blingfire 0.1.2 → 0.1.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 130b126e317ef7c923db7f6ef2b53384d1983cbfc25894c27c6e5b59a4ca2523
-  data.tar.gz: 2d121c5b405a760889a14977119d605e87375abcf2dc623151ce03c0b164ea2f
+  metadata.gz: af6d8ad14f2c3f56fb148eca079a81bb48b3cec80790bc184fc99afd03572f12
+  data.tar.gz: 92f351ac35b186b54b2d2c261427b3827ccb40613d3dd3fc1f57ce6eec83aa57
 SHA512:
-  metadata.gz: 15ecb71b9875542281cfe9167bd410abb58ba24d9ff57181afe2d942101ca335b7aa95c0c4fcdede39090dd3ec8ab57f7f1ed83c341a4b2505cfc9711e86637e
-  data.tar.gz: 23771de5c189aa3d7750cc2b09b2dd0bbe02512bf05fb8ccdad7f8664e63bbbcac1419c1d57f7137c16d9c57e8d36b5fac2d3cae96ea5979081d89bad6737ddd
+  metadata.gz: 5714bc6333a94669d1b6b69682dd3671933ffc0611107eeb2ce174263c329b42b57c92205afac8519b6626dcae631f724682fccfde57c6514c9cac9233653c62
+  data.tar.gz: 46a6de8f8a081c893db5d446b6e981a170efc4cdcfdaac158f58ba423592f0274990345066a703a687e67d2f2496cb46b2a46cfe8a0f476f19b30afd37fd4697

data/CHANGELOG.md

@@ -1,3 +1,10 @@
+## 0.1.3 (2020-10-01)
+
+- Added `text_to_words_with_offsets` method
+- Added `text_to_sentences_with_offsets` method
+- Added `text_to_ids_with_offsets` method
+- Added `normalize_spaces` method
+
 ## 0.1.2 (2020-06-25)
 
 - Updated Bling Fire to 0.1.3

data/LICENSE.txt

@@ -1,22 +1,22 @@
-Copyright (c) 2020 Andrew Kane
-
 MIT License
 
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
+Copyright (c) Microsoft Corporation. All rights reserved.
+Copyright (c) 2020 Andrew Kane
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
 
-The above copyright notice and this permission notice shall be
-included in all copies or substantial portions of the Software.
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE

data/README.md

@@ -32,6 +32,18 @@ Tokenize sentences
 model.text_to_sentences(text)
 ```
 
+Get offsets for words
+
+```ruby
+words, start_offsets, end_offsets = model.text_to_words_with_offsets(text)
+```
+
+Get offsets for sentences
+
+```ruby
+sentences, start_offsets, end_offsets = model.text_to_sentences_with_offsets(text)
+```
+
 ## Pre-trained Models
 
 BlingFire comes with a default model that follows the tokenization logic of NLTK with a few changes. You can also download other models:
@@ -60,6 +72,12 @@ Convert text to ids
 model.text_to_ids(text)
 ```
 
+Get offsets for ids
+
+```ruby
+ids, start_offsets, end_offsets = model.text_to_ids_with_offsets(text)
+```
+
 ## History
 
