Topic models are a popular tool for understanding text collections, but their evaluation has been a point of contention. Automated evaluation metrics such as coherence are often used, however, their validity has been questioned for neural topic models (NTMs) and can overlook a model’s benefits in real-world applications. To this end, we conduct the first evaluation of neural, supervised and classical topic models in an interactive task-based setting. We combine topic models with a classifier and test their ability to help humans conduct content analysis and document annotation. From simulated, real user and expert pilot studies, the Contextual Neural Topic Model does the best on cluster evaluation metrics and human evaluations; however, LDA is competitive with two other NTMs under our simulated experiment and user study results, contrary to what coherence scores suggest. We show that current automated metrics do not provide a complete picture of topic modeling capabilities, but the right choice of NTMs can be better than classical models on practical tasks.

Scientific papers and slides are two different representations of the same underlying information, but both require substantial work to prepare. While there had been prior efforts on automating document-to-slides generation, there is still a pressing need of customizing the presentation of content aligning with the persona of target audience or duration of presentation. This paper first introduces the concept of end-user specification-aware document to slides conversion that incorporates end-user specifications into the conversion process. For this, we initially introduce a new dataset reuse the existing SciDuet dataset consisting of pairs of papers and corresponding slides decks from recent years’ *ACL conferences to create four persona-aware configurations. Secondly, we present Persona-Aware-D2S, a novel approach by finetuning LLMs using target audience feedback to create persona-aware slides from scientific documents. Our evaluation on both automated metrics and qualitative human evaluation suggests that by incorporating end-user specifications into the conversion process, our model can create presentations that are not only informative but also tailored to expectations and cognitive abilities of target audience.

Simultaneous interpretation is a cognitively taxing task, and even seasoned professionals benefit from real-time assistance. However, both recruiting professional interpreters and evaluating new assistance techniques are difficult. We present a novel, realistic simultaneous interpretation task that mimics the cognitive load of interpretation with crowdworker surrogates. Our task tests different real-time assistance methods in a Wizard-of-Oz experiment with a large pool of proxy users and compares against professional interpreters. Both professional and proxy participants respond similarly to changes in interpreting conditions, including improvement with two assistance interventions—translation of specific terms and of numbers—compared to a no-assistance control.

Large Language Models (LLMs) are increasingly used for accessing information on the web. Their truthfulness and factuality are thus of great interest. To help users make the right decisions about the information they get, LLMs should not only provide information but also help users fact-check it. We conduct human experiments with 80 crowdworkers to compare language models with search engines (information retrieval systems) at facilitating fact-checking. We prompt LLMs to validate a given claim and provide corresponding explanations. Users reading LLM explanations are significantly more efficient than those using search engines while achieving similar accuracy. However, they over-rely on the LLMs when the explanation is wrong. To reduce over-reliance on LLMs, we ask LLMs to provide contrastive information—explain both why the claim is true and false, and then we present both sides of the explanation to users. This contrastive explanation mitigates users’ over-reliance on LLMs, but cannot significantly outperform search engines. Further, showing both search engine results and LLM explanations offers no complementary benefits compared to search engines alone. Taken together, our study highlights that natural language explanations by LLMs may not be a reliable replacement for reading the retrieved passages, especially in high-stakes settings where over-relying on wrong AI explanations could lead to critical consequences.

Questions posed by information-seeking users often contain implicit false or potentially harmful assumptions. In a high-risk domain such as maternal and infant health, a question-answering system must recognize these pragmatic constraints and go beyond simply answering user questions, examining them in context to respond helpfully. To achieve this, we study assumptions and implications, or pragmatic inferences, made when mothers ask questions about pregnancy and infant care by collecting a dataset of 2,727 inferences from 500 questions across three diverse sources. We study how health experts naturally address these inferences when writing answers, and illustrate that informing existing QA pipelines with pragmatic inferences produces responses that are more complete, mitigating the propagation of harmful beliefs.

We present a summary of the efforts to improve conference peer review that were implemented at ACL’23. This includes work with the goal of improving review quality, clearer workflow and decision support for the area chairs, as well as our efforts to improve paper-reviewer matching for various kinds of non- mainstream NLP work, and improve the overall incentives for all participants of the peer review process. We present analysis of the factors affecting peer review, identify the most problematic issues that the authors complained about, and provide suggestions for the future chairs. We hope that publishing such reports would (a) improve transparency in decision-making, (b) help the people new to the field to understand how the *ACL conferences work, (c) provide useful data for the future chairs and workshop organizers, and also academic work on peer review, and (d) provide useful context for the final program, as a source of information for meta-research on the structure and trajectory of the field of NLP.

