL || token_scores2 == NULL) return 0.0;
if (num_tokens1 > num_tokens2) {
@@ -205,19 +206,23 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
}
double jaro_winkler_min = options.jaro_winkler_min;
size_t jaro_winkler_min_length = options.jaro_winkler_min_length;
size_t damerau_levenshtein_max = options.damerau_levenshtein_max;
size_t damerau_levenshtein_min_length = options.damerau_levenshtein_min_length;
bool use_damerau_levenshtein = damerau_levenshtein_max > 0 && len1 >= damerau_levenshtein_min_length;
bool use_abbreviations = options.use_abbreviations;
bool possible_affine_gap_abbreviations = options.possible_affine_gap_abbreviations;
size_t matched_tokens = 0;
for (size_t i = 0; i < len1; i++) {
uint32_array *t1u = t1_tokens_unicode[i];
uint32_array *t2u;
double t1_score = token_scores1[i];
size_t t1_len = t1u->n;
double max_sim = 0.0;
size_t min_dist = t1u->n;
size_t min_dist = t1_len;
size_t argmax_sim = 0;
size_t argmin_dist = 0;
double argmin_dist_sim = 0.0;
@@ -230,6 +235,10 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
int64_t pm1 = phrase_memberships1 != NULL ? phrase_memberships1[i] : NULL_PHRASE_MEMBERSHIP;
phrase_t p1 = pm1 >= 0 ? phrases1->a[pm1] : NULL_PHRASE;
phrase_t argmax_phrase = NULL_PHRASE;
bool use_jaro_winkler = t1_len >= jaro_winkler_min_length;
bool use_damerau_levenshtein = damerau_levenshtein_max > 0 && t1_len >= damerau_levenshtein_min_length;
log_debug("use_jaro_winkler = %d, use_damerau_levenshtein=%d\n", use_jaro_winkler, use_damerau_levenshtein);
canonical_match_t best_canonical_phrase_response = CANONICAL_NO_MATCH;
@@ -243,6 +252,7 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
canonical_match_t canonical_response = CANONICAL_NO_MATCH;
if (p1.len > 0 && p2.len > 0 && phrases_have_same_canonical(num_tokens1, tokens1, num_tokens2, tokens2, p1, p2, &canonical_response)) {
if (canonical_response > best_canonical_phrase_response) {
log_debug("canonical_response = %d\n", canonical_response);
best_canonical_phrase_response = canonical_response;
argmax_sim = j;
argmax_phrase = p2;
@@ -275,7 +285,7 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
}
double jaro_winkler = jaro_winkler_distance_unicode(t1u, t2u);
if (jaro_winkler > max_sim) {
if (use_jaro_winkler && jaro_winkler > max_sim) {
max_sim = jaro_winkler;
argmax_sim = j;
}
@@ -291,7 +301,7 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
}
}
if (use_abbreviations) {
if (possible_affine_gap_abbreviations) {
bool is_abbreviation = possible_abbreviation_unicode(t1u, t2u);
if (is_abbreviation) {
last_abbreviation = j;
@@ -310,28 +320,39 @@ double soft_tfidf_similarity_with_phrases_and_acronyms(size_t num_tokens1, char
log_debug("have max sim = %f\n", max_sim);
t2_score = token_scores2[argmax_sim];
total_sim += max_sim * t1_score * t2_score;
matched_tokens++;
} else if (use_damerau_levenshtein && min_dist <= damerau_levenshtein_max) {
log_debug("levenshtein\n");
t2_score = token_scores2[argmin_dist];
total_sim += argmin_dist_sim * t1_score * t2_score;
} else if (use_abbreviations && have_abbreviation) {
matched_tokens++;
} else if (possible_affine_gap_abbreviations && have_abbreviation) {
log_debug("have abbreviation\n");
t2_score = token_scores2[last_abbreviation];
total_sim += last_abbreviation_sim * t1_score * t2_score;
matched_tokens++;
}
} else if (have_phrase_match) {
for (size_t p = argmax_phrase.start; p < argmax_phrase.start + argmax_phrase.len; p++) {
t2_score = token_scores2[p];
total_sim += max_sim * t1_score * t2_score;
}
matched_tokens++;
log_debug("have_phrase_match\n");
} else {
for (size_t p = acronym_phrase.start; p < acronym_phrase.start + acronym_phrase.len; p++) {
t2_score = token_scores2[p];
total_sim += max_sim * t1_score * t2_score;
}
log_debug("have_acronym_match\n");
matched_tokens++;
}
}
log_debug("matched_tokens = %zu\n", matched_tokens);
*num_matches = matched_tokens;
return_soft_tfidf_score:
if (t1_tokens_unicode != NULL) {
for (size_t i = 0; i < len1; i++) {
@@ -369,6 +390,6 @@ return_soft_tfidf_score:
}
double soft_tfidf_similarity(size_t num_tokens1, char **tokens1, double *token_scores1, size_t num_tokens2, char **tokens2, double *token_scores2, soft_tfidf_options_t options) {
return soft_tfidf_similarity_with_phrases_and_acronyms(num_tokens1, tokens1, token_scores1, NULL, num_tokens2, tokens2, token_scores2, NULL, NULL, options);
double soft_tfidf_similarity(size_t num_tokens1, char **tokens1, double *token_scores1, size_t num_tokens2, char **tokens2, double *token_scores2, soft_tfidf_options_t options, size_t *num_matches) {
return soft_tfidf_similarity_with_phrases_and_acronyms(num_tokens1, tokens1, token_scores1, NULL, num_tokens2, tokens2, token_scores2, NULL, NULL, options, num_matches);
}