Lillian Lee

Jack Hessel, Ana Marasović, Jena D. Hwang et al. · allen-ai, uw

Large neural networks can now generate jokes, but do they really "understand" humor? We challenge AI models with three tasks derived from the New Yorker Cartoon Caption Contest: matching a joke to a cartoon, identifying a winning caption, and explaining why a winning caption is funny. These tasks encapsulate progressively more sophisticated aspects of "understanding" a cartoon; key elements are the complex, often surprising relationships between images and captions and the frequent inclusion of indirect and playful allusions to human experience and culture. We investigate both multimodal and language-only models: the former are challenged with the cartoon images directly, while the latter are given multifaceted descriptions of the visual scene to simulate human-level visual understanding. We find that both types of models struggle at all three tasks. For example, our best multimodal models fall 30 accuracy points behind human performance on the matching task, and, even when provided ground-truth visual scene descriptors, human-authored explanations are preferred head-to-head over the best machine-authored ones (few-shot GPT-4) in more than 2/3 of cases. We release models, code, leaderboard, and corpus, which includes newly-gathered annotations describing the image's locations/entities, what's unusual in the scene, and an explanation of the joke.

Tianze Shi, Carlos Gómez-Rodríguez, Lillian Lee

We generalize Cohen, Gómez-Rodríguez, and Satta's (2011) parser to a family of non-projective transition-based dependency parsers allowing polynomial-time exact inference. This includes novel parsers with better coverage than Cohen et al. (2011), and even a variant that reduces time complexity to $O(n^6)$, improving over the known bounds in exact inference for non-projective transition-based parsing. We hope that this piece of theoretical work inspires design of novel transition systems with better coverage and better run-time guarantees. Code available at https://github.com/tzshi/nonproj-dp-variants-naacl2018

Vivian Nguyen, Lillian Lee, Cristian Danescu-Niculescu-Mizil

During a conversation, there can come certain moments where its outcome hangs in the balance. In these pivotal moments, how one responds can put the conversation on substantially different trajectories leading to significantly different outcomes. Systems that can detect when such moments arise could assist conversationalists in domains with highly consequential outcomes, such as mental health crisis counseling. In this work, we introduce an unsupervised computational method for detecting such pivotal moments as they happen, in an online fashion. Our approach relies on the intuition that a moment is pivotal if our expectation of the outcome varies widely depending on what might be said next. By applying our method to crisis counseling conversations, we first validate it by showing that it aligns with human perception -- counselors take significantly longer to respond during moments detected by our method -- and with the eventual conversational trajectory -- which is more likely to change course at these times. We then use our framework to explore the relation of the counselor's response during pivotal moments with the eventual outcome of the session.

Tianze Shi, Lillian Lee

We present our contribution to the IWPT 2021 shared task on parsing into enhanced Universal Dependencies. Our main system component is a hybrid tree-graph parser that integrates (a) predictions of spanning trees for the enhanced graphs with (b) additional graph edges not present in the spanning trees. We also adopt a finetuning strategy where we first train a language-generic parser on the concatenation of data from all available languages, and then, in a second step, finetune on each individual language separately. Additionally, we develop our own complete set of pre-processing modules relevant to the shared task, including tokenization, sentence segmentation, and multiword token expansion, based on pre-trained XLM-R models and our own pre-training of character-level language models. Our submission reaches a macro-average ELAS of 89.24 on the test set. It ranks top among all teams, with a margin of more than 2 absolute ELAS over the next best-performing submission, and best score on 16 out of 17 languages.

Tianze Shi, Lillian Lee

We propose a transition-based bubble parser to perform coordination structure identification and dependency-based syntactic analysis simultaneously. Bubble representations were proposed in the formal linguistics literature decades ago; they enhance dependency trees by encoding coordination boundaries and internal relationships within coordination structures explicitly. In this paper, we introduce a transition system and neural models for parsing these bubble-enhanced structures. Experimental results on the English Penn Treebank and the English GENIA corpus show that our parsers beat previous state-of-the-art approaches on the task of coordination structure prediction, especially for the subset of sentences with complex coordination structures.

Tianze Shi, Ozan İrsoy, Igor Malioutov et al.

