commit 5dc5e1192353ecf8b0a0e24848f8ebe199fe5f31
author Tony Mak <tonymak@google.com> Fri Feb 01 14:52:10 2019 +0000
committer Tony Mak <tonymak@google.com> Fri Feb 01 14:52:10 2019 +0000
tree 05b73ff34be98308fae905f1617db20d6b6a5e98
parent 1faaa0729e038c4a52ff1a0eea1db4f6ac2e98fa

Import libtextclassifier

Major change: Support intent config for action generation.


BUG: 123745079
Test: atest TextClassifierTest
Change-Id: Ia6dded6065ff37e6736abc1749a4078593fff87d

annotator/annotator.cc

diff --git a/annotator/annotator.cc b/annotator/annotator.cc
index 330cf0b..bd5f06f 100644
--- a/annotator/annotator.cc
+++ b/annotator/annotator.cc
@@ -385,7 +385,8 @@
 }
 
 bool Annotator::InitializeContactEngine(const std::string& serialized_config) {
-  std::unique_ptr<ContactEngine> contact_engine(new ContactEngine());
+  std::unique_ptr<ContactEngine> contact_engine(
+      new ContactEngine(selection_feature_processor_.get()));
   if (!contact_engine->Initialize(serialized_config)) {
     TC3_LOG(ERROR) << "Failed to initialize the contact engine.";
     return false;
@@ -670,7 +671,7 @@
 inline bool ClassifiedAsOther(
     const std::vector<ClassificationResult>& classification) {
   return !classification.empty() &&
-         classification[0].collection == Collections::kOther;
+         classification[0].collection == Collections::Other();
 }
 
 float GetPriorityScore(
@@ -936,7 +937,7 @@
   if (model_->classification_options()->max_num_tokens() > 0 &&
       model_->classification_options()->max_num_tokens() <
           selection_num_tokens) {
-    *classification_results = {{Collections::kOther, 1.0}};
+    *classification_results = {{Collections::Other(), 1.0}};
     return true;
   }
 
@@ -976,7 +977,7 @@
 
   if (!classification_feature_processor_->HasEnoughSupportedCodepoints(
           tokens, extraction_span)) {
-    *classification_results = {{Collections::kOther, 1.0}};
+    *classification_results = {{Collections::Other(), 1.0}};
     return true;
   }
 
@@ -1030,22 +1031,22 @@
 
   // Phone class sanity check.
   if (!classification_results->empty() &&
-      classification_results->begin()->collection == Collections::kPhone) {
+      classification_results->begin()->collection == Collections::Phone()) {
     const int digit_count = CountDigits(context, selection_indices);
     if (digit_count <
             model_->classification_options()->phone_min_num_digits() ||
         digit_count >
             model_->classification_options()->phone_max_num_digits()) {
-      *classification_results = {{Collections::kOther, 1.0}};
+      *classification_results = {{Collections::Other(), 1.0}};
     }
   }
 
   // Address class sanity check.
   if (!classification_results->empty() &&
-      classification_results->begin()->collection == Collections::kAddress) {
+      classification_results->begin()->collection == Collections::Address()) {
     if (selection_num_tokens <
         model_->classification_options()->address_min_num_tokens()) {
-      *classification_results = {{Collections::kOther, 1.0}};
+      *classification_results = {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1091,6 +1092,20 @@
   return false;
 }
 
+namespace {
+std::string PickCollectionForDatetime(
+    const DatetimeParseResult& datetime_parse_result) {
+  switch (datetime_parse_result.granularity) {
+    case GRANULARITY_HOUR:
+    case GRANULARITY_MINUTE:
+    case GRANULARITY_SECOND:
+      return Collections::DateTime();
+    default:
+      return Collections::Date();
+  }
+}
+}  // namespace
+
 bool Annotator::DatetimeClassifyText(
     const std::string& context, CodepointSpan selection_indices,
     const ClassificationOptions& options,
@@ -1118,7 +1133,8 @@
         selection_indices) {
       for (const DatetimeParseResult& parse_result : datetime_span.data) {
         classification_results->emplace_back(
-            Collections::kDate, datetime_span.target_classification_score);
+            PickCollectionForDatetime(parse_result),
+            datetime_span.target_classification_score);
         classification_results->back().datetime_parse_result = parse_result;
       }
       return true;
@@ -1157,7 +1173,7 @@
     if (!FilteredForClassification(knowledge_result)) {
       return {knowledge_result};
     } else {
-      return {{Collections::kOther, 1.0}};
+      return {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1168,7 +1184,7 @@
     if (!FilteredForClassification(contact_result)) {
       return {contact_result};
     } else {
-      return {{Collections::kOther, 1.0}};
+      return {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1178,7 +1194,7 @@
     if (!FilteredForClassification(regex_result)) {
       return {regex_result};
     } else {
-      return {{Collections::kOther, 1.0}};
+      return {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1196,7 +1212,7 @@
     if (!datetime_results.empty()) {
       return datetime_results;
     } else {
-      return {{Collections::kOther, 1.0}};
+      return {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1211,7 +1227,7 @@
     if (!FilteredForClassification(model_result[0])) {
       return model_result;
     } else {
-      return {{Collections::kOther, 1.0}};
+      return {{Collections::Other(), 1.0}};
     }
   }
 
@@ -1747,7 +1763,8 @@
       AnnotatedSpan annotated_span;
       annotated_span.span = datetime_span.span;
       annotated_span.classification = {
-          {Collections::kDate, datetime_span.target_classification_score,
+          {PickCollectionForDatetime(parse_result),
+           datetime_span.target_classification_score,
            datetime_span.priority_score}};
       annotated_span.classification[0].datetime_parse_result = parse_result;