arXiv subjects

Robert West

Publications and source records attributed to Robert West.

At least 109 records · Page 6Linked to original sources

arXiv2022

Evolutionary Dynamics in a Varying Environment: Continuous versus Discrete Noise

Environmental variations can significantly influence how populations compete for resources, and hence shape their evolution. Here, we study population dynamics subject to a fluctuating environment modeled by a varying carrying capacity changing continuously in time according to either binary random switches, or by being driven by a noise of continuous range. We consider a prototypical example of two competing strains, one growing slightly slower than the other, and consider also the scenario where the slow strain is a public goods producer. By systematically comparing the effect of binary- versus continuously-varying environment, we study how different noise statistics (mean, variance) influence the population size and fixation properties. We show that the slow strain fixation probability can be greatly enhanced for a continuously-varying environment compared to binary switches, even when the first two moments of the carrying capacity coincide.

q-bio.PE↗

arXiv2022

Invariant Language Modeling

Large pretrained language models are critical components of modern NLP pipelines. Yet, they suffer from spurious correlations, poor out-of-domain generalization, and biases. Inspired by recent progress in causal machine learning, in particular the invariant risk minimization (IRM) paradigm, we propose invariant language modeling, a framework for learning invariant representations that generalize better across multiple environments. In particular, we adapt a game-theoretic formulation of IRM (IRM-games) to language models, where the invariance emerges from a specific training schedule in which all the environments compete to optimize their own environment-specific loss by updating subsets of the model in a round-robin fashion. We focus on controlled experiments to precisely demonstrate the ability of our method to (i) remove structured noise, (ii) ignore specific spurious correlations without affecting global performance, and (iii) achieve better out-of-domain generalization. These benefits come with a negligible computational overhead compared to standard training, do not require changing the local loss, and can be applied to any language model. We believe this framework is promising to help mitigate spurious correlations and biases in language models.

cs.CL↗

arXiv2022

The Glass Ceiling of Automatic Evaluation in Natural Language Generation

Automatic evaluation metrics capable of replacing human judgments are critical to allowing fast development of new methods. Thus, numerous research efforts have focused on crafting such metrics. In this work, we take a step back and analyze recent progress by comparing the body of existing automatic metrics and human metrics altogether. As metrics are used based on how they rank systems, we compare metrics in the space of system rankings. Our extensive statistical analysis reveals surprising findings: automatic metrics -- old and new -- are much more similar to each other than to humans. Automatic metrics are not complementary and rank systems similarly. Strikingly, human metrics predict each other much better than the combination of all automatic metrics used to predict a human metric. It is surprising because human metrics are often designed to be independent, to capture different aspects of quality, e.g. content fidelity or readability. We provide a discussion of these findings and recommendations for future work in the field of evaluation.

cs.CL↗

arXiv2022

Characterizing Alternative Monetization Strategies on YouTube

One of the key emerging roles of the YouTube platform is providing creators the ability to generate revenue from their content and interactions. Alongside tools provided directly by the platform, such as revenue-sharing from advertising, creators co-opt the platform to use a variety of off-platform monetization opportunities. In this work, we focus on studying and characterizing these alternative monetization strategies. Leveraging a large longitudinal YouTube dataset of popular creators, we develop a taxonomy of alternative monetization strategies and a simple methodology to detect their usage automatically. We then proceed to characterize the adoption of these strategies. First, we find that the use of external monetization is expansive and increasingly prevalent, used in 18% of all videos, with 61% of channels using one such strategy at least once. Second, we show that the adoption of these strategies varies substantially among channels of different kinds and popularity, and that channels that establish these alternative revenue streams often become more productive on the platform. Lastly, we investigate how potentially problematic channels -- those that produce Alt-lite, Alt-right, and Manosphere content -- leverage alternative monetization strategies, finding that they employ a more diverse set of such strategies significantly more often than a carefully chosen comparison set of channels. This finding complicates YouTube's role as a gatekeeper, since the practice of excluding policy-violating content from its native on-platform monetization may not be effective. Overall, this work provides an important step toward broadening the understanding of the monetary incentives behind content creation on YouTube.

