Artificial voice and speech recognition are gaining on human ears and vocal chords, especially with the debut of tools that may make Internet chats verbal, not textual.
Recorders that automatically transcribe your memos, headsets that make it easier to bypass the telephone company with Internet phone calls, and the promise of a widely voice-enabled Web were among the advances showcased at SpeechTEK 2000 here last week.
While digital speech is sounding better, it still has a way to go, industry experts acknowledge. “Speech technology is not perfect, not by a long shot,” says Dr. Janet Baker, chair of Lernout & Hauspie Speech Products NV’s Technology Advisory Board, and a cofounder of Dragon Systems, which Lernout & Hauspie acquired. “The fact is, speech recognizers are not as good as humans, and they’ll never be.”
Yet the slew of advances in the voice arena is impressive: Computerized voices through text-to-speech sound less scary; dictation packages are increasingly accurate; digital recorders are appearing; and voice portals are catching on. The industry is still awaiting the first implementation of VoiceXML, an emerging standard for voice-enabling the Internet. Someday, you’ll be able to pick up the phone and get Web-based information anytime.
Sony announced two mobile packages. The company is bundling two digital voice recorders with Dragon NaturallySpeaking 5.0 by Lernout & Hauspie, which is a first for Sony. Both pocket size recorders weigh about two ounces and let you record your spoken thoughts on the fly; later, you can download your ramblings as editable text on screen. Both devices have small LCDs so you can see what you’re doing and locate recording files. If you have chubby fingers, watch out – the buttons (for Record, Stop, Rewind, and so on) on these things are tiny.
The $200 ICD-R100VTP recorder holds roughly 150 minutes of dictation and lets you organize your recorded files into two file folders. You hook the device to your PC using a PC Link cable, locate the recording you want, convert it into a .wav file, and fire up Dragon NaturallySpeaking. It transcribes the file before your eyes.
The more sophisticated $300 ICD-MS1VTP bundle uses Sony’s Memory Stick media. The device comes with a 16MB Memory Stick, which gives you roughly 130 minutes of recording time. You can bump that up to just under 8 hours with a 64MB card. The ICD-MS1 has more than 300 file folders, and you don’t have to go through as many steps to transcribe the recordings. As soon as you pop the Memory Stick into your PC (using a dedicated drive or a PC Card), the Memory Stick Voice Editor automatically converts the recording you want into a .wav file and starts transcribing it. Both products will be on the shelves by the end of November.
Norcom Electronics announced its pricey Norcom 2700 recorder. The $399 handheld unit uses Sandisk media and comes with an 8MB card. It’s longer and heavier than Sony’s recorders, weighing about five ounces. The 2700 is voice-activated – as soon as you start yakking, it will turn on; when it stops hearing your voice, the silence forces it to switch off. You get up to two and a half hours of dictation. The LCD lets you see how you organized your recordings using the file manager, and the device connects to the computer through the parallel port. The 2700 will work with any version speech-recognition package.
Telex Communications announced two consumer headsets, the H-831 Deluxe USB Headworn Microphone, and the H-341 Deluxe Computer Headset. Available in January 2001, the $50 H-831 has a built-in Universal Serial Bus adapter, or pod, the size of a matchbox. USB headsets can be great for laptop users because the headset will bypass the laptop’s sound system (which is sometimes inferior or subject to interference) and use its own bundled, separate sound card. The H-831 uses a soft and flexible headband that spans your entire head and feels very lightweight. One downside: The H-831 is good for input only; that is, it won’t work for two-way audio, such as making phone calls over the Internet.
The $20 H-341 is Telex’s latest analog microphone headset and uses the same soft headband. It is full duplex, so you can use it for dictation and Internet telephony. The H-341 is available now.
GN Netcom showed its Bluetooth prototype, the GN 9000 Bluetooth headset, scheduled to ship in the first quarter of 2001. The wireless headset attaches to a base unit about the size of a small book. It’s designed to connect with any Bluetooth-enabled device, such as personal digital assistants, cell phones, PCs, and even cars within a 30-foot radius. Early adopters won’t have it easy – the GN 9000 will set you back $499.
Also, representatives of Emkay Innovative Products talked about its preliminary plans for a Bluetooth wireless headset, which it will market to other vendors that will repackage, price, and sell it to consumers. Instead of having a regular headset with a headband, this Bluetooth headset will take the form of a single earpiece – a small gadget with a loop that will sit over your ear. It will connect with a base station. Emkay expects to ship it early next year.