Class: Spacy::Doc

Inherits:
Object
  • Object
show all
Includes:
Enumerable
Defined in:
lib/ruby-spacy.rb

Overview

See also spaCy Python API document for Doc.

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(nlp, py_doc: nil, text: nil, max_retrial: MAX_RETRIAL) ⇒ Doc

It is recommended to use Language#read method to create a doc. If you need to create one using #initialize, there are two method signatures: Spacy::Doc.new(nlp_id, py_doc: Object) and Spacy::Doc.new(nlp_id, text: String).

Parameters:

  • nlp (Language) —

    an instance of Language class

  • py_doc (Object) (defaults to: nil) —

    an instance of Python Doc class

  • text (String) (defaults to: nil) —

    the text string to be analyzed



94
95
96
97
98
99
100
101
102
103
104
105
106
107
 
# File 'lib/ruby-spacy.rb', line 94

def initialize(nlp, py_doc: nil, text: nil, max_retrial: MAX_RETRIAL)
  @py_nlp = nlp
  retrial = 0
  begin
    @py_doc = py_doc || nlp.call(text)
    @text = @py_doc.text
  rescue StandardError
    retrial += 1
    raise "Error: Failed to construct a Doc object" unless retrial <= max_retrial

    sleep 0.5
    retry
  end
end

Dynamic Method Handling

This class handles dynamic methods through the method_missing method

#method_missing(name, *args) ⇒ Object

Methods defined in Python but not wrapped in ruby-spacy can be called by this dynamic method handling mechanism.



490
491
492
 
# File 'lib/ruby-spacy.rb', line 490

def method_missing(name, *args)
  @py_doc.send(name, *args)
end

Instance Attribute Details

#py_doc ⇒ Object (readonly)

Returns a Python Doc instance accessible via PyCall.

Returns:

  • (Object) —

    a Python Doc instance accessible via PyCall



77
78
79
 
# File 'lib/ruby-spacy.rb', line 77

def py_doc
  @py_doc
end

#py_nlp ⇒ Object (readonly)

Returns a Python Language instance accessible via PyCall.

Returns:

  • (Object) —

    a Python Language instance accessible via PyCall



74
75
76
 
# File 'lib/ruby-spacy.rb', line 74

def py_nlp
  @py_nlp
end

#text ⇒ String (readonly)

Returns a text string of the document.

Returns:

  • (String) —

    a text string of the document



80
81
82
 
# File 'lib/ruby-spacy.rb', line 80

def text
  @text
end

Class Method Details

.from_bytes(nlp, byte_string) ⇒ Doc

Restores a doc from binary data created by #to_bytes. This is useful for caching processed documents to avoid re-processing.

Examples:

Load doc from file

bytes = File.binread("doc.bin")
doc = Spacy::Doc.from_bytes(nlp, bytes)

Parameters:

  • byte_string (String) —

    binary data from #to_bytes

Returns:

  • (Doc) —

    the restored doc



238
239
240
241
242
243
 
# File 'lib/ruby-spacy.rb', line 238

def self.from_bytes(nlp, byte_string)
  b64 = Base64.strict_encode64(byte_string)
  py_bytes = PyCall.eval("__import__('base64').b64decode('#{b64}')")
  py_doc = nlp.py_nlp.call("").from_bytes(py_bytes)
  new(nlp.py_nlp, py_doc: py_doc)
end

Instance Method Details

#[](range) ⇒ Object

Returns a span if given a range object; or returns a token if given an integer representing a position in the doc.

Parameters:

  • range (Range) —

    an ordinary Ruby's range object such as 0..3, 1...4, or 3 .. -1



204
205
206
207
208
209
210
211
 
# File 'lib/ruby-spacy.rb', line 204

def [](range)
  if range.is_a?(Range)
    py_span = @py_doc[range]
    Span.new(self, start_index: py_span.start, end_index: py_span.end - 1)
  else
    Token.new(@py_doc[range])
  end
end

#displacy(style: "dep", compact: false) ⇒ String

Visualize the document in one of two styles: "dep" (dependencies) or "ent" (named entities).