Large Language Models (LLMs) are increasingly being deployed in interactive contexts that involve direct user engagement, such as chatbots and writing assistants. These deployments are increasingly plagued by prompt injection and jailbreaking (collectively, prompt hacking), in which models are manipulated to ignore their original instructions and instead follow potentially malicious ones. Although widely acknowledged as a significant security threat, there is a dearth of a large-scale resource and quantitative study on prompt hacking. To address this lacuna, we launch a global prompt hacking competition, which allows for free-form human input attacks. We elicit 600K+ adversarial prompts against three state-of-the-art LLMs. We describe the dataset, which empirically verifies that current LLMs can indeed be manipulated via prompt hacking. We also present a comprehensive ontology of the types of adversarial prompts.

Translations help people understand content written in another language. However, even correct literal translations do not fulfill that goal when people lack the necessary background to understand them. Professional translators incorporate explicitations to explain the missing context by considering cultural differences between source and target audiences. Despite its potential to help users, NLP research on explicitation is limited because of the dearth of adequate evaluation methods. This work introduces techniques for automatically generating explicitations, motivated by WikiExpl: a dataset that we collect from Wikipedia and annotate with human translators. The resulting explicitations are useful as they help answer questions more accurately in a multilingual question answering framework.

Polarization and the marketplace for impressions have conspired to make navigating information online difficult for users, and while there has been a significant effort to detect false or misleading text, multimodal datasets have received considerably less attention. To complement existing resources, we present multimodal Video Misleading Headline (VMH), a dataset that consists of videos and whether annotators believe the headline is representative of the video’s contents. After collecting and annotating this dataset, we analyze multimodal baselines for detecting misleading headlines. Our annotation process also focuses on why annotators view a video as misleading, allowing us to better understand the interplay of annotators’ background and the content of the videos.

While recent large language models (LLMs) improve on various question answering (QA) datasets, it remains difficult for a single model to generalize across question types that require distinct reasoning abilities. We provide empirical evidence that state-of-the-art LLMs suffer from poor generalizability on reasoning types beyond those seen in the prompt. To remedy this, we propose a Mixture-of-Reasoning-Experts (MORE) framework that ensembles diverse specialized language models. We specialize the backbone language model with prompts optimized for different reasoning categories, including factual, multihop, mathematical, and commonsense reasoning. Our key insight is to leverage agreement among the specialized experts to select the best answer for each question, or to abstain from answering. This gives MORE higher accuracy than any single specialized model on a collection of 12 QA datasets from four reasoning types. Beyond generalizability, the interpretable design of MORE improves selective question answering results compared to baselines without incorporating inter-expert agreement. This framework is also more interpretable and useful to human consumers of QA outputs. Our human study confirms that presenting expert predictions and the answer selection process helps annotators more accurately calibrate when to trust the system’s output. We release all code and data to facilitate future work.

This paper develops automatic song translation (AST) for tonal languages and addresses the unique challenge of aligning words’ tones with melody of a song in addition to conveying the original meaning. We propose three criteria for effective AST—preserving meaning, singability and intelligibility—and design metrics for these criteria. We develop a new benchmark for English–Mandarin song translation and develop an unsupervised AST system, Guided AliGnment for Automatic Song Translation (GagaST), which combines pre-training with three decoding constraints. Both automatic and human evaluations show GagaST successfully balances semantics and singability.

For users to trust model predictions, they need to understand model outputs, particularly their confidence — calibration aims to adjust (calibrate) models’ confidence to match expected accuracy. We argue that the traditional calibration evaluation does not promote effective calibrations: for example, it can encourage always assigning a mediocre confidence score to all predictions, which does not help users distinguish correct predictions from wrong ones. Building on those observations, we propose a new calibration metric, MacroCE, that better captures whether the model assigns low confidence to wrong predictions and high confidence to correct predictions. Focusing on the practical application of open-domain question answering, we examine conventional calibration methods applied on the widely-used retriever-reader pipeline, all of which do not bring significant gains under our new MacroCE metric. Toward better calibration, we propose a new calibration method (ConsCal) that uses not just final model predictions but whether multiple model checkpoints make consistent predictions. Altogether, we provide an alternative view of calibration along with a new metric, re-evaluation of existing calibration methods on our metric, and proposal of a more effective calibration method.

