1 /* 2 * Copyright (C) 2013 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17 #ifndef LATINIME_TYPING_SCORING_H 18 #define LATINIME_TYPING_SCORING_H 19 20 #include "defines.h" 21 #include "suggest/core/dictionary/error_type_utils.h" 22 #include "suggest/core/policy/scoring.h" 23 #include "suggest/core/session/dic_traverse_session.h" 24 #include "suggest/policyimpl/typing/scoring_params.h" 25 26 namespace latinime { 27 28 class DicNode; 29 class DicTraverseSession; 30 31 class TypingScoring : public Scoring { 32 public: getInstance()33 static const TypingScoring *getInstance() { return &sInstance; } 34 getMostProbableString(const DicTraverseSession * const traverseSession,const float weightOfLangModelVsSpatialModel,SuggestionResults * const outSuggestionResults)35 AK_FORCE_INLINE void getMostProbableString(const DicTraverseSession *const traverseSession, 36 const float weightOfLangModelVsSpatialModel, 37 SuggestionResults *const outSuggestionResults) const {} 38 getAdjustedWeightOfLangModelVsSpatialModel(DicTraverseSession * const traverseSession,DicNode * const terminals,const int size)39 AK_FORCE_INLINE float getAdjustedWeightOfLangModelVsSpatialModel( 40 DicTraverseSession *const traverseSession, DicNode *const terminals, 41 const int size) const { 42 return 1.0f; 43 } 44 calculateFinalScore(const float compoundDistance,const int inputSize,const ErrorTypeUtils::ErrorType containedErrorTypes,const bool forceCommit,const bool boostExactMatches,const bool hasProbabilityZero)45 AK_FORCE_INLINE int calculateFinalScore(const float compoundDistance, const int inputSize, 46 const ErrorTypeUtils::ErrorType containedErrorTypes, const bool forceCommit, 47 const bool boostExactMatches, const bool hasProbabilityZero) const { 48 const float maxDistance = ScoringParams::DISTANCE_WEIGHT_LANGUAGE 49 + static_cast<float>(inputSize) * ScoringParams::TYPING_MAX_OUTPUT_SCORE_PER_INPUT; 50 float score = ScoringParams::TYPING_BASE_OUTPUT_SCORE - compoundDistance / maxDistance; 51 if (forceCommit) { 52 score += ScoringParams::AUTOCORRECT_OUTPUT_THRESHOLD; 53 } 54 if (hasProbabilityZero) { 55 // Previously, when both legitimate 0-frequency words (such as distracters) and 56 // offensive words were encoded in the same way, distracters would never show up 57 // when the user blocked offensive words (the default setting, as well as the 58 // setting for regression tests). 59 // 60 // When b/11031090 was fixed and a separate encoding was used for offensive words, 61 // 0-frequency words would no longer be blocked when they were an "exact match" 62 // (where case mismatches and accent mismatches would be considered an "exact 63 // match"). The exact match boosting functionality meant that, for example, when 64 // the user typed "mt" they would be suggested the word "Mt", although they most 65 // probably meant to type "my". 66 // 67 // For this reason, we introduced this change, which does the following: 68 // * Defines the "perfect match" as a really exact match, with no room for case or 69 // accent mismatches 70 // * When the target word has probability zero (as "Mt" does, because it is a 71 // distracter), ONLY boost its score if it is a perfect match. 72 // 73 // By doing this, when the user types "mt", the word "Mt" will NOT be boosted, and 74 // they will get "my". However, if the user makes an explicit effort to type "Mt", 75 // we do boost the word "Mt" so that the user's input is not autocorrected to "My". 76 if (boostExactMatches && ErrorTypeUtils::isPerfectMatch(containedErrorTypes)) { 77 score += ScoringParams::PERFECT_MATCH_PROMOTION; 78 } 79 } else { 80 if (boostExactMatches && ErrorTypeUtils::isExactMatch(containedErrorTypes)) { 81 score += ScoringParams::EXACT_MATCH_PROMOTION; 82 if ((ErrorTypeUtils::MATCH_WITH_WRONG_CASE & containedErrorTypes) != 0) { 83 score -= ScoringParams::CASE_ERROR_PENALTY_FOR_EXACT_MATCH; 84 } 85 if ((ErrorTypeUtils::MATCH_WITH_MISSING_ACCENT & containedErrorTypes) != 0) { 86 score -= ScoringParams::ACCENT_ERROR_PENALTY_FOR_EXACT_MATCH; 87 } 88 if ((ErrorTypeUtils::MATCH_WITH_DIGRAPH & containedErrorTypes) != 0) { 89 score -= ScoringParams::DIGRAPH_PENALTY_FOR_EXACT_MATCH; 90 } 91 } 92 } 93 return static_cast<int>(score * SUGGEST_INTERFACE_OUTPUT_SCALE); 94 } 95 getDoubleLetterDemotionDistanceCost(const DicNode * const terminalDicNode)96 AK_FORCE_INLINE float getDoubleLetterDemotionDistanceCost( 97 const DicNode *const terminalDicNode) const { 98 return 0.0f; 99 } 100 autoCorrectsToMultiWordSuggestionIfTop()101 AK_FORCE_INLINE bool autoCorrectsToMultiWordSuggestionIfTop() const { 102 return true; 103 } 104 sameAsTyped(const DicTraverseSession * const traverseSession,const DicNode * const dicNode)105 AK_FORCE_INLINE bool sameAsTyped(const DicTraverseSession *const traverseSession, 106 const DicNode *const dicNode) const { 107 return traverseSession->getProximityInfoState(0)->sameAsTyped( 108 dicNode->getOutputWordBuf(), dicNode->getNodeCodePointCount()); 109 } 110 111 private: 112 DISALLOW_COPY_AND_ASSIGN(TypingScoring); 113 static const TypingScoring sInstance; 114 TypingScoring()115 TypingScoring() {} ~TypingScoring()116 ~TypingScoring() {} 117 }; 118 } // namespace latinime 119 #endif // LATINIME_TYPING_SCORING_H 120