Naturally-occurring bracketings, such as answer fragments to natural language questions and hyperlinks on webpages, can reflect human syntactic intuition regarding phrasal boundaries. Their availability and approximate correspondence to syntax make them appealing as distant information sources to incorporate into unsupervised constituency parsing. But they are noisy and incomplete; to address this challenge, we develop a partial-brackets-aware structured ramp loss in learning. Experiments demonstrate that our distantly-supervised models trained on naturally-occurring bracketing data are more accurate in inducing syntactic structures than competing unsupervised systems. On the English WSJ corpus, our models achieve an unlabeled F1 score of 68.9 for constituency parsing.

Tianze Shi, Chen Zhao, Jordan Boyd-Graber et al.

Large-scale semantic parsing datasets annotated with logical forms have enabled major advances in supervised approaches. But can richer supervision help even more? To explore the utility of fine-grained, lexical-level supervision, we introduce Squall, a dataset that enriches 11,276 WikiTableQuestions English-language questions with manually created SQL equivalents plus alignments between SQL and question fragments. Our annotation enables new training possibilities for encoder-decoder models, including approaches from machine translation previously precluded by the absence of alignments. We propose and test two methods: (1) supervised attention; (2) adopting an auxiliary objective of disambiguating references in the input queries to table columns. In 5-fold cross validation, these strategies improve over strong baselines by 4.4% execution accuracy. Oracle experiments suggest that annotated alignments can support further accuracy gains of up to 23.9%.

Jack Hessel, Lillian Lee

Modeling expressive cross-modal interactions seems crucial in multimodal tasks, such as visual question answering. However, sometimes high-performing black-box algorithms turn out to be mostly exploiting unimodal signals in the data. We propose a new diagnostic tool, empirical multimodally-additive function projection (EMAP), for isolating whether or not cross-modal interactions improve performance for a given model on a given task. This function projection modifies model predictions so that cross-modal interactions are eliminated, isolating the additive, unimodal structure. For seven image+text classification tasks (on each of which we set new state-of-the-art benchmarks), we find that, in many cases, removing cross-modal interactions results in little to no performance degradation. Surprisingly, this holds even when expressive models, with capacity to consider interactions, otherwise outperform less expressive models; thus, performance improvements, even when present, often cannot be attributed to consideration of cross-modal feature interactions. We hence recommend that researchers in multimodal machine learning report the performance not only of unimodal baselines, but also the EMAP of their best-performing model.

Tianze Shi, Lillian Lee

An interesting and frequent type of multi-word expression (MWE) is the headless MWE, for which there are no true internal syntactic dominance relations; examples include many named entities ("Wells Fargo") and dates ("July 5, 2020") as well as certain productive constructions ("blow for blow", "day after day"). Despite their special status and prevalence, current dependency-annotation schemes require treating such flat structures as if they had internal syntactic heads, and most current parsers handle them in the same fashion as headed constructions. Meanwhile, outside the context of parsing, taggers are typically used for identifying MWEs, but taggers might benefit from structural information. We empirically compare these two common strategies--parsing and tagging--for predicting flat MWEs. Additionally, we propose an efficient joint decoding algorithm that combines scores from both strategies. Experimental results on the MWE-Aware English Dependency Corpus and on six non-English dependency treebanks with frequent flat structures show that: (1) tagging is more accurate than parsing for identifying flat-structure MWEs, (2) our joint decoder reconciles the two different views and, for non-BERT features, leads to higher accuracies, and (3) most of the gains result from feature sharing between the parsers and taggers.

Kumar Bhargav Srinivasan, Cristian Danescu-Niculescu-Mizil, Lillian Lee et al.

Moderators of online communities often employ comment deletion as a tool. We ask here whether, beyond the positive effects of shielding a community from undesirable content, does comment removal actually cause the behavior of the comment's author to improve? We examine this question in a particularly well-moderated community, the ChangeMyView subreddit. The standard analytic approach of interrupted time-series analysis unfortunately cannot answer this question of causality because it fails to distinguish the effect of having made a non-compliant comment from the effect of being subjected to moderator removal of that comment. We therefore leverage a "delayed feedback" approach based on the observation that some users may remain active between the time when they posted the non-compliant comment and the time when that comment is deleted. Applying this approach to such users, we reveal the causal role of comment deletion in reducing immediate noncompliance rates, although we do not find evidence of it having a causal role in inducing other behavior improvements. Our work thus empirically demonstrates both the promise and some potential limits of content removal as a positive moderation strategy, and points to future directions for identifying causal effects from observational data.

