MaiNLP at SemEval-2024 Task 1: Analyzing Source Language Selection in Cross-Lingual Textual Relatedness

This paper presents our system developed for the SemEval-2024 Task 1: Semantic Textual Relatedness (STR), on Track C: Cross-lingual. The task aims to detect semantic relatedness of two sentences in a given target language without access to direct supervision (i.e. zero-shot cross-lingual transfer). To this end, we focus on different source language selection strategies on two different pre-trained languages models: XLM-R and Furina. We experiment with 1) single-source transfer and select source languages based on typological similarity, 2) augmenting English training data with the two nearest-neighbor source languages, and 3) multi-source transfer where we compare selecting on all training languages against languages from the same family. We further study machine translation-based data augmentation and the impact of script differences. Our submission achieved the first place in the C8 (Kinyarwanda) test set.

Vicinal Risk Minimization for Few-Shot Cross-lingual Transfer in Abusive Language Detection

Cross-lingual transfer learning from high-resource to medium and low-resource languages has shown encouraging results. However, the scarcity of resources in target languages remains a challenge. In this work, we resort to data augmentation and continual pre-training for domain adaptation to improve cross-lingual abusive language detection. For data augmentation, we analyze two existing techniques based on vicinal risk minimization and propose MIXAG, a novel data augmentation method which interpolates pairs of instances based on the angle of their representations. Our experiments involve seven languages typologically distinct from English and three different domains. The results reveal that the data augmentation strategies can enhance few-shot cross-lingual abusive language detection. Specifically, we observe that consistently in all target languages, MIXAG improves significantly in multidomain and multilingual environments. Finally, we show through an error analysis how the domain adaptation can favour the class of abusive texts (reducing false negatives), but at the same time, declines the precision of the abusive language detection model.

Establishing Trustworthiness: Rethinking Tasks and Model Evaluation

Language understanding is a multi-faceted cognitive capability, which the Natural Language Processing (NLP) community has striven to model computationally for decades. Traditionally, facets of linguistic intelligence have been compartmentalized into tasks with specialized model architectures and corresponding evaluation protocols. With the advent of large language models (LLMs) the community has witnessed a dramatic shift towards general purpose, task-agnostic approaches powered by generative models. As a consequence, the traditional compartmentalized notion of language tasks is breaking down, followed by an increasing challenge for evaluation and analysis. At the same time, LLMs are being deployed in more real-world scenarios, including previously unforeseen zero-shot setups, increasing the need for trustworthy and reliable systems. Therefore, we argue that it is time to rethink what constitutes tasks and model evaluation in NLP, and pursue a more holistic view on language, placing trustworthiness at the center. Towards this goal, we review existing compartmentalized approaches for understanding the origins of a model's functional capacity, and provide recommendations for more multi-faceted evaluation protocols.

Donkii: Can Annotation Error Detection Methods Find Errors in Instruction-Tuning Datasets?

Instruction-tuning has become an integral part of training pipelines for Large Language Models (LLMs) and has been shown to yield strong performance gains. In an orthogonal line of research, Annotation Error Detection (AED) has emerged as a tool for detecting quality issues of gold-standard labels. But so far, the application of AED methods is limited to discriminative settings. It is an open question how well AED methods generalize to generative settings which are becoming widespread via generative LLMs. In this work, we present a first and new benchmark for AED on instruction-tuning data: Donkii. It encompasses three instruction-tuning datasets enriched with annotations by experts and semi-automatic methods. We find that all three datasets contain clear-cut errors that sometimes directly propagate into instruction-tuned LLMs. We propose four AED baselines for the generative setting and evaluate them comprehensively on the newly introduced dataset. Our results demonstrate that choosing the right AED method and model size is indeed crucial, thereby deriving practical recommendations. To gain insights, we provide a first case-study to examine how the quality of the instruction-tuning datasets influences downstream performance.

A General-Purpose Multilingual Document Encoder

Massively multilingual pretrained transformers (MMTs) have tremendously pushed the state of the art on multilingual NLP and cross-lingual transfer of NLP models in particular. While a large body of work leveraged MMTs to mine parallel data and induce bilingual document embeddings, much less effort has been devoted to training general-purpose (massively) multilingual document encoder that can be used for both supervised and unsupervised document-level tasks. In this work, we pretrain a massively multilingual document encoder as a hierarchical transformer model (HMDE) in which a shallow document transformer contextualizes sentence representations produced by a state-of-the-art pretrained multilingual sentence encoder. We leverage Wikipedia as a readily available source of comparable documents for creating training data, and train HMDE by means of a cross-lingual contrastive objective, further exploiting the category hierarchy of Wikipedia for creation of difficult negatives. We evaluate the effectiveness of HMDE in two arguably most common and prominent cross-lingual document-level tasks: (1) cross-lingual transfer for topical document classification and (2) cross-lingual document retrieval. HMDE is significantly more effective than (i) aggregations of segment-based representations and (ii) multilingual Longformer. Crucially, owing to its massively multilingual lower transformer, HMDE successfully generalizes to languages unseen in document-level pretraining. We publicly release our code and models at https://github.com/ogaloglu/pre-training-multilingual-document-encoders .

