Since 1992, the Speech Recognition in Noise Test, or SPRINT, has been the standard speech-in-noise test for assessing auditory fitness-for-duty of US Army Soldiers with hearing loss. The original SPRINT test consisted of 200 monosyllabic words presented at a Signal-to-Noise Ratio (SNR) of +9 dB in the presence of a six-talker babble noise. Normative data for the test was collected on 319 hearin...

Phonotactic constraints are language-specific patterns in the sequencing of speech sounds. Are these constraints represented at the syllable level (ng cannot begin syllables in English) or at the word level (ng cannot begin words)? In a continuous recognition-memory task, participants more often falsely recognized novel test items that followed than violated the training constraints, whether tr...

This paper presents some of the recent research on speaker-independent continuous speech recognition at LIMSI including efforts in phone and word recognition for both French and English. Evaluation of an HMMbased phone recognizer on a subset of the BREF corpus, gives a phone accuracy of 67.1% with 35 context-independent phone models and 74.2% with 428 context-dependent phone models. The word ac...

In recent years, word embeddings have been surprisingly effective at capturing intuitive characteristics of the words they represent. These vectors achieve the best results when training corpora are extremely large, sometimes billions of words. Clinical natural language processing datasets, however, tend to be much smaller. Even the largest publicly-available dataset of medical notes is three o...

Word sense disambiguation (WSD) is the problem of determining the right sense of a polysemous word in a certain context. This paper investigates the use of unlabeled data for WSD within a framework of semi-supervised learning, in which labeled data is iteratively extended from unlabeled data. Focusing on this approach, we first explicitly identify and analyze three problems inherently occurred ...

Annotating large numbers of sentences with senses is the heaviest requirement of current Word Sense Disambiguation. We present Train-O-Matic, a languageindependent method for generating millions of sense-annotated training instances for virtually all meanings of words in a language’s vocabulary. The approach is fully automatic: no human intervention is required and the only type of human knowle...

We propose a sequence-alignment based method for detecting and disambiguating coordinate conjunctions. In this method, averaged perceptron learning is used to adapt the substitution matrix to the training data drawn from the target language and domain. To reduce the cost of training data construction, our method accepts training examples in which complete word-by-word alignment labels are missi...

We present an unsupervised method for word sense disambiguation that exploits translation correspondences in parallel corpora. The technique takes advantage of the fact that cross-language lexicalizations of the same concept tend to be consistent, preserving some core element of its semantics, and yet also variable, reeecting diier-ing translator preferences and the in-uence of context. Working...

Most word sense disambiguation (WSD) methods require large quantities of manually annotated training data and/or do not exploit fully the semantic relations of thesauri. We propose a new unsupervised WSD algorithm, which is based on generating Spreading Activation Networks (SANs) from the senses of a thesaurus and the relations between them. A new method of assigning weights to the networks’ li...