Predicate Argument Alignment using a Global Coherence Model

Abstract

We present a joint model for predicate argument alignment. We leverage multiple sources of semantic information, including temporal ordering constraints between events. These are combined in a max-margin framework to find a globally consistent view of entities and events across multiple documents, which leads to improvements over a very strong local baseline.

Conceptual and Practical Steps in Event Coreference Analysis of Large-scale Data

“…In particular, it would be interesting to see whether exact specification of event head arguments would outperform the vague attribution with related entities. The state-of-the-art result in the supervised predicate alignment approach is a hint for rich linguistic features to be helpful (Wolfe et al, 2013). On the other hand, depending on the adapted event identity definition, coreferential events might not really share identical arguments (Hasler and Orasan, 2009).…”

“…Since in this approach decisions are independently made for every pair of event mentions, a clear formalism is needed to determine exactly what types of coreference are possible and how they are detected by looking at textual mentions (Chen and Ji, 2009;Hovy et al, 2013b). Some related work on predicate alignment also fit into this category of research (Roth and Frank, 2012;Wolfe et al, 2013). Alternatively, in automatic event clustering, the objective is basically discovering event instances: all we know about an event in the world is the collective information obtained from mentions referring to that in a text corpus.…”

A Consolidated Open Knowledge Representation for Multiple Texts

“…A secondary line of work deals with aligning predicates across document pairs, as done in Roth and Frank (2012). PARMA (Wolfe et al, 2013) treated the task as a token-alignment problem, aligning also arguments, while Wolfe et al (2015) added joint constraints to align predicates and their arguments.…”

A Hierarchical Distance-dependent Bayesian Model for Event Coreference Resolution

“…Liu et al (2014) demonstrated the benefits of propagating information between event arguments and event mentions during a post-processing step. Other work modeled event coreference as a predicate argument alignment problem between pairs of sentences, and trained classifiers for making alignment decisions (Roth and Frank, 2012;Wolfe et al, 2015). Our model also leverages event argument information into the decisions of event coreference but incorporates it into Bayesian clustering priors.…”