Jack Hessel, Lillian Lee, David Mimno

Images and text co-occur constantly on the web, but explicit links between images and sentences (or other intra-document textual units) are often not present. We present algorithms that discover image-sentence relationships without relying on explicit multimodal annotation in training. We experiment on seven datasets of varying difficulty, ranging from documents consisting of groups of images captioned post hoc by crowdworkers to naturally-occurring user-generated multimodal documents. We find that a structured training objective based on identifying whether collections of images and sentences co-occur in documents can suffice to predict links between specific sentences and specific images within the same document at test time.

Jack Hessel, Lillian Lee

Controversial posts are those that split the preferences of a community, receiving both significant positive and significant negative feedback. Our inclusion of the word "community" here is deliberate: what is controversial to some audiences may not be so to others. Using data from several different communities on reddit.com, we predict the ultimate controversiality of posts, leveraging features drawn from both the textual content and the tree structure of the early comments that initiate the discussion. We find that even when only a handful of comments are available, e.g., the first 5 comments made within 15 minutes of the original post, discussion features often add predictive capacity to strong content-and-rate only baselines. Additional experiments on domain transfer suggest that conversation-structure features often generalize to other communities better than conversation-content features do.

Carlos Gómez-Rodríguez, Tianze Shi, Lillian Lee

Shi, Huang, and Lee (2017) obtained state-of-the-art results for English and Chinese dependency parsing by combining dynamic-programming implementations of transition-based dependency parsers with a minimal set of bidirectional LSTM features. However, their results were limited to projective parsing. In this paper, we extend their approach to support non-projectivity by providing the first practical implementation of the MH_4 algorithm, an $O(n^4)$ mildly nonprojective dynamic-programming parser with very high coverage on non-projective treebanks. To make MH_4 compatible with minimal transition-based feature sets, we introduce a transition-based interpretation of it in which parser items are mapped to sequences of transitions. We thus obtain the first implementation of global decoding for non-projective transition-based parsing, and demonstrate empirically that it is more effective than its projective counterpart in parsing a number of highly non-projective languages

Jack Hessel, David Mimno, Lillian Lee

Multimodal machine learning algorithms aim to learn visual-textual correspondences. Previous work suggests that concepts with concrete visual manifestations may be easier to learn than concepts with abstract ones. We give an algorithm for automatically computing the visual concreteness of words and topics within multimodal datasets. We apply the approach in four settings, ranging from image captions to images/text scraped from historical books. In addition to enabling explorations of concepts in multimodal datasets, our concreteness scores predict the capacity of machine learning algorithms to learn textual/visual relationships. We find that 1) concrete concepts are indeed easier to learn; 2) the large number of algorithms we consider have similar failure cases; 3) the precise positive relationship between concreteness and performance varies between datasets. We conclude with recommendations for using concreteness scores to facilitate future multimodal research.

Tianze Shi, Liang Huang, Lillian Lee

We first present a minimal feature set for transition-based dependency parsing, continuing a recent trend started by Kiperwasser and Goldberg (2016a) and Cross and Huang (2016a) of using bi-directional LSTM features. We plug our minimal feature set into the dynamic-programming framework of Huang and Sagae (2010) and Kuhlmann et al. (2011) to produce the first implementation of worst-case O(n^3) exact decoders for arc-hybrid and arc-eager transition systems. With our minimal features, we also present O(n^3) global training methods. Finally, using ensembles including our new parsers, we achieve the best unlabeled attachment score reported (to our knowledge) on the Chinese Treebank and the "second-best-in-class" result on the English Penn Treebank.

