Digital Commons Network^™

Hypothesis Only Baselines In Natural Language Inference, Adam Poliak, Jason Naradowsky, Aparajita Haldar, Rachel Rudinger, Benjamin Van Durme

Computer Science Faculty Research and Scholarship

We propose a hypothesis only baseline for diagnosing Natural Language Inference (NLI). Especially when an NLI dataset assumes inference is occurring based purely on the relationship between a context and a hypothesis, it follows that assessing entailment relations while ignoring the provided context is a degenerate solution. Yet, through experiments on 10 distinct NLI datasets, we find that this approach, which we refer to as a hypothesis-only model, is able to significantly outperform a majority-class baseline across a number of NLI datasets. Our analysis suggests that statistical irregularities may allow a model to perform NLI in some datasets beyond what …

Collecting Diverse Natural Language Inference Problems For Sentence Representation Evaluation, Adam Poliak, Aparajita Haldar, Rachel Rudinger, J Edward Hu, Ellie Pavlick, Aaron Steven White, Benjamin Van Durme

Computer Science Faculty Research and Scholarship

We present a large-scale collection of diverse natural language inference (NLI) datasets that help provide insight into how well a sentence representation captures distinct types of reasoning. The collection results from recasting 13 existing datasets from 7 semantic phenomena into a common NLI structure, resulting in over half a million labeled context-hypothesis pairs in total. We refer to our collection as the DNC: Diverse Natural Language Inference Collection. The DNC is available online at https://www.decomp.net, and will grow over time as additional resources are recast and added from novel sources.

Exploring The Functional And Geometric Bias Of Spatial Relations Using Neural Language Models, Simon Dobnik, Mehdi Ghanimifard, John D. Kelleher

Conference papers

The challenge for computational models of spatial descriptions for situated dialogue systems is the integration of information from different modalities. The semantics of spatial descriptions are grounded in at least two sources of information: (i) a geometric representation of space and (ii) the functional interaction of related objects that. We train several neural language models on descriptions of scenes from a dataset of image captions and examine whether the functional or geometric bias of spatial descriptions reported in the literature is reflected in the estimated perplexity of these models. The results of these experiments have implications for the creation of …

Construction Of A Turkish Proposition Bank, Koray Ak, Cansu Toprak, Volkan Esgel, Olcay Taner Yildiz

Turkish Journal of Electrical Engineering and Computer Sciences

This paper describes our approach to developing the Turkish PropBank by adopting the semantic role-labeling guidelines of the original PropBank and using the translation of the English Penn-TreeBank as a resource. We discuss the semantic annotation process of the PropBank and language-specific cases for Turkish, the tools we have developed for annotation, and quality control for multiuser annotation. In the current phase of the project, more than 9500 sentences are semantically analyzed and predicate-argument information is extracted for 1330 verbs and 1914 verb senses. Our plan is to annotate 17,000 sentences by the end of 2017.

Assistive Technology Guidelines For Higher Education Disability Support Staff, Brenda Delee

CCAC Theses and Dissertations

With the changing laws and effective integration of assistive technology into the classroom environment, students can have the provision of multiple means to complete their work with greater independence. In post secondary education, any student who discloses a sensory, cognitive, or physical disability is eligible to request and receive assistive technology and other services. When used correctly, assistive technology can help students with reading, writing, math, and communication skills. With a possible influx of students, disability support staff must be prepared and willing to meet the needs and address issues relating to students with disabilities. If their needs are not …

Clinical Information Extraction From Unstructured Free-Texts, Mingzhe Tao

Legacy Theses & Dissertations (2009 - 2024)

Information extraction (IE) is a fundamental component of natural language processing (NLP) that provides a deeper understanding of the texts. In the clinical domain, documents prepared by medical experts (e.g., discharge summaries, drug labels, medical history records) contain a significant amount of clinically-relevant information that is crucial to the overall well-being of patients. Unfortunately, in many cases, clinically-relevant information is presented in an unstructured format, predominantly consisting of free-texts, making it inaccessible to computerized methods. Automatic extraction of this information can improve accessibility. However, the presence of synonymous expressions, medical acronyms, misspellings, negated phrases, and ambiguous terminologies make automatic extraction …