For humans and computers, the first step in answering an open-domain question is retrieving a set of relevant documents from a large corpus. However, the strategies that computers use fundamentally differ from those of humans. To better understand these differences, we design a gamified interface for data collection—Cheater’s Bowl—where a human answers complex questions with access to both traditional and modern search tools. We collect a dataset of human search sessions, analyze human search strategies, and compare them to state-of-the-art multi-hop QA models. Humans query logically, apply dynamic search chains, and use world knowledge to boost searching. We demonstrate how human queries can improve the accuracy of existing systems and propose improving the future design of QA models.

Detractors of neural machine translation admit that while its translations are fluent, it sometimes gets key facts wrong. This is particularly important in simultaneous interpretation where translations have to be provided as fast as possible: before a sentence is complete. Yet, evaluations of simultaneous machine translation (SimulMT) fail to capture if systems correctly translate the most salient elements of a question: people, places, and dates. To address this problem, we introduce a downstream word-by-word question answering evaluation task (SimQA): given a source language question, translate the question word by word into the target language, and answer as soon as possible. SimQA jointly measures whether the SimulMT models translate the question quickly and accurately, and can reveal shortcomings in existing neural systems—hallucinating or omitting facts.

Explanations promise to bridge the gap between humans and AI, yet it remains difficult to achieve consistent improvement in AI-augmented human decision making. The usefulness of AI explanations depends on many factors, and always showing the same type of explanation in all cases is suboptimal—so is relying on heuristics to adapt explanations for each scenario. We propose learning to explain”selectively”: for each decision that the user makes, we use a model to choose the best explanation from a set of candidates and update this model with feedback to optimize human performance. We experiment on a question answering task, Quizbowl, and show that selective explanations improve human performance for both experts and crowdworkers.

The NLP community are increasingly interested in providing explanations for NLP models to help people make sense of model behavior and potentially improve human interaction with models. In addition to computational challenges in generating these explanations, evaluations of the generated explanations require human-centered perspectives and approaches. This tutorial will provide an overview of human-centered evaluations of explanations. First, we will give a brief introduction to the psychological foundation of explanations as well as types of NLP model explanations and their corresponding presentation, to provide the necessary background. We will then present a taxonomy of human-centered evaluation of explanations and dive into depth in the two categories: 1) evaluation based on human-annotated explanations; 2) evaluation with human-subjects studies. We will conclude by discussing future directions. We will also adopt a flipped format to maximize the in- teractive components for the live audience.

Tamarian, a fictional language introduced in the Star Trek episode Darmok, communicates meaning through utterances of metaphorical references, such as “Darmok and Jalad at Tanagra” instead of “We should work together.” This work assembles a Tamarian-English dictionary of utterances from the original episode and several follow-on novels, and uses this to construct a parallel corpus of 456 English-Tamarian utterances. A machine translation system based on a large language model (T5) is trained using this parallel corpus, and is shown to produce an accuracy of 76% when translating from English to Tamarian on known utterances.

Pretrained multilingual models enable zero-shot learning even for unseen languages, and that performance can be further improved via adaptation prior to finetuning. However, it is unclear how the number of pretraining languages influences a model’s zero-shot learning for languages unseen during pretraining. To fill this gap, we ask the following research questions: (1) How does the number of pretraining languages influence zero-shot performance on unseen target languages? (2) Does the answer to that question change with model adaptation? (3) Do the findings for our first question change if the languages used for pretraining are all related? Our experiments on pretraining with related languages indicate that choosing a diverse set of languages is crucial. Without model adaptation, surprisingly, increasing the number of pretraining languages yields better results up to adding related languages, after which performance plateaus. In contrast, with model adaptation via continued pretraining, pretraining on a larger number of languages often gives further improvement, suggesting that model adaptation is crucial to exploit additional pretraining languages.

Neural coreference resolution models trained on one dataset may not transfer to new, low-resource domains. Active learning mitigates this problem by sampling a small subset of data for annotators to label. While active learning is well-defined for classification tasks, its application to coreference resolution is neither well-defined nor fully understood. This paper explores how to actively label coreference, examining sources of model uncertainty and document reading costs. We compare uncertainty sampling strategies and their advantages through thorough error analysis. In both synthetic and human experiments, labeling spans within the same document is more effective than annotating spans across documents. The findings contribute to a more realistic development of coreference resolution models.

We release FoolMeTwice (FM2 for short), a large dataset of challenging entailment pairs collected through a fun multi-player game. Gamification encourages adversarial examples, drastically lowering the number of examples that can be solved using “shortcuts” compared to other popular entailment datasets. Players are presented with two tasks. The first task asks the player to write a plausible claim based on the evidence from a Wikipedia page. The second one shows two plausible claims written by other players, one of which is false, and the goal is to identify it before the time runs out. Players “pay” to see clues retrieved from the evidence pool: the more evidence the player needs, the harder the claim. Game-play between motivated players leads to diverse strategies for crafting claims, such as temporal inference and diverting to unrelated evidence, and results in higher quality data for the entailment and evidence retrieval tasks. We open source the dataset and the game code.

