1 /*
2 * Copyright (C) 2018 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 #include "lang_id/lang-id-wrapper.h"
18
19 #include <fcntl.h>
20
21 #include "lang_id/fb_model/lang-id-from-fb.h"
22 #include "lang_id/lang-id.h"
23
24 namespace libtextclassifier3 {
25
26 namespace langid {
27
LoadFromPath(const std::string & langid_model_path)28 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> LoadFromPath(
29 const std::string& langid_model_path) {
30 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> langid_model =
31 libtextclassifier3::mobile::lang_id::GetLangIdFromFlatbufferFile(langid_model_path);
32 return langid_model;
33 }
34
LoadFromDescriptor(const int langid_fd)35 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> LoadFromDescriptor(
36 const int langid_fd) {
37 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> langid_model =
38 libtextclassifier3::mobile::lang_id::GetLangIdFromFlatbufferFileDescriptor(
39 langid_fd);
40 return langid_model;
41 }
42
LoadFromUnownedBuffer(const char * buffer,int size)43 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> LoadFromUnownedBuffer(
44 const char* buffer, int size) {
45 std::unique_ptr<libtextclassifier3::mobile::lang_id::LangId> langid_model =
46 libtextclassifier3::mobile::lang_id::GetLangIdFromFlatbufferBytes(buffer, size);
47 return langid_model;
48 }
49
GetPredictions(const libtextclassifier3::mobile::lang_id::LangId * model,const std::string & text)50 std::vector<std::pair<std::string, float>> GetPredictions(
51 const libtextclassifier3::mobile::lang_id::LangId* model, const std::string& text) {
52 return GetPredictions(model, text.data(), text.size());
53 }
54
GetPredictions(const libtextclassifier3::mobile::lang_id::LangId * model,const char * text,int text_size)55 std::vector<std::pair<std::string, float>> GetPredictions(
56 const libtextclassifier3::mobile::lang_id::LangId* model, const char* text,
57 int text_size) {
58 std::vector<std::pair<std::string, float>> prediction_results;
59 if (model == nullptr) {
60 return prediction_results;
61 }
62
63 const float noise_threshold =
64 model->GetFloatProperty("text_classifier_langid_noise_threshold", -1.0f);
65
66 // Speed up the things by specifying the max results we want. For example, if
67 // the noise threshold is 0.1, we don't need more than 10 results.
68 const int max_results =
69 noise_threshold < 0.01
70 ? -1 // -1 means FindLanguages returns all predictions
71 : static_cast<int>(1 / noise_threshold) + 1;
72
73 libtextclassifier3::mobile::lang_id::LangIdResult langid_result;
74 model->FindLanguages(text, text_size, &langid_result, max_results);
75 for (int i = 0; i < langid_result.predictions.size(); i++) {
76 const auto& prediction = langid_result.predictions[i];
77 if (prediction.second >= noise_threshold && prediction.first != "und") {
78 prediction_results.push_back({prediction.first, prediction.second});
79 }
80 }
81 return prediction_results;
82 }
83
GetLanguageTags(const libtextclassifier3::mobile::lang_id::LangId * model,const std::string & text)84 std::string GetLanguageTags(const libtextclassifier3::mobile::lang_id::LangId* model,
85 const std::string& text) {
86 const std::vector<std::pair<std::string, float>>& predictions =
87 GetPredictions(model, text);
88 const float threshold =
89 model->GetFloatProperty("text_classifier_langid_threshold", -1.0f);
90 std::string detected_language_tags = "";
91 bool first_accepted_language = true;
92 for (int i = 0; i < predictions.size(); i++) {
93 const auto& prediction = predictions[i];
94 if (threshold >= 0.f && prediction.second < threshold) {
95 continue;
96 }
97 if (first_accepted_language) {
98 first_accepted_language = false;
99 } else {
100 detected_language_tags += ",";
101 }
102 detected_language_tags += prediction.first;
103 }
104 return detected_language_tags;
105 }
106
107 } // namespace langid
108
109 } // namespace libtextclassifier3
110