Investigating Numeracy Learning Ability of a Text-to-Text Transfer Model

The transformer-based pre-trained language models have been tremendously successful in most of the conventional NLP tasks. But they often struggle in those tasks where numerical understanding is required. Some possible reasons can be the tokenizers and pre-training objectives which are not specifically designed to learn and preserve numeracy. Here we investigate the ability of text-to-text transfer learning model (T5), which has outperformed its predecessors in the conventional NLP tasks, to learn numeracy. We consider four numeracy tasks: numeration, magnitude order prediction, finding minimum and maximum in a series, and sorting. We find that, although T5 models perform reasonably well in the interpolation setting, they struggle considerably in the extrapolation setting across all four tasks.

Probing Pre-trained Language Models for Semantic Attributes and their Values

Pretrained language models (PTLMs) yield state-of-the-art performance on many natural language processing tasks, including syntax, semantics and commonsense. In this paper, we focus on identifying to what extent do PTLMs capture semantic attributes and their values, e.g., the correlation between rich and high net worth. We use PTLMs to predict masked tokens using patterns and lists of items from Wikidata in order to verify how likely PTLMs encode semantic attributes along with their values. Such inferences based on semantics are intuitive for humans as part of our language understanding. Since PTLMs are trained on large amount of Wikipedia data we would assume that they can generate similar predictions, yet our findings reveal that PTLMs are still much worse than humans on this task. We show evidence and analysis explaining how to exploit our methodology to integrate better context and semantics into PTLMs using knowledge bases.

Shaking Syntactic Trees on the Sesame Street: Multilingual Probing with Controllable Perturbations

Recent research has adopted a new experimental field centered around the concept of text perturbations which has revealed that shuffled word order has little to no impact on the downstream performance of Transformer-based language models across many NLP tasks. These findings contradict the common understanding of how the models encode hierarchical and structural information and even question if the word order is modeled with position embeddings. To this end, this paper proposes nine probing datasets organized by the type of controllable text perturbation for three Indo-European languages with a varying degree of word order flexibility: English, Swedish and Russian. Based on the probing analysis of the M-BERT and M-BART models, we report that the syntactic sensitivity depends on the language and model pre-training objectives. We also find that the sensitivity grows across layers together with the increase of the perturbation granularity. Last but not least, we show that the models barely use the positional information to induce syntactic trees from their intermediate self-attention and contextualized representations.

Exploring the Role of BERT Token Representations to Explain Sentence Probing Results

Several studies have been carried out on revealing linguistic features captured by BERT. This is usually achieved by training a diagnostic classifier on the representations obtained from different layers of BERT. The subsequent classification accuracy is then interpreted as the ability of the model in encoding the corresponding linguistic property. Despite providing insights, these studies have left out the potential role of token representations. In this paper, we provide a more in-depth analysis on the representation space of BERT in search for distinct and meaningful subspaces that can explain the reasons behind these probing results. Based on a set of probing tasks and with the help of attribution methods we show that BERT tends to encode meaningful knowledge in specific token representations (which are often ignored in standard classification setups), allowing the model to detect syntactic and semantic abnormalities, and to distinctively separate grammatical number and tense subspaces.

Probing Across Time: What Does RoBERTa Know and When?

Models of language trained on very large corpora have been demonstrated useful for natural language processing. As fixed artifacts, they have become the object of intense study, with many researchers probing'' the extent to which they acquire and readily demonstrate linguistic abstractions, factual and commonsense knowledge, and reasoning abilities. Recent work applied several probes to intermediate training stages to observe the developmental process of a large-scale model (Chiang et al., 2020). Following this effort, we systematically answer a question: for various types of knowledge a language model learns, when during (pre)training are they acquired? Using RoBERTa as a case study, we find: linguistic knowledge is acquired fast, stably, and robustly across domains. Facts and commonsense are slower and more domain-sensitive. Reasoning abilities are, in general, not stably acquired. As new datasets, pretraining protocols, and probes emerge, we believe that probing-across-time analyses can help researchers understand the complex, intermingled learning that these models undergo and guide us toward more efficient approaches that accomplish necessary learning faster.

Investigating Post-pretraining Representation Alignment for Cross-Lingual Question Answering

Human knowledge is collectively encoded in the roughly 6500 languages spoken around the world, but it is not distributed equally across languages. Hence, for information-seeking question answering (QA) systems to adequately serve speakers of all languages, they need to operate cross-lingually. In this work we investigate the capabilities of multilingually pretrained language models on cross-lingual QA. We find that explicitly aligning the representations across languages with a post-hoc finetuning step generally leads to improved performance. We additionally investigate the effect of data size as well as the language choice in this fine-tuning step, also releasing a dataset for evaluating cross-lingual QA systems.

Investigating Dominant Word Order on Universal Dependencies with Graph Rewriting

This paper details experiments we performed on the Universal Dependencies 2.7 corpora in order to investigate the dominant word order in the available languages. For this purpose, we used a graph rewriting tool, GREW, which allowed us to go beyond the surface annotations and identify the implicit subjects. We first measured the distribution of the six different word orders (SVO, SOV, VSO, VOS, OVS, OSV) in the corpora and investigated when there was a significant difference in the corpora within a given language. Then, we compared the obtained results with information provided in the WALS database (Dryer and Haspelmath, 2013) and in ( ̈Ostling, 2015). Finally, we examined the impact of using a graph rewriting tool for this task. The tools and resources used for this research are all freely available.

``Don't discuss'': Investigating Semantic and Argumentative Features for Supervised Propagandist Message Detection and Classification

One of the mechanisms through which disinformation is spreading online, in particular through social media, is by employing propaganda techniques. These include specific rhetorical and psychological strategies, ranging from leveraging on emotions to exploiting logical fallacies. In this paper, our goal is to push forward research on propaganda detection based on text analysis, given the crucial role these methods may play to address this main societal issue. More precisely, we propose a supervised approach to classify textual snippets both as propaganda messages and according to the precise applied propaganda technique, as well as a detailed linguistic analysis of the features characterising propaganda information in text (e.g., semantic, sentiment and argumentation features). Extensive experiments conducted on two available propagandist resources (i.e., NLP4IF'19 and SemEval'20-Task 11 datasets) show that the proposed approach, leveraging different language models and the investigated linguistic features, achieves very promising results on propaganda classification, both at sentence- and at fragment-level.

Probing Multilingual Language Models for Discourse

Pre-trained multilingual language models have become an important building block in multilingual Natural Language Processing. In the present paper, we investigate a range of such models to find out how well they transfer discourse-level knowledge across languages. This is done with a systematic evaluation on a broader set of discourse-level tasks than has been previously been assembled. We find that the XLM-RoBERTa family of models consistently show the best performance, by simultaneously being good monolingual models and degrading relatively little in a zero-shot setting. Our results also indicate that model distillation may hurt the ability of cross-lingual transfer of sentence representations, while language dissimilarity at most has a modest effect. We hope that our test suite, covering 5 tasks with a total of 22 languages in 10 distinct families, will serve as a useful evaluation platform for multilingual performance at and beyond the sentence level.

Investigating Active Learning in Interactive Neural Machine Translation

Interactive-predictive translation is a collaborative iterative process and where human translators produce translations with the help of machine translation (MT) systems interactively. Various sampling techniques in active learning (AL) exist to update the neural MT (NMT) model in the interactive-predictive scenario. In this paper and we explore term based (named entity count (NEC)) and quality based (quality estimation (QE) and sentence similarity (Sim)) sampling techniques -- which are used to find the ideal candidates from the incoming data -- for human supervision and MT model's weight updation. We carried out experiments with three language pairs and viz. German-English and Spanish-English and Hindi-English. Our proposed sampling technique yields 1.82 and 0.77 and 0.81 BLEU points improvements for German-English and Spanish-English and Hindi-English and respectively and over random sampling based baseline. It also improves the present state-of-the-art by 0.35 and 0.12 BLEU points for German-English and Spanish-English and respectively. Human editing effort in terms of number-of-words-changed also improves by 5 and 4 points for German-English and Spanish-English and respectively and compared to the state-of-the-art.