Relation Extraction Via One-Shot Dependency Parsing On Intersentential, Higher-Order, And Nested Relations, Gözde Gül Şahi̇n, Erdem Emekli̇gi̇l, Seçi̇l Arslan, Onur Ağin, Gülşen Eryi̇ği̇t

Turkish Journal of Electrical Engineering and Computer Sciences

Despite the emergence of digitalization, people still interact with institutions via traditional means such as submitting free formatted petitions, orders, or applications. These noisy documents generally consist of complex relations that are nested, higher-order, and intersentential. Most of the current approaches address extraction of only sentence-level and binary relations from grammatically correct text and generally require high-level linguistic features coming from preprocessors such as a parts-of-speech tagger, chunker, or syntactic parser. In this article, we focus on extracting complex relations in order to automate the task of understanding user intentions. We propose a novel language-agnostic and noise-immune approach that does …

Effect Of Intuitionistic Fuzzy Normalization In Microarray Gene Selection, Prema Ramasamy, Premalatha Kandhasamy

Turkish Journal of Electrical Engineering and Computer Sciences

Analysis of gene expression data is essential in microarray gene expression in order to retrieve the required information. Gene expression data generally contain a large number of genes but a small number of samples. The complicated relations among the different genes make analysis more difficult, and removing irrelevant genes improves the quality of results. This paper presents two fuzzy preprocessing techniques, using a fuzzy set (FS) and intuitionistic fuzzy set (IFS), to normalize datasets. In the feature selection part, four statistical methods were used. Using three publicly available gene expression datasets, the fuzzy normalization techniques were compared with two standard …

Implementing Universal Dependency, Morphology, And Multiword Expression Annotation Standards For Turkish Language Processing, Umut Sulubacak, Gülşen Eryi̇ği̇t

Turkish Journal of Electrical Engineering and Computer Sciences

Released only a year ago as the outputs of a research project (``Parsing Web 2.0 Sentences'', supported in part by a TÜBİTAK 1001 grant (No. 112E276) and a part of the ICT COST Action PARSEME (IC1207)), IMST and IWT are currently the most comprehensive Turkish dependency treebanks in the literature. This article introduces the final states of our treebanks, as well as a newly integrated hierarchical categorization of the multiheaded dependencies and their organization in an exclusive deep dependency layer in the treebanks. It also presents the adaptation of recent studies on standardizing multiword expression and named entity annotation schemes …

Automated Citation Sentiment Analysis Using High Order N-Grams: A Preliminary Investigation, Muhammad Touseef Ikram, Muhammad Tanvir Afzal, Naveed Anwer Butt

Turkish Journal of Electrical Engineering and Computer Sciences

Scientific papers hold an association with previous research contributions (i.e. books, journals or conference papers, and web resources) in the form of citations. Citations are deemed as a link or relatedness of the previous work to the cited work. The nature of the cited material could be supportive (positive), contrastive (negative), or objective (neutral). Extraction of the author's sentiment towards the cited scientific articles is an emerging research discipline due to various linguistic differences between the citation sentences and other domains of sentiment analysis. In this paper, we propose a technique for the identification of the sentiment of the citing …

Large Vocabulary Recognition For Online Turkish Handwriting With Sublexical Units, Esma Fatima Bi̇lgi̇n Taşdemi̇r, Ayşe Berri̇n Yanikoğlu Yeşi̇lyurt

Turkish Journal of Electrical Engineering and Computer Sciences

We present a system for large vocabulary recognition of online Turkish handwriting, using hidden Markov models. While using a traditional approach for the recognizer, we have identified and developed solutions for the main problems specific to Turkish handwriting recognition. % First, since large amounts of Turkish handwriting samples are not available, the system is trained and optimized using the large UNIPEN dataset of English handwriting, before extending it to Turkish using a small Turkish dataset. % The delayed strokes, which pose a significant source of variation in writing order due to the large number of diacritical marks in Turkish, are …

Extended Correlated Principal Component Analysis With Svm-Puk In Opinion Mining, Kollimarla Anusha Devi, Deepak Chowdary Edara, Venkatrama Phani Kumar Sistla, Venkata Krishna Kishore Kolli