Boosting Zero-shot Cross-lingual Retrieval by Training on Artificially Code-Switched Data

Transferring information retrieval (IR) models from a high-resource language (typically English) to other languages in a zero-shot fashion has become a widely adopted approach. In this work, we show that the effectiveness of zero-shot rankers diminishes when queries and documents are present in different languages. Motivated by this, we propose to train ranking models on artificially code-switched data instead, which we generate by utilizing bilingual lexicons. To this end, we experiment with lexicons induced from (1) cross-lingual word embeddings and (2) parallel Wikipedia page titles. We use the mMARCO dataset to extensively evaluate reranking models on 36 language pairs spanning Monolingual IR (MoIR), Cross-lingual IR (CLIR), and Multilingual IR (MLIR). Our results show that code-switching can yield consistent and substantial gains of 5.1 MRR@10 in CLIR and 3.9 MRR@10 in MLIR, while maintaining stable performance in MoIR. Encouragingly, the gains are especially pronounced for distant languages (up to 2x absolute gain). We further show that our approach is robust towards the ratio of code-switched tokens and also extends to unseen languages. Our results demonstrate that training on code-switched data is a cheap and effective way of generalizing zero-shot rankers for cross-lingual and multilingual retrieval.

ZusammenQA: Data Augmentation with Specialized Models for Cross-lingual Open-retrieval Question Answering System

This paper introduces our proposed system for the MIA Shared Task on Cross-lingual Open-retrieval Question Answering (COQA). In this challenging scenario, given an input question the system has to gather evidence documents from a multilingual pool and generate from them an answer in the language of the question. We devised several approaches combining different model variants for three main components: Data Augmentation, Passage Retrieval, and Answer Generation. For passage retrieval, we evaluated the monolingual BM25 ranker against the ensemble of re-rankers based on multilingual pretrained language models (PLMs) and also variants of the shared task baseline, re-training it from scratch using a recently introduced contrastive loss that maintains a strong gradient signal throughout training by means of mixed negative samples. For answer generation, we focused on language- and domain-specialization by means of continued language model (LM) pretraining of existing multilingual encoders. Additionally, for both passage retrieval and answer generation, we augmented the training data provided by the task organizers with automatically generated question-answer pairs created from Wikipedia passages to mitigate the issue of data scarcity, particularly for the low-resource languages for which no training data were provided. Our results show that language- and domain-specialization as well as data augmentation help, especially for low-resource languages.

Parameter-Efficient Neural Reranking for Cross-Lingual and Multilingual Retrieval

State-of-the-art neural (re)rankers are notoriously data hungry which - given the lack of large-scale training data in languages other than English - makes them rarely used in multilingual and cross-lingual retrieval settings. Current approaches therefore typically transfer rankers trained on English data to other languages and cross-lingual setups by means of multilingual encoders: they fine-tune all the parameters of a pretrained massively multilingual Transformer (MMT, e.g., multilingual BERT) on English relevance judgments and then deploy it in the target language. In this work, we show that two parameter-efficient approaches to cross-lingual transfer, namely Sparse Fine-Tuning Masks (SFTMs) and Adapters, allow for a more lightweight and more effective zero-shot transfer to multilingual and cross-lingual retrieval tasks. We first train language adapters (or SFTMs) via Masked Language Modelling and then train retrieval (i.e., reranking) adapters (SFTMs) on top while keeping all other parameters fixed. At inference, this modular design allows us to compose the ranker by applying the task adapter (or SFTM) trained with source language data together with the language adapter (or SFTM) of a target language. Besides improved transfer performance, these two approaches offer faster ranker training, with only a fraction of parameters being updated compared to full MMT fine-tuning. We benchmark our models on the CLEF-2003 benchmark, showing that our parameter-efficient methods outperform standard zero-shot transfer with full MMT fine-tuning, while enabling modularity and reducing training times. Further, we show on the example of Swahili and Somali that, for low(er)-resource languages, our parameter-efficient neural re-rankers can improve the ranking of the competitive machine translation-based ranker.

On Cross-Lingual Retrieval with Multilingual Text Encoders

In this work we present a systematic empirical study focused on the suitability of the state-of-the-art multilingual encoders for cross-lingual document and sentence retrieval tasks across a number of diverse language pairs. We first treat these models as multilingual text encoders and benchmark their performance in unsupervised ad-hoc sentence- and document-level CLIR. In contrast to supervised language understanding, our results indicate that for unsupervised document-level CLIR -- a setup with no relevance judgments for IR-specific fine-tuning -- pretrained multilingual encoders on average fail to significantly outperform earlier models based on CLWEs. For sentence-level retrieval, we do obtain state-of-the-art performance: the peak scores, however, are met by multilingual encoders that have been further specialized, in a supervised fashion, for sentence understanding tasks, rather than using their vanilla 'off-the-shelf' variants. Following these results, we introduce localized relevance matching for document-level CLIR, where we independently score a query against document sections. In the second part, we evaluate multilingual encoders fine-tuned in a supervised fashion (i.e., we learn to rank) on English relevance data in a series of zero-shot language and domain transfer CLIR experiments. Our results show that supervised re-ranking rarely improves the performance of multilingual transformers as unsupervised base rankers. Finally, only with in-domain contrastive fine-tuning (i.e., same domain, only language transfer), we manage to improve the ranking quality. We uncover substantial empirical differences between cross-lingual retrieval results and results of (zero-shot) cross-lingual transfer for monolingual retrieval in target languages, which point to "monolingual overfitting" of retrieval models trained on monolingual data.