Easy Victories and Uphill Battles in Coreference Resolution
Citations
7,412 citations
1,696 citations
705 citations
Cites background or methods from "Easy Victories and Uphill Battles i..."
...The attention component is inspired by parser-derived head-word matching features from previous systems (Durrett and Klein, 2013), but is less susceptible to cascading errors....
[...]
..., 2012; Björkelund and Kuhn, 2014; Martschat and Strube, 2015), or (4) mention-ranking models (Durrett and Klein, 2013; Wiseman et al., 2015; Clark and Manning, 2016a)....
[...]
562 citations
Cites background or methods from "Easy Victories and Uphill Battles i..."
..., 2010), feature-based (Durrett and Klein, 2013; Peng et al., 2015a), and neuralnetwork based (Clark and Manning, 2016; Lee et al....
[...]
...In this section we evaluate of three representative systems: rule based, Rule, (Raghunathan et al., 2010), feature-rich, Feature, (Durrett and Klein, 2013), and end-to-end neural (the current state-ofthe-art), E2E, (Lee et al., 2017)....
[...]
...15 cal examples: the Stanford Deterministic Coreference System (Raghunathan et al., 2010), the Berkeley Coreference Resolution System (Durrett and Klein, 2013) and the current best published system: the UW End-to-end Neural Coreference Resolution System (Lee et al., 2017)....
[...]
..., 2010), feature-rich, Feature, (Durrett and Klein, 2013), and end-to-end neural (the current state-ofthe-art), E2E, (Lee et al....
[...]
..., 2010), the Berkeley Coreference Resolution System (Durrett and Klein, 2013) and the current best published system: the UW End-to-end Neural Coreference Resolution System (Lee et al....
[...]
366 citations
References
7,244 citations
6,984 citations
"Easy Victories and Uphill Battles i..." refers methods in this paper
...001 and optimize the objective using AdaGrad (Duchi et al., 2011)....
[...]
...We set (αFA, αFN, αWL) = (0.1, 3.0, 1.0) and λ = 0.001 and optimize the objective using AdaGrad (Duchi et al., 2011)....
[...]
1,994 citations
"Easy Victories and Uphill Battles i..." refers background in this paper
...And finally, rather than targeting centering theory (Grosz et al., 1995) with rule-based features identifying syntactic positions (Stoyanov et al., 2010; Haghighi and Klein, 2010), our features on word context can identify configurational clues like whether a mention is preceded or followed by a…...
[...]
1,059 citations
"Easy Victories and Uphill Battles i..." refers background or methods in this paper
...However, the semantic information contained even in a coreference corpus of thousands of documents is insufficient to generalize to unseen data,8 so system designers have turned to external resources such as semantic classes derived from WordNet (Soon et al., 2001), WordNet hypernymy or synonymy (Stoyanov et al., 2010), semantic similarity computed from online resources (Ponzetto and Strube, 2006), named entity type features, gender and number match using the dataset of Bergsma and Lin (2006), and features from unsupervised clusters (Hendrickx and Daelemans, 2007; Durrett et al., 2013)....
[...]
...…is insufficient to generalize to unseen data,8 so system designers have turned to external resources such as semantic classes derived from WordNet (Soon et al., 2001), WordNet hypernymy or synonymy (Stoyanov et al., 2010), semantic similarity computed from online resources (Ponzetto and Strube,…...
[...]
...In this section, we consider the following subset of these information sources: • WordNet hypernymy and synonymy • Number and gender data for nominals and propers from Bergsma and Lin (2006) • Named entity types • Latent clusters computed from English Gigaword (Graff et al., 2007), where a latent cluster label generates each nominal head (excluding pronouns) and a conjunction of its verbal governor and semantic role, if any (Durrett et al., 2013)....
[...]
...Unlike binary classification-based coreference systems where independent binary decisions are made about each pair (Soon et al., 2001; Bengtson and Roth, 2008; Versley et al., 2008; Stoyanov et al., 2010), we use a log-linear model to select at most one antecedent for...
[...]
...However, the semantic information contained even in a coreference corpus of thousands of documents is insufficient to generalize to unseen data,8 so system designers have turned to external resources such as semantic classes derived from WordNet (Soon et al., 2001), WordNet hypernymy or synonymy (Stoyanov et al....
[...]
931 citations
"Easy Victories and Uphill Battles i..." refers methods in this paper
...Throughout this work, we use the datasets from the CoNLL 2011 shared task2 (Pradhan et al., 2011), which is derived from the OntoNotes corpus (Hovy et al., 2006)....
[...]
...which is derived from the OntoNotes corpus (Hovy et al., 2006)....
[...]