This paper describes the Recognizing Textual Entailment (RTE) system that our teams developed for TAC 2011. Our system combines the entailment score calculated by lexicallevel matching with the machine-learningbased filtering mechanism using various features obtained from lexical-level, chunk-level and predicate argument structure-level information. In the filtering mechanism, we try to discard...

This paper discusses a computational model of language generation, based on work in Phase Theory, that attempts to shed light on how the human mind generates sentences. This model presents explicit algorithms that a) determine selection and merger of Lexical Items, b) determine the labels of Merged elements, c) account for movement of Lexical Items within a derivation, and d) account for when c...

Research on paraphrase has mostly focussed on lexical or syntactic variation within individual sentences. Our concern is with larger-scale paraphrases, from multiple sentences or paragraphs to entire documents. In this paper we address the problem of generating paraphrases of large chunks of texts. We ground our discussion through a worked example of extending an existing NLG system to accept a...

research on multiword clusters (chunks) is based on the assumption that native speakers use plenty of chunks in their everyday language and they are considered as fluent speakers of language. therefore the present study was an attempt to investigate the impact of using chunks on speaking fluency of iranian efl learners. in the first phase of the study, the students of two intermediate classes s...

Human knowledge consists of static and dynamic knowledge chunks. The static ones include the so called lexical knowledge or the ability to sense similarities between facts and between predicates. Through dynamic attainments one can make deductions or one can give answers to a question. There are several and very different approaches to make a model of human knowledge, but one of the most common...

We present an approach for Semantic Role Labeling (SRL) using Conditional Random Fields in a joint identification/classification step. The approach is based on shallow syntactic information (chunks) and a number of lexicalized features such as selectional preferences and automatically inferred similar words, extracted using lexical databases and distributional similarity metrics. We use semanti...

There are conflicting views in the literature as to the role of listener-adaptive processes in language production in general and articulatory reduction in particular. We present two novel pieces of corpus evidence that corroborate the hypothesis that non-lexical variation of durations is related to the speed of retrieval of stored motor code chunks and durational reduction is the result of fac...

We present a method for improving statistical machine translation performance by using linguistically motivated syntactic information. Our algorithm recursively decomposes source language sentences into syntactically simpler and shorter chunks, and recomposes their translation to form target language sentences. This improves both the word order and lexical selection of the translation. We repor...

The task of our research is to form phone-like models and a phoneme-like set from spoken word samples without using any transcriptions except for the lexical identi cation of each word in a vocabulary. This framework is derived from two motivations: 1) automatic design of optimal speech recognition units and structures of phone models, and 2) multi-lingual speech recognition based on languagein...