The rule is: ÔiÕ before ÔeÕ except after Ôc.Õ
But that rule gets broken all the time. (Think Òweird,” Òchief,” Òbeige,” etc.) Not to mention that syllables donÕt always break between double letters or begin with a consonant. The English language is full of exceptions.
Which is why rule-based text-to-speech computer programs tend to talk funny.
Computer scientist Connie Adsett is investigating how to minimize that problem by finding the best tools for automatically breaking words into their proper syllables instead of relying on a rules-based system.
Her research has been recognized with a top prize awarded by the Natural Sciences and Engineering Research Council (NSERC). Each year, the federal funding agency awards two AndrŽ Hamer Postgraduate Prizes, and this year, both are going to Dalhousie-affiliated researchers.
The other prize goes to PhD student Erin Johnson, who came to ϳԹ to continue working with her supervisor, Dr. Axel Becke, one of the most-cited chemists in the world. As a computational chemist, the QueenÕs University PhD student predicts how chemicals behave using modeling methods on the computer instead of doing experiments in the lab. Her current research focuses on refining existing modeling methods and developing new ones to increase their accuracy and efficiency.
ÒThere are a lot of good reasons for figuring out how to make the computer pronounce words better,” continues Ms. Adsett, reached in England where she is a visiting student.
Text-to-speech technology is useful for people with certain disabilities. Blind people, for example, can have the contents of electronic documents read to them and those unable to speak can enter words that the computer will speak for them. And, itÕs not too far off to imagine that soon weÕll be asking our handheld computers to read our email for us.
ÒI guess IÕve always been curious,” says Ms. Adsett, who does her work at the NRC Institute for Biodiagnostics (NRC-IBD) located at the Halifax Infirmary. Her supervisor is Dr. Yannick Marchand.
ÒWhen youÕre in computer science, youÕre always dealing with programs that donÕt work and need fixing. ItÕs such a high when you solve a problem and then go on to the next one.”
Besides working on text-to-speech programs for the English language, the 23-year-old graduate student says her research also focuses on programs for German, Dutch, French, Italian, Spanish, Norwegian and lesser-known languages, Basque, from northern Spain, and Frisian, spoken in the north of Holland and Germany. The languages were chosen because they have databases of words in which syllabification is known.
In the case of Frisian and Basque, databases exist thanks to the efforts of people trying to preserve traditional dialects.
ÒItÕs because people are really dedicated to making these languages survive,” she explains.