Parameters:

  • style (String) (defaults to: "dep") —

    either dep or ent

  • compact (Boolean) (defaults to: false) —

    only relevant to the `dep' style

Returns:

  • (String) —

    in the case of dep, the output text will be an SVG, whereas in the ent style, the output text will be an HTML.



249
250
251
 
# File 'lib/ruby-spacy.rb', line 249

def displacy(style: "dep", compact: false)
  PyDisplacy.render(py_doc, style: style, options: { compact: compact }, jupyter: false)
end

#each ⇒ Object

Iterates over the elements in the doc yielding a token instance each time.



144
145
146
147
148
 
# File 'lib/ruby-spacy.rb', line 144

def each
  PyCall::List.call(@py_doc).each do |py_token|
    yield Token.new(py_token)
  end
end

#ents ⇒ Array<Span>

Returns an array of spans each representing a named entity.

Returns:



196
197
198
199
200
 
# File 'lib/ruby-spacy.rb', line 196

def ents
  PyCall::List.call(@py_doc.ents).map do |py_span|
    Span.new(self, py_span: py_span)
  end
end

#instance_variables_to_inspect ⇒ Object 



498
499
500
 
# File 'lib/ruby-spacy.rb', line 498

def instance_variables_to_inspect
  [:@text]
end

#linguistic_summary(sections: [:text, :tokens, :entities, :noun_chunks], token_attributes: [:text, :lemma, :pos, :dep, :head]) ⇒ String

Generates a JSON string summarizing the linguistic analysis of the document. Designed to be passed as context to an LLM (e.g., via OpenAIHelper#chat).

Parameters:

  • sections (Array<Symbol>) (defaults to: [:text, :tokens, :entities, :noun_chunks]) —

    which sections to include (:text, :tokens, :entities, :noun_chunks, :sentences)

  • token_attributes (Array<Symbol>) (defaults to: [:text, :lemma, :pos, :dep, :head]) —

    which token attributes to include (:text, :lemma, :pos, :tag, :dep, :head, :ent_type, :morphology)

Returns:

  • (String) —

    a JSON string of the linguistic summary



261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
 
# File 'lib/ruby-spacy.rb', line 261

def linguistic_summary(sections: [:text, :tokens, :entities, :noun_chunks],
                       token_attributes: [:text, :lemma, :pos, :dep, :head])
  result = {}

  sections.each do |section|
    case section
    when :text
      result[:text] = @text
    when :tokens
      result[:tokens] = tokens.map do |token|
        token_hash = {}
        token_attributes.each do |attr|
          case attr
          when :head
            token_hash[:head] = token.head.text
          when :morphology
            # Use string form and parse to ensure a plain Ruby Hash for JSON serialization
            morph_str = token.morphology(hash: false)
            token_hash[:morphology] = if morph_str.empty?
                                        {}
                                      else
                                        morph_str.split("|").each_with_object({}) do |pair, h|
                                          k, v = pair.split("=", 2)
                                          h[k] = v
                                        end
                                      end
          else
            token_hash[attr] = token.send(attr)
          end
        end
        token_hash
      end
    when :entities
      ent_list = ents
      result[:entities] = ent_list.map do |ent|
        { text: ent.text, label: ent.label }
      end
    when :noun_chunks
      result[:noun_chunks] = noun_chunks.map do |chunk|
        { text: chunk.text, root: chunk.root.text }
      end
    when :sentences
      result[:sentences] = sents.map(&:text)
    end
  end

  result.to_json
end

#noun_chunks ⇒ Array<Span>

Returns an array of spans representing noun chunks.

Returns:



180
181
182
183
184
 
# File 'lib/ruby-spacy.rb', line 180

def noun_chunks
  PyCall::List.call(@py_doc.noun_chunks).map do |py_chunk|
    Span.new(self, start_index: py_chunk.start, end_index: py_chunk.end - 1)
  end
end

#openai_completion(access_token: nil, max_completion_tokens: nil, max_tokens: nil, temperature: nil, model: OpenAIClient::DEFAULT_MODEL) ⇒ String?