Turkish Journal of Electrical Engineering and Computer Sciences

With the rapid growth of microblogs and online sites, an inordinate number of product reviews are available on the Internet. They not only help in analyzing, but also assist in making informed decisions about product quality. In the proposed work, an extended correlated principal component analysis (ECPCA) is used for dimensionality reduction. A comparative analysis is conducted on movie reviews (DB-1) and Twitter datasets (DB-2 and DB-3) in opinion mining extraction. The performance of naive Bayes, CHIRP, and support vector machine (SVM) with kernel methods such as radial basis function (RBF), polynomial, and Pearson (PUK) are compared and analyzed on …

Enlarging Multiword Expression Dataset By Co-Training, Senem Kumova Meti̇n

Turkish Journal of Electrical Engineering and Computer Sciences

In multiword expressions (MWEs), multiple words unite to build a new unit in language. When MWE identification is accepted as a binary classification task, one of the most important factors in performance is to train the classifier with enough number of labelled samples. Since manual labelling is a time-consuming task, the performances of MWE recognition studies are limited with the size of the training sets. In this study, we propose the comparison-based and common-decision co-training approaches in order to enlarge the MWE dataset. In the experiments, the performances of the proposed approaches were compared to those of the standard co-training …

Design Of A Multiagent-Based Smart Microgrid System For Building Energy And Comfort Management, Batyrkhan Omarov, Aigerim Altayeva

Turkish Journal of Electrical Engineering and Computer Sciences

Modern intelligent control system design methodology in the electric power industry and their features are considered when providing the required comfort in multizone buildings with the use of multiagent power consumption and comfort management systems. Such a control system covers all the monitored zones of a building and, if necessary, can provide the greatest possible overall comfort in the building while reducing the required electric power. The purpose of this study was to develop a comfort management system in a multizoned building that can provide comfort while reducing the required electric power.

Blood Glucose Control Using An Abc Algorithm-Based Fuzzy-Pid Controller, Seli̇m Soylu, Kenan Danişman

Turkish Journal of Electrical Engineering and Computer Sciences

In this paper, a Mamdani-type fuzzy controller is proposed as the controller part of an artificial pancreas. The controller is optimized with the artificial bee colony optimization algorithm. The glucose{insulin regulatory system, based on a nonlinear differential model in the presence of delay, is used both for virtual patient and healthy person data. The main target of the controller is to mimic a blood glucose concentration profile of the healthy person with exogenous insulin infusion. Simulations are performed to assess the control function in terms of tracking the blood glucose concentration profile of the healthy person and minimizing errors. To …

Management Strategies For Adopting Agile Methods Of Software Development In Distributed Teams, Igor A. Schtein

Walden Dissertations and Doctoral Studies

Between 2003 and 2015, more than 61% of U.S. software development teams failed to satisfy project requirements, budgets, or timelines. Failed projects cost the software industry an estimated 60 billion dollars. Lost opportunities and misused resources are often the result of software development leaders failing to implement appropriate methods for managing software projects. The purpose of this qualitative multiple case study was to explore strategies software development managers use in adopting Agile methodology in the context of distributed teams. The tenets of Agile approach are individual interaction over tools, working software over documentation, and collaboration over a contract. The conceptual …

Exploratory Data Analysis And Crime Prediction In San Francisco, Isha Pradhan

Master's Projects

Crime has been prevalent in our society for a very long time and it continues to be so even today. The San Francisco Police Department has continued to register numerous such crime cases daily and has released this data to the public as a part of the open data initiative. In this paper, Big Data analysis is used on this dataset and a tool that predicts crime in San Francisco is provided. The focus of the project is to perform an in-depth analysis of the major types of crimes that occurred in the city, observe the trend over the years, …

"What's Ur Type?" Contextualized Classification Of User Types In Marijuana-Related Communications Using Compositional Multiview Embedding, Ugur Kursuncu, Manas Gaur, Usha Lokala, Anurag Illendula, Krishnaprasad Thirunarayan, Raminta Daniulaityte, Amit P. Sheth, Budak Arpinar

