#### DMCA

## Joint Information Extraction and Reasoning: A Scalable Statistical Relational Learning Approach

### Citations

3269 | The PageRank citation ranking: Bringing order to the Web. Work in progress. URL: http://google.stanford.edu/backrub/pageranksub.ps
- Page, Brin, et al.
(Show Context)
Citation Context ...ll define a stochastic process on the graph, which will generate a score for each node, and hence a score for each answer to the query. The stochastic process used in ProPPR is personalized PageRank (=-=Page et al., 1998-=-; Csalogny et al., 2005), also known as random-walkwith-restart. Intuitively, this process upweights solution nodes that are reachable by many short proofs (i.e., short paths from the query node.) For... |

2753 |
Evaluating structural equation models with unobservable variables and measurement error
- Fornell, Lacker
- 1981
(Show Context)
Citation Context ...ed latent variable modeling in probabilistic graphical models and statistics (Skrondal and Rabe-Hesketh, 2004): modeling hidden variables helps take into account the measurement (observation) errors (=-=Fornell and Larcker, 1981-=-) and results in a more robust model. 6 Discussions Compared to state-of-the-art joint models (Riedel et al., 2013) that learn the latent factor representations, our method gives strong improvements i... |

816 | Markov logic networks
- Richardson, Domingos
- 2006
(Show Context)
Citation Context ...elation learning templates (a-c) from Table 2. Information Extraction (IE) includes only templates (d) and (e). Markov Logic Networks (MLN) is the Alchemy’s implementation10 of Markov Logic Networks (=-=Richardson and Domingos, 2006-=-), using the first-order clauses learned from SL method11. We used conjugate gradient weight learning (Lowd and Domingos, 2007) with 10 iterations. Finally, Universal Schema is a state-of-the-art matr... |

339 |
Introduction to Statistical Relational Learning
- Getoor, Taskar
- 2007
(Show Context)
Citation Context ...nce of the full system1. In this paper, we address this issue, and propose a joint model system for IE and KB completion in a statistical relational learning (SRL) setting (Sutton and McCallum, 2006; =-=Getoor and Taskar, 2007-=-). In particular, we outline a system which takes as input a partially-populated KB and a set of relation mentions in context, and jointly learns: 1) how to extract new KB facts from the relation ment... |

323 |
An introduction to Conditional Random Fields for relational learning, chapter 4,
- Sutton, McCallum
- 2006
(Show Context)
Citation Context ...des, which reduces performance of the full system1. In this paper, we address this issue, and propose a joint model system for IE and KB completion in a statistical relational learning (SRL) setting (=-=Sutton and McCallum, 2006-=-; Getoor and Taskar, 2007). In particular, we outline a system which takes as input a partially-populated KB and a set of relation mentions in context, and jointly learns: 1) how to extract new KB fac... |

265 | Constructing biological knowledge bases by extracting information from text sources. - Craven, Kumlien - 1999 |

257 |
Generalized Latent Variable Modeling: Multilevel, Longitudinal, and Structural Equation Models.
- Skrondal, Rabe-Hesketh
- 2004
(Show Context)
Citation Context ...s inspired by predicate invention in inductive logic programming, our result is also consistent with theories of generalized latent variable modeling in probabilistic graphical models and statistics (=-=Skrondal and Rabe-Hesketh, 2004-=-): modeling hidden variables helps take into account the measurement (observation) errors (Fornell and Larcker, 1981) and results in a more robust model. 6 Discussions Compared to state-of-the-art joi... |

150 | Autonomously semantifying wikipedia. In: - Wu, Weld - 2007 |

104 | Knowledgebased weak supervision for information extraction of overlapping relations.
- Hoffmann, Zhang, et al.
- 2011
(Show Context)
Citation Context ...ext may be associated with one, none, or several of the possible relation labels, a property which complicates the application of distant supervision methods (Mintz et al., 2009; Riedel et al., 2010; =-=Hoffmann et al., 2011-=-; Surdeanu et al., 2012). Lao et al. (2012) learned syntactic rules for finding relations defined by “lexico-semantic” paths spanning KB relations and text data. Wang et al. (2015) extends the methods... |

87 | Efficient weight learning for Markov logic networks.
- Lowd, Domingos
- 2007
(Show Context)
Citation Context ...s (MLN) is the Alchemy’s implementation10 of Markov Logic Networks (Richardson and Domingos, 2006), using the first-order clauses learned from SL method11. We used conjugate gradient weight learning (=-=Lowd and Domingos, 2007-=-) with 10 iterations. Finally, Universal Schema is a state-of-the-art matrix factorization based universal method for jointly learning surface patterns and relations. We used the code and parameter se... |

77 | Relation extraction with matrix factorization and universal schemas.
- Riedel, Yao, et al.
- 2013
(Show Context)
Citation Context ...t relation-learning and IE. To summarize our contributions: • We present a joint model for IE and relational learning in a statistical relational learning setting which outperforms universal schemas (=-=Riedel et al., 2013-=-), a state-of-theart joint method; • We incorporate latent context into the joint SRL model, bringing additional improvements. In next section, we discuss related work. We describe our approach in Sec... |

75 | Modeling relations and their mentions without labeled text.
- Riedel, Yao, et al.
- 2010
(Show Context)
Citation Context ...r of entities in context may be associated with one, none, or several of the possible relation labels, a property which complicates the application of distant supervision methods (Mintz et al., 2009; =-=Riedel et al., 2010-=-; Hoffmann et al., 2011; Surdeanu et al., 2012). Lao et al. (2012) learned syntactic rules for finding relations defined by “lexico-semantic” paths spanning KB relations and text data. Wang et al. (20... |

