Then, we describe how hypernymy can be represented in FDS in 3. In 4, we discuss how Functional Distributional Semantics (FDS; Emerson existential and universal quantifications support or and Copestake, 2016; Emerson, 2018) suggests undermine the Distributional Inclusion Hypothesis that the meaning of a word can be modelled as a (DIH), how FDS can handle both quantifications, truth-conditional function, whose parameters can and how FDS models can learn hypernymy under be learnt using the distributional information in a the DIH, and the reverse of it when equipped with corpus (Emerson, 2020a; Lo et al., 2023).

Language models (LMs) excel in in-distribution (ID) scenarios where train and test data are independent and identically distributed. However, their performance often degrades in real-world applications like argument mining. Such degradation happens when new topics emerge, or other text domains and languages become relevant. To assess LMs' generalization abilities in such out-of-distribution (OOD) scenarios, we simulate such distribution shifts by deliberately withholding specific instances for testing, as from the social media domain or the topic Solar Energy. Unlike prior studies focusing on specific shifts and metrics in isolation, we comprehensively analyze OOD generalization. We define three metrics to pinpoint generalization flaws and propose eleven classification tasks covering topic, domain, and language shifts. Overall, we find superior performance of prompt-based fine-tuning, notably when train and test splits primarily differ semantically. Simultaneously, in-context learning is more effective than prompt-based or vanilla fine-tuning for tasks when training data embodies heavy discrepancies in label distribution compared to testing data. This reveals a crucial drawback of gradient-based learning: it biases LMs regarding such structural obstacles.

This thesis focuses on improving the pre-training of natural language models using unsupervised raw data to make them more efficient and aligned with downstream applications. In the first part, we introduce three alternative pre-training objectives to BERT's Masked Language Modeling (MLM), namely Random Token Substitution (RTS), Cluster-based Random Token Substitution (C-RTS), and Swapped Language Modeling (SLM). These objectives involve token swapping instead of masking, with RTS and C-RTS aiming to predict token originality and SLM predicting the original token values. Results show that RTS and C-RTS require less pre-training time while maintaining performance comparable to MLM. Surprisingly, SLM outperforms MLM on certain tasks despite using the same computational budget. In the second part, we proposes self-supervised pre-training tasks that align structurally with downstream applications, reducing the need for labeled data. We use large corpora like Wikipedia and CC-News to train models to recognize if text spans originate from the same paragraph or document in several ways. By doing continuous pre-training, starting from existing models like RoBERTa, ELECTRA, DeBERTa, BART, and T5, we demonstrate significant performance improvements in tasks like Fact Verification, Answer Sentence Selection, and Summarization. These improvements are especially pronounced when limited annotation data is available. The proposed objectives also achieve state-of-the-art results on various benchmark datasets, including FEVER (dev set), ASNQ, WikiQA, and TREC-QA, as well as enhancing the quality of summaries. Importantly, these techniques can be easily integrated with other methods without altering the internal structure of Transformer models, making them versatile for various NLP applications.

Visual object recognition in unseen and cluttered indoor environments is a challenging problem for mobile robots. Toward this goal, we extend our previous work to propose the TOPS2 descriptor, and an accompanying recognition framework, THOR2, inspired by a human reasoning mechanism known as object unity. We interleave color embeddings obtained using the Mapper algorithm for topological soft clustering with the shape-based TOPS descriptor to obtain the TOPS2 descriptor. THOR2, trained using synthetic data, achieves substantially higher recognition accuracy than the shape-based THOR framework and outperforms RGB-D ViT on two real-world datasets: the benchmark OCID dataset and the UW-IS Occluded dataset. Therefore, THOR2 is a promising step toward achieving robust recognition in low-cost robots.

Among the flourishing research of weakly supervised learning (WSL), we recognize the lack of a unified interpretation of the mechanism behind the weakly supervised scenarios, let alone a systematic treatment of the risk rewrite problem, a crucial step in the empirical risk minimization approach. In this paper, we introduce a framework providing a comprehensive understanding and a unified methodology for WSL. The formulation component of the framework, leveraging a contamination perspective, provides a unified interpretation of how weak supervision is formed and subsumes fifteen existing WSL settings. The induced reduction graphs offer comprehensive connections over WSLs. The analysis component of the framework, viewed as a decontamination process, provides a systematic method of conducting risk rewrite. In addition to the conventional inverse matrix approach, we devise a novel strategy called marginal chain aiming to decontaminate distributions. We justify the feasibility of the proposed framework by recovering existing rewrites reported in the literature.

