• Research data close
• Research software close
• Other research products close
• DK close
• Digital Humanities and Cultural Her... close
• Knowmad Institut close

Read the paper on the folowing link: https://www.aclweb.org/anthology/2021.naacl-main.204/ Abstract: Supervised learning assumes that a ground truth label exists. However, the reliability of this ground truth depends on human annotators, who often disagree. Prior work has shown that this disagreement can be helpful in training models. We propose a novel method to incorporate this disagreement as information: in addition to the standard error computation, we use soft labels (i.e., probability distributions over the annotator labels) as an auxiliary task in a multi-task neural network. We measure the divergence between the predictions and the target soft labels with several loss-functions and evaluate the models on various NLP tasks. We find that the soft-label prediction auxiliary task reduces the penalty for errors on ambiguous entities and thereby mitigates overfitting. It significantly improves performance across tasks beyond the standard approach and prior work.

In this project, we discuss the phenomenon of UAV warfare. By conducting a discourse analysis of two speeches, one conducted by then counterterrorism advisor John Brennan and one by President Barack Obama, we investigate how UAV use is justified. We briefly discuss the historical background and contemporary public opinion in order to contextualize the discourse presented in the two speeches. The discourse analysis is structured in three analytical categories: how are representations of identity are articulated, how the speakers make claims about the future and finally the specific nature of the justifications of UAV use. Finally, we discuss how our empirical findings relate to the discussion of the changing nature of warfare, as well as we present a brief critique of a position in the current UAV debate. Our main argument in this discussion is that UAVs should be discussed within the social, discursive practice they are used, and not regarded merely as technological objects distinct from the context they exist in.

ROOTS is a 1.6TB multilingual text corpus developed for the training of BLOOM, currently the largest language model explicitly accompanied by commensurate data governance efforts. In continuation of these efforts, we present the ROOTS Search Tool: a search engine over the entire ROOTS corpus offering both fuzzy and exact search capabilities. ROOTS is the largest corpus to date that can be investigated this way. The ROOTS Search Tool is open-sourced and available on Hugging Face Spaces: https://huggingface.co/spaces/bigscience-data/roots-search. We describe our implementation and the possible use cases of our tool.

Detecting offensive language is a challenging task. Generalizing across different cultures and languages becomes even more challenging: besides lexical, syntactic and semantic differences, pragmatic aspects such as cultural norms and sensitivities, which are particularly relevant in this context, vary greatly. In this paper, we target Chinese offensive language detection and aim to investigate the impact of transfer learning using offensive language detection data from different cultural backgrounds, specifically Korean and English. We find that culture-specific biases in what is considered offensive negatively impact the transferability of language models (LMs) and that LMs trained on diverse cultural data are sensitive to different features in Chinese offensive language detection. In a few-shot learning scenario, however, our study shows promising prospects for non-English offensive language detection with limited resources. Our findings highlight the importance of cross-cultural transfer learning in improving offensive language detection and promoting inclusive digital spaces.

In the paper we describe a new EU infrastructure project dedicated to lexicography. The project is part of the Horizon 2020 program, with a duration of four years (2018-2022). The result of the project will be an infrastructure which will (1) enable efficient access to high quality lexicographic data, and (2) bridge the gap between more advanced and less-resourced scholarly communities working on lexicographic resources. One of the main issues addressed by the project is the fact that current lexicographic resources have different levels of (incompatible) structuring, and are not equally suitable for application in in Natural Language Processing and other fields. The project will therefore develop strategies, tools and standards for extracting, structuring and linking lexicographic resources to enable their inclusion in Linked Open Data and the Semantic Web, as well as their use in the context of digital humanities.

Read the paper on the folowing link: https://www.aclweb.org/anthology/2021.naacl-main.320/ Abstract: In order to interpret the communicative intents of an utterance, it needs to be grounded in something that is outside of language; that is, grounded in world modalities. In this paper we argue that dialogue clarification mechanisms make explicit the process of interpreting the communicative intents of the speaker's utterances by grounding them in the various modalities in which the dialogue is situated. This paper frames dialogue clarification mechanisms as an understudied research problem and a key missing piece in the giant jigsaw puzzle of natural language understanding. We discuss both the theoretical background and practical challenges posed by this problem, and propose a recipe for obtaining grounding annotations. We conclude by highlighting ethical issues that need to be addressed in future work.

This project was set out to explore the role of the Turing Test in the development of Artificial Intelligence (AI), with emphasis on the historical perspective. This report contains an introductory presentation of the Turing Test and Artificial Intelligence. Furthermore, it presents two methods for analysis. The first method is a quantitative search in extracting the number of results from Google Scholars for search range between 1950 and 2019. The searched terms are ‘Turing Test’ and ‘Artificial Intelligence’. The second method is the one used for the analysis of two case studies, ELIZA and Google Duplex. In exploring the historical development, ELIZA is an early research topic from 1966 and Google Duplex is a contemporary project from 2018. This report concludes that the Turing Test appears to have played a role in the historical development of AI. Results from the quantitative search show that there is an exponential growth, followed by a short stabilisation, before it begins to decay towards the last decade. Both case studies failed when subjected to a strict Turing Test. Though when subjected to the Total Turing Test, Google Duplex seems to surpass it. Finally, this report also concludes that the Turing Test may no longer be relevant, as mediums for AI have evolved beyond text-based and most developments are no longer concerned with tricking humans.

Read the paper on the folowing link: https://www.aclweb.org/anthology/2021.naacl-main.10/ Abstract: We analyze if large language models are able to predict patterns of human reading behavior. We compare the performance of language-specific and multilingual pretrained transformer models to predict reading time measures reflecting natural human sentence processing on Dutch, English, German, and Russian texts. This results in accurate models of human reading behavior, which indicates that transformer models implicitly encode relative importance in language in a way that is comparable to human processing mechanisms. We find that BERT and XLM models successfully predict a range of eye tracking features. In a series of experiments, we analyze the cross-domain and cross-language abilities of these models and show how they reflect human sentence processing.

Additional file 2: Table S1. Description of the studies included in the analyses.

Read the paper on the folowing link: https://www.aclweb.org/anthology/2021.naacl-main.22/ Abstract: Interpretability or explainability is an emerging research field in NLP. From a user-centric point of view, the goal is to build models that provide proper justification for their decisions, similar to those of humans, by requiring the models to satisfy additional constraints. To this end, we introduce a new application on legal text where, contrary to mainstream literature targeting word-level rationales, we conceive rationales as selected paragraphs in multi-paragraph structured court cases. We also release a new dataset comprising European Court of Human Rights cases, including annotations for paragraph-level rationales. We use this dataset to study the effect of already proposed rationale constraints, i.e., sparsity, continuity, and comprehensiveness, formulated as regularizers. Our findings indicate that some of these constraints are not beneficial in paragraph-level rationale extraction, while others need re-formulation to better handle the multi-label nature of the task we consider. We also introduce a new constraint, singularity, which further improves the quality of rationales, even compared with noisy rationale supervision. Experimental results indicate that the newly introduced task is very challenging and there is a large scope for further research.