Learning Entailment Relations by Global Graph Structure Optimization

Abstract

Identifying entailment relations between predicates is an important part of applied semantic inference. In this article we propose a global inference algorithm that learns such entailment rules. First, we define a graph structure over predicates that represents entailment relations as directed edges. Then, we use a global transitivity constraint on the graph to learn the optimal set of edges, formulating the optimization problem as an Integer Linear Program. The algorithm is applied in a setting where, given a target concept, the algorithm learns on the fly all entailment rules between predicates that co-occur with this concept. Results show that our global algorithm improves performance over baseline algorithms by more than 10%.

A Consolidated Open Knowledge Representation for Multiple Texts

“…Finally, while the performance of the predicate entailment component reflects the current stateof-the-art (Berant et al, 2012;Han and Sun, 2016), the performance on entity entailment is much worse than the current state-of-the-art in this task as measured on common lexical inference test sets. We conjecture that this stems from the nature of the entities in our dataset, consisting of both named entities and common nouns, many of which are multi-word expressions, whereas most work in entity entailment is focused on single word common nouns.…”

“…A threshold for the binary entailment decision was then calibrated on a held out development set. Finally, for Predicate Entailment we used the entailment rules extracted by Berant et al (2012).…”

Unsupervised acquisition of entailment relations from the Web

“…This is typically done under a supervised framework (Berant, Dagan and Goldberger 2012;Weisman et al 2012). First, while the focus of this article is to increase the scalability of Web-based rule acquisition, a following step would be to further improve the quality of the acquired rules.…”

Multi-Context Term Embeddings: the Use Case of Corpus-based Term Set Expansion

“…To combine multi-context embeddings, we follow the general idea of Berant et al (2012) who train an SVM to combine different similarity score features to learn textual entailment relations. Similarly, we train a Multilayer Perceptron (MLP) binary classifier that predicts whether a candidate term should be part of the expanded set based on 10 similarity scores (considered as input features), using the above 2 different scoring methods for each of the 5 context types.…”