Application of Fuzzy Transcription to Automatic Speech Recognition
Preethi Jyothi, Beckman Postdoctoral Fellow
Transcribed speech is a critical resource for building speech recognition systems. Recent work has looked towards soliciting transcriptions for large speech corpora from native speakers of a target language using crowdsourcing techniques. However, for low resource languages, native speakers may not be readily available. We examine whether humans unfamiliar with the target language can help provide a “fuzzy" transcription accurate enough for building an ASR system. We describe our techniques used for fuzzy transcription and show how it can be used to significantly improve ASR systems for various languages.
Activational and Cellular Changes in the Adult Auditory Cortex Following Developmental Exposure to PCBs in Rats
Renee Sadowski, Beckman Postdoctoral Fellow
Polychlorinated biphenyls, or PCBs, are persistent environmental toxicants that are found in high concentrations in many areas of the United States, including the Great Lakes of the Midwest. Developmental exposure to PCBs causes hearing impairments in both humans and in animal models, and recent evidence from the Schantz laboratory has also demonstrated long-lasting increases in audiogenic seizures after exposure to this toxicant. My research utilizes imaging and electrophysiological techniques in rodents to determine how an environmental relevant mixture of PCBs alters inhibition and excitation within the auditory cortex, a brain structure important for the processing of auditory information. My work has revealed that developmental PCB exposure increases activation within the auditory cortex when GABA, the primary inhibitory neurotransmitter, is antagonized. Additionally, preliminary analysis shows indications for increased cellular inhibition in the auditory cortex of PCB-exposed animals. These findings suggest that inhibition and excitation in the auditory cortex is disrupted by developmental exposure to PCBs.