In this paper we explore a learning-based approach to the problem of predicting language impairment in children. We analyzed spontaneous narratives of children and extracted features measuring different aspects of language including morphology, speech fluency, language productivity and vocabulary. Then, we evaluated a learning-based approach and compared its predictive accuracy against a method based on language models. Empirical results on monolingual English-speaking children and bilingual Spanish-English speaking children show the learning-based approach is a promising direction for automatic language assessment.

Previous research has studied the representation and portrayal of Hispanics in mainstream media and media popular with minority groups. This study adds to previous findings by looking at both the English and Spanish versions of women's interest magazines. It investigates the frequency and portrayal of Hispanics in magazine advertisements, as well as the proportion of advertisers that advertise in both the English and Spanish versions and the level of adaptation they are using. Hispanics were found to be under-represented in advertisements in English language magazines. Results also revealed that few advertisers are advertising in both the English and Spanish versions of magazines. INTRODUCTION The U.S. Hispanic population has surged and is now more accepted and embraced; Hispanic culture Hispanic is in. And, Hispanic pride is growing and causing bilingual and English-dominant Hispanics to return to their roots (Faura 114-115). U.S. Hispanic market is large, consisting of 42.4 million people, and is growing. The Hispanic population makes up half of the total population increase since the 2000 Census and is predicted to arrive at one in six by 2010 (Connecting With Hispanic Magazine Readers). This high growth rate can be attributed to immigration and the fact that Hispanics tend to have larger households, thus a higher birth rate (Faura 5). In addition, their buying power is growing at a rate of 118%. Despite these figures, most advertisers' spending targeted at the Hispanic market is still below what is necessary to be effective. On average, corporations are only spending about 3.2% of their advertising budget on Hispanic advertising, when the optimal amount to be effective is about 8%. The Association of Hispanic Advertising Agencies has investigated which industries are leading in their investment in the Hispanic market and those that are lagging behind. The study revealed that food and beverage products, food services, general merchandise, telecommunications, personal care, and insurance industries are investing the most in the Hispanic market, while pharmaceuticals, the U.S. government, auto industries, travel and entertainment, software, computer makers, securities and financial services, and specialty retail are spending the least (Missed Opportunities). It is critical for long-term success that advertisers understand and appeal to this important market. This study is aimed at investigating the level of Hispanic representation and portrayals in magazine advertisements in the English and Spanish versions, in addition to examining the proportion of advertisers that advertise in both the English and Spanish versions of magazines and level of adaptation they are employing. Information on the U.S. Hispanic market and recommendations will be provided so that advertisers better understand and are therefore able to appeal more to

In this paper, we present a hidden Markov model (HMM) approach to segment meeting transcripts into topics. To learn the model, we use unsupervised learning to cluster the text segments obtained from topic boundary information. Using modified WinDiff and P k metrics, we demonstrate that an HMM outperforms LCSeg, a state-of-the-art lexical chain based method for topic segmentation using the ICSI meeting corpus. We evaluate the effect of language model order, the number of hidden states, and the use of stop words. Our experimental results show that a unigram LM is better than a trigram LM, using too many hidden states degrades topic segmentation performance, and that removing the stop words from the transcripts does not improve segmentation performance.

AorTech 2A A/F’ Society of Cardiology & The Victorian Heart Center 49A Asia Publishing Exchange (APEX) 9A Asian Cardiovnscirlnr & Tliorncic Annnls 1 OA CarboMedics 48A Cosmotec 49A Datascope 1 IA Elsevier Science 45A Geister 17A Hewlett-Packard 16A ICR Publishers 55A J. Mitra & Bros. Inside Front Cover Medtronic 12A-13A Mitroflow 7A Philips Medical Systems 4A-5A Picker International 50A Schneider 46A Shiley Heart Valve Research Center 51A Shimadzu 47A Silent Partners 17A St. Jude Medical Special Insert and Back Cover Terumo 56A Thermal International Inside Back Cover Vascutek 52A