Jack Hessel, Lillian Lee, David Mimno

The content of today's social media is becoming more and more rich, increasingly mixing text, images, videos, and audio. It is an intriguing research question to model the interplay between these different modes in attracting user attention and engagement. But in order to pursue this study of multimodal content, we must also account for context: timing effects, community preferences, and social factors (e.g., which authors are already popular) also affect the amount of feedback and reaction that social-media posts receive. In this work, we separate out the influence of these non-content factors in several ways. First, we focus on ranking pairs of submissions posted to the same community in quick succession, e.g., within 30 seconds, this framing encourages models to focus on time-agnostic and community-specific content features. Within that setting, we determine the relative performance of author vs. content features. We find that victory usually belongs to "cats and captions," as visual and textual features together tend to outperform identity-based features. Moreover, our experiments show that when considered in isolation, simple unigram text features and deep neural network visual features yield the highest accuracy individually, and that the combination of the two modalities generally leads to the best accuracies overall.

Liye Fu, Lillian Lee, Cristian Danescu-Niculescu-Mizil

Group discussions are a way for individuals to exchange ideas and arguments in order to reach better decisions than they could on their own. One of the premises of productive discussions is that better solutions will prevail, and that the idea selection process is mediated by the (relative) competence of the individuals involved. However, since people may not know their actual competence on a new task, their behavior is influenced by their self-estimated competence --- that is, their confidence --- which can be misaligned with their actual competence. Our goal in this work is to understand the effects of confidence-competence misalignment on the dynamics and outcomes of discussions. To this end, we design a large-scale natural setting, in the form of an online team-based geography game, that allows us to disentangle confidence from competence and thus separate their effects. We find that in task-oriented discussions, the more-confident individuals have a larger impact on the group's decisions even when these individuals are at the same level of competence as their teammates. Furthermore, this unjustified role of confidence in the decision-making process often leads teams to under-perform. We explore this phenomenon by investigating the effects of confidence on conversational dynamics.

Chenhao Tan, Lillian Lee

In meetings where important decisions get made, what items receive more attention may influence the outcome. We examine how different types of rhetorical (de-)emphasis -- including hedges, superlatives, and contrastive conjunctions -- correlate with what gets revisited later, controlling for item frequency and speaker. Our data consists of transcripts of recurring meetings of the Federal Reserve's Open Market Committee (FOMC), where important aspects of U.S. monetary policy are decided on. Surprisingly, we find that words appearing in the context of hedging, which is usually considered a way to express uncertainty, are more likely to be repeated in subsequent meetings, while strong emphasis indicated by superlatives has a slightly negative effect on word recurrence in subsequent meetings. We also observe interesting patterns in how these effects vary depending on social factors such as status and gender of the speaker. For instance, the positive effects of hedging are more pronounced for female speakers than for male speakers.

Liye Fu, Cristian Danescu-Niculescu-Mizil, Lillian Lee

Gender bias is an increasingly important issue in sports journalism. In this work, we propose a language-model-based approach to quantify differences in questions posed to female vs. male athletes, and apply it to tennis post-match interviews. We find that journalists ask male players questions that are generally more focused on the game when compared with the questions they ask their female counterparts. We also provide a fine-grained analysis of the extent to which the salience of this bias depends on various factors, such as question type, game outcome or player rank.

Chenhao Tan, Vlad Niculae, Cristian Danescu-Niculescu-Mizil et al.

Changing someone's opinion is arguably one of the most important challenges of social interaction. The underlying process proves difficult to study: it is hard to know how someone's opinions are formed and whether and how someone's views shift. Fortunately, ChangeMyView, an active community on Reddit, provides a platform where users present their own opinions and reasoning, invite others to contest them, and acknowledge when the ensuing discussions change their original views. In this work, we study these interactions to understand the mechanisms behind persuasion. We find that persuasive arguments are characterized by interesting patterns of interaction dynamics, such as participant entry-order and degree of back-and-forth exchange. Furthermore, by comparing similar counterarguments to the same opinion, we show that language factors play an essential role. In particular, the interplay between the language of the opinion holder and that of the counterargument provides highly predictive cues of persuasiveness. Finally, since even in this favorable setting people may not be persuaded, we investigate the problem of determining whether someone's opinion is susceptible to being changed at all. For this more difficult task, we show that stylistic choices in how the opinion is expressed carry predictive power.

