Anushree Hede

Projects

Below is a brief summary of my research and academic projects. They cover areas in natural language processing, machine learning, deep learning and distributed systems. For access to any of the code repositories, please reach out to me over email!

A Data-Driven Error Analysis of Toxicity Detection Models

Developed a data-driven method to analyze cause of errors in toxicity detection models related to marginalized groups and identities and developed a simple solution that tackles such errors.
[Read More] [Report]

PennSearch - Mini Search Engine with Distributed Components

Developed a mini-search engine for a wide range of domains with a distributed crawler, indexer and PageRanker. The crawler functionality was inspired from the Mercator-style crawler and the core functionality of the indexer and PageRanker was implemented using MapReduce (Hadoop).
[Read More] [Report]

PennCloud - Distributed Cloud platform with Mail (SMTP) and File Storage Services

Developed a small cloud platform with user accounts, email and file storage. Frontend and backend servers employ core principles of distributed systems such as consistency, replication, fault tolerance and recovery.
[Read More] [Report]

Explainability for Multiple-Choice Science Question-Answering

Designed a method for the explainability of Multiple-Choice QA, where a trained SOTA model is probed with different subsets of the surrounding context sentences, in order to determine which sentences are most important in deciding the correct answer. We found that the model performs well even when a smaller subset of context sentences are provided to it.
[Read More] [Report]

Visual Question-Answering: Using Object Information as Metadata

Designed a deep learning architecture for visual question-answering using GRU and a pre-trained ResNet. As an enhancement, we included a module that captures Faster-rCNN features of object positions in the image. This gave us a 5% increase in accuracy for the task.
[Read More] [Report]

Predicting Success of Kickstarter Projects

Performed a detailed feature selection for the Kickstarter projects, and applied a variety of classical machine learning techniques to predict the success of a Kickstarter projects.
[Read More] [Report]

Classification of SSH Compromises and Attacks

Extracted the flow features of network traffic using Python libraries to read packet capture files. Used the flow features and implemented two approaches in Java: (i) a machine learning method to detect SSH dictionary attacks and (ii) a method to identify a compromise by matching it to a possible action generally taken during a system compromise
Advised by: Dr. Gokul Kannan (BITS Pilani)
[Code A][Code B]

Software Aging Prediction Using Artificial Intelligence

Used a variety of class-balancing methods on the software aging datasets and applied preprocessing techniques to select the significant and uncorrelated features for further analysis. We then applied machine learning classification models on the data using Python (Scikit-learn) and compared their results using accuracy and Area Under Curve metrics
Advised by: Dr. N Bhanumurthy (BITS Pilani)
[Code]

Application for a Leap Motion device for ease of communication for stroke victims

Developed a "virtual keyboard" desktop application for the Leap Motion controller especially for stroke victims who have difficulty in typing on traditional keyboards. Used the Leap Motion SDK for Python
Advised by: Dr. Tathagata Ray (BITS Pilani)
[Code]

Experience

A compilation of my research, teaching and educational experiences.

[Resume]

Research and Teaching

Teaching Assistant

January 2021 - May 2021

"Deep Learning for Data Science" taught by Dr. Lyle Ungar and Dr. Konrad Kording at University of Pennsylvania (Philadelphia, USA)

· Course Material
· Prepared teaching materials and homework for Recurrent Neural Networks and NLP topics using PyTorch on Google Colab
· Mentored a pod of 9 students for 5 hours/week with deep learning concepts, homework and the final project

Student at Seminar-Type Course

January 2020 - May 2020

"Reasoning for Natural Language Understanding" taught by Dr. Dan Roth at University of Pennsylvania (Philadelphia, USA)

· Course Website
· Read and discussed research papers throughout the semester with the goal of understanding theories of reasoning in the context of natural language understanding
· Presented two papers to the class and facilitated discussion
· The first paper was What Can Neural Networks Reason About?, and my slides can be found here
· The second paper was Graph-Based Reasoning over Heterogeneous External Knowledge for Commonsense Question Answering, and my slides can be found here

Graduate Research Assistant (NLP, Computational Social Science)

October 2019 - December 2020

University of Pennsylvania (Philadelphia, USA)

· Demonstrated that the popular toxicity detection tool (Jigsaw’s Perspective API) is unable to relatively rank incivility (hostility/agitation/quarrelsomeness/rudeness) among three American news shows, in a manner similar to humans
· Deduced that erroneous Perspective scores are spuriously correlated with presence of non-offensive ‘error’ words
· Curated a dataset of video clips and transcript segments of American news shows with annotations of incivility
· Advised by: Dr. Ani Nenkova

Teaching Assistant

September 2019 - May 2020

"Senior Design Project" taught by Dr. Ani Nenkova at University of Pennsylvania (Philadelphia, USA)

· Course Website
· Facilitated and monitored project team discussions within class, and graded technical updates and progress of 140 students
· Managed scheduling of monthly consultations of 36 student teams with relevant faculty members

Research Intern

January 2019 - June 2019

Robert Bosch Engineering and Business Solutions Private Limited - Research and Technology Center (Bangalore, India)

· Developed a novel word-embedding-based trend detection algorithm for time-stamped automobile consumer complaints
· Demonstrated proof-of-concept through quantitative comparisons with a popular topic modelling algorithm (online-LDA)
· Advised by: Dr. Rajesh N. Rao and Rishabh Gupta

Teaching Assistant

August 2018 - December 2018

"Principles of Programming Languages" taught by Dr. Aruna Malapati at BITS Pilani (Hyderabad, India)

