BEST PAPER AWARD

CiteSee provides a personalized paper reading experience by visually augmenting inline citations based on their connections to the current user based on their reading history, paper library, and publication records. During literature review sessions, CiteSee allows users to prioritize their exploration to focus first on inline citations most relevant to their recent readings. It also highlights inline citations to familiar papers, allowing users to keep track of their exploration and better contexturalize the current paper. Lab and field studies showed CiteSee can improve better paper discovery via inline citations, and allowed participants to have better situational awareness when conducting real-world literature reviews.

Scientific breakthroughs often rely upon scholars synthesizing many published work into broad overviews and rich insights to identify gaps in the current literature. However, survey articles require significant time and effort to synthesize and can quickly become outdated. Researchers in fast-paced disciplines also rely on the related work section to better understand the background when reading a paper. While related work sections also summarizes multiple prior work, they typically provide partial views of the larger research domain. Relatedly explore how a system in which users can quickly explore and read related work sections extracted across many papers can help scholars gain richer and more comprehensive overviews of fast-paced domains.

Significant research has been devoted to creating systems that help scholars discover relevant papers. While recent approaches have also shown benefits in highlighting relevant authors to improve paper discovery, these systems do not capture and utilize users’ evolving knowledge of authors. We introduce ComLittee, a literature discovery system that supports author-centric exploration. ComLittee supports author discovery to curation of a comittee of authors relevant to a research topic of interests based on existing paper recommendation systems. In our evaluation, we demonstrate how ComLittee leads to a higher efficiency, quality, and novelty in author discovery that also improves paper discovery.

Reviewing the literature to understand relevant past work is a critical part of research. However, as the scientific literature grows the challenges for users to find and make sense of the many different threads of research grow as well. In this work we explore a tool integrated into users’ reading process that helps them with leveraging authors’ existing summarization of prior research threads such as in related work sections. We developed Threddy that supports efficient extraction and organization of threads along with supporting evidence as scientists read research articles. The system then recommends further relevant articles based on user-created threads. Our lab study showed that Threddy helps scientists to follow and curate research threads without breaking out of their flow of reading, collect relevant papers and clips, and discover interesting new articles to further grow threads.

While people can fluidly collect, make sense of and organize information across many online sources in their minds to make informed decisions, the amount of online information can be overwhelming and exceed people’s working memory. We introduce Fuse, a browser extension that combined low-cost collection with lightweight organization of web content in a compact card-based sidebar. Fuse helps users simultaneously extract key web content and structure it in a lightweight and visual way. A 22-month public deployment and interviews provide longitudinal insights into the collecting, externalization and structuring behaviors of real-world users conducting information foraging tasks.

Users conducting online sensemaking tasks face the challenge of capturing the information they find for later use without interrupting their flow. Specifically, as people collect information, they also need to triaging support such as how urgent a topic is to follow up on, or rating a piece of evidence as a “pro” or “con,” which helps scaffold subsequent deeper exploration. However, current approaches incur a high cost, often requiring users to select, copy, context switch, paste, and annotate information. In this work, we explore a new interaction technique called “wiggling,” which can be used to fluidly collect, organize, and rate information during early sensemaking stages with a single gesture. Through implementation and user evaluation, we found that wiggling helped participants accurately collect information and encode their mental context with a 58% reduction in operational cost while being 24% faster compared to a common baseline.

The nature of people’s online activities has gone through dramatic changes in past decades, yet browser interfaces have stayed largely the same since tabs were introduced nearly 20 years ago. The divide between browser interfaces that only provide affordances for managing individual tabs and users who manage their attention at the task level can lead to an “tab overload” problem. We explored a task-centric tab manager called which enables users to save their open tabs to manage them with task structures and affordances that better reflect their mental models. To lower the cost of importing, Tabs.do uses a neural network in the browser to make suggestions for grouping users’ open tabs by tasks.

BEST PAPER NOMINATION

Tabs have become integral to Web browsing yet have changed little since their introduction nearly 20 years ago. In contrast, the internet has gone through dramatic changes and increasingly used to support complex sensemaking tasks. This paper investigates how tabs today are overloaded with a diverse set of functionalities and issues users face when managing them. We uncovered competing pressures pushing for keeping tabs open (ranging from interaction to emotional costs) versus pushing for closing them (such as limited attention and resources). We further developed rich design implications for future browser interfaces.