 View the [changelog](https://github.com/ankane/blingfire/blob/master/CHANGELOG.md)
@@ -79,6 +97,6 @@ To get started with development:
 git clone https://github.com/ankane/blingfire.git
 cd blingfire
 bundle install
-bundle exec rake vendor:all
+bundle exec rake vendor:all download:models
 bundle exec rake test
 ```

data/lib/blingfire.rb

@@ -46,6 +46,18 @@ module BlingFire
       end
     end
 
+    def text_to_words_with_offsets(text)
+      text_to_with_offsets(text, " ") do |t, out, start_offsets, end_offsets|
+        FFI.TextToWordsWithOffsets(t, t.bytesize, out, start_offsets, end_offsets, out.size)
+      end
+    end
+
+    def text_to_words_with_offsets_with_model(model, text)
+      text_to_with_offsets(text, " ") do |t, out, start_offsets, end_offsets|
+        FFI.TextToWordsWithOffsetsWithModel(t, t.bytesize, out, start_offsets, end_offsets, out.size, model)
+      end
+    end
+
     def text_to_sentences(text)
       text_to(text, "\n") do |t, out|
         FFI.TextToSentences(t, t.bytesize, out, out.size)
@@ -58,6 +70,18 @@ module BlingFire
       end
     end
 
+    def text_to_sentences_with_offsets(text)
+      text_to_with_offsets(text, "\n") do |t, out, start_offsets, end_offsets|
+        FFI.TextToSentencesWithOffsets(t, t.bytesize, out, start_offsets, end_offsets, out.size)
+      end
+    end
+
+    def text_to_sentences_with_offsets_with_model(model, text)
+      text_to_with_offsets(text, "\n") do |t, out, start_offsets, end_offsets|
+        FFI.TextToSentencesWithOffsetsWithModel(t, t.bytesize, out, start_offsets, end_offsets, out.size, model)
+      end
+    end
+
     def text_to_ids(model, text, max_len = nil, unk_id = 0)
       text = encode_utf8(text.dup) unless text.encoding == Encoding::UTF_8
       ids = Fiddle::Pointer.malloc((max_len || text.size) * Fiddle::SIZEOF_INT)
@@ -66,10 +90,34 @@ module BlingFire
       ids[0, (max_len || out_size) * Fiddle::SIZEOF_INT].unpack("i!*")
     end
 
+    def text_to_ids_with_offsets(model, text, max_len = nil, unk_id = 0)
+      text = encode_utf8(text.dup) unless text.encoding == Encoding::UTF_8
+      ids = Fiddle::Pointer.malloc((max_len || text.size) * Fiddle::SIZEOF_INT)
+
+      start_offsets = Fiddle::Pointer.malloc(Fiddle::SIZEOF_INT * ids.size)
+      end_offsets = Fiddle::Pointer.malloc(Fiddle::SIZEOF_INT * ids.size)
+
+      out_size = FFI.TextToIdsWithOffsets(model, text, text.bytesize, ids, start_offsets, end_offsets, ids.size, unk_id)
+
+      check_status out_size, ids
+
+      result = ids[0, (max_len || out_size) * Fiddle::SIZEOF_INT].unpack("i!*")
+      [result].concat(unpack_offsets(start_offsets, end_offsets, result, text))
+    end
+
     def free_model(model)
       FFI.FreeModel(model)
     end
 
+    def normalize_spaces(text)
+      u_space = 0x20
+      text = encode_utf8(text.dup) unless text.encoding == Encoding::UTF_8
+      out = Fiddle::Pointer.malloc([text.bytesize * 1.5, 20].max)
+      out_size = FFI.NormalizeSpaces(text, text.bytesize, out, out.size, u_space)
+      check_status out_size, out
+      encode_utf8(out.to_str(out_size))
+    end
+
     private
 
     def check_status(ret, ptr)
@@ -79,14 +127,52 @@ module BlingFire
     def text_to(text, sep)
       text = encode_utf8(text.dup) unless text.encoding == Encoding::UTF_8
       # TODO allocate less, and try again if needed
-      out = Fiddle::Pointer.malloc([text.bytesize * 1.5, 20].max)
+      out = Fiddle::Pointer.malloc([text.bytesize * 3, 20].max)
       out_size = yield(text, out)
       check_status out_size, out
       encode_utf8(out.to_str(out_size - 1)).split(sep)
     end
 
+    def text_to_with_offsets(text, sep)
+      text = encode_utf8(text.dup) unless text.encoding == Encoding::UTF_8
+      # TODO allocate less, and try again if needed
+      out = Fiddle::Pointer.malloc([text.bytesize * 3, 20].max)
+
+      start_offsets = Fiddle::Pointer.malloc(Fiddle::SIZEOF_INT * out.size)
+      end_offsets = Fiddle::Pointer.malloc(Fiddle::SIZEOF_INT * out.size)
+
+      out_size = yield(text, out, start_offsets, end_offsets)
+
+      check_status out_size, out
+
+      result = encode_utf8(out.to_str(out_size - 1)).split(sep)
+      [result].concat(unpack_offsets(start_offsets, end_offsets, result, text))
+    end
+
     def encode_utf8(text)
       text.force_encoding(Encoding::UTF_8)
     end
+
+    def unpack_offsets(start_offsets, end_offsets, result, text)
+      start_bytes = start_offsets.to_s(Fiddle::SIZEOF_INT * result.size).unpack("i*")
+      end_bytes = end_offsets.to_s(Fiddle::SIZEOF_INT * result.size).unpack("i*")
+      starts = []
+      ends = []
+
+      # convert byte offsets to character offsets
