Joint Bootstrapping Machines for High Confidence Relation Extraction

Proceedings of NAACL-HLT 2018, pages 26–36 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Joint Bootstrapping Machines for High Confidence Relation Extraction Pankaj Gupta1,2, Benjamin Roth2, Hinrich Sch¨utze2 1Corporate Technology, Machine-Intelligence (MIC-DE), Siemens AG Munich, Germany 2CIS, University of Munich (LMU) Munich, Germany pankaj.gupta@siemens.com | pankaj.gupta@campus.lmu.de {beroth, inquiries}@cis.lmu.de Abstract Semi-supervised bootstrapping techniques for relationship extraction from text iteratively ex- pand a set of initial seed instances. Due to the lack of labeled data, a key challenge in boot- strapping is semantic drift: if a false positive instance is added during an iteration, then all following iterations are contaminated. We in- troduce BREX, a new bootstrapping method that protects against such contamination by highly effective confidence assessment. This is achieved by using entity and template seeds jointly (as opposed to just one as in previous work), by expanding entities and templates in parallel and in a mutually constraining fash- ion in each iteration and by introducing higher- quality similarity measures for templates. Ex- perimental results show that BREX achieves an F1 that is 0.13 (0.87 vs. 0.74) better than the state of the art for four relationships. 1 Introduction Traditional semi-supervised bootstrapping rela- tion extractors (REs) such as BREDS (Batista et al., 2015), SnowBall (Agichtein and Gravano, 2000) and DIPRE (Brin, 1998) require an initial set of seed entity pairs for the target binary rela- tion. They find occurrences of positive seed en- tity pairs in the corpus, which are converted into extraction patterns, i.e., extractors, where we de- fine an extractor as a cluster of instances generated from the corpus. The initial seed entity pair set is expanded with the relationship entity pairs newly extracted by the extractors from the text iteratively. The augmented set is then used to extract new re- lationships until a stopping criterion is met. Due to lack of sufficient labeled data, rule- based systems dominate commercial use (Chiti- cariu et al., 2013). Rules are typically defined by creating patterns around the entities (entity ex- traction) or entity pairs (relation extraction). Re- cently, supervised machine learning, especially deep learning techniques (Gupta et al., 2015; Nguyen and Grishman, 2015; Vu et al., 2016a,b; Gupta et al., 2016), have shown promising results in entity and relation extraction; however, they need sufficient hand-labeled data to train models, which can be costly and time consuming for web- scale extractions. Bootstrapping machine-learned rules can make extractions easier on large corpora. Thus, open information extraction systems (Carl- son et al., 2010; Fader et al., 2011; Mausam et al., 2012; Mesquita et al., 2013; Angeli et al., 2015) have recently been popular for domain specific or independent pattern learning. Hearst (1992) used hand written rules to gen- erate more rules to extract hypernym-hyponym pairs, without distributional similarity. For en- tity extraction, Riloff (1996) used seed entities to generate extractors with heuristic rules and scored them by counting positive extractions. Prior work (Lin et al., 2003; Gupta et al., 2014) investigated different extractor scoring measures. Gupta and Manning (2014) …

Content truncated to 500 words