Decomposing Semantic Inference

Decomposing Semantic Inferences Elena Cabrio1 and Bernardo Magnini2 Beside formal approaches to semantic inference that rely on logical representation of meaning, the notion of Textual Entailment (TE) has been proposed as an applied framework to capture major semantic inference needs across applications in Computational Linguistics. Al- though several approaches have been tried and evaluation campaigns have shown improvements in TE, a renewed interest is rising in the research community towards a deeper and better understanding of the core phenomena involved in textual inference. Pursuing this direction, we are convinced that crucial progress will derive from a focus on de- composing the complexity of the TE task into basic phenomena and on their combination. In this paper, we carry out a deep analysis on TE data sets, investigating the relations among two relevant aspects of semantic inferences: the logical dimension, i.e. the capacity of the inference to prove the conclusion from its premises, and the linguistic dimension, i.e. the linguistic devices used to accomplish the goal of the inference. We propose a decomposition approach over TE pairs, where single linguistic phenomena are isolated in what we have called atomic inference pairs, and we show that at this granularity level the actual correlation between the linguistic and the logical dimensions of semantic inferences emerges and can be empirically observed. 1 Introduction The ability to carry out semantic inferences is pervasive in our capacity to understand natural languages. In particular, we show a crucial skill in establishing meaningful relations among di↵erent pieces of text in order to reconstruct their connections: as an example, the meaning of one portion of text can be expressed by another portion of text (i.e. paraphrasing), it can be contained (i.e. entailed) by the other, it can 1INRIA Sophia Antipolis, France. 2Fondazione Bruno Kessler, Trento, Italy. 79 LiLT Volume 9 Perspectives on Semantic Representations for Textual Inference. Copyright c ⃝ 2014, CSLI Publications. 80 / Elena Cabrio and Bernardo Magnini be interpreted as the cause or the e↵ect, or it can express the fact that it temporally precedes or follows the other. From a computational perspective, it seems difficult for any automatic system not to aim at replicating some degree of human semantic inferencing. While the logical nature of such semantic inferences has been the subject of a huge amount of literature in the area of Philosophy of Language, it is only in the recent years that this topic has produced new trends of investigation in Computational Linguistics. A relevant achievement has been the focus on automatically recognizing “textual inferences” as the main research goal, which has let to the set-up of a general framework of research, independent from the actual methods used to address the problem. Focusing on the discovery of semantic relations among two portions of text has in fact opened the way to a number of new approaches and techniques, as well as to the develop- ment of several annotated data sets. The renaissance of interest around semantic inferences in Computa- tional Linguistics is well shown …

Content truncated to 500 words