Entity Relation Extraction from News Articles in Portuguese for Competitive Intelligence Based on BERT

• Daniel De Los Reyes PUCRS https://orcid.org/0000-0003-0495-5982
• Douglas Trajano PUCRS https://orcid.org/0000-0001-7801-8938
• Isabel Harb Manssour PUCRS http://orcid.org/0000-0001-9446-6757
• Renata Vieira University of Évora http://orcid.org/0000-0003-2449-5477
• Rafael H. Bordini PUCRS https://orcid.org/0000-0001-8688-9901

Resumo

Competitive intelligence (CI) is a relevant area of a corporation and can support the strategic business area by showing those responsible, helping decision making on how to position an organization in the market. This work uses the Bidirectional Transformer Encoding Representations (BERT) to process a sentence and its named entities and extract the parts of the sentences that represent or describe the semantic relationship between these named entities. The approach was developed for the Portuguese language, considering the financial domain and exploring deep linguistic representations without using other lexical-semantic resources. The results of the experiments show a precision of 73.5% using the Jaccard metric that measures the similarity between sentences. A second contribution of this work is the manually constructed dataset with more than 4.500 tuples (phrase, entity, entity) annotated.

Palavras-chave: Competitive intelligence, Entity relation classification, Relation extraction