Bipol: A Novel Multi-Axes Bias Evaluation Metric with Explainability for NLP

We introduce bipol, a new metric with explainability, for estimating social bias in text data. Harmful bias is prevalent in many online sources of data that are used for training machine learning (ML) models. In a step to address this challenge we create a novel metric that involves a two-step process: corpus-level evaluation based on model classification and sentence-level evaluation based on (sensitive) term frequency (TF). After creating new models to detect bias along multiple axes using SotA architectures, we evaluate two popular NLP datasets (COPA and SQUAD). As additional contribution, we created a large dataset (with almost 2 million labelled samples) for training models in bias detection and make it publicly available. We also make public our codes.

Adapting to the Low-Resource Double-Bind: Investigating Low-Compute Methods on Low-Resource African Languages

Many natural language processing (NLP) tasks make use of massively pre-trained language models, which are computationally expensive. However, access to high computational resources added to the issue of data scarcity of African languages constitutes a real barrier to research experiments on these languages. In this work, we explore the applicability of low-compute approaches such as language adapters in the context of this low-resource double-bind. We intend to answer the following question: do language adapters allow those who are doubly bound by data and compute to practically build useful models? Through fine-tuning experiments on African languages, we evaluate their effectiveness as cost-effective approaches to low-resource African NLP. Using solely free compute resources, our results show that language adapters achieve comparable performances to massive pre-trained language models which are heavy on computational resources. This opens the door to further experimentation and exploration on full-extent of language adapters capacities.

Bipol: Multi-axes Evaluation of Bias with Explainability in Benchmark Datasets

We evaluate five English NLP benchmark datasets (available on the superGLUE leaderboard) for bias, along multiple axes. The datasets are the following: Boolean Question (Boolq), CommitmentBank (CB), Winograd Schema Challenge (WSC), Winogender diagnostic (AXg), and Recognising Textual Entailment (RTE). Bias can be harmful and it is known to be common in data, which ML models learn from. In order to mitigate bias in data, it is crucial to be able to estimate it objectively. We use bipol, a novel multi-axes bias metric with explainability, to quantify and explain how much bias exists in these datasets. Multilingual, multi-axes bias evaluation is not very common. Hence, we also contribute a new, large labelled Swedish bias-detection dataset, with about 2 million samples; translated from the English version. In addition, we contribute new multi-axes lexica for bias detection in Swedish. We train a SotA model on the new dataset for bias detection. We make the codes, model, and new dataset publicly available.

Separating Grains from the Chaff: Using Data Filtering to Improve Multilingual Translation for Low-Resourced African Languages

We participated in the WMT 2022 Large-Scale Machine Translation Evaluation for the African Languages Shared Task. This work describes our approach, which is based on filtering the given noisy data using a sentence-pair classifier that was built by fine-tuning a pre-trained language model. To train the classifier, we obtain positive samples (i.e. high-quality parallel sentences) from a gold-standard curated dataset and extract negative samples (i.e. low-quality parallel sentences) from automatically aligned parallel data by choosing sentences with low alignment scores. Our final machine translation model was then trained on filtered data, instead of the entire noisy dataset. We empirically validate our approach by evaluating on two common datasets and show that data filtering generally improves overall translation quality, in some cases even significantly.

T5 for Hate Speech, Augmented Data and Ensemble

We conduct relatively extensive investigations of automatic hate speech (HS) detection using different state-of-the-art (SoTA) baselines over 11 subtasks of 6 different datasets. Our motivation is to determine which of the recent SoTA models is best for automatic hate speech detection and what advantage methods like data augmentation and ensemble may have on the best model, if any. We carry out 6 cross-task investigations. We achieve new SoTA on two subtasks - macro F1 scores of 91.73% and 53.21% for subtasks A and B of the HASOC 2020 dataset, where previous SoTA are 51.52% and 26.52%, respectively. We achieve near-SoTA on two others - macro F1 scores of 81.66% for subtask A of the OLID 2019 dataset and 82.54% for subtask A of the HASOC 2021 dataset, where SoTA are 82.9% and 83.05%, respectively. We perform error analysis and use two explainable artificial intelligence (XAI) algorithms (IG and SHAP) to reveal how two of the models (Bi-LSTM and T5) make the predictions they do by using examples. Other contributions of this work are 1) the introduction of a simple, novel mechanism for correcting out-of-class (OOC) predictions in T5, 2) a detailed description of the data augmentation methods, 3) the revelation of the poor data annotations in the HASOC 2021 dataset by using several examples and XAI (buttressing the need for better quality control), and 4) the public release of our model checkpoints and codes to foster transparency.