Sends a text completion request to OpenAI's chat API.

Parameters:

  • access_token (String, nil) (defaults to: nil) —

    OpenAI API key (defaults to OPENAI_API_KEY env var)

  • max_completion_tokens (Integer) (defaults to: nil) —

    Maximum tokens in the response

  • max_tokens (Integer) (defaults to: nil) —

    Alias for max_completion_tokens (deprecated, for backward compatibility)

  • temperature (Float, nil) (defaults to: nil) —

    Sampling temperature (omitted from requests when nil; models that reject it are retried without it)

  • model (String) (defaults to: OpenAIClient::DEFAULT_MODEL) —

    The model to use (default: gpt-5-mini)

Returns:

  • (String, nil) —

    The completed text



440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
 
# File 'lib/ruby-spacy.rb', line 440

def openai_completion(access_token: nil, max_completion_tokens: nil, max_tokens: nil, temperature: nil, model: OpenAIClient::DEFAULT_MODEL)
  # Support both max_completion_tokens and max_tokens for backward compatibility
  max_completion_tokens ||= max_tokens || 1000

  messages = [
    { role: "system", content: "Complete the text input by the user." },
    { role: "user", content: @text }
  ]

  client = openai_client(access_token)
  response = client.chat(
    model: model,
    messages: messages,
    max_completion_tokens: max_completion_tokens,
    temperature: temperature
  )
  response.dig("choices", 0, "message", "content")
rescue OpenAIClient::APIError => e
  warn "Error: OpenAI API call failed - #{e.message}"
  nil
end

#openai_embeddings(access_token: nil, model: OpenAIClient::DEFAULT_EMBEDDINGS_MODEL, dimensions: nil) ⇒ Array<Float>?

Generates text embeddings using OpenAI's embeddings API.

Parameters:

  • access_token (String, nil) (defaults to: nil) —

    OpenAI API key (defaults to OPENAI_API_KEY env var)

  • model (String) (defaults to: OpenAIClient::DEFAULT_EMBEDDINGS_MODEL) —

    The embeddings model (default: text-embedding-3-small)

  • dimensions (Integer, nil) (defaults to: nil) —

    The number of dimensions for the output embeddings (nil uses model default)

Returns:

  • (Array<Float>, nil) —

    The embedding vector



468
469
470
471
472
473
474
475
 
# File 'lib/ruby-spacy.rb', line 468

def openai_embeddings(access_token: nil, model: OpenAIClient::DEFAULT_EMBEDDINGS_MODEL, dimensions: nil)
  client = openai_client(access_token)
  response = client.embeddings(model: model, input: @text, dimensions: dimensions)
  response.dig("data", 0, "embedding")
rescue OpenAIClient::APIError => e
  warn "Error: OpenAI API call failed - #{e.message}"
  nil
end

#openai_query(access_token: nil, max_completion_tokens: nil, max_tokens: nil, temperature: nil, model: OpenAIClient::DEFAULT_MODEL, messages: [], prompt: nil, response_format: nil, max_tool_call_depth: 5, _tool_call_depth: 0) ⇒ String?

Sends a query to OpenAI's chat completion API with optional tool support. The get_tokens tool allows the model to request token-level linguistic analysis.

Parameters:

  • access_token (String, nil) (defaults to: nil) —

    OpenAI API key (defaults to OPENAI_API_KEY env var)

  • max_completion_tokens (Integer) (defaults to: nil) —

    Maximum tokens in the response

  • max_tokens (Integer) (defaults to: nil) —

    Alias for max_completion_tokens (deprecated, for backward compatibility)

  • temperature (Float, nil) (defaults to: nil) —

    Sampling temperature (omitted from requests when nil; models that reject it are retried without it)

  • model (String) (defaults to: OpenAIClient::DEFAULT_MODEL) —

    The model to use (default: gpt-5-mini)

  • messages (Array<Hash>) (defaults to: []) —

    Conversation history (for recursive tool calls). Note: this array is modified in place when tool calls occur.