cs.CY↗

arXiv2022

On the Choice of Databases in Differential Privacy Composition

Differential privacy (DP) is a widely applied paradigm for releasing data while maintaining user privacy. Its success is to a large part due to its composition property that guarantees privacy even in the case of multiple data releases. Consequently, composition has received a lot of attention from the research community: there exist several composition theorems for adversaries with different amounts of flexibility in their choice of mechanisms. But apart from mechanisms, the adversary can also choose the databases on which these mechanisms are invoked. The classic tool for analyzing the composition of DP mechanisms, the so-called composition experiment, neither allows for incorporating constraints on databases nor for different assumptions on the adversary's prior knowledge about database membership. We therefore propose a generalized composition experiment (GCE), which has this flexibility. We show that composition theorems that hold with respect to the classic composition experiment also hold with respect to the worst case of the GCE. This implies that existing composition theorems give a privacy guarantee for more cases than are explicitly covered by the classic composition experiment. Beyond these theoretical insights, we demonstrate two practical applications of the GCE: the first application is to give better privacy bounds in the presence of restrictions on the choice of databases; the second application is to reason about how the adversary's prior knowledge influences the privacy leakage. In this context, we show a connection between adversaries with an uninformative prior and subsampling, an important primitive in DP. To the best of our knowledge, this paper is the first to analyze the interplay between the databases in DP composition, and thereby gives both a better understanding of composition and practical tools for obtaining better composition bounds.

cs.CR↗

arXiv2022

Distribution inference risks: Identifying and mitigating sources of leakage

A large body of work shows that machine learning (ML) models can leak sensitive or confidential information about their training data. Recently, leakage due to distribution inference (or property inference) attacks is gaining attention. In this attack, the goal of an adversary is to infer distributional information about the training data. So far, research on distribution inference has focused on demonstrating successful attacks, with little attention given to identifying the potential causes of the leakage and to proposing mitigations. To bridge this gap, as our main contribution, we theoretically and empirically analyze the sources of information leakage that allows an adversary to perpetrate distribution inference attacks. We identify three sources of leakage: (1) memorizing specific information about the $\mathbb{E}[Y|X]$ (expected label given the feature values) of interest to the adversary, (2) wrong inductive bias of the model, and (3) finiteness of the training data. Next, based on our analysis, we propose principled mitigation techniques against distribution inference attacks. Specifically, we demonstrate that causal learning techniques are more resilient to a particular type of distribution inference risk termed distributional membership inference than associative learning methods. And lastly, we present a formalization of distribution inference that allows for reasoning about more general adversaries than was previously possible.

cs.CR↗

arXiv2022

Anticipated versus Actual Effects of Platform Design Change: A Case Study of Twitter's Character Limit

The design of online platforms is both critically important and challenging, as any changes may lead to unintended consequences, and it can be hard to predict how users will react. Here we conduct a case study of a particularly important real-world platform design change: Twitter's decision to double the character limit from 140 to 280 characters to soothe users' need to ''cram'' or ''squeeze'' their tweets, informed by modeling of historical user behavior. In our analysis, we contrast Twitter's anticipated pre-intervention predictions about user behavior with actual post-intervention user behavior: Did the platform design change lead to the intended user behavior shifts, or did a gap between anticipated and actual behavior emerge? Did different user groups react differently? We find that even though users do not ''cram'' as much under 280 characters as they used to under 140 characters, emergent ``cramming'' at the new limit seems to not have been taken into account when designing the platform change. Furthermore, investigating textual features, we find that, although post-intervention ''crammed'' tweets are longer, their syntactic and semantic characteristics remain similar and indicative of ''squeezing''. Applying the same approach as Twitter policy-makers, we create updated counterfactual estimates and find that the character limit would need to be increased further to reduce cramming that re-emerged at the new limit. We contribute to the rich literature studying online user behavior with an empirical study that reveals a dynamic interaction between platform design and user behavior, with immediate policy and practical implications for the design of socio-technical systems.

