CLingDing - The IU Computational Linguistics Discussion Group
Time: 01:00pm - 02:00pm
Place: Memorial Hall 401
Alex RudnickParaguay has a unique position in the Americas in that Paraguayans are typically bilingual in Spanish and Guarani, the largest indigenous language from the region. Guarani is at a significant disadvantage, however, in public life. Politics, science, etc., are conducted almost entirely in Spanish, even though Guarani is co-official.We just got back from a week in Paraguay, meeting local language activists and planning our computer-assisted translation tool. I'll talk about Paraguay, Guarani, the tool we plan to build, and about our cunning plan to help bilingual Paraguayans help us collect enough bitext to train a good MT/CAT system for their local language pair.
In category: Computational linguistics
Interpretation and reinterpretation of phonological cues in psycholinguistic research: Assessing distractor neutrality
Time: 12:15pm - 01:00pm
Place: Ballantine 004
Wallace SalkauskiExperimental research employs several techniques designed to control for confounding variables during testing, e.g. distractors. Of these, the neutral distractor has become controversial. While a primary purpose of the neutral (unrelated) distractor is to serve as a base for comparison, recent research suggests that, while the neutral distractor has been assumed to be impartial and unbiased, this may not be the case. Several questions arise regarding the actual effect of neutral elements and the role they play in experimental research. Within this, the type or mode of distractor also has a significant impact on the outcome of a study. In particular, the use of phonological elements may have an unforeseen consequence on experimental tasks. Therefore, my study provides empirical evidence regarding the lack of partiality of the phonological neutral distractor. Specifically, the distractor in the neutral condition resulted in significant inhibitory effects on L1 and L2 production rather than serving as an impartial variable as intended.In a study on language environment and cue interpretation, L2 learners of varying levels of Spanish proficiency (N=57) performed a picture naming and a translation task (in both English and Spanish). During each task, participants were aurally presented with phonological distractors (initial segments of target language stimuli, e.g. participants heard [wu] for woman or [mu] for mujer). The distractors were presented simultaneously with target stimuli and were of three conditions: the related condition (i.e., English onset when speaking English), the unrelated condition (English onset when speaking in Spanish), or a neutral condition, with segments from Arabic (i.e. [fi], for the Arabic elephant). A goal was to examine how participants interpreted various phonological cues in English and Spanish language environments. Results suggest that we need to reevaluate the function of ‘neutral’ distractors and question whether unambiguous neutrality is obtainable. In this presentation, I assess and discuss distractor neutrality and consider its implications for models of bilingual production, I propose ways to avoid such methodological pitfalls, and argue that psycholinguistic research consider phonologically-based theories of processing in addition to conventional models.
In category: Phonetics and phonology
An overview of the development and results obtained with CDT’s Speech Perception Assessment and Training System (SPATS) with ESL Learners, Cochlear-Implant Users, Hearing-Aid Users, and Aphasics
Time: 01:30pm - 03:00pm
Place: Psychology 128 (conference room)
James Miller, Charles Watson, and Roy Sillings (Communication Disorders Technology, Inc.)A study of over 200 ESL learners with basic English literacy leads to following conclusions:1) All have serious deficiencies in the identification of the syllable constituents of English and the identification of words in sentences especially in the presence of background multi-talker babble.2) After 15 to 25 hours of distributed practice with the SPATS-ESL program, almost all ESL learners achieve near-native or native-like accuracy in the identification of syllable constituents and of words in sentences presented in multi-talker babble.3) Most ESL learners studied are best at onset identification, worst at nucleus identification, and find codas to be of intermediate difficulty.4) Subjective ratings of the trainees indicated that they believe program was useful.Studies of 8 cochlear-implant users and 32 hearing-aid users with 12 to 30 hours of SPATS-HI training leads to the following conclusions:1) SPATS-Hi Training leads to improvements in the users’ abilities to identify syllable constituents in and to identify words in sentences in quiet and noise. However, it is only the identification of syllable nuclei that approaches the performance of normally hearing young adults.2) Statistically significant, but modest, improvements on non-SPATS measures of speech perception such as the QuickSIN, W22, CST tests have been observed. The magnitude of the improvements may be related to cognitive and the efficiency of the hearing aids and implants as fitted to individual clients.3) Typical trainees value SPATS training and think that SPATS training may help them in everyday situations and almost all believe that it helps them understand the nature of their hearing problem.4) The evaluation of SPATS-HI with hearing aid users is a continuing effort and improvements in the training curricula are being examined.Preliminary Observations with Mild to Moderate Aphasics:We are engaged in a study of feasibility of SPATS-like training for clients with mild-to-moderate aphasia. To date observations have been made with only three such clients: One suffered traumatic brain injury in an auto accident; another suffered injury from a stroke, while the third suffered damage from encephalitis. All three believe that SPATS-like training “is what I need.” All three have difficulty in recognizing syllable constituents and holding them in short-term memory. All three have difficulty in segmenting the sounds of naturally spoken sentences into syllable and words. All show improvement in these skills with practice. Interesting anecdotal observations will be reported that are suggestive of the problems these clients encounter.References will be available. Several related papers and documents can be found at References at www.comdistec.com/new/HI.html and www.comdistec.com/new/ESL.html.
JEvents v3.0.9 Stable
Copyright © 2006-2013
Powered by Joomla!®
Copyright © 2008 The Trustees of Indiana University | Copyright Complaints