Multilingual semantic search is the task of retrieving relevant contents to a query expressed in different language combinations. This requires a better semantic understanding of the user's intent and its contextual meaning. Multilingual semantic search is less explored and more challenging than its monolingual or bilingual counterparts, due to the lack of multilingual parallel resources for this task and the need to circumvent "language bias". In this work, we propose an alignment approach: MAML-Align, specifically for low-resource scenarios. Our approach leverages meta-distillation learning based on MAML, an optimization-based Model-Agnostic Meta-Learner. MAML-Align distills knowledge from a Teacher meta-transfer model T-MAML, specialized in transferring from monolingual to bilingual semantic search, to a Student model S-MAML, which meta-transfers from bilingual to multilingual semantic search. To the best of our knowledge, we are the first to extend meta-distillation to a multilingual search application. Our empirical results show that on top of a strong baseline based on sentence transformers, our meta-distillation approach boosts the gains provided by MAML and significantly outperforms naive fine-tuning methods. Furthermore, multilingual meta-distillation learning improves generalization even to unseen languages.

In this paper we present the first investigation into the effectiveness of Large Language Models (LLMs) for Failure Mode Classification (FMC). FMC, the task of automatically labelling an observation with a corresponding failure mode code, is a critical task in the maintenance domain as it reduces the need for reliability engineers to spend their time manually analysing work orders. We detail our approach to prompt engineering to enable an LLM to predict the failure mode of a given observation using a restricted code list. We demonstrate that the performance of a GPT-3.5 model (F1=0.80) fine-tuned on annotated data is a significant improvement over a currently available text classification model (F1=0.60) trained on the same annotated data set. The fine-tuned model also outperforms the out-of-the box GPT-3.5 (F1=0.46). This investigation reinforces the need for high quality fine-tuning data sets for domain-specific tasks using LLMs.

Due to the imbalanced nature of networked observational data, the causal effect predictions for some individuals can severely violate the positivity/overlap assumption, rendering unreliable estimations. Nevertheless, this potential risk of individual-level treatment effect estimation on networked data has been largely under-explored. To create a more trustworthy causal effect estimator, we propose the uncertainty-aware graph deep kernel learning (GraphDKL) framework with Lipschitz constraint to model the prediction uncertainty with Gaussian process and identify unreliable estimations. To the best of our knowledge, GraphDKL is the first framework to tackle the violation of positivity assumption when performing causal effect estimation with graphs. With extensive experiments, we demonstrate the superiority of our proposed method in uncertainty-aware causal effect estimation on networked data.

The need for improved network situational awareness has been highlighted by the growing complexity and severity of cyber-attacks. Mobile phones pose a significant risk to network situational awareness due to their dynamic behaviour and lack of visibility on a network. Machine learning techniques enhance situational awareness by providing administrators insight into the devices and activities which form their network. Developing machine learning techniques for situational awareness requires a testbed to generate and label network traffic. Current testbeds, however, are unable to automate the generation and labelling of realistic network traffic. To address this, we describe a testbed which automates applications on mobile devices to generate and label realistic traffic. From this testbed, two labelled datasets of network traffic have been created. We provide an analysis of the testbed automation reliability and benchmark the datasets for the task of application classification.

In recent years, large language models (LLMs) have shown remarkable capabilities at scale, particularly at generating text conditioned on a prompt. In our work, we investigate the use of LLMs to augment training data of small language models~(SLMs) with automatically generated counterfactual~(CF) instances -- i.e. minimally altered inputs -- in order to improve out-of-domain~(OOD) performance of SLMs in the extractive question answering~(QA) setup. We show that, across various LLM generators, such data augmentation consistently enhances OOD performance and improves model calibration for both confidence-based and rationale-augmented calibrator models. Furthermore, these performance improvements correlate with higher diversity of CF instances in terms of their surface form and semantic content. Finally, we show that CF augmented models which are easier to calibrate also exhibit much lower entropy when assigning importance, indicating that rationale-augmented calibrators prefer concise explanations.