Complex question answering often requires finding a reasoning chain that consists of multiple evidence pieces. Current approaches incorporate the strengths of structured knowledge and unstructured text, assuming text corpora is semi-structured. Building on dense retrieval methods, we propose a new multi-step retrieval approach (BeamDR) that iteratively forms an evidence chain through beam search in dense representations. When evaluated on multi-hop question answering, BeamDR is competitive to state-of-the-art systems, without using any semi-structured information. Through query composition in dense space, BeamDR captures the implicit relationships between evidence in the reasoning chain. The code is available at https://github.com/henryzhao5852/BeamDR.

Question answering (QA) models use retriever and reader systems to answer questions. Reliance on training data by QA systems can amplify or reflect inequity through their responses. Many QA models, such as those for the SQuAD dataset, are trained and tested on a subset of Wikipedia articles which encode their own biases and also reproduce real-world inequality. Understanding how training data affects bias in QA systems can inform methods to mitigate inequity. We develop two sets of questions for closed and open domain questions respectively, which use ambiguous questions to probe QA models for bias. We feed three deep-learning-based QA systems with our question sets and evaluate responses for bias via the metrics. Using our metrics, we find that open-domain QA models amplify biases more than their closed-domain counterparts and propose that biases in the retriever surface more readily due to greater freedom of choice.

How would you explain Bill Gates to a German? He is associated with founding a company in the United States, so perhaps the German founder Carl Benz could stand in for Gates in those contexts. This type of translation is called adaptation in the translation community. Until now, this task has not been done computationally. Automatic adaptation could be used in natural language processing for machine translation and indirectly for generating new question answering datasets and education. We propose two automatic methods and compare them to human results for this novel NLP task. First, a structured knowledge base adapts named entities using their shared properties. Second, vector-arithmetic and orthogonal embedding mappings methods identify better candidates, but at the expense of interpretable features. We evaluate our methods through a new dataset of human adaptations.

The goal of question answering (QA) is to answer _any_ question. However, major QA datasets have skewed distributions over gender, profession, and nationality. Despite that skew, an analysis of model accuracy reveals little evidence that accuracy is lower for people based on gender or nationality; instead, there is more variation on professions (question topic) and question ambiguity. But QA’s lack of representation could itself hide evidence of bias, necessitating QA datasets that better represent global diversity.

Open-domain question answering answers a question based on evidence retrieved from a large corpus. State-of-the-art neural approaches require intermediate evidence annotations for training. However, such intermediate annotations are expensive, and methods that rely on them cannot transfer to the more common setting, where only question–answer pairs are available. This paper investigates whether models can learn to find evidence from a large corpus, with only distant supervision from answer labels for model training, thereby generating no additional annotation cost. We introduce a novel approach (DistDR) that iteratively improves over a weak retriever by alternately finding evidence from the up-to-date model and encouraging the model to learn the most likely evidence. Without using any evidence labels, DistDR is on par with fully-supervised state-of-the-art methods on both multi-hop and single-hop QA benchmarks. Our analysis confirms that DistDR finds more accurate evidence over iterations, which leads to model improvements. The code is available at https://github.com/henryzhao5852/DistDR.

A flaw in QA evaluation is that annotations often only provide one gold answer. Thus, model predictions semantically equivalent to the answer but superficially different are considered incorrect. This work explores mining alias entities from knowledge bases and using them as additional gold answers (i.e., equivalent answers). We incorporate answers for two settings: evaluation with additional answers and model training with equivalent answers. We analyse three QA benchmarks: Natural Questions, TriviaQA, and SQuAD. Answer expansion increases the exact match score on all datasets for evaluation, while incorporating it helps model training over real-world datasets. We ensure the additional answers are valid through a human post hoc evaluation.

Question answering (QA) primarily descends from two branches of research: (1) Alan Turing’s investigation of machine intelligence at Manchester University and (2) Cyril Cleverdon’s comparison of library card catalog indices at Cranfield University. This position paper names and distinguishes these paradigms. Despite substantial overlap, subtle but significant distinctions exert an outsize influence on research. While one evaluation paradigm values creating more intelligent QA systems, the other paradigm values building QA systems that appeal to users. By better understanding the epistemic heritage of QA, researchers, academia, and industry can more effectively accelerate QA research.

