Spain

This paper describes the submission by the NILC Computational Linguistics research group of the University of São Paulo/Brazil to the English Track 2 (closed sub-track) at the Surface Realisation Shared Task 2020. The success of the current pre-trained models like BERT or GPT-2 in several tasks is well-known, however, this is not the case for data-to-text generation tasks and just recently some initiatives focused on it. This way, we explore how a pre-trained model (GPT-2) performs on the UD-to-text generation task. In general, the achieved results were poor, but there are some interesting ideas to explore. Among the learned lessons we may note that it is necessary to study strategies to represent UD inputs and to introduce structural knowledge into these pre-trained models.

COLING 2020

NILC at SR&#39;20: Exploring Pre-Trained Models in Surface Realisation

stages 

COLING, the International Conference on Computational Linguistics, is one of the premier conferences for natural language processing and computational linguistics. Often grouped within the field of artificial intelligence, but actually pre-dating the development of artificial intelligence, advances in computational linguistics and natural language processing are now some of the major drivers behind the use of artificial intelligence for commercial and social applications – for example, on-line search, machine translation and with voice-assisted conversational devices.

First established in 1965, the biennial COLING conference is held in diverse parts of the globe and attracts participants from both top-ranked research centers and emerging countries. Today, the most important developments in our field are taking place not only in universities and academic research institutes, but also in industrial research departments and in technological startups. COLING conferences provide opportunities for all these communities to showcase their exciting developments.

COLING, the International Conference on Computational Linguistics, is one of the premier conferences for natural language processing and computational linguistics. Often grouped within the field of artificial intelligence, but actually pre-dating the development of artificial intelligence, advances in computational linguistics and natural language processing are now some of the major drivers behind the use of artificial intelligence for commercial and social applications – for example, on-line search, machine translation and with voice-assisted conversational devices.

NILC at SR'20: Exploring Pre-Trained Models in Surface Realisation

technical paper

In the context of Natural Language Generation, surface realization is the task of generating the linear form of a text following a given grammar. Surface realization models usually consist of a cascade of complex sub-modules, either rule-based or neural network-based, each responsible for a specific sub-task. In this work, we show that a single encoder-decoder language model can be used in an end-to-end fashion for all sub-tasks of surface realization. The model is designed based on the BART language model that receives a linear representation of unordered and non-inflected tokens in a sentence along with their corresponding Universal Dependency information and produces the linear sequence of inflected tokens along with the missing words. The model was evaluated on the shallow and deep tracks of the 2020 Surface Realization Shared Task (SR'20) using both human and automatic evaluation. The results indicate that despite its simplicity, our model achieves competitive results among all participants in the shared task.

Surface Realization Using Pretrained Language Models

Gender bias has been identified in many models for Natural Language Processing, stemming from implicit biases in the text corpora used to train the models. Such corpora are too large to closely analyze for biased or stereotypical content. Thus, we argue for a combination of quantitative and qualitative methods, where the quantitative part produces a view of the data of a size suitable for qualitative analysis. We investigate the usefulness of semi-supervised topic modeling for the detection and analysis of gender bias in three corpora (mainstream news articles in English and Swedish, and LGBTQ+ web content in English). We compare differences in topic models for three gender categories (masculine, feminine, and nonbinary or neutral) in each corpus. We find that in all corpora, genders are treated differently and that these differences tend to correspond to hegemonic ideas of gender.

Semi-Supervised Topic Modeling for Gender Bias Discovery in English and Swedish

There is a growing collection of work analyzing and mitigating societal biases in language understanding, generation, and retrieval tasks, though examining biases in creative tasks remains underexplored. Creative language applications are meant for direct interaction with users, so it is important to quantify and mitigate societal biases in these applications. We introduce a novel study on a pipeline to mitigate societal biases when retrieving next verse suggestions in a poetry composition system. Our results suggest that data augmentation through sentiment style transfer has potential for mitigating societal biases.

Investigating Societal Biases in a Poetry Composition System

Misrepresentation of certain communities in datasets is causing big disruptions in artificial intelligence applications. In this presentation, we review the related work on this field, explain the methodology that has been followed and present the results obtained.

Fine-tuning Neural Machine Translation Systems on Gender-Balanced Datasets

Neural Machine Translation (NMT) has been shown to struggle with grammatical gender that is dependent on the gender of human referents, which can cause gender bias effects. Many existing approaches to this problem seek to control gender inflection in the target language by explicitly or implicitly adding a gender feature to the source sentence, usually at the sentence level. 

