"X raise sanctions against Y" - Infer Sentiment Relations with AREkit and DeepPavlov [part 2/2]

alt text

In this post we apply fine-tuned model for inferring sentiment relations from Mass-Media texts.

COMPLETE EXAMPLE

The code snippets provided below are related to the:

Serialization stage, organized by AREkit.

BertTextsSerializationPipelineItem(
  synonyms=read_synonyms_collection(synonyms_filepath=synonyms_filepath, 
                                    stemmer=stemmer),
  terms_per_context=terms_per_context,
  entities_parser=BertOntonotesNERPipelineItem(
      lambda s_obj: s_obj.ObjectType in ["ORG", "PERSON", "LOC", "GPE"]),
  entity_fmt=create_entity_formatter(EntityFormatterTypes.HiddenBertStyled),
  name_provider=ExperimentNameProvider(name="example-bert", suffix="infer"),
  text_b_type=text_b_type,
  output_dir=output_dir,
  opin_annot=DefaultAnnotator(
      PairBasedAnnotationAlgorithm(
          dist_in_terms_bound=None,
          label_provider=ConstantLabelProvider(label_instance=NoLabel()))),
  data_folding=NoFolding(doc_ids_to_fold=list(range(texts_count)),
                         supported_data_types=[DataType.Test])),

Sentiment Relation Extraction, provided by DeepPavlov library.

BertInferencePipelineItem(
  data_type=DataType.Test,
  predict_writer=TsvPredictWriter(),
  bert_config_file=bert_config_path,
  model_checkpoint_path=bert_finetuned_ckpt_path,
  vocab_filepath=bert_vocab_path,
  max_seq_length=max_seq_length,
  do_lowercase=do_lowercase,
  labels_scaler=labels_scaler),

Backend. Visualization, provided by Brat library.

BratBackendContentsPipelineItem(label_to_rel={
  str(labels_scaler.label_to_uint(ExperimentPositiveLabel())): "POS",
  str(labels_scaler.label_to_uint(ExperimentNegativeLabel())): "NEG"
},
  obj_color_types={"ORG": '#7fa2ff', "GPE": "#7fa200", 
                   "PERSON": "#7f00ff", "Frame": "#00a2ff"},
  rel_color_types={"POS": "GREEN", "NEG": "RED"},
)

Share on: