Disagreement between coders is ubiquitous in virtually all datasets annotated with human judgements in both natural language processing and computer vision. However, most supervised machine learning methods assume that a single preferred interpretation exists for each item, which is at best an idealization. The aim of the SemEval-2021 shared task on learning with disagreements (Le-Wi-Di) was to provide a unified testing framework for methods for learning from data containing multiple and possibly contradictory annotations covering the best-known datasets containing information about disagreements for interpreting language and classifying images. In this paper we describe the shared task and its results.

In this work, we design an end-to-end model for poetry generation based on conditioned recurrent neural network (RNN) language models whose goal is to learn stylistic features (poem length, sentiment, alliteration, and rhyming) from examples alone. We show this model successfully learns the ‘meaning’ of length and sentiment, as we can control it to generate longer or shorter as well as more positive or more negative poems. However, the model does not grasp sound phenomena like alliteration and rhyming, but instead exploits low-level statistical cues. Possible reasons include the size of the training data, the relatively low frequency and difficulty of these sublexical phenomena as well as model biases. We show that more recent GPT-2 models also have problems learning sublexical phenomena such as rhyming from examples alone.

The inability to quantify key aspects of creative language is a frequent obstacle to natural language understanding. To address this, we introduce novel tasks for evaluating the creativeness of language—namely, scoring and ranking text by humorousness and metaphor novelty. To sidestep the difficulty of assigning discrete labels or numeric scores, we learn from pairwise comparisons between texts. We introduce a Bayesian approach for predicting humorousness and metaphor novelty using Gaussian process preference learning (GPPL), which achieves a Spearman’s ρ of 0.56 against gold using word embeddings and linguistic features. Our experiments show that given sparse, crowdsourced annotation data, ranking using GPPL outperforms best–worst scaling. We release a new dataset for evaluating humour containing 28,210 pairwise comparisons of 4,030 texts, and make our software freely available.

The study of argumentation and the development of argument mining tools depends on the availability of annotated data, which is challenging to obtain in sufficient quantity and quality. We present a method that breaks down a popular but relatively complex discourse-level argument annotation scheme into a simpler, iterative procedure that can be applied even by untrained annotators. We apply this method in a crowdsourcing setup and report on the reliability of the annotations obtained. The source code for a tool implementing our annotation method, as well as the sample data we obtained (4909 gold-standard annotations across 982 documents), are freely released to the research community. These are intended to serve the needs of qualitative research into argumentation, as well as of data-driven approaches to argument mining.

The translation of wordplay is one of the most extensively researched problems in translation studies, but it has attracted little attention in the fields of natural language processing and machine translation. This is because today’s language technologies treat anomalies and ambiguities in the input as things that must be resolved in favour of a single “correct” interpretation, rather than preserved and interpreted in their own right. But if computers cannot yet process such creative language on their own, can they at least provide specialized support to translation professionals? In this paper, I survey the state of the art relevant to computational processing of humorous wordplay and put forth a vision of how existing theories, resources, and technologies could be adapted and extended to support interactive, computer-assisted translation.

Argument mining is a core technology for enabling argument search in large corpora. However, most current approaches fall short when applied to heterogeneous texts. In this paper, we present an argument retrieval system capable of retrieving sentential arguments for any given controversial topic. By analyzing the highest-ranked results extracted from Web sources, we found that our system covers 89% of arguments found in expert-curated lists of arguments from an online debate portal, and also identifies additional valid arguments.

Argument mining is a core technology for automating argument search in large document collections. Despite its usefulness for this task, most current approaches are designed for use only with specific text types and fall short when applied to heterogeneous texts. In this paper, we propose a new sentential annotation scheme that is reliably applicable by crowd workers to arbitrary Web texts. We source annotations for over 25,000 instances covering eight controversial topics. We show that integrating topic information into bidirectional long short-term memory networks outperforms vanilla BiLSTMs by more than 3 percentage points in F1 in two- and three-label cross-topic settings. We also show that these results can be further improved by leveraging additional data for topic relevance using multi-task learning.

