UTEP study findings could help make AI voices more natural-sounding

EL PASO, Texas (Jan. 20, 2026) – A speech study by a research team from The University of Texas at El Paso has identified an underappreciated aspect of speech in English and Spanish speakers that could lead to improvements in artificial intelligence (AI) spoken dialogue systems.

People tend to articulate their words less precisely when expressing a positive feeling, according to new research led by Nigel Ward, Ph.D., a professor in UTEP’s Department of Computer Science in the College of Engineering. Reduced articulation in speech, called phonetic reduction, has not previously been considered meaningful by speech technology researchers, Ward said.

“People often perceive AI systems as cold and unaware and don’t trust them, even if their task performance is superb,” Ward said. “Current AI voices are consistently highly intelligible, so there’s space to sacrifice a little articulatory precision when needed. We plan to build voices that do this, so that AI systems may be finally able to escape the cold, robotic stereotype and become more communicative, more trustable, and more useful.”

Ward details his findings in a peer-reviewed paper titled Phonetic Reduction is Associated with Positive Assessment and other Pragmatic Functions, which was recently published in the journal Speech Communication. Undergraduate research assistants Raul O. Gomez, Carlos A. Ortega and Georgina Bugarini coauthored the paper.

In the study, conducted in fall 2023 and spring 2024, researchers recorded speakers saying one of six phrases twice – once in a neutral tone of voice and once in a positive tone. An independent set of non-expert judges then compared each phoneme, or meaningful unit of sound, across the two versions.

Speakers were found to be a third more likely to cut corners in their pronunciation in the positive tone than the neutral tone, with 30% judged to be “reduced” and 9% judged to be “highly reduced” in English. The team also found evidence for the same pattern in Spanish, with 35% reduced and 4% highly reduced.

Phonetic reduction, which includes mumbled, sloppy or generally less articulated speech, also occurs in phrases that serve other social purposes, such as self-talk, expressing uncertainty or closing out a conversation topic. In the study, researchers sought to isolate the correlation between reduction and positive speech.

“When you're speaking in a more positive way, there is a higher pitch and you tend to speak a little bit faster,” said Gomez. “We made sure to eliminate that by making speakers do the positive one first and then the neutral reenactment, so in case the neutral one was longer, they could shorten it down. So we’re only viewing the reduction via the positiveness in the voice rather than the length.”

Ward hopes the study will inspire further exploration of phonetic reduction, especially as it relates to spoken dialogue systems like Siri, Alex and Google Assistant that converse with humans using a human-sounding voice.

The researcher also pointed out that his study focuses only on American English and Mexican Spanish, two languages salient to the Paso del Norte region but not representative of all languages worldwide.

“Even though we might not be able to conduct research on those languages, other people can pick up the torch and then maybe corroborate the findings – or find something else,” Ward said.

About The University of Texas at El Paso

The University of Texas at El Paso is America’s leading Hispanic-serving university. Located at the westernmost tip of Texas, where three states and two countries converge along the Rio Grande, 84% of our 26,000 students are Hispanic, and more than half are the first in their families to go to college. UTEP offers 169 bachelor’s, master’s and doctoral degree programs at the only open-access, top-tier research university in America.