Leaderboards are widely used in NLP and push the field forward. While leaderboards are a straightforward ranking of NLP models, this simplicity can mask nuances in evaluation items (examples) and subjects (NLP models). Rather than replace leaderboards, we advocate a re-imagining so that they better highlight if and where progress is made. Building on educational testing, we create a Bayesian leaderboard model where latent subject skill and latent item difficulty predict correct responses. Using this model, we analyze the ranking reliability of leaderboards. Afterwards, we show the model can guide what to annotate, identify annotation errors, detect overfitting, and identify informative examples. We conclude with recommendations for future benchmark tasks.

Cross-lingual word embeddings transfer knowledge between languages: models trained on high-resource languages can predict in low-resource languages. We introduce CLIME, an interactive system to quickly refine cross-lingual word embeddings for a given classification problem. First, CLIME ranks words by their salience to the downstream task. Then, users mark similarity between keywords and their nearest neighbors in the embedding space. Finally, CLIME updates the embeddings using the annotations. We evaluate CLIME on identifying health-related text in four low-resource languages: Ilocano, Sinhalese, Tigrinya, and Uyghur. Embeddings refined by CLIME capture more nuanced word semantics and have higher test accuracy than the original embeddings. CLIME often improves accuracy faster than an active learning baseline and can be easily combined with active learning to improve results.

Active learning strives to reduce annotation costs by choosing the most critical examples to label. Typically, the active learning strategy is contingent on the classification model. For instance, uncertainty sampling depends on poorly calibrated model confidence scores. In the cold-start setting, active learning is impractical because of model instability and data scarcity. Fortunately, modern NLP provides an additional source of information: pre-trained language models. The pre-training loss can find examples that surprise the model and should be labeled for efficient fine-tuning. Therefore, we treat the language modeling loss as a proxy for classification uncertainty. With BERT, we develop a simple strategy based on the masked language modeling loss that minimizes labeling costs for text classification. Compared to other baselines, our approach reaches higher accuracy within less sampling iterations and computation time.

Text representations are critical for modern natural language processing. One form of text representation, sense-specific embeddings, reflect a word’s sense in a sentence better than single-prototype word embeddings tied to each type. However, existing sense representations are not uniformly better: although they work well for computer-centric evaluations, they fail for human-centric tasks like inspecting a language’s sense inventory. To expose this discrepancy, we propose a new coherence evaluation for sense embeddings. We also describe a minimal model (Gumbel Attention for Sense Induction) optimized for discovering interpretable sense representations that are more coherent than existing sense embeddings.

Verb prediction is important for understanding human processing of verb-final languages, with practical applications to real-time simultaneous interpretation from verb-final to verb-medial languages. While previous approaches use classical statistical models, we introduce an attention-based neural model to incrementally predict final verbs on incomplete sentences in Japanese and German SOV sentences. To offer flexibility to the model, we further incorporate synonym awareness. Our approach both better predicts the final verbs in Japanese and German and provides more interpretable explanations of why those verbs are selected.

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 encoderdecoder 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%.

Cross-lingual word embeddings (CLWE) are often evaluated on bilingual lexicon induction (BLI). Recent CLWE methods use linear projections, which underfit the training dictionary, to generalize on BLI. However, underfitting can hinder generalization to other downstream tasks that rely on words from the training dictionary. We address this limitation by retrofitting CLWE to the training dictionary, which pulls training translation pairs closer in the embedding space and overfits the training dictionary. This simple post-processing step often improves accuracy on two downstream tasks, despite lowering BLI test accuracy. We also retrofit to both the training dictionary and a synthetic dictionary induced from CLWE, which sometimes generalizes even better on downstream tasks. Our results confirm the importance of fully exploiting training dictionary in downstream tasks and explains why BLI is a flawed CLWE evaluation.

Trust is implicit in many online text conversations—striking up new friendships, or asking for tech support. But trust can be betrayed through deception. We study the language and dynamics of deception in the negotiation-based game Diplomacy, where seven players compete for world domination by forging and breaking alliances with each other. Our study with players from the Diplomacy community gathers 17,289 messages annotated by the sender for their intended truthfulness and by the receiver for their perceived truthfulness. Unlike existing datasets, this captures deception in long-lasting relationships, where the interlocutors strategically combine truth with lies to advance objectives. A model that uses power dynamics and conversational contexts can predict when a lie occurs nearly as well as human players.