cs.SI↗

arXiv2022

Biased Bytes: On the Validity of Estimating Food Consumption from Digital Traces

Given that measuring food consumption at a population scale is a challenging task, researchers have begun to explore digital traces (e.g., from social media or from food-tracking applications) as potential proxies. However, it remains unclear to what extent digital traces reflect real food consumption. The present study aims to bridge this gap by quantifying the link between dietary behaviors as captured via social media (Twitter) v.s. a food-tracking application (MyFoodRepo). We focus on the case of Switzerland and contrast images of foods collected through the two platforms, by designing and deploying a novel crowdsourcing framework for estimating biases with respect to nutritional properties and appearance. We find that the food type distributions in social media v.s. food tracking diverge; e.g., bread is 2.5 times more frequent among consumed and tracked foods than on Twitter, whereas cake is 12 times more frequent on Twitter. Controlling for the different food type distributions, we contrast consumed and tracked foods of a given type with foods shared on Twitter. Across food types, food posted on Twitter is perceived as tastier, more caloric, less healthy, less likely to have been consumed at home, more complex, and larger-portioned, compared to consumed and tracked foods. The fact that there is a divergence between food consumption as measured via the two platforms implies that at least one of the two is not a faithful representation of the true food consumption in the general Swiss population. Thus, researchers should be attentive and aim to establish evidence of validity before using digital traces as a proxy for the true food consumption of a general population. We conclude by discussing the potential sources of these biases and their implications, outlining pitfalls and threats to validity, and proposing actionable ways for overcoming them.

cs.SI↗

arXiv2022

ESC-Rules: Explainable, Semantically Constrained Rule Sets

We describe a novel approach to explainable prediction of a continuous variable based on learning fuzzy weighted rules. Our model trains a set of weighted rules to maximise prediction accuracy and minimise an ontology-based 'semantic loss' function including user-specified constraints on the rules that should be learned in order to maximise the explainability of the resulting rule set from a user perspective. This system fuses quantitative sub-symbolic learning with symbolic learning and constraints based on domain knowledge. We illustrate our system on a case study in predicting the outcomes of behavioural interventions for smoking cessation, and show that it outperforms other interpretable approaches, achieving performance close to that of a deep learning model, while offering transparent explainability that is an essential requirement for decision-makers in the health domain.

cs.AI↗

arXiv2022

United States Politicians' Tone Became More Negative with 2016 Primary Campaigns

There is a widespread belief that the tone of US political language has become more negative recently, in particular when Donald Trump entered politics. At the same time, there is disagreement as to whether Trump changed or merely continued previous trends. To date, data-driven evidence regarding these questions is scarce, partly due to the difficulty of obtaining a comprehensive, longitudinal record of politicians' utterances. Here we apply psycholinguistic tools to a novel, comprehensive corpus of 24 million quotes from online news attributed to 18,627 US politicians in order to analyze how the tone of US politicians' language evolved between 2008 and 2020. We show that, whereas the frequency of negative emotion words had decreased continuously during Obama's tenure, it suddenly and lastingly increased with the 2016 primary campaigns, by 1.6 pre-campaign standard deviations, or 8% of the pre-campaign mean, in a pattern that emerges across parties. The effect size drops by 40% when omitting Trump's quotes, and by 50% when averaging over speakers rather than quotes, implying that prominent speakers, and Trump in particular, have disproportionately, though not exclusively, contributed to the rise in negative language. This work provides the first large-scale data-driven evidence of a drastic shift toward a more negative political tone following Trump's campaign start as a catalyst, with important implications for the debate about the state of US politics.

