The Small Business Innovation Research (SBIR) research project will create software that gives beginning and intermediate English language learners (ELLs) a jumpstart in listening comprehension. A major problem for ELLs is simply understanding fluent speech. Even though they may have no trouble with well-enunciated speech, they are at a loss with the contractions and phonetic transformations of casual speech. To address this, the project will automatically transform well-articulated utterances into more casual, real-life speech using innovative voice transformation technology. This controlled variation of speaking style will help users attach new mental representations to casual speech, enlarging their statistical representations of style variability in English. Better listening comprehension lets users understand their teachers, employers, customers and peers and better participate in class discussions, handle real world situations and assimilate into an English-speaking society. Casual speech differs from clear in more than just duration or speaking rate. It has differences in segmental and phonetic/acoustic properties. The product outcomes will show that presenting all of these variations by transforming clear speech into truly casual form is much more effective for long term, real-world relevant learning outcomes.
The market for English language learning products in targeted sectors represents a massive revenue opportunity for providers of scalable, effective and affordable language training solutions. This market includes: business process outsourcing (BPO)/call center: $2B, US education: $260M, global education: $2B to $3B and consumer market of about $100B. Although there are many products that deal with reading and writing and some with speaking, there are few products that target listening comprehension skills in a methodical way. This presents first-movers into this sector with an immediate and significant business opportunity. The project will create a novel listening comprehension product and market it alongside companion products so that a complete suite of answers are available to students who wish to converse with a native speaker of English. The algorithms developed will be language-independent so that other products can be developed for other languages as well.
