Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2017 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 | |
Lukas Zilka | b23e212 | 2018-02-09 10:25:19 +0100 | [diff] [blame] | 17 | #ifndef LIBTEXTCLASSIFIER_TOKENIZER_H_ |
| 18 | #define LIBTEXTCLASSIFIER_TOKENIZER_H_ |
Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 19 | |
| 20 | #include <string> |
| 21 | #include <vector> |
| 22 | |
| 23 | #include "model_generated.h" |
| 24 | #include "types.h" |
| 25 | #include "util/base/integral_types.h" |
Lukas Zilka | b23e212 | 2018-02-09 10:25:19 +0100 | [diff] [blame] | 26 | #include "util/utf8/unicodetext.h" |
Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 27 | |
| 28 | namespace libtextclassifier2 { |
| 29 | |
| 30 | const int kInvalidScript = -1; |
| 31 | const int kUnknownScript = -2; |
| 32 | |
| 33 | // Tokenizer splits the input string into a sequence of tokens, according to the |
| 34 | // configuration. |
| 35 | class Tokenizer { |
| 36 | public: |
| 37 | explicit Tokenizer( |
| 38 | const std::vector<const TokenizationCodepointRange*>& codepoint_ranges, |
| 39 | bool split_on_script_change); |
| 40 | |
| 41 | // Tokenizes the input string using the selected tokenization method. |
Lukas Zilka | b23e212 | 2018-02-09 10:25:19 +0100 | [diff] [blame] | 42 | std::vector<Token> Tokenize(const std::string& text) const; |
| 43 | |
| 44 | // Same as above but takes UnicodeText. |
| 45 | std::vector<Token> Tokenize(const UnicodeText& text_unicode) const; |
Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 46 | |
| 47 | protected: |
| 48 | // Finds the tokenization codepoint range config for given codepoint. |
| 49 | // Internally uses binary search so should be O(log(# of codepoint_ranges)). |
Lukas Zilka | ba849e7 | 2018-03-08 14:48:21 +0100 | [diff] [blame] | 50 | const TokenizationCodepointRangeT* FindTokenizationRange(int codepoint) const; |
Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 51 | |
| 52 | // Finds the role and script for given codepoint. If not found, DEFAULT_ROLE |
| 53 | // and kUnknownScript are assigned. |
| 54 | void GetScriptAndRole(char32 codepoint, |
| 55 | TokenizationCodepointRange_::Role* role, |
| 56 | int* script) const; |
| 57 | |
| 58 | private: |
| 59 | // Codepoint ranges that determine how different codepoints are tokenized. |
| 60 | // The ranges must not overlap. |
Lukas Zilka | ba849e7 | 2018-03-08 14:48:21 +0100 | [diff] [blame] | 61 | std::vector<std::unique_ptr<const TokenizationCodepointRangeT>> |
| 62 | codepoint_ranges_; |
Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 63 | |
| 64 | // If true, tokens will be additionally split when the codepoint's script_id |
| 65 | // changes. |
| 66 | bool split_on_script_change_; |
| 67 | }; |
| 68 | |
| 69 | } // namespace libtextclassifier2 |
| 70 | |
Lukas Zilka | b23e212 | 2018-02-09 10:25:19 +0100 | [diff] [blame] | 71 | #endif // LIBTEXTCLASSIFIER_TOKENIZER_H_ |