In addition to the traditional task of machines answering questions, question answering (QA) research creates interesting, challenging questions that help systems how to answer questions and reveal the best systems. We argue that creating a QA dataset—and the ubiquitous leaderboard that goes with it—closely resembles running a trivia tournament: you write questions, have agents (either humans or machines) answer the questions, and declare a winner. However, the research community has ignored the hard-learned lessons from decades of the trivia community creating vibrant, fair, and effective question answering competitions. After detailing problems with existing QA datasets, we outline the key lessons—removing ambiguity, discriminating skill, and adjudicating disputes—that can transfer to QA research and how they might be implemented.

Adversarial evaluation stress-tests a model’s understanding of natural language. Because past approaches expose superficial patterns, the resulting adversarial examples are limited in complexity and diversity. We propose human- in-the-loop adversarial generation, where human authors are guided to break models. We aid the authors with interpretations of model predictions through an interactive user interface. We apply this generation framework to a question answering task called Quizbowl, where trivia enthusiasts craft adversarial questions. The resulting questions are validated via live human–computer matches: Although the questions appear ordinary to humans, they systematically stump neural and information retrieval models. The adversarial questions cover diverse phenomena from multi-hop reasoning to entity type distractors, exposing open challenges in robust question answering.

Multilingual topic models (MTMs) learn topics on documents in multiple languages. Past models align topics across languages by implicitly assuming the documents in different languages are highly comparable, often a false assumption. We introduce a new model that does not rely on this assumption, particularly useful in important low-resource language scenarios. Our MTM learns weighted topic links and connects cross-lingual topics only when the dominant words defining them are similar, outperforming LDA and previous MTMs in classification tasks using documents’ topic posteriors as features. It also learns coherent topics on documents with low comparability.

Question answering is an AI-complete problem, but existing datasets lack key elements of language understanding such as coreference and ellipsis resolution. We consider sequential question answering: multiple questions are asked one-by-one in a conversation between a questioner and an answerer. Answering these questions is only possible through understanding the conversation history. We introduce the task of question-in-context rewriting: given the context of a conversation’s history, rewrite a context-dependent into a self-contained question with the same answer. We construct, CANARD, a dataset of 40,527 questions based on QuAC (Choi et al., 2018) and train Seq2Seq models for incorporating context into standalone questions.

Understanding common sense is important for effective natural language reasoning. One type of common sense is how two objects compare on physical properties such as size and weight: e.g., ‘is a house bigger than a person?’. We probe whether pre-trained representations capture comparisons and find they, in fact, have higher accuracy than previous approaches. They also generalize to comparisons involving objects not seen during training. We investigate how such comparisons are made: models learn a consistent ordering over all the objects in the comparisons. Probing models have significantly higher accuracy than those baseline models which use dataset artifacts: e.g., memorizing some words are larger than any other word.

Topic models are typically evaluated with respect to the global topic distributions that they generate, using metrics such as coherence, but without regard to local (token-level) topic assignments. Token-level assignments are important for downstream tasks such as classification. Even recent models, which aim to improve the quality of these token-level topic assignments, have been evaluated only with respect to global metrics. We propose a task designed to elicit human judgments of token-level topic assignments. We use a variety of topic model types and parameters and discover that global metrics agree poorly with human assignments. Since human evaluation is expensive we propose a variety of automated metrics to evaluate topic models at a local level. Finally, we correlate our proposed metrics with human judgments from the task on several datasets. We show that an evaluation based on the percent of topic switches correlates most strongly with human judgment of local topic quality. We suggest that this new metric, which we call consistency, be adopted alongside global metrics such as topic coherence when evaluating new topic models.

Cross-lingual word embeddings (CLWE) underlie many multilingual natural language processing systems, often through orthogonal transformations of pre-trained monolingual embeddings. However, orthogonal mapping only works on language pairs whose embeddings are naturally isomorphic. For non-isomorphic pairs, our method (Iterative Normalization) transforms monolingual embeddings to make orthogonal alignment easier by simultaneously enforcing that (1) individual word vectors are unit length, and (2) each language’s average vector is zero. Iterative Normalization consistently improves word translation accuracy of three CLWE methods, with the largest improvement observed on English-Japanese (from 2% to 44% test accuracy).

Cross-lingual word embeddings encode the meaning of words from different languages into a shared low-dimensional space. An important requirement for many downstream tasks is that word similarity should be independent of language—i.e., word vectors within one language should not be more similar to each other than to words in another language. We measure this characteristic using modularity, a network measurement that measures the strength of clusters in a graph. Modularity has a moderate to strong correlation with three downstream tasks, even though modularity is based only on the structure of embeddings and does not require any external resources. We show through experiments that modularity can serve as an intrinsic validation metric to improve unsupervised cross-lingual word embeddings, particularly on distant language pairs in low-resource settings.

