Veteran journalist Carl Hoffman believes he’s solved one of the great mysteries of the 20th century. In 1961 at the age of 23, Michael Rockefeller – son of New York Governor Nelson Rockefeller and a member of one of the richest and most powerful families in America ¬– travelled to remote New Guinea in search of primitive art for his father’s new museum.
Columbus Residents Talk for Science
Researchers at Ohio State University have created a database of Columbus residents talking. Researchers say the recordings of people’s conversations will be a valuable resource for anyone who studies spoken communication. Eventually, it may even lead to computers that can understand human speech.
It’s called the Buckeye Speech Corpus – a collection of interviews with 40 Columbus residents on everyday topics, like sports, politics, and schools. OSU scientists recorded their speech, and carefully transcribed every sound, syllable, cough, and laugh. After several years of transcribing, they’ve created a database that’s the largest of its kind, with 40 hours of sound and nearly 307,000 words. In March of this year, they released half the database for use by anyone who studies speech and language.
OSU Psychology Professor Mark Pitt is one of the leaders of the project. He says this work is important because it helps people understand the fundamental human property of language.
“Language – how you and I communicate right now – language is an inherent property of being a social creature, and humans are super-social creatures,” says Pitt.
OSU postdoctoral researcher Laura Dilley also helped lead the project. She says spoken communication is a lot more complex and subtle than people realize.
“What most people miss on a daily basis is that that process is far from trivial,” Dilley says. “The fact that we’re able to vibrate molecules in the air and on the receiving end pick those vibrations up as spoken language which we can subsequently understand is amazing. And we’re interested in helping to further the effort of scientists around the world in understanding that process.”
Dilley says people often take liberties in pronunciation, leading many variations in how people talk.
These variations in speech sometimes appear to be quite significant – yet, for the most part, people understand what they’re saying to each other. Pitt and Dilley study how people interpret spoken language. They’re interested in what Pitt calls “acoustic cues” that help the listener understand what’s being said. A listener can usually understand a spoken sentence in its entirety even if it’s spoken quickly. But if the listener only hears a word or two without the rest of the sentence, it’s almost impossible to understand.
Here’s an excerpt of a woman talking from the database: “They have ways of wording those things that you have to read it very carefully to know whether to say yes or no “
The woman slurs the words “carefully to know” so much it comes across as gibberish. But with the whole sentence, the listener can interpret those words, even though isolated, it’s incomprehensible.
Pitt says one major application of the corpus – which is Latin for a body of work – is in developing computers that understand human speech. Cell phones may understand single words like “home,” but full sentences are another story. Because people slur words and talk with different accents, it’s difficult for computers to really know what people are talking about. ut learning how spoken language works will allow scientists to eventually program computers to understand human speech.
Pitt says one of the strengths of this database is it’s available to researchers from around the world in a variety of fields.
He says, “this is valuable for research across multiple disciplines, in psychology, computer science, speech and hearing sciences, and of course linguistics.”
OSU Professor of Linguistics Beth Hume is also affiliated with this project. She uses the corpus to learn what factors influence certain pronunciations of words like butter. Hume wants to know whether things like preceding vowels make people replace a “t” with a “d” and say “budder,” instead of “butter.” But she says this is more than just a neat exercise.
“We’re looking at language to give us a window into human cognition,” she says.
So far, the researchers have received many requests to use the corpus. In some cases, people have requested to use it in unexpected areas. One such person is an English Professor at OSU, David Herman. Herman’s a part of Project Narrative. The project promotes research in what’s called narrative theory – the study of what distinguishes a story from a mere collection of words.
Using scientific tools, he wants to search the database for clues that reveal the nature of storytelling. Herman says narrative research is a great way to build bridges between the arts and sciences.
“I think actually just the existence of the corpus, the fact that I’m in the English Department and came across it and connect with somebody in Psychology in itself the corpus has the potential to create these kinds of bridges and connections. So, I’m pleased to have access to it,” he says.
Scientists say they’ll officially release the second half of the Buckeye Speech Corpus in September.
Marcus Woo, WOSU news.