In this paper, we propose using metaheuristics—in particular, simulated annealing and the new D-Bees algorithm—to solve word sense disambiguation as an optimization problem within a knowledge-based lexical substitution system. We are the first to perform such an extrinsic evaluation of metaheuristics, for which we use two standard lexical substitution datasets, one English and one German. We find that D-Bees has robust performance for both languages, and performs better than simulated annealing, though both achieve good results. Moreover, the D-Bees–based lexical substitution system outperforms state-of-the-art systems on several evaluation metrics. We also show that D-Bees achieves competitive performance in lexical simplification, a variant of lexical substitution.

A pun is a form of wordplay in which a word suggests two or more meanings by exploiting polysemy, homonymy, or phonological similarity to another word, for an intended humorous or rhetorical effect. Though a recurrent and expected feature in many discourse types, puns stymie traditional approaches to computational lexical semantics because they violate their one-sense-per-context assumption. This paper describes the first competitive evaluation for the automatic detection, location, and interpretation of puns. We describe the motivation for these tasks, the evaluation methods, and the manually annotated data set. Finally, we present an overview and discussion of the participating systems’ methodologies, resources, and results.

We describe the construction of GLASS, a newly sense-annotated version of the German lexical substitution data set used at the GermEval 2015: LexSub shared task. Using the two annotation layers, we conduct the first known empirical study of the relationship between manually applied word senses and lexical substitutions. We find that synonymy and hypernymy/hyponymy are the only semantic relations directly linking targets to their substitutes, and that substitutes in the target’s hypernymy/hyponymy taxonomy closely align with the synonyms of a single GermaNet synset. Despite this, these substitutes account for a minority of those provided by the annotators. The results of our analysis accord with those of a previous study on English-language data (albeit with automatically induced word senses), leading us to suspect that the sense―substitution relations we discovered may be of a universal nature. We also tentatively conclude that relatively cheap lexical substitution annotations can be used as a knowledge source for automatic WSD. Also introduced in this paper is Ubyline, the web application used to produce the sense annotations. Ubyline presents an intuitive user interface optimized for annotating lexical sample data, and is readily adaptable to sense inventories other than GermaNet.

When processing arguments in online user interactive discourse, it is often necessary to determine their bases of support. In this paper, we describe a supervised approach, based on deep neural networks, for classifying the claims made in online arguments. We conduct experiments using convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) on two claim data sets compiled from online user comments. Using different types of distributional word embeddings, but without incorporating any rich, expensive set of features, we achieve a significant improvement over the state of the art for one data set (which categorizes arguments as factual vs. emotional), and performance comparable to the state of the art on the other data set (which categorizes propositions according to their verifiability). Our approach has the advantages of using a generalized, simple, and effective methodology that works for claim categorization on different data sets and tasks.

The coverage and quality of conceptual information contained in lexical semantic resources is crucial for many tasks in natural language processing. Automatic alignment of complementary resources is one way of improving this coverage and quality; however, past attempts have always been between pairs of specific resources. In this paper we establish some set-theoretic conventions for describing concepts and their alignments, and use them to describe a method for automatically constructing n-way alignments from arbitrary pairwise alignments. We apply this technique to the production of a three-way alignment from previously published WordNet-Wikipedia and WordNet-Wiktionary alignments. We then present a quantitative and informal qualitative analysis of the aligned resource. The three-way alignment was found to have greater coverage, an enriched sense representation, and coarser sense granularity than both the original resources and their pairwise alignments, though this came at the cost of accuracy. An evaluation of the induced word sense clusters in a word sense disambiguation task showed that they were no better than random clusters of equivalent granularity. However, use of the alignments to enrich a sense inventory with additional sense glosses did significantly improve the performance of a baseline knowledge-based WSD algorithm.