Speech recognition on cellphones is no longer about saying a name and then waiting and hoping that the right number is dialled, many experts say. With most early versions, users trained their phones to understand commands. But the accuracy of the function in real-world use was sketchy at best and nearly zero if the training was too noisy.
Most new cellphones have voice-recognition software already included; on some others the software can be downloaded. With the most advanced software, users can dictate a text or email message, find a calendar item on the phone or jump directly to a ring tone and buy it with a simple command like "Madonna ringtone." This last possibility is especially appealing for carriers, which have content on their mobile portals they are trying to sell clients, most of whom cannot be bothered to click through multiple menus to find what might interest them.
"The challenge has been getting people to realise they can do more with voice applications then just dial a phone number," said Collin Holmes, director of product marketing for V-Enable, a San Diego-based company that makes voice-recognition search software. V-Enable’s software, which in some cases is loaded on the handset directly by the manufacturer but can also be downloaded onto many new phones, is used in the United States by Verizon, US Cellular and other carriers. The company also has a recommendation engine that works like Amazon’s: If you buy a U2 ring tone, for instance, you will receive a list of other products bought by people who acquired that ringtone.
The most compelling market for voice-recognition software might be Asia, because typing ideograms on a cellular phone is more laborious than using a Western alphabet. Many companies, including NEC, are busily developing products.Cellphone users will dedicate 30 seconds to finding the content they are looking for, according to a study published in 2002 by Barry Smyth, the chief technical officer and co-founder of ChangingWorlds, an Irish maker of speech-recognition software to improve the usability of cellphone portals. Smyth said the research findings were still valid despite the passing of three years.
For every second of navigation time the average user saves when searching for content, he or she will spend another three seconds browsing, he found. Three seconds: big deal, right? Indeed it is for cellphone carriers trying to get clients to do more than just use their phones for calls and text messages. "In a lot of cases it takes 15 clicks for people to find the content they want," said Ms Mira Genser, vice-president of marketing communications at VoiceSignal, a Massachusetts-based developer of speech-recognition software.
"That is way too long, because studies show that with every click you lose 10 per cent of the people browsing." VoiceSignal’s core product, VSpeak, is available in 22 languages, including most European languages, Mandarin, Japanese and Korean. It is loaded directly onto phones made by Samsung, Panasonic, Motorola and Nokia. VSpeak does not require that the phone be trained, and it can be used to address messages, open applications and look up contact information.
Another product, VoiceMode, is available only in English and can be used to dictate messages and create email. VoiceMode, which has a vocabulary of more than 30,000 words and can accept new words into its dictionary much as a word processor does, will soon be made available in other languages, with the first addition probably being an Asian language, Genser said.
Most new cellphones have voice-recognition software already included; on some others the software can be downloaded. With the most advanced software, users can dictate a text or email message, find a calendar item on the phone or jump directly to a ring tone and buy it with a simple command like "Madonna ringtone." This last possibility is especially appealing for carriers, which have content on their mobile portals they are trying to sell clients, most of whom cannot be bothered to click through multiple menus to find what might interest them.
"The challenge has been getting people to realise they can do more with voice applications then just dial a phone number," said Collin Holmes, director of product marketing for V-Enable, a San Diego-based company that makes voice-recognition search software. V-Enable’s software, which in some cases is loaded on the handset directly by the manufacturer but can also be downloaded onto many new phones, is used in the United States by Verizon, US Cellular and other carriers. The company also has a recommendation engine that works like Amazon’s: If you buy a U2 ring tone, for instance, you will receive a list of other products bought by people who acquired that ringtone.
The most compelling market for voice-recognition software might be Asia, because typing ideograms on a cellular phone is more laborious than using a Western alphabet. Many companies, including NEC, are busily developing products.Cellphone users will dedicate 30 seconds to finding the content they are looking for, according to a study published in 2002 by Barry Smyth, the chief technical officer and co-founder of ChangingWorlds, an Irish maker of speech-recognition software to improve the usability of cellphone portals. Smyth said the research findings were still valid despite the passing of three years.
For every second of navigation time the average user saves when searching for content, he or she will spend another three seconds browsing, he found. Three seconds: big deal, right? Indeed it is for cellphone carriers trying to get clients to do more than just use their phones for calls and text messages. "In a lot of cases it takes 15 clicks for people to find the content they want," said Ms Mira Genser, vice-president of marketing communications at VoiceSignal, a Massachusetts-based developer of speech-recognition software.
"That is way too long, because studies show that with every click you lose 10 per cent of the people browsing." VoiceSignal’s core product, VSpeak, is available in 22 languages, including most European languages, Mandarin, Japanese and Korean. It is loaded directly onto phones made by Samsung, Panasonic, Motorola and Nokia. VSpeak does not require that the phone be trained, and it can be used to address messages, open applications and look up contact information.
Another product, VoiceMode, is available only in English and can be used to dictate messages and create email. VoiceMode, which has a vocabulary of more than 30,000 words and can accept new words into its dictionary much as a word processor does, will soon be made available in other languages, with the first addition probably being an Asian language, Genser said.