Consumers can choose from many different products and base their decisions on the tens of thousands of online evidence about each of their options. However, to synthesize this information into confident decisions can incur high interaction and cognitive costs. Online information is scattered across different sources, and evidence such as reviews can be subjective and conflicting, requiring users to interpret them under their personal context. We introduce Mesh, which scaffolds users in iteratively building up a better understanding of both their choices by evaluating evidence gathered across sources. Lab and field deployment studies found that Mesh significantly reduces the costs of gathering and evaluating evidence and scaffolds decision-making through personalized criteria enabling users to gain deeper insights from data to make confident purchase decisions.

Whether figuring out where to eat in an unfamiliar city or deciding which apartment to live in, reviews and forum posts are often a significant factor in online decision making. However, making sense of these rich repositories of diverse opinions can be prohibitively effortful, searchers need to sift through a large number of reviews to characterize each item based on aspects that they care about. We introduce a novel system, SearchLens, where searchers build up a collection of composable and reusable “Lenses” that reflect their different latent interests. Also, the Lenses allowed the system to generate personalized interfaces with visual explanations that promote transparency and enable in-depth exploration.

Analogies in distant domains often lead to scientific discoveries. However, it can be prohibitively difficult for researchers to find useful analogies from unfamiliar domains as search engines poorly support it. We introduce Solvent, a mixed-initiative system where annotators structure abstracts of academic papers into different aspects and use a semantic model to find analogies among research papers and across different domains. These results demonstrate a new path towards computationally supported knowledge sharing in research communities.

Complex searches can be overwhelming, leading to lots of opened tabs. This tab overload can make conducting searches on mobile devices especially difficult where screen real-estate is limited, and progress can often be interrupted. Rather than using tabs to manage information, we introduce browsing through scaffolding. Search result lists serve as mutable workspaces where progress can be suspended and resumed. BentoBrowser is available for download from the iPhone AppStore.

BEST PAPER NOMINATION
Crowd-powered chatbots are robust than current pure AI approach, but can be slower and more expensive at runtime. We attempted to combine the two approaches for high quality, low latency, and low cost. We introduce Evorus, a crowd-powered chatbot that automate itself over time by learning to integrate AI chatbots, reusing responses, and assess response quality. A 5-month-long public deployment study shows promising results. You can try talking to Evorus today.

Generating comprehensive labeling guidelines for crowdworkers can be challenging for complex datasets. Revolt harnesses crowd disagreements to identify ambiguous concepts in the data and coordinates the crowd to collaboratively create rich structures for requesters to make posthoc decisions, removing the need for comprehensive guidelines and enabling dynamic label boundaries.

Work done during internship at Microsoft Research, Redmond.

Highlighting can be mentally taxing for learners who are often unsure about how much information they needed to include. We introduce the idea of intentionally uncertain input in the context of highlighting on mobile devices. We present a system that uses force touch and fuzzy bounding boxes to support saving information while users are uncertain about where to highlight.

BEST PAPER NOMINATION
HCOMP 2016 INVITED ENCORE TALK
Many crowd clustering approaches have difficulties providing global context to workers in order to generate meaningful categories. Alloy uses a sample-and-search technique to provide a better understanding of the global context. It also combines the in-depth semantic knowledge from human computation and the scalability of machine learning models to create rich structures from unorganized documents with high quality and efficiency.

BEST PAPER NOMINATION
Answering complex questions such as “How do I grow better tomatoes?” often requires individuals to conduct extensive online research and synthesis. Can we crowdsource this complex, high context process with 100 crowdworkers conducting microtasks distributedly? The Knowledge Accelerator uses crowdworkers to extract and synthesize text clips across web pages into coherent articles without a centralized coordinator.

Code-switching behavior is a common phenomenon on social media to express solidarity or establish authority. While past work on automatic code-switching detection depends on dictionary look-up or named-entity recognition, our recurrent neural network model that relies on only raw features outperformed the top systems in the EMNLP’14 Code-Switching Workshop by 17% in error rate reduction.

Final project for the Deep Learning course at CMU.

TermMine is an information extraction system that can automatically mine translation pairs of terms from the web. We used a small set of terms and translations to gather mixed-code text from the web to train a CRF model that can identify translation pairs at run-time.

We introduced a method for classifying named-entities into broad semantic categories in WordNet. We extracted rich features from Wikipedia, allowing us to classify named-entities with high precision and coverage. The result is a large scale named-entity semantic database with 1.2 million entries and over 95% accuracy, covering 80% of all named-entities found on Wikipedia.