In this paper we propose schemes for incorporating explicit word-level gender inflection tags into NMT. We explore the potential of this gender-inflection controlled translation when the gender feature can be determined from a human reference, or when a test sentence can be automatically gender-tagged, assessing on English-to-Spanish and English-to-German translation.

We find that simple existing approaches can over-generalize a gender-feature to multiple entities in a sentence, and suggest effective alternatives in the form of tagged coreference adaptation data. We also propose an extension to assess translations of gender-neutral entities from English given a corresponding linguistic convention, such as a non-binary inflection, in the target language. 


Neural Machine Translation Doesn't Translate Gender Coreference Right Unless You Make It

In this paper, we present an approach for sentence-level gender reinflection using linguistically enhanced sequence-to-sequence models. Our system takes an Arabic sentence and a given target gender as input and generates a gender-reinflected sentence based on the target gender. We formulate the problem as a user-aware grammatical error correction task and build an encoder-decoder architecture to jointly model reinflection for both masculine and feminine grammatical genders. We also show that adding linguistic features to our model leads to better reinflection results. The results on a blind test set using our best system show improvements over previous work, with a 3.6% absolute increase in M2 F0.5.

Gender-Aware Reinflectionusing Linguistically Enhanced Neural Models

This presentation introduces a bias-aware methodology we propose for engaging with power relations in natural language processing (NLP) research. NLP research rarely engages with bias in social contexts, limiting its ability to mitigate bias. While researchers have recommended actions, technical methods, and documentation practices, no methodology exists to integrate critical reflections on bias with technical NLP methods. In this presentation we provide an overview of our proposed methodology to mitigate bias in NLP, a focused discussion on one activity of the methodology, and a brief discussion of how we are applying the methodology in our own research on bias in archival metadata descriptions.

Situated Data, Situated Systems: A Methodology to Engage with Power Relations in Natural Language Processing Research

Is it possible to draw a line between workforce statistics and gender bias in contextualized word embeddings? Focusing on BERT (Devlin et al., 2018), we measure gender bias by studying associations between gender-denoting target words and names of professions (Kurita et al., 2019), offering two further perspectives: a comparison with U.S. labor statistics as well as a cross-lingual approach. We mitigate bias by fine-tuning BERT on the GAP corpus (Webster et al., 2018), after applying Counterfactual Data Substitution (CDS; Maudslay et al., 2019). While our method of measuring bias is appropriate for languages such as English, it is not suitable for languages with gender-marking, such as German. Our results highlight the importance of investigating bias and mitigation techniques cross-linguistically and connect large-scale language models to real-world data.

Unmasking Contextual Stereotypes: Measuring and Mitigating BERT'S Gender Bias

Recent years have seen a surge in research on the biases in word embeddings with respect to gender and, to a lesser extent, race. Few of these studies, however, have given attention to the critical intersection of race and gender. In this case study, we analyze the dimensions of gender and race in contextualized word embeddings of given names, taken from BERT, and investigate the nature and nuance of their interaction. We find that these demographic axes, though typically treated as physically and conceptually separate, are in fact interdependent and thus inadvisable to consider in isolation. Further, we show that demographic dimensions predicated on default settings in language, such as in pronouns, may risk rendering groups with multiple marginalized identities invisible. We conclude by discussing the importance and implications of intersectionality for future studies on bias and debiasing in NLP.

Interdependencies of Gender and Race in Contextualized Word Embeddings

Recent research in Natural Language Processing has revealed that word embeddings can encode social biases present in the training data which can affect minorities in real world applications. This paper explores the gender bias implicit in Dutch embeddings while investigating whether English language based approaches can also be used in Dutch. We implement the Word Embeddings Association Test (WEAT), Clustering and Sentence Embeddings Association Test (SEAT) methods to quantify the gender bias in Dutch word embeddings, then we proceed to reduce the bias with Hard-Debias and Sent-Debias mitigation methods and finally we evaluate the performance of the debiased embeddings in downstream tasks. The results suggest that, among others, gender bias is present in traditional and contextualized Dutch word embeddings. We highlight how techniques used to measure and reduce bias created for English can be used in Dutch embeddings by adequately translating the data and taking into account the unique characteristics of the language. Furthermore, we analyze the effect of the debiasing techniques on downstream tasks which show a negligible impact on traditional embeddings and a 2% decrease in performance in contextualized embeddings.

Downloads

Next from COLING 2020

Surface Realization Using Pretrained Language Models

Similar lecture

Reference to Discourse Topics: Introducing “Global” Shell Nouns

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from COLING 2020

Surface Realization Using Pretrained Language Models

Similar lecture

Reference to Discourse Topics: Introducing “Global” Shell Nouns

Downloads