Recent work establishes dataset difficulty and removes annotation artifacts via partial-input baselines (e.g., hypothesis-only model for SNLI or question-only model for VQA). A successful partial-input baseline indicates that the dataset is cheatable. But the converse is not necessarily true: failures of partial-input baselines do not mean the dataset is free of artifacts. We first design artificial datasets to illustrate how the trivial patterns that are only visible in the full input can evade any partial-input baseline. Next, we identify such artifacts in the SNLI dataset—a hypothesis-only model augmented with trivial patterns in the premise can solve 15% of previously-thought “hard” examples. Our work provides a caveat for the use and creation of partial-input baselines for datasets.

To address the lack of comparative evaluation of Human-in-the-Loop Topic Modeling (HLTM) systems, we implement and evaluate three contrasting HLTM modeling approaches using simulation experiments. These approaches extend previously proposed frameworks, including constraints and informed prior-based methods. Users should have a sense of control in HLTM systems, so we propose a control metric to measure whether refinement operations’ results match users’ expectations. Informed prior-based methods provide better control than constraints, but constraints yield higher quality topics.

Previous work on question-answering systems mainly focuses on answering individual questions, assuming they are independent and devoid of context. Instead, we investigate sequential question answering, asking multiple related questions. We present QBLink, a new dataset of fully human-authored questions. We extend existing strong question answering frameworks to include previous questions to improve the overall question-answering accuracy in open-domain question answering. The dataset is publicly available at http://sequential.qanta.org.

One way to interpret neural model predictions is to highlight the most important input features—for example, a heatmap visualization over the words in an input sentence. In existing interpretation methods for NLP, a word’s importance is determined by either input perturbation—measuring the decrease in model confidence when that word is removed—or by the gradient with respect to that word. To understand the limitations of these methods, we use input reduction, which iteratively removes the least important word from the input. This exposes pathological behaviors of neural models: the remaining words appear nonsensical to humans and are not the ones determined as important by interpretation methods. As we confirm with human experiments, the reduced examples lack information to support the prediction of any label, but models still make the same predictions with high confidence. To explain these counterintuitive results, we draw connections to adversarial examples and confidence calibration: pathological behaviors reveal difficulties in interpreting neural models trained with maximum likelihood. To mitigate their deficiencies, we fine-tune the models by encouraging high entropy outputs on reduced examples. Fine-tuned models become more interpretable under input reduction, without accuracy loss on regular examples.

Multilingual topic models enable document analysis across languages through coherent multilingual summaries of the data. However, there is no standard and effective metric to evaluate the quality of multilingual topics. We introduce a new intrinsic evaluation of multilingual topic models that correlates well with human judgments of multilingual topic coherence as well as performance in downstream applications. Importantly, we also study evaluation for low-resource languages. Because standard metrics fail to accurately measure topic quality when robust external resources are unavailable, we propose an adaptation model that improves the accuracy and reliability of these metrics in low-resource settings.

Automatic colorization is the process of adding color to greyscale images. We condition this process on language, allowing end users to manipulate a colorized image by feeding in different captions. We present two different architectures for language-conditioned colorization, both of which produce more accurate and plausible colorizations than a language-agnostic version. Furthermore, we demonstrate through crowdsourced experiments that we can dramatically alter colorizations simply by manipulating descriptive color words in captions.

Local model interpretation methods explain individual predictions by assigning an importance value to each input feature. This value is often determined by measuring the change in confidence when a feature is removed. However, the confidence of neural networks is not a robust measure of model uncertainty. This issue makes reliably judging the importance of the input features difficult. We address this by changing the test-time behavior of neural networks using Deep k-Nearest Neighbors. Without harming text classification accuracy, this algorithm provides a more robust uncertainty metric which we use to generate feature importance values. The resulting interpretations better align with human perception than baseline methods. Finally, we use our interpretation method to analyze model predictions on dataset annotation artifacts.

Simultaneous interpretation, translation of the spoken word in real-time, is both highly challenging and physically demanding. Methods to predict interpreter confidence and the adequacy of the interpreted message have a number of potential applications, such as in computer-assisted interpretation interfaces or pedagogical tools. We propose the task of predicting simultaneous interpreter performance by building on existing methodology for quality estimation (QE) of machine translation output. In experiments over five settings in three language pairs, we extend a QE pipeline to estimate interpreter performance (as approximated by the METEOR evaluation metric) and propose novel features reflecting interpretation strategy and evaluation measures that further improve prediction accuracy.

