In this paper we present preliminary experiments that aim to reduce lexical data sparsity in statistical parsing by exploiting information about named entities. Words in the WSJ corpus are mapped to named entity clusters and a latent variable constituency parser is trained and tested on the transformed corpus. We explore two different methods for mapping words to entities, and look at the effec...

Wednesday, June 15th 8:00 Conference Registration (Registration desk) 8:45 Session 1: Large-Scale Online Linguistic Resources (I) Chair: "Text Categorization Based on Subtopic Clusters" Francis Chik, Robert Luk, Korris Chung "Automatic Filtering of Bilingual Corpora for Statistical Machine Translation" Shahram Khadivi, Hermann Ney "The Role of Word Sense Disambiguation in Automated Text Categor...

This paper presents a model of the mental lexicon and its formation, based on the self-organizing neural network. When exposed to raw text, the model clusters words according to their semantic relatedness to form a semantic network 7]. Simulations using artiicial data are described that show how co-occurrence information can be used to create a low-dimensional representation of lexical semantic...

Ambiguous person names are a problem in many forms of written text, including that which is found on the Web. In this paper we explore the use of unsupervised clustering techniques to discriminate among entities named in Web pages. We examine three main issues via an extensive experimental study. First, the effect of using a held–out set of training data for feature selection versus using the d...

Generative probabilistic models have been used for content modelling and template induction, and are typically trained on small corpora in the target domain. In contrast, vector space models of distributional semantics are trained on large corpora, but are typically applied to domaingeneral lexical disambiguation tasks. We introduce Distributional Semantic Hidden Markov Models, a novel variant ...

This paper reports on a pilot study of aspect marker generation in an English-to-Chinese translation scenario. Our classifier combines a number of linguistic features in a Maximum Entropy framework and achieves an overall accuracy of 78%. We also investigate the impact of different clusters of linguistic features; we find that syntactic features have the highest utility and lexical aspectual pr...

We investigate the use of generalized representations (POS, morphological analysis and word clusters) in phrase-based models and the N-gram-based Operation Sequence Model (OSM). Our integration enables these models to learn richer lexical and reordering patterns, consider wider contextual information and generalize better in sparse data conditions. When interpolating generalized OSM models on t...

This thesis introduces the concept of syntactic cross-priming: priming between nonidentical syntactic rules, analogous to lexical association. I also present the hypothesis that cross-priming is related to similarity, to the effect that pairs more similar rules will have higher cross-priming strengths. I introduce a variant of adaptation, crossadaptation as a measure for cross-priming in a corp...

PURPOSE To study the role of visual perception of phonemes in visual perception of sentences and words among normal-hearing individuals. METHOD Twenty-four normal-hearing adults identified consonants, words, and sentences, spoken by either a human or a synthetic talker. The synthetic talker was programmed with identical parameters within phoneme groups, hypothetically resulting in simplified ...

Inter-consonantal cohesion in French word-initial CC clusters is investigated in light of recent proposals of gestural coordination. Specifically, the timing of lip and tongue movements of C1/l/ and C1/n/ productions, with C1 being one of the consonants /p, f, k/, of two speakers were studied using electromagnetic articulography (EMA). In French, C/l/ clusters occur frequently in word-initial p...