Chandler May

Samuel Barham, Chandler May, Benjamin Van Durme

We introduce MegaWika 2, a large, multilingual dataset of Wikipedia articles with their citations and scraped web sources; articles are represented in a rich data structure, and scraped source texts are stored inline with precise character offsets of their citations in the article text. MegaWika 2 is a major upgrade from the original MegaWika, spanning six times as many articles and twice as many fully scraped citations. Both MegaWika and MegaWika 2 support report generation research ; whereas MegaWika also focused on supporting question answering and retrieval applications, MegaWika 2 is designed to support fact checking and analyses across time and language.

Jiefu Ou, William Gantt Walden, Kate Sanders et al.

A core part of scientific peer review involves providing expert critiques that directly assess the scientific claims a paper makes. While it is now possible to automatically generate plausible (if generic) reviews, ensuring that these reviews are sound and grounded in the papers' claims remains challenging. To facilitate LLM benchmarking on these challenges, we introduce CLAIMCHECK, an annotated dataset of NeurIPS 2023 and 2024 submissions and reviews mined from OpenReview. CLAIMCHECK is richly annotated by ML experts for weakness statements in the reviews and the paper claims that they dispute, as well as fine-grained labels of the validity, objectivity, and type of the identified weaknesses. We benchmark several LLMs on three claim-centric tasks supported by CLAIMCHECK, requiring models to (1) associate weaknesses with the claims they dispute, (2) predict fine-grained labels for weaknesses and rewrite the weaknesses to enhance their specificity, and (3) verify a paper's claims with grounded reasoning. Our experiments reveal that cutting-edge LLMs, while capable of predicting weakness labels in (2), continue to underperform relative to human experts on all other tasks.

William Walden, Kathryn Ricci, Miriam Wanner et al.

Wikipedia is a critical resource for modern NLP, serving as a rich repository of up-to-date and citation-backed information on a wide variety of subjects. The reliability of Wikipedia -- its groundedness in its cited sources -- is vital to this purpose. This work analyzes both how grounded Wikipedia is and how readily fine-grained grounding evidence can be retrieved. To this end, we introduce PeopleProfiles -- a large-scale, multi-level dataset of claim support annotations on biographical Wikipedia articles. We show that: (1) ~22% of claims in Wikipedia lead sections are unsupported by the article body; (2) ~30% of claims in the article body are unsupported by their publicly accessible sources; and (3) real-world Wikipedia citation practices often differ from documented standards. Finally, we show that complex evidence retrieval remains a challenge -- even for recent reasoning rerankers.

Michelle Yuan, Patrick Xia, Chandler May et al.

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.

Patrick Xia, Guanghui Qin, Siddharth Vashishtha et al.

We present LOME, a system for performing multilingual information extraction. Given a text document as input, our core system identifies spans of textual entity and event mentions with a FrameNet (Baker et al., 1998) parser. It subsequently performs coreference resolution, fine-grained entity typing, and temporal relation prediction between events. By doing so, the system constructs an event and entity focused knowledge graph. We can further apply third-party modules for other types of annotation, like relation extraction. Our (multilingual) first-party modules either outperform or are competitive with the (monolingual) state-of-the-art. We achieve this through the use of multilingual encoders like XLM-R (Conneau et al., 2020) and leveraging multilingual training data. LOME is available as a Docker container on Docker Hub. In addition, a lightweight version of the system is accessible as a web demo.

Chandler May, Alex Wang, Shikha Bordia et al.

The Word Embedding Association Test shows that GloVe and word2vec word embeddings exhibit human-like implicit biases based on gender, race, and other social constructs (Caliskan et al., 2017). Meanwhile, research on learning reusable text representations has begun to explore sentence-level texts, with some sentence encoders seeing enthusiastic adoption. Accordingly, we extend the Word Embedding Association Test to measure bias in sentence encoders. We then test several sentence encoders, including state-of-the-art methods such as ELMo and BERT, for the social biases studied in prior work and two important biases that are difficult or impossible to test at the word level. We observe mixed results including suspicious patterns of sensitivity that suggest the test's assumptions may not hold in general. We conclude by proposing directions for future work on measuring bias in sentence encoders.

Chandler May, Kevin Duh, Benjamin Van Durme et al.

We develop a streaming (one-pass, bounded-memory) word embedding algorithm based on the canonical skip-gram with negative sampling algorithm implemented in word2vec. We compare our streaming algorithm to word2vec empirically by measuring the cosine similarity between word pairs under each algorithm and by applying each algorithm in the downstream task of hashtag prediction on a two-month interval of the Twitter sample stream. We then discuss the results of these experiments, concluding they provide partial validation of our approach as a streaming replacement for word2vec. Finally, we discuss potential failure modes and suggest directions for future work.

Chandler May, Ryan Cotterell, Benjamin Van Durme

Topic models are typically represented by top-$m$ word lists for human interpretation. The corpus is often pre-processed with lemmatization (or stemming) so that those representations are not undermined by a proliferation of words with similar meanings, but there is little public work on the effects of that pre-processing. Recent work studied the effect of stemming on topic models of English texts and found no supporting evidence for the practice. We study the effect of lemmatization on topic models of Russian Wikipedia articles, finding in one configuration that it significantly improves interpretability according to a word intrusion metric. We conclude that lemmatization may benefit topic models on morphologically rich languages, but that further investigation is needed.