Story segmentation plays a critical role in spoken document processing. Spoken documents often come in a continuous audio stream without explicit boundaries related to stories or topics. It is important to be able to automatically segment these audio streams into coherent units. This work is an initial attempt to make use of informative lexical terms (or key terms) in recognition transcripts of...

We will argue that the beginnings of words are perceptual “islands of reliability” in connected speech, and that their perceptual and temporal properties allow them to drive critical aspects of spoken word recognition including lexical segmentation. This argument rests on three generalizations derived from research in speech science, phonology, and psycholinguistics. We suggest that word onsets...

This paper presents a unified approach for Chinese lexical analysis using hierarchical hidden Markov model (HHMM), which aims to incorporate Chinese word segmentation, Part-Of-Speech tagging, disambiguation and unknown words recognition into a whole theoretical frame. A class-based HMM is applied in word segmentation, and in this level unknown words are treated in the same way as common words l...

In this paper we compare the performance of three distinct approaches to lexical cohesion based text segmentation. Most work in this area has focused on the discovery of textual units that discuss subtopic structure within documents. In contrast our segmentation task requires the discovery of topical units of text i.e. distinct news stories from broadcast news programmes. Our approach to news s...

Two experiments sought to extend the demonstration of English-learning infants' abilities to segment nouns from fluent speech to a new lexical class: verbs. Moreover, we explored whether two factors previously shown to influence noun segmentation, stress pattern (strong-weak or weak-strong) and type of initial phoneme (consonant or vowel), also influence verb segmentation. Our results establish...

This dissertation project examined the influence of language typology on the use of segmentation cues by second language (L2) learners of English. Previous research has shown that native English speakers rely more on sentence context and lexical knowledge than segmental (i.e. phonotactics or acoustic-phonetics) or prosodic cues (e.g., word stresss) in native language (L1) segmentation. However,...

This study investigated the role of phrase-level prosodic boundary information in word segmentation in Korean with two word-spotting experiments. In experiment 1, it was found that intonational cues alone helped listeners with lexical segmentation. Listeners paid more attention to local intonational cues (...H#L...) across the prosodic boundary than the intonational information within a prosodi...

We present a probabilistic model that uses both prosodic and lexical cues for the automatic segmentation of speech into topically coherent units. We propose two methods for combining lexical and prosodic information using hiddenMarkov models and decision trees. Lexical information is obtained from a speech recognizer, and prosodic features are extracted automatically from speech waveforms. We e...

In this article we present a new method for text segmentation. The method relies on the number of lexical chains (LCs) which end in a sentence, which begin in the following sentence and which traverse the two successive sentences. The lexical chains are based on Roget’s thesaurus (the 1987 and the 1911 version). We evaluate the method on ten texts from the DUC 2002 conference and on twenty text...