Modern question answering systems have been touted as approaching human performance. However, existing question answering datasets are imperfect tests. Questions are written with humans in mind, not computers, and often do not properly expose model limitations. To address this, we develop an adversarial writing setting, where humans interact with trained models and try to break them. This annotation process yields a challenge set, which despite being easy for trivia players to answer, systematically stumps automated question answering systems. Diagnosing model errors on the evaluation data provides actionable insights to explore in developing robust and generalizable question answering systems.

Annotated corpora enable supervised machine learning and data analysis. To reduce the cost of manual annotation, tasks are often assigned to internet workers whose judgments are reconciled by crowdsourcing models. We approach the problem of crowdsourcing using a framework for learning from rich prior knowledge, and we identify a family of crowdsourcing models with the novel ability to combine annotations with differing structures: e.g., document labels and word labels. Annotator judgments are given in the form of the predicted expected value of measurement functions computed over annotations and the data, unifying annotation models. Our model, a specific instance of this framework, compares favorably with previous work. Furthermore, it enables active sample selection, jointly selecting annotator, data item, and annotation structure to reduce annotation effort.

Probabilistic topic models are important tools for indexing, summarizing, and analyzing large document collections by their themes. However, promoting end-user understanding of topics remains an open research problem. We compare labels generated by users given four topic visualization techniques—word lists, word lists with bars, word clouds, and network graphs—against each other and against automatically generated labels. Our basis of comparison is participant ratings of how well labels describe documents from the topic. Our study has two phases: a labeling phase where participants label visualized topics and a validation phase where different participants select which labels best describe the topics’ documents. Although all visualizations produce similar quality labels, simple visualizations such as word lists allow participants to quickly understand topics, while complex visualizations take longer but expose multi-word expressions that simpler visualizations obscure. Automatic labels lag behind user-created labels, but our dataset of manually labeled topics highlights linguistic patterns (e.g., hypernyms, phrases) that can be used to improve automatic topic labeling algorithms.

Online topic modeling, i.e., topic modeling with stochastic variational inference, is a powerful and efficient technique for analyzing large datasets, and ADAGRAD is a widely-used technique for tuning learning rates during online gradient optimization. However, these two techniques do not work well together. We show that this is because ADAGRAD uses accumulation of previous gradients as the learning rates’ denominators. For online topic modeling, the magnitude of gradients is very large. It causes learning rates to shrink very quickly, so the parameters cannot fully converge until the training ends

Machine translation is a natural candidate problem for reinforcement learning from human feedback: users provide quick, dirty ratings on candidate translations to guide a system to improve. Yet, current neural machine translation training focuses on expensive human-generated reference translations. We describe a reinforcement learning algorithm that improves neural machine translation systems from simulated human feedback. Our algorithm combines the advantage actor-critic algorithm (Mnih et al., 2016) with the attention-based neural encoder-decoder architecture (Luong et al., 2015). This algorithm (a) is well-designed for problems with a large action space and delayed rewards, (b) effectively optimizes traditional corpus-level machine translation metrics, and (c) is robust to skewed, high-variance, granular feedback modeled after actual human behaviors.

Models work best when they are optimized taking into account the evaluation criteria that people care about. For topic models, people often care about interpretability, which can be approximated using measures of lexical association. We integrate lexical association into topic optimization using tree priors, which provide a flexible framework that can take advantage of both first order word associations and the higher-order associations captured by word embeddings. Tree priors improve topic interpretability without hurting extrinsic performance.

Interactive topic models are powerful tools for those seeking to understand large collections of text. However, existing sampling-based interactive topic modeling approaches scale poorly to large data sets. Anchor methods, which use a single word to uniquely identify a topic, offer the speed needed for interactive work but lack both a mechanism to inject prior knowledge and lack the intuitive semantics needed for user-facing applications. We propose combinations of words as anchors, going beyond existing single word anchor algorithms—an approach we call “Tandem Anchors”. We begin with a synthetic investigation of this approach then apply the approach to interactive topic modeling in a user study and compare it to interactive and non-interactive approaches. Tandem anchors are faster and more intuitive than existing interactive approaches.

Adaptor grammars are a flexible, powerful formalism for defining nonparametric, unsupervised models of grammar productions. This flexibility comes at the cost of expensive inference. We address the difficulty of inference through an online algorithm which uses a hybrid of Markov chain Monte Carlo and variational inference. We show that this inference strategy improves scalability without sacrificing performance on unsupervised word segmentation and topic modeling tasks.