In certain contexts, maximum entropy (ME) modeling can be viewed as maximum likelihood training for exponential models, and like other maximum likelihood methods is prone to overfitting of training data. Several smoothing methods for maximum entropy models have been proposed to address this problem, but previous results do not make it clear how these smoothing methods compare with smoothing methods for other types of related models. In this work, we survey previous work in maximum entropy smoothing and compare the performance of several of these algorithms with conventional techniques for smoothing n-gram language models. Because of the mature body of research in n-gram model smoothing and the close connection between maximum entropy and conventional n-gram models, this domain is well-suited to gauge the performance of maximum entropy smoothing methods. Over a large number of data sets, we find that an ME smoothing method proposed to us by Lafferty [1] performs as well as or better tha...

Abstract—In certain contexts, maximum entropy (ME) modeling can be viewed as maximum likelihood (ML) training for exponential models, and like other ML methods is prone to overfitting of training data. Several smoothing methods for ME models have been proposed to address this problem, but previous results do not make it clear how these smoothing methods compare with smoothing methods for other types of related models. In this work, we survey previous work in ME smoothing and compare the performance of several of these algorithms with conventional techniques for smoothing-gram language models. Because of the mature body of research in-gram model smoothing and the close connection between ME and conventional-gram models, this domain is well-suited to gauge the performance of ME smoothing methods. Over a large number of data sets, we find that fuzzy ME smoothing performs as well as or better than all other algorithms under consideration. We contrast this method with previous-gram smoothing methods to explain its superior performance. Index Terms—Exponential models, language modeling, maximum entropy, minimum divergence,-gram models, smoothing.

Sources of training data suitable for language modeling of conversational speech are limited. In this paper, we show how training data can be supplemented with text from the web filtered to match the style and/or topic of the target recognition task, but also that it is possible to get bigger performance gains from the data by using class-dependent interpolation of N-grams.

Speech recognition performance is severely affected when the lexical, syntactic, or semantic characteristics of the discourse in the training and recognition tasks differ. The aim of language model adaptation is to exploit specific, albeit limited, knowledge about the recognition task to compensate for this mismatch. More generally, an adaptive language model seeks to maintain an adequate representation of the current task domain under changing conditions involving potential variations in vocabulary, syntax, content, and style. This paper presents an overview of the major approaches proposed to address this issue, and offers some perspectives regarding their comparative merits and associated tradeoffs. Ó 2003 Elsevier B.V. All rights reserved. 1.

We propose a novel method for using the World Wide Web to acquire trigram estimates for statistical language modeling. We submit an N-gram as a phrase query to web search engines. The search engines return the number of web pages containing the phrase, from which the N-gram count is estimated. The N-gram counts are then used to form web-based trigram probability estimates. We discuss the properties of such estimates, and methods to interpolate them with traditional corpus based trigram estimates. We show that the interpolated models improve speech recognition word error rate significantly over a small test set. 1.

Conditional Maximum Entropy models have been successfully

Copyright by Roger Levy 2005 ii

In this paper we report on new developments in the automatic meeting transcription task. Unlike other types of speech (such as those found in Broadcast News and Switchboard), meetings are unique in their richer dynamics of human-to-human interaction. An intuitive "thumbnail" plot is proposed to visualize such turntaking behavior. We will also show how recognition of short turns can be improved by building a language model tailored specifically for short turns. Out-Of-Vocabulary (OOV) words become a more salient problem in the meeting transcription task, as they are mostly topic words and proper names, lack of which not only causes Word Error Rate (WER) increase, but also limits further use of recognition hypotheses. We describe a prototype system which uses the Web as a source for vocabulary expansion, and present preliminary OOV retrieval results.

Time is an important dimension of relevance for a large number of searches, such as over blogs and news archives. So far, research on searching over such collections has largely focused on locating topically similar documents for a query. Unfortunately, topic similarity alone is not always sufficient for document ranking. In this paper, we observe that, for an important class of queries that we call time-sensitive queries, the publication time of the documents in a news archive is important and should be considered in conjunction with the topic similarity to derive the final document ranking. Earlier work has focused on improving retrieval for “recency ” queries that target recent documents. We propose a more general framework for handling time-sensitive queries and we automatically identify the important time intervals that are likely to be of interest for a query. Then, we build scoring techniques that seamlessly integrate the temporal aspect into the overall ranking mechanism. We extensively evaluated our techniques using a variety of news article data sets, including TREC data as well as real web data analyzed using the Amazon Mechanical Turk. We examined several alternatives for detecting the important time intervals for a query over a news archive and for incorporating this information in the retrieval process. Our techniques are robust and significantly improve result quality for time-sensitive queries compared to state-of-the-art retrieval techniques.

This paper addresses a critical problem in deploying a spoken dialog system (SDS). One of the main bottlenecks of SDS deployment for a new domain is data sparseness in building a statistical language model. Our goal is to devise a method to efficiently build a reliable language model for a new SDS. We consider the worst yet quite common scenario where only a small amount (∼1.7K utterances) of domain specific data is available for the target domain. We present a new method that exploits external static text resources that are collected for other speech recognition tasks as well as dynamic text resources acquired from World Wide Web (WWW). We show that language models built using external resources can jointly be used with limited in–domain (baseline) language model to obtain significant improvements in speech recognition accuracy. Combining language models built using external resources with the in–domain language model provides over 20 % reduction in WER over the baseline in–domain language model. Equivalently, we achieve almost the same level of performance by having ten times as much in–domain data (17K utterances). 1.