A5059946729- Topic Modeling
- Natural Language Processing Techniques
- Speech and dialogue systems
- Text Readability and Simplification
- Advanced Text Analysis Techniques
- Semantic Web and Ontologies
- Computational and Text Analysis Methods
- Mental Health via Writing
- Sentiment Analysis and Opinion Mining
- Handwritten Text Recognition Techniques
- Digital Mental Health Interventions
- Algorithms and Data Compression
- Biomedical Text Mining and Ontologies
- Social Media and Politics
- Multimodal Machine Learning Applications
- Text and Document Classification Technologies
- Speech Recognition and Synthesis
- Software Engineering Research
- Complex Network Analysis Techniques
- Hate Speech and Cyberbullying Detection
- Authorship Attribution and Profiling
- Neuroscience and Music Perception
- Syntax, Semantics, Linguistic Variation
- Misinformation and Its Impacts
- Data Mining Algorithms and Applications
University of Maryland, College Park
2016-2025
Research Institute for Advanced Computer Science
2001-2023
Johns Hopkins University
2022
Amazon (Germany)
2021
Key Laboratory of Nuclear Radiation and Nuclear Energy Technology
2021
Adobe Systems (United States)
2020-2021
Paderborn University
2021
Google (Switzerland)
2019
Institute of Linguistics
2014-2016
University of Arkansas at Little Rock
2014
This article presents a measure of semantic similarity in an IS-A taxonomy based on the notion shared information content. Experimental evaluation against benchmark set human judgments demonstrates that performs better than traditional edge-counting approach. The algorithms take advantage taxonomic resolving syntactic and ambiguity, along with experimental results demonstrating their effectiveness.
This paper presents a new measure of semantic similarity in an IS-A taxonomy, based on the notion information content. Experimental evaluation suggests that performs encouragingly well (a correlation r = 0.79 with benchmark set human judgments, upper bound 0.90 for subjects performing same task), and significantly better than traditional edge counting approach (r 0.66).
Parallel corpora have become an essential resource for work in multilingual natural language processing. In this article, we report on our using the STRAND system mining parallel text World Wide Web, first reviewing original algorithm and results then presenting a set of significant enhancements. These enhancements include use supervised learning based structural features documents to improve classification performance, new content-based measure translational equivalence, adaptation take...
Broad coverage, high quality parsers are available for only a handful of languages. A prerequisite developing broad coverage more languages is the annotation text with desired linguistic representations (also known as "treebanking"). However, syntactic labor intensive and time-consuming process, it difficult to find linguistically annotated in sufficient quantities. In this article, we explore using parallel help solving problem creating The central idea annotate English side corpus, project...
An individual's words often reveal their political ideology.Existing automated techniques to identify ideology from text focus on bags of or wordlists, ignoring syntax.Taking inspiration recent work in sentiment analysis that successfully models the compositional aspect language, we apply a recursive neural network (RNN) framework task identifying position evinced by sentence.To show importance modeling subsentential elements, crowdsource annotations at phrase and sentence level.Our model...
Dallas Card, Amber E. Boydstun, Justin H. Gross, Philip Resnik, Noah A. Smith. Proceedings of the 53rd Annual Meeting Association for Computational Linguistics and 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers). 2015.
Philip Resnik, William Armstrong, Leonardo Claudino, Thang Nguyen, Viet-An Jordan Boyd-Graber. Proceedings of the 2nd Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Reality. 2015.
Han-Chin Shing, Suraj Nair, Ayah Zirikly, Meir Friedenberg, Hal Daumé III, Philip Resnik. Proceedings of the Fifth Workshop on Computational Linguistics and Clinical Psychology: From Keyboard to Clinic. 2018.
Even though human experience unfolds continuously in time, it is not strictly linear; instead, entails cascading processes building hierarchical cognitive structures. For instance, during speech perception, humans transform a varying acoustic signal into phonemes, words, and meaning, these levels all have distinct but interdependent temporal Time-lagged regression using response functions (TRFs) has recently emerged as promising tool for disentangling electrophysiological brain responses...
STRAND (Resnik, 1998) is a language-independent system for automatic discovery of text in parallel translation on the World Wide Web. This paper extends preliminary results by adding language identification, scaling up orders magnitude, and formally evaluating performance. The most recent end-product an automatically acquired corpus comprising 2491 English-French document pairs, approximately 1.5 million words per language.
Minimum-error-rate training (MERT) is a bottleneck for current development in statistical machine translation because it limited the number of weights can reliably optimize. Building on work Watanabe et al., we explore use MIRA algorithm Crammer al. as an alternative to MERT. We first show that by parallel processing and exploiting more parse forest, obtain results using match or surpass MERT terms both quality computational cost. then test method two classes features address deficiencies...
Resnik and Yarowsky (1997) made a set of observations about the state-of-the-art in automatic word sense disambiguation and, motivated by those observations, offered several specific proposals regarding improved evaluation criteria, common training testing resources, definition inventories. Subsequent discussion resulted SENSEVAL , first exercise for (Kilgarriff Palmer 2000). This article is revised extended version our 1997 workshop paper, reviewing its discussing them light exercise. It...
We present an unsupervised method for word sense disambiguation that exploits translation correspondences in parallel corpora. The technique takes advantage of the fact cross-language lexicalizations same concept tend to be consistent, preserving some core element its semantics, and yet also variable, reflecting differing translator preferences influence context. Working with corpora introduces extra complication evaluation, since it is difficult find a corpus both tagged another language;...
article Share on Challenges in information retrieval and language modeling: report of a workshop held at the center for intelligent retrieval, University Massachusetts Amherst, September 2002 Authors: James Allan View Profile , Jay Aslam Nicholas Belkin Chris Buckley Jamie Callan Bruce Croft Sue Dumais Norbert Fuhr Donna Harman David J. Harper Djoerd Hiemstra Thomas Hofmann Eduard Hovy Wessel Kraaij John Lafferty Victor Lavrenko Lewis Liz Liddy R. Manmatha Andrew McCallum Ponte Prager...
In this paper, we describe a new corpus-based approach to prepositional phrase attachment disambiguation, and present results comparing performance of algorithm with other approaches problem.
Work on sentiment analysis often focuses the words and phrases that people use in overtly opinionated text. In this paper, we introduce a new approach to problem not lexical indicators, but syntactic "packaging" of ideas, which is well suited investigating identification implicit sentiment, or perspective. We establish strong predictive connection between linguistically motivated features then show how computational approximations these can be used improve existing state-of-the-art...
Untranslated words still constitute a major problem for Statistical Machine Translation (SMT), and current SMT systems are limited by the quantity of parallel training texts. Augmenting data with paraphrases generated pivoting through other languages alleviates this problem, especially so-called "low density" languages. But requires additional We address deriving monolingually, using distributional semantic similarity measures, thus providing access to larger resources, such as comparable...