Models of lexical semantics are core paradigms in most NLP applications, such as dialogue, information extraction and document understanding. Unfortunately, the coverage of currently available resources (e.g. FrameNet) is still unsatisfactory. This paper presents a largely applicable approach for extending frame semantic resources, combining word sense information derived from WordNet and corpus-based distributional information. We report a large scale evaluation over the English FrameNet, and results on extending FrameNet to the Italian language, as the basis of the development of

Paraphrasing methods recognize, generate, or extract phrases, sentences, or longer natural language expressions that convey almost the same information. Textual entailment methods, on the other hand, recognize, generate, or extract pairs of natural language expressions, such that a human who reads (and trusts) the first element of a pair would most likely infer that the other element is also true. Paraphrasing can be seen as bidirectional textual entailment and methods from the two areas are often similar. Both kinds of methods are useful, at least in principle, in a wide range of natural language processing applications, including question answering, summarization, text generation, and machine translation. We summarize key ideas from the two areas by considering in turn recognition, generation, and extraction methods, also pointing to prominent articles and resources.

This paper presents a discourse processing framework based on weighted abduction. We elaborate on ideas described in Hobbs et al. (1993) and implement the abductive inference procedure in a system called Mini-TACITUS. Particular attention is paid to constructing a large and reliable knowledge base for supporting inferences. For this purpose we exploit such lexical-semantic resources as WordNet and FrameNet. We test the proposed procedure and the obtained knowledge base on the Recognizing Textual Entailment task using the data sets from the RTE-2 challenge for evaluation. In addition, we provide an evaluation of the semantic role labeling produced by the system taking the Frame-Annotated Corpus for Textual Entailment as a gold standard. 1

For this year's RTE challenge we have continued to pursue a (somewhat) "logical" approach to recognizing entailment, in which our system, called BLUE (Boeing Language Understanding Engine) first creates a logic-based representation of a text T and then performs simple inference (using WordNet and the DIRT inference rule database) to try and infer a hypothesis H. The overall system can be viewed as comprising of three main elements: parsing, WordNet, and DIRT, built on top of a simple baseline of bag-of-words comparison. Ablation studies suggest that WordNet substantially improves the accuracy scores, while, somewhat suprisingly, parsing and DIRT only marginally improve the accuracy scores. We illustrate and discuss these results. Overall, BLUE's reasoning is sometimes insightful but sometimes nonsensical, the primary challenges being noise in the knowledge sources, lack of world knowledge, and the difficulty of accurate syntactic and semantic analysis. Despite these challenges, we argue that forming semantic representations is a necessary first step towards the larger goal of machine reading, and worthy of further exploration. Our best scores were 61.5 % (2 way), 54.7 % (3 way), and F=0.29 (Search Pilot). 1.

Supervised semantic role labeling (SRL) systems are generally claimed to have accuracies in the range of 80 % and higher (Erk and Padó, 2006). These numbers, though, are the result of highly-restricted evaluations, i.e., typically evaluating on hand-picked lemmas for which training data is available. In this paper we consider performance of such systems when we evaluate at the document level rather than on the lemma level. While it is wellknown that coverage gaps exist in the resources available for training supervised SRL systems, what we have been lacking until now is an understanding of the precise nature of this coverage problem and its impact on the performance of SRL systems. We present a typology of five different types of coverage gaps in FrameNet. We then analyze the impact of the coverage gaps on performance of a supervised semantic role labeling system on full texts, showing an average oracle upper bound of 46.8%.

Discourse references, notably coreference and bridging, play an important role in many text understanding applications, but their impact on textual entailment is yet to be systematically understood. On the basis of an in-depth analysis of entailment instances, we argue that discourse references have the potential of substantially improving textual entailment recognition, and identify a number of research directions towards this goal. 1

This paper focuses on the improvement of the conceptual structure of FrameNet for the sake of applying this resource to knowledgeintensive NLP tasks requiring reasoning, such as question answering, information extraction etc. Ontological analysis supported by data-driven methods is used for axiomatizing, enriching and cleaning up frame relations. The impact of the achieved axiomatization is investigated on recognizing textual entailment. 1.

Paraphrasing methods recognize, generate, or extract phrases, sentences, or longer natural language expressions that convey almost the same information. Textual entailment methods, on the other hand, recognize, generate, or extract pairs of natural language expressions, such that a human who reads (and trusts) the first element of a pair would most likely infer that the other element is also true. Paraphrasing can be seen as bidirectional textual entailment and methods from the two areas are often very similar. Both kinds of methods are useful, at least in principle, in a wide range of natural language processing applications, including question answering, summarization, text generation, and machine translation. In this thesis, we focus on paraphrase and textual entailment recognition, as well as paraphrase generation. We propose three paraphrase and textual entailment recognition methods, experimentally evaluated on existing benchmarks. The key idea is that by capturing similarities at various abstractions of the inputs, we can recognize paraphrases and textual entailment reasonably well. Additionally, we exploit WordNet and use features that operate on the syntactic level of the language expressions. The best

We present a method and a software tool, the FrameNet Transformer, for deriving customized versions of the FrameNet database based on frame and frame element relations. The FrameNet Transformer allows users to iteratively coarsen the FrameNet sense inventory in two ways. First, the tool can merge entire frames that are related by user-specified relations. Second, it can merge word senses that belong to frames related by specified relations. Both methods can be interleaved. The Transformer automatically outputs format-compliant FrameNet versions, including modified corpus annotation files that can be used for automatic processing. The customized FrameNet versions can be used to determine which granularity is suitable for particular applications. In our evaluation of the tool, we show that our method increases accuracy of statistical semantic parsers by reducing the number of word-senses (frames) per lemma, and increasing the number of annotated sentences per lexical unit and frame. We further show in an experiment on the FATE corpus that by coarsening FrameNet we do not incur a significant loss of information that is relevant to the Recognizing Textual Entailment task. 1.

Systems designed to recognize textual entailment typically employ both syntax and semantics. Our goal in this paper is to explore the degree to which semantics alone can be used to accurately detect entailment so that we can gain a better understanding of this single component within an entailment system. This paper reports the knowledge-bases considered and selected for person names, locations and organizations and the results of the system when used on the Recognizing