Nested Propositions in Open Information Extraction

Abstract

The challenges of Machine Reading and Knowledge Extraction at a web scale require a system capable of extracting diverse information from large, heterogeneous corpora. The Open Information Extraction (OIE) paradigm aims at extracting assertions from large corpora without requiring a vocabulary or relation-specific training data. Most systems built on this paradigm extract binary relations from arbitrary sentences, ignoring the context under which the assertions are correct and complete. They lack the expressiveness needed to properly represent and extract complex assertions commonly found in the text. To address the lack of representation power, we propose NESTIE, which uses a nested representation to extract higher-order relations, and complex, interdependent assertions. Nesting the extracted propositions allows NESTIE to more accurately reflect the meaning of the original sentence. Our experimental study on real-world datasets suggests that NESTIE obtains comparable precision with better minimality and informativeness than existing approaches. NESTIE produces 1.7-1.8 times more minimal extractions and achieves 1.1-1.2 times higher informativeness than CLAUSIE.

Supervised Open Information Extraction

“…While most Open IE systems aim to extract the common case of verbal binary propositions (i.e, subject-verb-object tuples), some systems specialize in other syntactic constructions, including noun-mediated relations (Yahya et al, 2014;Pal and Mausam, 2016), n-ary relations (Akbik and Löser, 2012), or nested propositions (Bhutani et al, 2016).…”

Supervising Unsupervised Open Information Extraction Models

“…Early Open IE systems apply handcrafted rules or self-supervised learning paradigm, where the extraction results which satisfy a set of syntactic constraints are considered as positive examples and the results which do not satisfy the con- (Angeli et al, 2015) and OpenIE5 are developed by combining several different approaches as described in details in Section 3.1. Further, several systems focusing on a specialized constructs were developed, including noun-mediated relations (Pal and Mausam, 2016), n-ary relations (Akbik and Löser, 2012), nested propositions (Bhutani et al, 2016) and numerical Open IE (Saha et al, 2017a). Recently, there have been efforts to apply deep learning methods to Open IE.…”

MinIE: Minimizing Facts in Open Information Extraction

“…Earlier systems-e.g., Fader et al (2011)-relied mostly on shallower NLP techniques such as POS tagging and chunking, while later systems often use dependency parsing in addition (Gamallo et al, 2012;Wu and Weld, 2010). Most OIE systems represent extractions in the form of triples, although some also produce n-ary extractions (Akbik and Löser, 2012;Del Corro and Gemulla, 2013) or nested representations (Bast and Haussmann, 2013;Bhutani et al, 2016). Some systems focus on non-verb-mediated relations (Yahya et al, 2014).…”

“…CSD-IE (Bast and Haussmann, 2013) introduced the notion of nested facts (termed "minimal" in their paper) and produce extractions with "pointers" to other extractions. NestIE (Bhutani et al, 2016) takes up this idea. OLLIE's clause modifier has a similar purpose.…”