This paper discusses the development of trainable statistical models for extracting content from television and radio news broadcasts. In particular we concentrate on statistical finite state models for identifying proper names and other named entities in broadcast speech. Two models are presented: the first models name class information as a word attribute; the second explicitly models both word-word and class-class transitions. A common n-gram based formulation is used for both models. The task of named entity identification is characterized by relatively sparse training data and issues related to smoothing are discussed. Experiments are reported using the DARPA/NIST Hub--4E evaluation for North American Broadcast News.

Speech is the primary and most convenient means of communication between individuals [1]. In the future network era, the digital content over the network will include all the information activities for human life, from real-time information to knowledge archives, from working environments to private services. Apparently, the most attractive form of the network content will be in multimedia, including speech information. Such speech information usually provides insight concerning the subjects, topics, and concepts of the multimedia content. As a result, the spoken documents associated with the network content will become key for retrieval and browsing. On the other hand, the rapid development of network and wireless technologies is making it possible for people to access the network content not only from the office/home, but from anywhere, at any time, via small handheld devices such as personal digital assistants (PDAs) or cell phones. Today, network access is primarily text based. The users enter the instructions by words or texts, and the network or search engine offers text materials from which the user can select. The users interact with the network or search engine and obtain the desired information via text-based media. In the future, it can be imagined that almost all such functions of text can also be performed with speech. The user’s instructions can be entered not only by text but possibly through speech as well since speech is a convenient user interface for a variety of user terminals, especially for small handheld devices. The network content may be indexed/retrieved and browsed not only by text but possibly also by the associated spoken documents as mentioned above. The users may also interact with the network or the search engine via either text-based media or spoken/multimodal dialogs. Text-to-speech synthesis can be used to transform the text information in the content into speech when required. This is the general environment of retrieval/browsing applications for multimedia content with associated spoken documents.

In this paper we investigate the utility of three aspects of named entity processing: detection, localization and value extraction. We corroborate this task categorization by providing examples of practical applications for each of these subtasks. We also suggest methods for tackling these subtasks, giving particular attention to working with speech data. We employ Support Vector Machines to solve the detection task and show how localization and value extraction can successfully be dealt with using a combination of grammar-based and statistical methods. 1.

Audio comprehension tests are designed to help evaluate a listbner's understanding of a spoken passage and are frequentl) a' key component of language competency exams. Just as reading comprehension exams are proving useful in evaluating text-based language processing technology, audio comprehension exams can be used to evaluate spoken language processing systems. In this paper we discuss some of the challenges of developing automated systems for taking audio comprehension exams.