GEMv2: Multilingual NLG Benchmarking in a Single Line of Code

Evaluation in machine learning is usually informed by past choices, for example which datasets or metrics to use. This standardization enables the comparison on equal footing using leaderboards, but the evaluation choices become sub-optimal as better alternatives arise. This problem is especially pertinent in natural language generation which requires ever-improving suites of datasets, metrics, and human evaluation to make definitive claims. To make following best model evaluation practices easier, we introduce GEMv2. The new version of the Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers to benefit from each others work. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.

Vector Representations of Idioms in Conversational Systems

We demonstrate, in this study, that an open-domain conversational system trained on idioms or figurative language generates more fitting responses to prompts containing idioms. Idioms are part of everyday speech in many languages, across many cultures, but they pose a great challenge for many Natural Language Processing (NLP) systems that involve tasks such as Information Retrieval (IR) and Machine Translation (MT), besides conversational AI. We utilize the Potential Idiomatic Expression (PIE)-English idioms corpus for the two tasks that we investigate: classification and conversation generation. We achieve state-of-the-art (SoTA) result of 98% macro F1 score on the classification task by using the SoTA T5 model. We experiment with three instances of the SoTA dialogue model, Dialogue Generative Pre-trained Transformer (DialoGPT), for conversation generation. Their performances are evaluated using the automatic metric perplexity and human evaluation. The results show that the model trained on the idiom corpus generates more fitting responses to prompts containing idioms 71.9% of the time, compared to a similar model not trained on the idioms corpus. We contribute the model checkpoint/demo and code on the HuggingFace hub for public access.

State-of-the-art in Open-domain Conversational AI: A Survey

We survey SoTA open-domain conversational AI models with the purpose of presenting the prevailing challenges that still exist to spur future research. In addition, we provide statistics on the gender of conversational AI in order to guide the ethics discussion surrounding the issue. Open-domain conversational AI are known to have several challenges, including bland responses and performance degradation when prompted with figurative language, among others. First, we provide some background by discussing some topics of interest in conversational AI. We then discuss the method applied to the two investigations carried out that make up this study. The first investigation involves a search for recent SoTA open-domain conversational AI models while the second involves the search for 100 conversational AI to assess their gender. Results of the survey show that progress has been made with recent SoTA conversational AI, but there are still persistent challenges that need to be solved, and the female gender is more common than the male for conversational AI. One main take-away is that hybrid models of conversational AI offer more advantages than any single architecture. The key contributions of this survey are 1) the identification of prevailing challenges in SoTA open-domain conversational AI, 2) the unusual discussion about open-domain conversational AI for low-resource languages, and 3) the discussion about the ethics surrounding the gender of conversational AI.

Ìtàkúròso: Exploiting Cross-Lingual Transferability for Natural Language Generation of Dialogues in Low-Resource, African Languages

We investigate the possibility of cross-lingual transfer from a state-of-the-art (SoTA) deep monolingual model (DialoGPT) to 6 African languages and compare with 2 baselines (BlenderBot 90M, another SoTA, and a simple Seq2Seq). The languages are Swahili, Wolof, Hausa, Nigerian Pidgin English, Kinyarwanda & Yor\`ub\'a. Generation of dialogues is known to be a challenging task for many reasons. It becomes more challenging for African languages which are low-resource in terms of data. Therefore, we translate a small portion of the English multi-domain MultiWOZ dataset for each target language. Besides intrinsic evaluation (i.e. perplexity), we conduct human evaluation of single-turn conversations by using majority votes and measure inter-annotator agreement (IAA). The results show that the hypothesis that deep monolingual models learn some abstractions that generalise across languages holds. We observe human-like conversations in 5 out of the 6 languages. It, however, applies to different degrees in different languages, which is expected. The language with the most transferable properties is the Nigerian Pidgin English, with a human-likeness score of 78.1%, of which 34.4% are unanimous. The main contributions of this paper include the representation (through the provision of high-quality dialogue data) of under-represented African languages and demonstrating the cross-lingual transferability hypothesis for dialogue systems. We also provide the datasets and host the model checkpoints/demos on the HuggingFace hub for public access.