  • prompt (String, nil) (defaults to: nil) —

    System prompt for the query

Returns:

  • (String, nil) —

    The model's response content



322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
 
# File 'lib/ruby-spacy.rb', line 322

def openai_query(access_token: nil,
                 max_completion_tokens: nil,
                 max_tokens: nil,
                 temperature: nil,
                 model: OpenAIClient::DEFAULT_MODEL,
                 messages: [],
                 prompt: nil,
                 response_format: nil,
                 max_tool_call_depth: 5,
                 _tool_call_depth: 0)
  # Support both max_completion_tokens and max_tokens for backward compatibility
  max_completion_tokens ||= max_tokens || 1000
  if messages.empty?
    messages = [
      { role: "system", content: prompt },
      { role: "user", content: @text }
    ]
  end

  client = openai_client(access_token)

  # Tool definition for token analysis (GPT-5 tools API format)
  tools = nil
  tool_choice = nil
  if _tool_call_depth < max_tool_call_depth
    tools = [
      {
        type: "function",
        function: {
          name: "get_tokens",
          description: "Tokenize given text and return a list of tokens with their attributes: surface, lemma, tag, pos (part-of-speech), dep (dependency), ent_type (entity type), and morphology",
          parameters: {
            type: "object",
            properties: {
              text: {
                type: "string",
                description: "text to be tokenized"
              }
            },
            required: ["text"]
          }
        }
      }
    ]
    tool_choice = "auto"
  end

  response = client.chat(
    model: model,
    messages: messages,
    max_completion_tokens: max_completion_tokens,
    temperature: temperature,
    tools: tools,
    tool_choice: tool_choice,
    response_format: response_format
  )

  message = response.dig("choices", 0, "message")

  # Handle tool calls (GPT-5 format)
  if message["tool_calls"] && !message["tool_calls"].empty?
    messages << message

    message["tool_calls"].each do |tool_call|
      function_name = tool_call.dig("function", "name")
      tool_call_id = tool_call["id"]

      case function_name
      when "get_tokens"
        result = tokens.map do |t|
          {
            surface: t.text,
            lemma: t.lemma,
            pos: t.pos,
            tag: t.tag,
            dep: t.dep,
            ent_type: t.ent_type,
            morphology: t.morphology
          }
        end.to_json

        messages << {
          role: "tool",
          tool_call_id: tool_call_id,
          content: result
        }
      end
    end

    # Recursive call to get final response after tool execution
    openai_query(
      access_token: access_token,
      max_completion_tokens: max_completion_tokens,
      temperature: temperature,
      model: model,
      messages: messages,
      prompt: prompt,
      response_format: response_format,
      max_tool_call_depth: max_tool_call_depth,
      _tool_call_depth: _tool_call_depth + 1
    )
  else
    message["content"]
  end
rescue OpenAIClient::APIError => e
  warn "Error: OpenAI API call failed - #{e.message}"
  nil
end

#respond_to_missing?(sym, include_private = false) ⇒ Boolean

Returns:

  • (Boolean) 


494
495
496
 
# File 'lib/ruby-spacy.rb', line 494

def respond_to_missing?(sym, include_private = false)
  Spacy.py_hasattr?(@py_doc, sym) || super
end

#retokenize(start_index, end_index, attributes = {}) ⇒ Object

Retokenizes the text merging a span into a single token.

Parameters:

  • start_index (Integer) —

    the start position of the span to be retokenized in the document

  • end_index (Integer) —

    the end position of the span to be retokenized in the document

  • attributes (Hash) (defaults to: {}) —

    attributes to set on the merged token



113
114
115
116
117
 
# File 'lib/ruby-spacy.rb', line 113

def retokenize(start_index, end_index, attributes = {})
  PyCall.with(@py_doc.retokenize) do |retokenizer|
    retokenizer.merge(@py_doc[start_index..end_index], attrs: attributes)
  end
end

#retokenize_split(pos_in_doc, split_array, head_pos_in_split, ancestor_pos, attributes = {}) ⇒ Object

Retokenizes the text splitting the specified token.

