Automatic literature review generation
Time: 12:00pm - 01:00pm
Place: Wells Library, Room E174
Xiaozhong LiuIn this talk, I will present an innovative method of automatically generating (1) a textual review paper, given a scientific topic (i.e., keyword); and (2) a literature review section, based on a piece of working text (i.e., a publication abstract or introduction). We proposed a four-step solution for generating an automatic review paper or literature review section: Citation Recommendation (generating a list of citations) given a user information need, Citation Management (generating clusters of citations), Review Generation (summarizing citation clusters into collections of candidate sentences), and Readability Enhancement (optimizing sentence representation). The aim of this proposed project is to give scholars easier access to decent-quality literature reviews relevant to their information needs (working text or a keyword represented topic). Automatically generated textual research reviews could be an important supplementary information access or provide additional feature to a classical academic search system, i.e., Google Scholar.
In category: Computational linguistics
Do good recipes need butter? Predicting user ratings of online recipes
Time: 03:00pm - 04:00pm
Place: Memorial Hall 401
Can LiuIn this talk, we will present our work in automatic prediction of user ratings for recipes. We use information from Epicurious - a recipe website, to answer the following questions:* Can ingredients predict ratings?* Are instructions indicative?* How well can reviews predict ratings?* Do Metadata help?* And, among the above sets of information, which is the most reliable ?We will describe the data set, feature extraction, experimental set-up, results analysis and future work.
Telepresence robots and robot-mediated communication
Time: 02:00pm - 03:30pm
Place: Wells Library Room LI030
Susan HerringTelepresence robotics is a sophisticated form of robotic remote control in which a human operator has a sense of being on location – that is, of being telepresent. Telepresence robots are being used in a growing number of organizational and occupational contexts. In this talk, I describe research that I carried out recently on telepresence robots for use by academic professionals, including a user study and a review of commercially-available telepresence robots. Robot-mediated communication (RMC) is introduced as a type of computer-mediated communication (CMC), and a program of RMC research is outlined that focuses on language and social interaction.
In category: Sociolinguistics and pragmatics
Feature reassembly and forming syntactic ties: The acquisition of non-canonical agreement in Arabic L2
Time: 02:30pm - 04:00pm
Place: Ballantine Hall 205
Boshra El-Ghazoly (Dissertation defense)Agreement phenomena are central and crucial to the description, analysis and acquisition of language. Building on Lardiere's (2009) insights on feature reassembly, this study investigates whether English-Arabic learners can successfully acquire Gender and Number morphosyntactic features of Arabic canonical and non-canonical agreement (CA and NCA). This provides a relevant test between the claims of Hawkins & Chan's (1997) Failed Functional Features Hypothesis (FFFH) and Schwartz & Sprouse's (1996) Full Transfer/Full Access Hypothesis (FTFA), which propose different approaches to the acquirability of morphosyntactic features unattested in the first language (L1). Taking a selective deficiency view of L2ers' accessibility to UG, FFFH predicts English-Arabic learners' failure in acquiring CA and NCA as it involves the acquisition of a feature absent from English (Gender). Contrary to FFFH, FTFA (and other FA accounts) predicts the possibility of acquiring Gender morphosyntactic forms. Difficulty might arise in NCA constructions due to the inability to detect feature constellations that are complex in nature. In the same vein, Lardiere (2008, 2009) emphasizes that difficulty in formal feature acquisition does not necessarily relate to a lack of access to the universal feature inventory or to the inability to activate or select the relevant features but to the challenge of detecting associations between morphological forms and feature distributions and combinations. The thesis is a cross-sectional study where four groups of participants were recruited; a native Arabic speaker-group (N=15), three learner-groups at three levels of proficiency in Arabic L2 (Intermediate (N=15), Upper-intermediate (N=15), and Advanced (N=15). The study employs three tasks to obtain comprehension and production data: a computer delivered Grammaticality Judgment Task, a Picture Description task and a Word and Picture Game. The results are argued to be most consistent with FTFA and Lardiere's Feature Reassembly approach.
In category: Second language acquisition
JEvents v3.0.9 Stable
Copyright © 2006-2013
Powered by Joomla!®
Copyright © 2008 The Trustees of Indiana University | Copyright Complaints