cs.CL↗

arXiv2022

Quote Erat Demonstrandum: A Web Interface for Exploring the Quotebank Corpus

The use of attributed quotes is the most direct and least filtered pathway of information propagation in news. Consequently, quotes play a central role in the conception, reception, and analysis of news stories. Since quotes provide a more direct window into a speaker's mind than regular reporting, they are a valuable resource for journalists and researchers alike. While substantial research efforts have been devoted to methods for the automated extraction of quotes from news and their attribution to speakers, few comprehensive corpora of attributed quotes from contemporary sources are available to the public. Here, we present an adaptive web interface for searching Quotebank, a massive collection of quotes from the news, which we make available at https://quotebank.dlab.tools.

cs.IR↗

arXiv2022

Strong Heuristics for Named Entity Linking

Named entity linking (NEL) in news is a challenging endeavour due to the frequency of unseen and emerging entities, which necessitates the use of unsupervised or zero-shot methods. However, such methods tend to come with caveats, such as no integration of suitable knowledge bases (like Wikidata) for emerging entities, a lack of scalability, and poor interpretability. Here, we consider person disambiguation in Quotebank, a massive corpus of speaker-attributed quotations from the news, and investigate the suitability of intuitive, lightweight, and scalable heuristics for NEL in web-scale corpora. Our best performing heuristic disambiguates 94% and 63% of the mentions on Quotebank and the AIDA-CoNLL benchmark, respectively. Additionally, the proposed heuristics compare favourably to the state-of-the-art unsupervised and zero-shot methods, Eigenthemes and mGENRE, respectively, thereby serving as strong baselines for unsupervised and zero-shot entity linking.

cs.CL↗

arXiv2022

Efficient Entity Candidate Generation for Low-Resource Languages

Candidate generation is a crucial module in entity linking. It also plays a key role in multiple NLP tasks that have been proven to beneficially leverage knowledge bases. Nevertheless, it has often been overlooked in the monolingual English entity linking literature, as naive approaches obtain very good performance. Unfortunately, the existing approaches for English cannot be successfully transferred to poorly resourced languages. This paper constitutes an in-depth analysis of the candidate generation problem in the context of cross-lingual entity linking with a focus on low-resource languages. Among other contributions, we point out limitations in the evaluation conducted in previous works. We introduce a characterization of queries into types based on their difficulty, which improves the interpretability of the performance of different methods. We also propose a light-weight and simple solution based on the construction of indexes whose design is motivated by more complex transfer learning based neural approaches. A thorough empirical analysis on 9 real-world datasets under 2 evaluation settings shows that our simple solution outperforms the state-of-the-art approach in terms of both quality and efficiency for almost all datasets and query types.

cs.CL↗

arXiv2022

Can online attention signals help fact-checkers fact-check?

Recent research suggests that not all fact-checking efforts are equal: when and what is fact-checked plays a pivotal role in effectively correcting misconceptions. In that context, signals capturing how much attention specific topics receive on the Internet have the potential to study (and possibly support) fact-checking efforts. This paper proposes a framework to study fact-checking with online attention signals. The framework consists of: 1) extracting claims from fact-checking efforts; 2) linking such claims with knowledge graph entities; and 3) estimating the online attention these entities receive. We use this framework to conduct a preliminary study of a dataset of 879 COVID-19-related fact-checks done in 2020 by 81 international organizations. Our findings suggest that there is often a disconnect between online attention and fact-checking efforts. For example, in around 40% of countries that fact-checked ten or more claims, half or more than half of the ten most popular claims were not fact-checked. Our analysis also shows that claims are first fact-checked after receiving, on average, 35% of the total online attention they would eventually receive in 2020. Yet, there is a considerable variation among claims: some were fact-checked before receiving a surge of misinformation-induced online attention; others are fact-checked much later. Overall, our work suggests that the incorporation of online attention signals may help organizations assess their fact-checking efforts and choose what and when to fact-check claims or stories. Also, in the context of international collaboration, where claims are fact-checked multiple times across different countries, online attention could help organizations keep track of which claims are "migrating" between countries.

