The best aspect of a research environment, in my opinion, is the abundance of bright people with whom you argue, discuss, and nurture your ideas. I thank all of the people at Penn and elsewhere who have given me the feedback that has helped me to separate the good ideas from the bad ideas. I hope that Ihave kept the good ideas in this thesis, and left the bad ideas out! Iwould like toacknowledge the following people for their contribution to my education: I thank my advisor Mitch Marcus, who gave me the intellectual freedom to pursue what I believed to be the best way to approach natural language processing, and also gave me direction when necessary. I also thank Mitch for many fascinating conversations, both personal and professional, over the last four years at Penn. I thank all of my thesis committee members: John La erty from Carnegie Mellon University, Aravind Joshi, Lyle Ungar, and Mark Liberman, for their extremely valuable suggestions and comments about my thesis research. I thank Mike Collins, Jason Eisner, and Dan Melamed, with whom I've had many stimulating and impromptu discussions in the LINC lab. Iowe them much gratitude for their valuable feedback onnumerous rough drafts of papers and thesis chapters.

We present a trainable model for identifying sentence boundaries in raw text. Given a corpus annotated with sentence boundaries, our model learns to classify each occurrence of., ?, and ! as either a valid or invalid sentence boundary. The training procedure requires no hand-crafted rules, lexica, part-of-speech tags, or domain-specific information. The model can therefore be trained easily on any genre of English, and should be trainable on any other Romanalphabet language. Performance is comparable to or better than the performance of similar systems, but we emphasize the simplicity of retraining for new domains.

Abstract not found

this article presents an efficient, trainable system for sentence boundary disambiguation. The system, called Satz, makes simple estimates of the parts of speech of the tokens immediately preceding and following each punctuation mark, and uses these estimates as input to a machine learning algorithm that then classifies the punctuation mark. Satz is very fast both in training and sentence analysis, and its combined robustness and accuracy surpass existing techniques. The system needs only a small lexicon and training corpus, and has been shown to transfer quickly and easily from English to other languages, as demonstrated on French and German.

prerequisite for many natural language processing tasks, including part-ofspeech tagging and sentence alignment. End-of-sentence punctuation marks are ambiguous; to disambiguate them most systems use brittle, special-purpose regular expression grammars and exception rules. As an alternative, we have developed an efficient, trainable algorithm that uses a lexicon with part-of-speech probabilities and a feed-forward neural network. This work demonstrates the feasibility of using prior probabilities of part-of-speech assignments, as opposed to words or definite part-ofspeech assignments, as contextual information. After training for less than one minute, the method correctly labels over 98.5% of sentence boundaries in a corpus of over 27,000 sentence-boundary marks. We show the method to be efficient and easily adaptable to different text genres, including single-case texts.

The automatic derivation of word pronunciations from input text is a central task for any text-to-speech system. For general English text at least, this is often thought to be a solved problem, with manually-derived linguistic rules assumed capable of handling `novel' words missing from the system dictionary. Data-driven methods, based on machine learning of the regularities implicit in a large pronouncing dictionary, have received considerable attention recently but are generally thought to perform less well. However, these tentative beliefs are at best uncertain without powerful methods for comparing text-to-phoneme subsystems. This paper contributes to the development of such methods by comparing the performance of four representative approaches to automatic phonemisation on the same test dictionary. As well as rule-based approaches, three data-driven techniques are evaluated: pronunciation by analogy (PbA), NETspeak and IB1-IG (a modified k-nearest neighbour method). Issues involved in comparative evaluation are detailed and elucidated. The data-driven techniques outperform rules in accuracy of letter-to-phoneme translation by a very significant margin but require aligned text-phoneme training data and are slower. Best translation results are obtained with PbA at approximately 72% words correct on a reasonably large pronouncing dictionary, compared to something like 26% words correct for the rules, indicating that automatic pronunciation of text is not a solved problem.

Text-to-Prosody systems based on the use of prosodic databases extracted from natural speech will be a key point for further development of new Text-to-Speech systems. This paper

This paper presents the prosodic component of a French text-to-speech synthesis system based on a metrical model of rhythm and intonation in which the prosodic well-formedness of utterances is governed by a set of rhythmic and morphosyntactic constraints. We first set out the theoretic basis of the generation of prosodic levels that correspond to the metrical and tonal structure of utterances. Then, we outline the implementation in our system, and, in particular, the prosodic module that produces a metrical interpretation of phrase-level parsed text, by computing relative prominence levels and generating the F0 patterns and segmental duration. This approach produces high quality results for text-tospeech synthesis at a very minimal implementation cost, and enables a realistic modelling of the prosodic variability observed in real speech. 1. INTRODUCTION We have undertaken a modular research program on the metrical, morpho-syntactic and semantico-pragmatic constraints which govern the...

Dictionary look-up is the primary strategy for deriving pronunciations for input words in a text-to-speech (TTS) system. This strategy is accurate for dictionary words, but it is not complete: it is impossible to list exhaustively all input words. The proper treatment of `unknown' words is currently an unsolved problem in TTS synthesis. There are many competing techniques for letter-to-sound conversion and the system developer must make a rational selection among them. However, it is unclear how different techniques should be properly compared. In this paper, we report a comparative assessment of the competitor methods of letter-to-sound rules, pronunciation by analogy, feedforward neural networks and a k-nearest neighbour method, with respect to their success at automatic phonemisation. This is achieved by using standardised scoring methods, test lexicon and phoneme inventories. The problem of standardising the phoneme set (`harmonisation') is deceptive: this is much harder than at fi...

This paper presents a stochastic model of intonation contours for use in text-to-speech synthesis. The model has two modules, a linguistic module that generates abstract prosodic labels from text, and a phonetic module that generates an F 0 curve from the abstract prosodic labels. This model differs from previous work in the abstract prosodic labels used, which can be automatically derived from the training corpus. This feature makes it possible to use large 1 This paper is based on a communication presented at Eurospeech'97 (Vronis et al. 1997) and has been recommended by the Editorial Board of Speech Communication. 2 corpora or several corpora of different speech styles, in addition to making it easy to adapt to new languages. The present paper focuses on the linguistic module, which does not require full syntactic analysis of the text but simply relies on part-of-speech tagging. The results were validated on French by means of a perception test. Listeners did not perceive a signif...