Language Technologies Institute
School of Computer Science
Carnegie Mellon University
5000 Forbes Avenue,
Office: 6403 Gates Hillman Complex
Phone: (617) 869-1844
Affiliations:Scone Knowledge Representation Group
Research InterestsSemantic language modeling, using semantics in automatic speech recognition (ASR), ASR, knowledge representation, construction grammars, linguistics, and natural language processing.
Current research work
Integrating semantics into ASR systems
My current research work is on how to incorporate high-level semantic information into automatic speech recognition (ASR). For example, how does one inform an ASR system that "the pulp will be used to produce newsprint" should be preferred over "the Pope will be used to produce newsprint"? Acoustically, these two are very similar, but to a person, one sounds clearly more plausible than the other. My research is on how to encode semantic information needed to prefer the former, and how to use that semantic information in an ASR system.
In December 2009, I proposed this thesis topic! Read more about it here: thesis proposal.
I am currently working as an organizing editor for the Journal of Negative Results in Speech and Audio Sciences. This journal is specifically tailored to negative results, which we believe are often even more useful and educational than positive results. Check out the journal here: jnrsas.org
- Benjamin Lambert, Bhiksha Raj, and Rita Singh, “Creating a semantic coherence dataset with non-expert annotators,” In submission to HLT-NAACL Workshop on Creating Speech and Language Data with Amazon’s Mechanical Turk, 2009.
- Benjamin Lambert, “A Knowledge-Based Architecture for using Semantics in Automatic Speech Recognition,” Ph.D. Thesis Proposal, Carnegie Mellon University, 2009.
- Benjamin Lambert and Scott E. Fahlman. “Knowledge-Driven Learning and Discovery.” Student-track poster session of the Twenty-Second AAAI Conference on Artificial Intelligence. Vancouver, British Columbia, 2007
- Alicia Tribble, Benjamin Lambert, and Scott E. Fahlman. “SconeEdit: A Text-Guided Domain Knowledge Editor.” Demonstration Sessions of HLT/NAACL-2006. New York, 2006.
- Benjamin Lambert. “Improving Information Retrieval with Natural Language Processing.” Master's thesis, University of Illinois Urbana-Champaign, 2006.
- Benjamin Lambert. “Classifying Entity Relations in Natural Language.” Senior undergraduate honors thesis. University of Massachusetts at Amherst. June 2003.