Speech Scientist

This job posting is no longer active

Responsibilities

 

Verizon’s OnCue Voice technologies team is looking for a Speech Scientist to help develop best in class ASR/NLU solutions for improving the TV watching experience. An ideal candidate will be an expert in Speech recognition techniques, machine learning and natural language understanding concepts.

 

                                                                                                                                                    

  • Research innovative approaches to ASR that can handle noisy and reverberant rooms.
  • Simulate new research ideas in Matlab and work with the ASR integration team in validating performance improvements.
  • Recommend and come up with comprehensive data sets for optimal acoustic and language model development.
  • Take a lead role in improving the ASR performance for various user profiles and varying room conditions.

              

Qualifications

  

  • Acoustic model: Understanding of the hidden Markov model and standard criteria for training and adaptation (e.g. MCE, MMI, MAP).
  • Detailed knowledge of Language/Acoustic model training and search algorithms.
  • Familiar with version control and bug tracking tools.
  • Knowledge of acoustic modeling, language modeling, statistical modeling, robust front-end processing.
  • Language model: Knowledge of smoothing techniques for N-grams and model regularization.  Develop novel methods for enhancing performance geared towards specific NLP tasks (e.g. Noun phrase detection).
  • Proficient in Matlab simulations.
  • Search Algorithms: A fundamental understanding of commonly used algorithms such as Viterbi decoding and language model rescoring. Develop and research approaches for combining multiple ASR systems to boost overall recognition accuracies.
  • 2+ years of industry experiencing with a proven track record of delivering at least one ASR solution.
  • Signal processing: Familiarity with standard acoustic features (e.g. MFCC, PLP) and transformations (e.g. LDA). Develop feature spaces that are discriminative, robust to noise, and speaker-adaptive.
  • Minimum Educational Qualifications: MS in Electrical Engineering, PhD preferred.

Equal Employment Opportunity

  • Verizon is a Federal Contractor
  • Verizon requests veteran priority referrals
  • Verizon is an equal opportunity and affirmative action employer M/F/Disability/Vet.
 

Your Careers Account

Create or update your career account

Related Careers