Kno.e.sis Publications

With 93% of pro-marijuana population in US favoring legalization of medical marijuana, high expectations of a greater return for Marijuana stocks, and public actively sharing information about medical, recreational and business aspects related to marijuana, it is no surprise that marijuana culture is thriving on Twitter. After the legalization of marijuana for recreational and medical purposes in 29 states, there has been a dramatic increase in the volume of drug-related communication on Twitter. Specifically, Twitter accounts have been established for promotional and informational purposes, some prominent among them being American Ganja, Medical Marijuana Exchange, and Cannabis Now. Identification and characterization …

Personalized Prediction Of Suicide Risk For Web-Based Intervention, Amanuel Alambo, Manas Gaur, Ugur Kursuncu, Krishnaprasad Thirunarayan, Jeremiah Schumm, Jyotishman Pathak, Amit P. Sheth

Kno.e.sis Publications

Across the United States, suicide is the second leading cause of death for people aged between 15 and 34, and younger people are more prone to mental health problems, suicidal thoughts, and behaviors. For instance, 80% of patients with Borderline Personality Disorder have suicide-related behaviors, and between 4-9% of them commit suicide. Moreover, the social stigma associated with mental health issues and suicide deter patients from sharing their experiences directly with others. In such a situation, social media that provides a free and open forum for voluntary expression can provide insights into suicide ideation and self-destructive behavior.

Reddit is a …

Slim Embedding Layers For Recurrent Neural Language Models, Zhongliang Li

Browse all Theses and Dissertations

Recurrent neural language (RNN) models are the state-of-the-art method for language modeling. When the vocabulary size is large, the space taken to store the model parameters becomes the bottleneck for the use of these type of models. We introduce a simple space compression method that stochastically shares the structured parameters at both the input and output embedding layers of RNN models to significantly reduce the size of model parameters, but still compactly represents the original input and the output embedding layers. The method is easy to implement and tune. Experiments on several data sets show that the new method achieves …

Generating Classification Rules From Training Samples, Arun D. Kulkarni

Computer Science Faculty Publications and Presentations

In this paper, we describe an algorithm to extract classification rules from training samples using fuzzy membership functions. The algorithm includes steps for generating classification rules, eliminating duplicate and conflicting rules, and ranking extracted rules. We have developed software to implement the algorithm using MATLAB scripts. As an illustration, we have used the algorithm to classify pixels in two multispectral images representing areas in New Orleans and Alaska. For each scene, we randomly selected 10 per cent of the samples from our training set data for generating an optimized rule set and used the remaining 90 per cent of samples …

Eight Observations And 24 Research Questions About Open Source Projects: Illuminating New Realities, Matt Germonprez, Georg J.P. Link, Kevin Lumbard, Sean Goggins

Information Systems and Quantitative Analysis Faculty Publications

The rapid acceleration of corporate engagement with open source projects is drawing out new ways for CSCW researchers to consider the dynamics of these projects. Research must now consider the complex ecosystems within which open source projects are situated, including issues of for-profit motivations, brokering foundations, and corporate collaboration. Localized project considerations cannot reveal broader workings of an open source ecosystem, yet much empirical work is constrained to a local context. In response, we present eight observations from our eight-year engaged field study about the changing nature of open source projects. We ground these observations through 24 research questions that …

Analysis Of Driver Behavior Modeling In Connected Vehicle Safety Systems Through High Fidelity Simulation, Ahura Jami

Electronic Theses and Dissertations

A critical aspect of connected vehicle safety analysis is understanding the impact of human behavior on the overall performance of the safety system. Given the variation in human driving behavior and the expectancy for high levels of performance, it is crucial for these systems to be flexible to various driving characteristics. However, design, testing, and evaluation of these active safety systems remain a challenging task, exacerbated by the lack of behavioral data and practical test platforms. Additionally, the need for the operation of these systems in critical and dangerous situations makes the burden of their evaluation very costly and time-consuming. …

A Recurrent Neural Network Architecture For Biomedical Event Trigger Classification, Jeevith Bopaiah

Theses and Dissertations--Computer Science

A “biomedical event” is a broad term used to describe the roles and interactions between entities (such as proteins, genes and cells) in a biological system. The task of biomedical event extraction aims at identifying and extracting these events from unstructured texts. An important component in the early stage of the task is biomedical trigger classification which involves identifying and classifying words/phrases that indicate an event. In this thesis, we present our work on biomedical trigger classification developed using the multi-level event extraction dataset. We restrict the scope of our classification to 19 biomedical event types grouped under four broad …