Chenhao Tan, Lillian Lee

Although analyzing user behavior within individual communities is an active and rich research domain, people usually interact with multiple communities both on- and off-line. How do users act in such multi-community environments? Although there are a host of intriguing aspects to this question, it has received much less attention in the research community in comparison to the intra-community case. In this paper, we examine three aspects of multi-community engagement: the sequence of communities that users post to, the language that users employ in those communities, and the feedback that users receive, using longitudinal posting behavior on Reddit as our main data source, and DBLP for auxiliary experiments. We also demonstrate the effectiveness of features drawn from these aspects in predicting users' future level of activity. One might expect that a user's trajectory mimics the "settling-down" process in real life: an initial exploration of sub-communities before settling down into a few niches. However, we find that the users in our data continually post in new communities; moreover, as time goes on, they post increasingly evenly among a more diverse set of smaller communities. Interestingly, it seems that users that eventually leave the community are "destined" to do so from the very beginning, in the sense of showing significantly different "wandering" patterns very early on in their trajectories; this finding has potentially important design implications for community maintainers. Our multi-community perspective also allows us to investigate the "situation vs. personality" debate from language usage across different communities.

Chenhao Tan, Lillian Lee

The strength with which a statement is made can have a significant impact on the audience. For example, international relations can be strained by how the media in one country describes an event in another; and papers can be rejected because they overstate or understate their findings. It is thus important to understand the effects of statement strength. A first step is to be able to distinguish between strong and weak statements. However, even this problem is understudied, partly due to a lack of data. Since strength is inherently relative, revisions of texts that make claims are a natural source of data on strength differences. In this paper, we introduce a corpus of sentence-level revisions from academic writing. We also describe insights gained from our annotation efforts for this task.

Chenhao Tan, Lillian Lee, Bo Pang

Consider a person trying to spread an important message on a social network. He/she can spend hours trying to craft the message. Does it actually matter? While there has been extensive prior work looking into predicting popularity of social-media content, the effect of wording per se has rarely been studied since it is often confounded with the popularity of the author and the topic. To control for these confounding factors, we take advantage of the surprising fact that there are many pairs of tweets containing the same url and written by the same user but employing different wording. Given such pairs, we ask: which version attracts more retweets? This turns out to be a more difficult task than predicting popular topics. Still, humans can answer this question better than chance (but far from perfectly), and the computational methods we develop can do better than both an average human and a strong competing method trained on non-controlled data.

Eunsol Choi, Chenhao Tan, Lillian Lee et al.

Understanding the ways in which participants in public discussions frame their arguments is important in understanding how public opinion is formed. In this paper, we adopt the position that it is time for more computationally-oriented research on problems involving framing. In the interests of furthering that goal, we propose the following specific, interesting and, we believe, relatively accessible question: In the controversy regarding the use of genetically-modified organisms (GMOs) in agriculture, do pro- and anti-GMO articles differ in whether they choose to adopt a "scientific" tone? Prior work on the rhetoric and sociology of science suggests that hedging may distinguish popular-science text from text written by professional scientists for their colleagues. We propose a detailed approach to studying whether hedge detection can be used to understanding scientific framing in the GMO debates, and provide corpora to facilitate this study. Some of our preliminary analyses suggest that hedges occur less frequently in scientific discourse than in popular text, a finding that contradicts prior assertions in the literature. We hope that our initial work and data will encourage others to pursue this promising line of inquiry.

Cristian Danescu-Niculescu-Mizil, Justin Cheng, Jon Kleinberg et al.

Understanding the ways in which information achieves widespread public awareness is a research question of significant interest. We consider whether, and how, the way in which the information is phrased --- the choice of words and sentence structure --- can affect this process. To this end, we develop an analysis framework and build a corpus of movie quotes, annotated with memorability information, in which we are able to control for both the speaker and the setting of the quotes. We find that there are significant differences between memorable and non-memorable quotes in several key dimensions, even after controlling for situational and contextual factors. One is lexical distinctiveness: in aggregate, memorable quotes use less common word choices, but at the same time are built upon a scaffolding of common syntactic patterns. Another is that memorable quotes tend to be more general in ways that make them easy to apply in new contexts --- that is, more portable. We also show how the concept of "memorable language" can be extended across domains.