aIBM Almaden Research Center,
650 Harry Road, San Jose, CA 95120, USA
bDepartment of Computer Science,
University of North Carolina, Chapel Hill, NC 27599, USA
Ticker has recently emerged as one of the dominant user interfaces in Internet information systems for displaying dynamic content. It is widely used by Webcasting or Internet push products such as PointCast  to highlight frequently updated information. It is also deployed by many popular Websites such as ESPN Sports Zone and ABC News  for similar purposes.
Typically, a ticker scrolls text items continuously either from left to right or from bottom to top displaying information to the user. This kind of on-screen visual ticker has its drawbacks. First, the user must pay close attention to the ticker's display, otherwise he/she would not be able to get the information content. Second, the ticker display is often visually distracting to users concentrating on other tasks using the same screen such as editing a program. Third, a ticker requires display real estate, making it inappropriate for devices with no or little display capability such as a mobile phone or a PDA. This paper presents a new technology called audio ticker that can eliminate the above problems while offering the advantages of a ticker and much more.
Generally speaking, an audio ticker presents a sequence of tokens from a content channel to the users using the audio medium. A channel can be a Webcasting channel, a frequently updated Web page, and any other source of information. A token is the unit of channel content. For example, a token in a stock channel can be the name and price of a stock, whereas a token of a weather channel typically includes the name of a place, its high and low temperatures, and a general weather description such as sunny or cloudy.
When applied to a channel, an audio ticker reads the content of each token to the user sequentially, accompanied by optional background music. In addition, each channel can have its own audio prelude and postlude played at the beginning and end respectively. The user interface of audio ticker provides similar functions as a normal audio player, including controls on volume and speed, mute, pause, forward and backward (playing tokens in reverse order). Its other features include playing once, playing repeatedly, and playing updated (playing only those tokens that have been updated). It also supports audio triggers, which cause a specific sound to be played when the content of a token satisfies a specified condition. For example, playing cheerful music when the value of a particular stock goes over a pre-set value, or saying "donāt forget to bring your umbrella" when the local weather forecast includes the word rain or raini .
The functions of an audio ticker can be realized in many different ways. This section describes how audio ticker is implemented in the Grand Central Station (GCS) [4,5] project at IBM Almaden Research Center. It is used by the GCS Webcasting client to present information from a variety of Web channels. Tokens in GCS Web channels are represented in text. The key enabling mechanisms for audio ticker is the process of mapping textual tokens into pre-recorded audio clips for words and phrases, and seamlessly assembling them with other audio components such as background music, prelude, postlude and audio triggers into a cohesive stream, as shown in Fig. 1. The rationale for using pre-recorded audio instead of a text-to-speech synthesizer is because most of the Web channels employ a small and fixed vocabulary with each token following a rigid pattern. For example, a sports channel only has a few dozen team names and a small number of different scores. A stock channel uses a few thousand stock names but only a couple hundred of distinct stock values. The main benefit of using pre-recorded voice is that it produces more natural sound effect than synthesized voice.
Fig. 1. Mapping text tokens into audio clips.
The text-to-audio mapping in GCS is performed at the client side, which can be done alternatively at the server side. Our design requires less network bandwidth because text tokens are much smaller in size than their audio counterpart. This is a very important advantage when audio ticker is used in a mobile environment via a wireless connection. Various techniques can be employed to map words or phrases unavailable in the audio database, such as using a channel-specific thesaurus, consulting a larger audio database on the servers. Note that even if a token cannot be completely mapped into audio presentation, producing an audio excerpt is often adequate because a tickerās main purpose is highlighting information.
The biggest advantage of audio ticker is not requiring a display screen. Users can listen to the information while working on other tasks such as editing a program or filling a form. It is also applicable for a wide variety of information as long as the subject involves a limited vocabulary, such as traffic report, weather forecast, cafeteria menu, recipes, and much more. It can also be extended to integrate with a speech synthesizer and external audio data.
One of the most important applications of audio ticker is enabling mobile users to access Internet information using digital cellular phones or other audio enabled portable computing devices. Users on a shopping trip or sunbathing on a beach can call a designated number to listen to their stock quotes and other information pertaining to their personal interests.
The audio ticker currently running on a GCS client can present information in four different channels, stock, sports, traffic and weather forecast. We are working on extending its functionality and scope to offer more audio channels in GCS.