· Prepared tutorial material and assignment questions for six core topics (data types, regular expressions, BNF, parameter passing & scope, logic programming, functional programming)
· Invigilated and prepared solutions for two mid-semester quiz evaluation components for a class of 160 students

Summer Research Intern

May 2018 - July 2018

Information Retrieval and Extraction Lab, IIIT Hyderabad (Hyderabad, India)

· Studied works in the domain of text summarization and measurement of text coherence
· Built a web-based user-interface tool using HTML, CSS, JavaScript and JSP for text summarization
· Developed a neural model to measure coherence for multi-document summarization
· Advised by: Dr. Vasudeva Varma and Dr. Litton J. Kurisinkel

Summer Intern

May 2017 - July 2017

Sensei Technologies (Bangalore, India)

· Built an Alexa Skill (for Amazon Echo) as an in-house voice-based assistant tool for managing the on-going projects in the company
· Used node.js and AWS Lambda for development
· Learnt basics of Natural Language Processing
· Advised by: N S Nagaraja

Education

MS in Computer and Information Science

August 2019 - May 2021

University of Pennsylvania (Philadelphia, PA)

Relevant Courses: Machine Learning, Deep Learning, Computational Linguistics, Software Systems, Internet & Web Systems, Reasoning for Natural Langugage Understanding, Analysis of Algorithms, Big Data Analytics
Master's Thesis: A Data-Driven Error Analysis of Toxicity Detection Models

BE in Computer Science

August 2015 - July 2019

Birla Institute of Technology and Science (BITS), Pilani (Hyderabad Campus, India)

Relevant Courses: Machine Learning, Data Structures, Algorithms, Information Retrieval, Databases, Data Mining, Probability, Linear Algebra, Operating Systems

Achievements

EACL Publication

2021

· "From Toxicity in Online Comments to Incivility in American News: Proceed with Caution", Anushree Hede, Oshin Agarwal, Linda Lu, Diana C. Mutz and Ani Nenkova
· [Paper][Slides][Poster]

Pre-doctoral Summer School

2020

· Cornell, Maryland, Max Planck Pre-doctoral Research School
· Selected for the competitive program, aimed at fostering academic interactions between pre-doctoral students and faculty
· Attended lectures in databases and data analysis, distributed systems, security and privacy, Internet measurement and network architecture, large-scale machine learning, and theory of deep learning

80% Merit Tuition Waiver

2015-2019

· Awarded by BITS Pilani
· Received this merit-based scholarship consistently for all eight semesters of the undergraduate degree

96.6% in Class 12 CBSE Board Exam

2015

· Awarded by: Government of India
· Felicitated by Mrs. Smriti Irani, the Union Minister for Human Resource Development, Government of India, for the excellent performance in the 12th board exams.

NTSE Scholarship

2011

· Awarded by: Government of India
· National Talent Search Exam, a nation-wide scholarship aimed at identifying and recognizing school students with high intellect and academic talent
· Attended the Nurtuance Programme organized at NIT Surathkal with the goal of understanding prospects of higher studies and careers in STEM

Anushree Hede

About

Projects

A Data-Driven Error Analysis of Toxicity Detection Models

PennSearch - Mini Search Engine with Distributed Components

PennCloud - Distributed Cloud platform with Mail (SMTP) and File Storage Services

Explainability for Multiple-Choice Science Question-Answering

Visual Question-Answering: Using Object Information as Metadata

Predicting Success of Kickstarter Projects

Classification of SSH Compromises and Attacks

Software Aging Prediction Using Artificial Intelligence

Application for a Leap Motion device for ease of communication for stroke victims

[Master's Thesis] A Data-Driven Error Analysis of Toxicity Detection Models

PennSearch - Mini Search Engine with Distributed Components

PennCloud - Distributed Cloud platform with Mail (SMTP) and File Storage Services

Explainability for Multiple-Choice Science Question-Answering

Visual Question-Answering: Using Object Information as Metadata

Predicting Success of Kickstarter Projects

Experience

Research and Teaching

Teaching Assistant

"Deep Learning for Data Science" taught by Dr. Lyle Ungar and Dr. Konrad Kording at University of Pennsylvania (Philadelphia, USA)

Student at Seminar-Type Course

"Reasoning for Natural Language Understanding" taught by Dr. Dan Roth at University of Pennsylvania (Philadelphia, USA)

Graduate Research Assistant (NLP, Computational Social Science)

University of Pennsylvania (Philadelphia, USA)

Teaching Assistant

"Senior Design Project" taught by Dr. Ani Nenkova at University of Pennsylvania (Philadelphia, USA)

Research Intern

Robert Bosch Engineering and Business Solutions Private Limited - Research and Technology Center (Bangalore, India)

Teaching Assistant

"Principles of Programming Languages" taught by Dr. Aruna Malapati at BITS Pilani (Hyderabad, India)

Summer Research Intern

Information Retrieval and Extraction Lab, IIIT Hyderabad (Hyderabad, India)

Summer Intern

Sensei Technologies (Bangalore, India)

Education

MS in Computer and Information Science

University of Pennsylvania (Philadelphia, PA)

BE in Computer Science

Birla Institute of Technology and Science (BITS), Pilani (Hyderabad Campus, India)

Achievements

EACL Publication

Pre-doctoral Summer School

80% Merit Tuition Waiver

96.6% in Class 12 CBSE Board Exam

NTSE Scholarship

Where to find me

Email Me At