Parameters:

  • pos_in_doc (Integer) —

    the position of the span to be retokenized in the document

  • split_array (Array<String>) —

    text strings of the split results

  • ancestor_pos (Integer) —

    the position of the immediate ancestor element of the split elements in the document

  • attributes (Hash) (defaults to: {}) —

    the attributes of the split elements



124
125
126
127
128
129
 
# File 'lib/ruby-spacy.rb', line 124

def retokenize_split(pos_in_doc, split_array, head_pos_in_split, ancestor_pos, attributes = {})
  PyCall.with(@py_doc.retokenize) do |retokenizer|
    heads = [[@py_doc[pos_in_doc], head_pos_in_split], @py_doc[ancestor_pos]]
    retokenizer.split(@py_doc[pos_in_doc], split_array, heads: heads, attrs: attributes)
  end
end

#sents ⇒ Array<Span>

Returns an array of spans each representing a sentence.

Returns:



188
189
190
191
192
 
# File 'lib/ruby-spacy.rb', line 188

def sents
  PyCall::List.call(@py_doc.sents).map do |py_sent|
    Span.new(self, start_index: py_sent.start, end_index: py_sent.end - 1)
  end
end

#similarity(other) ⇒ Float

Returns a semantic similarity estimate.

Parameters:

  • other (Doc) —

    the other doc to which a similarity estimation is made

Returns:

  • (Float) 


216
217
218
 
# File 'lib/ruby-spacy.rb', line 216

def similarity(other)
  py_doc.similarity(other.py_doc)
end

#span(range_or_start, optional_size = nil) ⇒ Span

Returns a span of the specified range within the doc. The method should be used either of the two ways: Doc#span(range) or Doc#span{start_pos, size_of_span}.

Parameters:

  • range_or_start (Range, Integer) —

    a range object, or, alternatively, an integer that represents the start position of the span

  • optional_size (Integer) (defaults to: nil) —

    an integer representing the size of the span

Returns:



155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
 
# File 'lib/ruby-spacy.rb', line 155

def span(range_or_start, optional_size = nil)
  doc_len = PyCall.len(@py_doc)

  if optional_size
    start_index = range_or_start
    start_index += doc_len if start_index < 0
    end_index = start_index + optional_size - 1
  else
    range = range_or_start
    start_index = range.first
    start_index += doc_len if start_index < 0
    end_val = range.end
    if end_val.nil?
      end_index = doc_len - 1
    else
      end_val += doc_len if end_val < 0
      end_index = range.exclude_end? ? end_val - 1 : end_val
    end
  end

  Span.new(self, start_index: start_index, end_index: end_index)
end

#to_bytes ⇒ String

Serializes the doc to a binary string. The binary data includes all annotations (tokens, entities, etc.) and can be used to restore the doc later without re-processing.

Examples:

Save doc to file

doc = nlp.read("Hello world")
File.binwrite("doc.bin", doc.to_bytes)

Returns:

  • (String) —

    binary representation of the doc



227
228
229
 
# File 'lib/ruby-spacy.rb', line 227

def to_bytes
  @py_doc.to_bytes.force_encoding(Encoding::BINARY)
end

#to_s ⇒ String

String representation of the document.

Returns:

  • (String) 


133
134
135
 
# File 'lib/ruby-spacy.rb', line 133

def to_s
  @text
end

#tokens ⇒ Array<Token>

Returns an array of tokens contained in the doc.

Returns:



139
140
141
 
# File 'lib/ruby-spacy.rb', line 139

def tokens
  PyCall::List.call(@py_doc).map { |py_token| Token.new(py_token) }
end