Bacteria Biotope Relation Extraction via Lexical Chains and Dependency Graphs

Abstract

In this article, we describe our approach for the Bacteria Biotopes relation extraction (BBrel) subtask in the BioNLP Shared Task 2019. This task aims to promote the development of text mining systems that extract relationships between Microorganism, Habitat and Phenotype entities. In this paper, we propose a novel approach for dependency graph construction based on lexical chains, so one dependency graph can represent one or multiple sentences. After that, we propose a neural network model which consists of the bidirectional long shortterm memories and an attention graph convolution neural network to learn relation extraction features from the graph. Our approach is able to extract both intra-and inter-sentence relations, and meanwhile utilize syntax information. The results show that our approach achieved the best F1 (66.3%) in the official evaluation participated by 7 teams. 1

Relation Extraction in Biomedical Texts Based on Multi-Head Attention Model With Syntactic Dependency Feature: Modeling Study

“…Table 4 lists the comparison between our method and other previous systems in BB-rel task. The first 3 lines in the table are the official top 3 systems (10 participated), among which Yuhang_Wu used a multilayer perceptron [35], AliAI [39] used a multitask architecture similar to BERT, and whunlp [37] achieves state-of-the-art performance by using dependency graph and attention graph convolution neural network. The fourth line is the baseline provided by the task organizer, which uses a co-occurrence method.…”

“…For instance, the Yuhang_Wu team used a multilayer perceptron and achieved an F 1 score of 60.49% on the test set. The highest F 1 score was 66.39%, which was submitted by the whunlp team [37]. They constructed a dependency graph based on lexical association, and used bidirectional LSTM (BiLSTM) [38] and an attention graph convolution neural network to detect the relation.…”