Neural Relation Extraction within and across Sentence Boundaries

Abstract

Past work in relation extraction mostly focuses on binary relation between entity pairs within single sentence. Recently, the NLP community has gained interest in relation extraction in entity pairs spanning multiple sentences. In this paper, we propose a novel architecture for this task: inter-sentential dependency-based neural networks (iDepNN). iDepNN models the shortest and augmented dependency paths via recurrent and recursive neural networks to extract relationships within (intra-) and across (inter-) sentence boundaries. Compared to SVM and neural network baselines, iDepNN is more robust to false positives in relationships spanning sentences. We evaluate our models on four datasets from newswire (MUC6) and medical (BioNLP shared task) domains that achieve state-of-the-art performance and show a better balance in precision and recall for inter-sentential relationships. We perform better than 11 teams participating in the BioNLP shared task 2016 and achieve a gain of 5.2% (0.587 vs 0.558) in F1 over the winning team. We also release the crosssentence annotations for MUC6.Paul Allen has started a company and named [Vern Raburnse1 its President. The company, to be called [Paul Allen Group] e2 will be based in Bellevue, Washington.

Neural Architectures for Fine-Grained Propaganda Detection in News

“…Additionally, we apply a decision function such that a sentence is tagged as propaganda if prediction probability of the classifier is greater than a threshold (τ ). We relax the binary decision boundary to boost recall, similar to Gupta et al (2019b).…”

Unsupervised inference of implicit biomedical events using context triggers

“…In particular, extracting cross-sentence events, where relevant entities are found across multiple sentences, is considered challenging because of the lack of clear linguistic evidence for how entities mentioned in different sentences are associated to form a single event. For this reason, few studies have looked into this issue [15][16][17][18][19][20].…”

“…Some follow-up studies have been conducted since then, and updated the state-of-the-art performance with diverse types of neural networks built over syntactic dependency paths between entities. Examples include bidirectional long-short term memory networks [23], gated recurrent unit networks [24], and recursive and recurrent neural networks [20]. However, they either did not extract cross-sentence events at all, or achieved the best results only when cross-sentence events are ignored.…”

“…By contrast, the BB-event task targets fine-grained mention-level relations with minimal training data, and no task-specific curated knowledge bases are available. Recent research results suggest that supervised systems based on high-dimensional representation learning would not be very effective at addressing the data sparsity issue in cross-sentence events in the BB-event datasets [20,21,23,24].…”

Approximate Definitional Constructs as Lightweight Evidence for Detecting Classes Among Wikipedia Articles

“…Motivation: Wikipedia and knowledge repositories derived from it are useful in a variety of tasks. They pertain to knowledge acquisition from text [12,14,18,[30][31][32], text analysis [22,23] and information retrieval [3,5,6,15,16,24,28,33] including commercial Web search, helping to potentially transform search results from sets of hyperlinks to relevant documents into sets of concepts directly relevant to users' queries [26]. Most Wikipedia articles correspond to concepts that are instances ("Wynnewood Valley Park Sensory Garden") as opposed to classes ("Garden").…”