Fixed exceptions from Annoy and nearest_neighbors

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'

[WilliamGustafson]

Fixes for Issues #19 and #18.

• In Process.query and Process.index self.nearest_neighbors does not expect a sparse matrix, now the vectors are converted to a dense numpy matrix
• The metric name 'cosine' is not used in Annoy, so the default metric in char_ngram_process was changed to 'angular' (the closest analogue)
  • sklearn's pairwise_distances function doesn't have an 'angular' metric so 'cosine' is used in process.ratio in place of 'angular'