+      # TODO see if more efficient to store next_pos in variable
+      pos = 0
+      text.each_char.with_index do |c, i|
+        while pos == start_bytes[starts.size]
+          starts << i
+        end
+        pos += c.bytesize
+        while pos - 1 == end_bytes[ends.size]
+          ends << i + 1
+        end
+      end
+
+      [starts, ends]
+    end
   end
 end

data/lib/blingfire/ffi.rb

@@ -10,13 +10,35 @@ module BlingFire
       raise e
     end
 
+    # https://github.com/microsoft/BlingFire/blob/master/blingfiretools/blingfiretokdll/blingfiretokdll.cpp
+
+    # version
     extern "int GetBlingFireTokVersion()"
-    extern "void* LoadModel(char * pszLdbFileName)"
-    extern "int FreeModel(void* ModelPtr)"
+
+    # text to sentences
+    extern "int TextToSentencesWithOffsetsWithModel(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int * pStartOffsets, int * pEndOffsets, int MaxOutUtf8StrByteCount, void * hModel)"
+    extern "int TextToSentencesWithOffsets(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int * pStartOffsets, int * pEndOffsets, int MaxOutUtf8StrByteCount)"
+    extern "int TextToSentencesWithModel(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount, void * hModel)"
+    extern "int TextToSentences(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount)"
+
+    # text to words
+    extern "int TextToWordsWithOffsetsWithModel(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int * pStartOffsets, int * pEndOffsets, int MaxOutUtf8StrByteCount, void * hModel)"
+    extern "int TextToWordsWithOffsets(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int * pStartOffsets, int * pEndOffsets, int MaxOutUtf8StrByteCount)"
     extern "int TextToWordsWithModel(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount, void * hModel)"
     extern "int TextToWords(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount)"
+
+    # misc
+    extern "int NormalizeSpaces(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount, int uSpace)"
+    extern "int TextToHashes(char * pInUtf8Str, int InUtf8StrByteCount, int32_t * pHashArr, int MaxHashArrLength, int wordNgrams, int bucketSize)"
+
+    # model
+    extern "void* LoadModel(char * pszLdbFileName)"
+
+    # text to ids
+    extern "int TextToIdsWithOffsets(void* ModelPtr, char * pInUtf8Str, int InUtf8StrByteCount, int32_t * pIdsArr, int * pStartOffsets, int * pEndOffsets, int MaxIdsArrLength, int UnkId)"
     extern "int TextToIds(void* ModelPtr, char * pInUtf8Str, int InUtf8StrByteCount, int32_t * pIdsArr, int MaxIdsArrLength, int UnkId)"
-    extern "int TextToSentences(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount)"
-    extern "int TextToSentencesWithModel(char * pInUtf8Str, int InUtf8StrByteCount, char * pOutUtf8Str, int MaxOutUtf8StrByteCount, void * hModel)"
+
+    # free model
+    extern "int FreeModel(void* ModelPtr)"
   end
 end

data/lib/blingfire/model.rb

@@ -1,6 +1,7 @@
 module BlingFire
   class Model
     def initialize(path = nil)
+      @handle = nil
       if path
         raise Error, "Model not found" unless File.exist?(path)
         @handle = FFI.LoadModel(path)
@@ -16,6 +17,14 @@ module BlingFire
       end
     end
 
+    def text_to_words_with_offsets(text)
+      if @handle
+        BlingFire.text_to_words_with_offsets_with_model(@handle, text)
+      else
+        BlingFire.text_to_words_with_offsets(text)
+      end
+    end
+
     def text_to_sentences(text)
       if @handle
         BlingFire.text_to_sentences_with_model(@handle, text)
@@ -24,6 +33,14 @@ module BlingFire
       end
     end
 
+    def text_to_sentences_with_offsets(text)
+      if @handle
+        BlingFire.text_to_sentences_with_offsets_with_model(@handle, text)
+      else
+        BlingFire.text_to_sentences_with_offsets(text)
+      end
+    end
+
     def text_to_ids(text, max_len = nil, unk_id = 0)
       if @handle
         BlingFire.text_to_ids(@handle, text, max_len, unk_id)
@@ -32,6 +49,14 @@ module BlingFire
       end
     end
 
+    def text_to_ids_with_offsets(text, max_len = nil, unk_id = 0)
+      if @handle
+        BlingFire.text_to_ids_with_offsets(@handle, text, max_len, unk_id)
+      else
+        raise "Not implemented"
+      end
+    end
+
     def to_ptr
       @handle
     end

data/lib/blingfire/version.rb

@@ -1,3 +1,3 @@
 module BlingFire
-  VERSION = "0.1.2"
+  VERSION = "0.1.3"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: blingfire
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.1.3
 platform: ruby
 authors:
 - Andrew Kane
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2020-06-25 00:00:00.000000000 Z
+date: 2020-10-02 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler