Citation Context

...ered into three main kinds of methods: (1) methods that directly model the correlation between the variables, using discrete Markov random fields (e.g., [23]) or continuous relaxations thereof (e.g., =-=[34]-=-); (2) methods that use latent variables to model the correlations indirectly, using either discrete factors (e.g., [48]) or continuous factors (e.g., [31, 11, 20, 37]); and (3) methods that approxima...

Citation Context

...ling both discrete and continuous data. The effectiveness of HL-MRFs and PSL has also been demonstrated in other publications on many problem domains, including automatic knowledge base construction (=-=Pujara et al., 2013-=-), high-level computer vision (London et al., 2013b), drug discovery (Fakhraei et al., 2014), natural language semantics (Beltagy et al., 2014; Sridhar et al., 2015), automobiletraffic modeling (Chen ...

Citation Context

...conds, while also supporting lazy inference to produce a broader set of 4M candidates in just 2 hours. In this extended abstract we summarize contributions from a longer paper presented at ISWC 2013 (=-=Pujara et al. 2013-=-). Our work: • Defines the knowledge graph identification problem; • Uses soft-logic values to leverage extractor confidences; • Formulates knowledge graph inference as convex optimization; • Evaluate...

Citation Context

...ntiments and +/-effect events to be provided as oracle information by manual annotations. Probabilistic Soft Logic. Probabilistic Soft Logic (PSL) is a variation of Markov Logic Networks, which is a framework for probabilistic logic that employs weighted formulas in firstorder logic to compactly encode complex undirected probabilistic graphical models (i.e., Markov networks) (Bach et al., 2015; Beltagy et al., 2014). PSL is a new statistical relational learning method that has been applied to many NLP and other machine learning tasks in recent years (Beltagy et al., 2014; London et al., 2013; Pujara et al., 2013; Bach et al., 2013; Huang et al., 2013; Memory et al., 2012). Previously, PSL has not been applied to entity/event-level sentiment analysis. 3 Task Definition In this section, we introduce the definition of the entity/event-level sentiment analysis task, followed by a description of the gold standard corpus. For each sentence s, we define a set E consisting of entities, events, and the writer of s, and sets P and N consisting of positive and negative sentiments, respectively. Each element in P is a tuple, representing a positive pair of two entities, (e1, e2) where e1, e2 ∈ E, and e1 is posit...

Citation Context

...mpany [3] “We define a Knowledge Graph as an RDF graph. An RDF graph consists of a set of RDF triples where each RDF triple (s, p, o) is an ordered set of the following RDF terms: a subject s ∈ U ∪ B, a predicate p ∈ U , and an object U ∪ B ∪ L. An RDF term is either a URI u ∈ U , a blank node b ∈ B, or a literal l ∈ L.” Farber et al. [7] “[...] systems exist, [...], which use a variety of techniques to extract new knowledge, in the form of facts, from the web. These facts are interrelated, and hence, recently this extracted knowledge has been referred to as a knowledge graph.” Pujara et al. [17] Table 1: Selected definitions of knowledge graph tion and do not even enforce a graph structure. In addition, size is highlighted as an essential characteristic, which is reflected by phrases such as “large networks” or “vast networks” [11], while it remains unclear what “large” means in this context. Farber et al. defined a knowledge graph as an Resource Description Framework (RDF) graph and stated that the term KG was coined by Google to describe any graph-based knowledge base (KB) [7]. Although this definition is the only formal one, it contradicts with more general definitions as it expl...

Citation Context

...Intuitively, an ideal bonding agent would reside on as many cross group pairwise shortest paths as possible, and meanwhile connect as large portion of two groups as possible. Such agents could best serve the message passing between two groups. The above two examples imply an essential need to efficiently discover the most valuable bonding agents between two sets of vertices, which could be given at ad hoc. As a matter of fact, such a demand is common in many real world applications, like ad hoc logistics planning [38], or extracting and querying correlations in knowledge graphs on input facts [41], and etc. There is a rich literature on structure oriented graph queries, such as shortest path, reachability, subgraph matching, influential maximization [28] and etc. However, we find that none of these existing graph queries could be directly applied to the scenarios we discussed above. We briefly review the most relevant graph queries and elaborate the novelty of the VSB query. As VSB is defined on the shortest path semantic, it makes betweenness centrality a reasonable metric to employ. Top-k betweenness computing or kbetweenness are classic queries employed to find important vertices in...

Citation Context

...a NELL dataset consisting of 1.7M extractions and 80K ontological constraints.We adopt the formulation of (Jiang et al., 2012) to relate extractions and ontological relationships to facts. Our model (=-=Pujara et al., 2013-=-) uses a a continuous-valued Markov random field, which we formulate using Probabilistic Soft Logic (PSL)(Broecheler et al., 2010). PSL relaxes the truth-value of logical atoms to the [0, 1] domain an...

Citation Context

...ships between these structured extractions, and representing both the extractions and relationships as a knowledge graph [16]. In conjunction with the knowledge graph representation, current research =-=[7, 14]-=- jointly reasons about millions of web-scale extractions, while leveraging ontological information, to produce a consistent knowledge graph. Jointly reasoning about noisy extractions from the web has ...

Citation Context

...), a recent method introduced in [13] to make probabilistic inference tractable on relational networks. It has been successfully used in collective classification [14], knowledge graph identification =-=[15]-=- or sentiment analysis in social networks [16]. HL-MRFs were motivated in Probabilistic Similarity Logic (PSL) [17], a general-purpose framework for probabilistic reasoning about similarity in relatio...

Citation Context

...using HL-MRFs. HLMRFs are a scalable class of continuous, conditional graphical models [2]. HL-MRFs have achieved state-of-the-art performance in many domains including knowledge graph identification =-=[14]-=-, understanding engagement in MOOCs [15], biomedicine and multi-relational link prediction [5], and modelling social trust [9]. 4.2.1 Probabilistic Soft Logic HL-MRF models can be specified using Prob...