Orhan Karaali

A high quality text-to-speech system composed of multiple neural networks

While neural networks have been employed to handle several different text-to-speech tasks, ours is the first system to use neural networks throughout, for both linguistic and acoustic processing. We divide the text-to-speech task into three subtasks, a linguistic module mapping from text to a linguistic representation, an acoustic module mapping from the linguistic representation to speech, and a video module mapping from the linguistic representation to animated images. The linguistic module employs a letter-to-sound neural network and postlexical neural network. The acoustic module employs a duration neural network and a phonetic neural network. The visual neural network is employed in parallel to the acoustic module to drive a talking head. The use of neural networks that can be retrained on the characteristics of different voices and languages affords our system a degree of adaptability and naturalness heretofore unavailable.

Modelling volatility derivatives using neural networks

Several papers over the past few years have addressed the pricing and potential usefulness of derivatives based on volatility. The Chicago Board Options Exchange (OEX) introduced an index on stock market volatility beginning in 1986. This volatility index is based on the implied volatility of eight different OEX options and is a measure that attempts to provide a reliable estimate of the markets consensus forecast of short-term volatility. It also provides a standard upon which volatility derivatives can be based. In this paper, we construct a similar measure for the markets short-term forecast of Deutschemark volatility. We evaluate the indexs time series properties, explore a closed-form solution for valuing derivative instruments based on the index, and employ neural network methods to price options based on the index. Lastly, we explore a practical example of using futures on the index to hedge the volatility risk of a portfolio of Deutschemark options. For this project, neural net technology has been applied in two different areas. One neural net has been used to forecast the volatility index. A second neural net has been used to obtain prices of options traded on this contract. It has been shown that commercial neural net-based systems can outperform the systems based on classical techniques in time series processing and digital signal processing applications. Also, neural nets have been applied to many financial applications, including mutual fund performance forecasting.