cs.CY↗

arXiv2022

Post Approvals in Online Communities

In many online communities, community leaders (i.e., moderators and administrators) can proactively filter undesired content by requiring posts to be approved before publication. But although many communities adopt post approvals, there has been little research on its impact on community behavior. Through a longitudinal analysis of 233,402 Facebook Groups, we examined 1) the factors that led to a community adopting post approvals and 2) how the setting shaped subsequent user activity and moderation in the group. We find that communities that adopted post approvals tended to do so following sudden increases in user activity (e.g., comments) and moderation (e.g., reported posts). This adoption of post approvals led to fewer but higher-quality posts. Though fewer posts were shared after adoption, not only did community members write more comments, use more reactions, and spend more time on the posts that were shared, they also reported these posts less. Further, post approvals did not significantly increase the average time leaders spent in the group, though groups that enabled the setting tended to appoint more leaders. Last, the impact of post approvals varied with both group size and how the setting was used, e.g.,, group size mediates whether leaders spent more or less time in the group following the adoption of the setting. Our findings suggest ways that proactive content moderation may be improved to better support online communities.

cs.SI↗

arXiv2022

GenIE: Generative Information Extraction

Structured and grounded representation of text is typically formalized by closed information extraction, the problem of extracting an exhaustive set of (subject, relation, object) triplets that are consistent with a predefined set of entities and relations from a knowledge base schema. Most existing works are pipelines prone to error accumulation, and all approaches are only applicable to unrealistically small numbers of entities and relations. We introduce GenIE (generative information extraction), the first end-to-end autoregressive formulation of closed information extraction. GenIE naturally exploits the language knowledge from the pre-trained transformer by autoregressively generating relations and entities in textual form. Thanks to a new bi-level constrained generation strategy, only triplets consistent with the predefined knowledge base schema are produced. Our experiments show that GenIE is state-of-the-art on closed information extraction, generalizes from fewer training data points than baselines, and scales to a previously unmanageable number of entities and relations. With this work, closed information extraction becomes practical in realistic scenarios, providing new opportunities for downstream tasks. Finally, this work paves the way towards a unified end-to-end approach to the core tasks of information extraction. Code, data and models available at https://github.com/epfl-dlab/GenIE.

cs.CL↗

arXiv2022

Homepage2Vec: Language-Agnostic Website Embedding and Classification

Currently, publicly available models for website classification do not offer an embedding method and have limited support for languages beyond English. We release a dataset of more than two million category-labeled websites in 92 languages collected from Curlie, the largest multilingual human-edited Web directory. The dataset contains 14 website categories aligned across languages. Alongside it, we introduce Homepage2Vec, a machine-learned pre-trained model for classifying and embedding websites based on their homepage in a language-agnostic way. Homepage2Vec, thanks to its feature set (textual content, metadata tags, and visual attributes) and recent progress in natural language representation, is language-independent by design and generates embedding-based representations. We show that Homepage2Vec correctly classifies websites with a macro-averaged F1-score of 0.90, with stable performance across low- as well as high-resource languages. Feature analysis shows that a small subset of efficiently computable features suffices to achieve high performance even with limited computational resources. We make publicly available the curated Curlie dataset aligned across languages, the pre-trained Homepage2Vec model, and libraries

cs.CL↗

arXiv2022

On the Context-Free Ambiguity of Emoji

Emojis come with prepacked semantics making them great candidates to create new forms of more accessible communications. Yet, little is known about how much of this emojis semantic is agreed upon by humans, outside of textual contexts. Thus, we collected a crowdsourced dataset of one-word emoji descriptions for 1,289 emojis presented to participants with no surrounding text. The emojis and their interpretations were then examined for ambiguity. We find that with 30 annotations per emoji, 16 emojis (1.2%) are completely unambiguous, whereas 55 emojis (4.3%) are so ambiguous that their descriptions are indistinguishable from randomly chosen descriptions. Most of studied emojis are spread out between the two extremes. Furthermore, investigating the ambiguity of different types of emojis, we find that an important factor is the extent to which an emoji has an embedded symbolical meaning drawn from an established code-book of symbols. We conclude by discussing design implications.

cs.CL↗