75 | Multi-instance multi-label learning for relation extraction.
- Surdeanu, Tibshirani, et al.
- 2012
(Show Context)
Citation Context ...ith one, none, or several of the possible relation labels, a property which complicates the application of distant supervision methods (Mintz et al., 2009; Riedel et al., 2010; Hoffmann et al., 2011; =-=Surdeanu et al., 2012-=-). Lao et al. (2012) learned syntactic rules for finding relations defined by “lexico-semantic” paths spanning KB relations and text data. Wang et al. (2015) extends the methods used by Lao et al. to ... |

55 | Reasoning With Neural Tensor Networks For Knowledge Base Completion.
- Socher, Chen, et al.
- 2013
(Show Context)
Citation Context ...) populated by IE techniques have also been used as an input to systems that learn rules allowing further inferences to be drawn from the KB (Lao et al., 2011), a task sometimes called KB completion (=-=Socher et al., 2013-=-; Wang et al., 2014; West et al., 2014). Pipelines of this sort frequently suffer from error cascades, which reduces performance of the full system1. In this paper, we address this issue, and propose ... |

43 | Fast and robust neural network joint models for statistical machine translation.
- Devlin, Zbib, et al.
- 2014
(Show Context)
Citation Context ...tasks. For example, Finkel and Manning (2009) work on the problem of joint IE and parsing, where they use tree representations to combine named entities and syntactic chunks. Recently, Devlin et al. (=-=Devlin et al., 2014-=-) use a joint neural network model for machine translation, and obtain an impressive 6.3 BLEU point improvement over a hierarchical phrase-based system. In information extraction, weak supervision (Cr... |

23 | Reading the web with learned syntactic-semantic inference rules. - Lao, Subramanya, et al. - 2012 |

16 |
Knowledge base completion via search-based question answering.
- West, Gabrilovich, et al.
- 2014
(Show Context)
Citation Context ...een used as an input to systems that learn rules allowing further inferences to be drawn from the KB (Lao et al., 2011), a task sometimes called KB completion (Socher et al., 2013; Wang et al., 2014; =-=West et al., 2014-=-). Pipelines of this sort frequently suffer from error cascades, which reduces performance of the full system1. In this paper, we address this issue, and propose a joint model system for IE and KB com... |

15 | Programming with personalized pagerank: a locally groundable first-order probabilistic logic.
- Wang, Mazaitis, et al.
- 2013
(Show Context)
Citation Context ...W),hasWord(Y,W), linkedBy(X,Y,W) # sim,word. linkedBy(X,Y,W) :- true # by(W). Table 1: A simple program in ProPPR. See text for explanation. newly proposed scalable probabilistic logic called ProPPR (=-=Wang et al., 2013-=-; Wang et al., 2014). Then, we describe the joint model for information extraction and relational learning. Finally, a latent context invention theory is proposed for enhancing the performance of the ... |

11 | Connecting language and knowledge bases with embedding models for relation extraction. - Weston, Bordes, et al. - 2013 |

7 | Combining distant and partial supervision for relation extraction,”
- Angeli, Tibshirani, et al.
- 2014
(Show Context)
Citation Context ...ntal results in Section 5, discuss in Section 6, and conclude in Section 7. 1For example, KBP slot filling is known for its complex pipeline, and the best overall F1 scores (Wiegand and Klakow, 2013; =-=Angeli et al., 2014-=-) for recent competitions are within the range of 30-40. 2 Related Work In NLP, our work clearly aligns with recent work on joint models of individual text processing tasks. For example, Finkel and Ma... |

7 | Weakly supervised user profile extraction from twitter
- Li, Ritter, et al.
- 2014
(Show Context)
Citation Context ... in a pipeline that contains non-trivial downstream tasks, such as question answering (Mollá et al., 2006), machine translation (Babych and Hartley, 2003), or other applications (Wang and Hua, 2014; =-=Li et al., 2014-=-). Knowledge bases (KBs) populated by IE techniques have also been used as an input to systems that learn rules allowing further inferences to be drawn from the KB (Lao et al., 2011), a task sometimes... |

7 | Efficient inference and learning in a large knowledge base: Reasoning with extracted information using a locally groundable first-order probabilistic logic. arXiv preprint arXiv:1404.3301 - Wang, Mazaitis, et al. - 2014 |

4 | Structure learning via parameter learning
- Wang, Mazaitis, et al.
- 2014
(Show Context)
Citation Context ...hniques have also been used as an input to systems that learn rules allowing further inferences to be drawn from the KB (Lao et al., 2011), a task sometimes called KB completion (Socher et al., 2013; =-=Wang et al., 2014-=-; West et al., 2014). Pipelines of this sort frequently suffer from error cascades, which reduces performance of the full system1. In this paper, we address this issue, and propose a joint model syste... |

3 | Dniel Fogaras, Balzs Rcz, and Tams Sarls. 2005. Towards scaling fully personalized PageRank: Algorithms, lower bounds, and experiments - Csalogny |

2 |
Can predicate invention in meta-interpretive learning compensate for incomplete background knowledge
- Cropper, Muggleton
- 2014
(Show Context)
Citation Context ... dataset includes 30 relations8. As for the Geo dataset, there are 497 6To give some background on this nomenclature, we note that the SL method is inspired by Cropper and Muggleton’s Metagol system (=-=Cropper and Muggleton, 2014-=-), which includes predicate invention. In principle predicates could be invented by SL, by extending the interpreter to consider “invented” predicate symbols as binding to its template variables (e.g.... |

2 | Walk Inference and Learning in A Large Scale Knowledge Base - Random |

2 | entity recognition for question answering - Named |

1 | Tassilo Barth Michael Wiegand and Mittul Singh Dietrich Klakow. 2013. Effective slot filling based on shallow distant supervision methods - Roth |