Predicting Perceived Age: Both Language Ability And Appearance Are Important, Sarah Plane, Ariel Marvasti, Tyler Egan, Casey Kennington

Computer Science Faculty Publications and Presentations

When interacting with robots in a situated spoken dialogue setting, human dialogue partners tend to assign anthropomorphic and social characteristics to those robots. In this paper, we explore the age and educational level that human dialogue partners assign to three different robotic systems, including an un-embodied spoken dialogue system. We found that how a robot speaks is as important to human perceptions as the way the robot looks. Using the data from our experiment, we derived prosodic, emotional, and linguistic features from the participants to train and evaluate a classifier that predicts perceived intelligence, age, and education level.

New Trends In Second Language Learning And Teaching Through The Lens Of Ict, Networked Learning, And Artificial Intelligence, Jaya Kannan, Pilar Munday

Languages Faculty Publications

In the last few decades, Information and Communications Technology (ICT) applications have been shaping the field of Computer Assisted Language Learning (CALL). Mobile Assisted Language Learning (MALL) paved the way for ubiquitous learning. The advent of new technologies in the early 21st century also added a social dimension to ICT that allowed for Networked Learning (NL). Given that language learning is fundamentally a socio-cultural experience, networked learning capabilities have provided the potential for language learning in community settings. This has revitalized the earlier frameworks provided by CALL. NL has empowered language learners today to connect globally, to access Open Educational …

Deep Neural Networks For Multi-Label Text Classification: Application To Coding Electronic Medical Records, Anthony Rios

Theses and Dissertations--Computer Science

Coding Electronic Medical Records (EMRs) with diagnosis and procedure codes is an essential task for billing, secondary data analyses, and monitoring health trends. Both speed and accuracy of coding are critical. While coding errors could lead to more patient-side financial burden and misinterpretation of a patient’s well-being, timely coding is also needed to avoid backlogs and additional costs for the healthcare facility. Therefore, it is necessary to develop automated diagnosis and procedure code recommendation methods that can be used by professional medical coders.

The main difficulty with developing automated EMR coding methods is the nature of the label space. The …

Assesing Completeness Of Solvency And Financial Condition Reports Through The Use Of Machine Learning And Text Classification, Ruairí Nugent

Dissertations

Text mining is a method for extracting useful information from unstructured data through the identification and exploration of large amounts of text. It is a valuable support tool for organisations. It enables a greater understanding and identification of relevant business insights from text. Critically it identifies connections between information within texts that would otherwise go unnoticed. Its application is prevalent in areas such as marketing and political science however, until recently it has been largely overlooked within economics. Central banks are beginning to investigate the benefits of machine learning, sentiment analysis and natural language processing in light of the large …

An Exploration Of Parliamentary Speeches In The Irish Parliament Using Topic Modeling, Fiona Leheny

Dissertations

The only resource available in the public domain which highlights parliamentary ac tivity is parliamentary questions. Up until the last ten years, manual content analysis was carried out to classify these. More recently, machine learning techniques have been used to automatically classify and analyse these data sets. This study analyses the verbal parliamentary speeches in the Irish Parliament (known as the D´ail) over a ten year period using unsupervised machine learning. It does so by applying a less utilised topic modeling technique, known as Non-negative Matrix Factorisation (NMF), to de tect the latent themes in these speeches. A two-layer dynamic …

Human-Intelligence And Machine-Intelligence Decision Governance Formal Ontology, Faisal Mahmud

Engineering Management & Systems Engineering Theses & Dissertations

Since the beginning of the human race, decision making and rational thinking played a pivotal role for mankind to either exist and succeed or fail and become extinct. Self-awareness, cognitive thinking, creativity, and emotional magnitude allowed us to advance civilization and to take further steps toward achieving previously unreachable goals. From the invention of wheels to rockets and telegraph to satellite, all technological ventures went through many upgrades and updates. Recently, increasing computer CPU power and memory capacity contributed to smarter and faster computing appliances that, in turn, have accelerated the integration into and use of artificial intelligence (AI) in …