Among Bayesian methods, Monte-Carlo dropout provides principled tools for evaluating the epistemic uncertainty of neural networks. Its popularity recently led to seminal works that proposed activating the dropout layers only during inference for evaluating uncertainty. This approach, which we call dropout injection, provides clear benefits over its traditional counterpart (which we call embedded dropout) since it allows one to obtain a post hoc uncertainty measure for any existing network previously trained without dropout, avoiding an additional, time-consuming training process. Unfortunately, no previous work compared injected and embedded dropout; therefore, we provide the first thorough investigation, focusing on regression problems. The main contribution of our work is to provide guidelines on the effective use of injected dropout so that it can be a practical alternative to the current use of embedded dropout. In particular, we show that its effectiveness strongly relies on a suitable scaling of the corresponding uncertainty measure, and we discuss the trade-off between negative log-likelihood and calibration error as a function of the scale factor. Experimental results on UCI data sets and crowd counting benchmarks support our claim that dropout injection can effectively behave as a competitive post hoc uncertainty quantification technique.

There has been growing interest in applying NLP techniques in the financial domain, however, resources are extremely limited. This paper introduces StockEmotions, a new dataset for detecting emotions in the stock market that consists of 10,000 English comments collected from StockTwits, a financial social media platform. Inspired by behavioral finance, it proposes 12 fine-grained emotion classes that span the roller coaster of investor emotion. Unlike existing financial sentiment datasets, StockEmotions presents granular features such as investor sentiment classes, fine-grained emotions, emojis, and time series data. To demonstrate the usability of the dataset, we perform a dataset analysis and conduct experimental downstream tasks. For financial sentiment/emotion classification tasks, DistilBERT outperforms other baselines, and for multivariate time series forecasting, a Temporal Attention LSTM model combining price index, text, and emotion features achieves the best performance than using a single feature.

We create a mixed-integer optimization (MIO) approach for doing cluster-aware regression, i.e. linear regression that takes into account the inherent clustered structure of the data. We compare to the linear mixed effects regression (LMEM) which is the most used current method, and design simulation experiments to show superior performance to LMEM in terms of both predictive and inferential metrics in silico. Furthermore, we show how our method is formulated in a very interpretable way; LMEM cannot generalize and make cluster-informed predictions when the cluster of new data points is unknown, but we solve this problem by training an interpretable classification tree that can help decide cluster effects for new data points, and demonstrate the power of this generalizability on a real protein expression dataset.

Grammatical forms are said to evolve via two main mechanisms. These are, respectively, the `descent' mechanism, where current forms can be seen to have descended (albeit with occasional modifications) from their roots in ancient languages, and the `contact' mechanism, where evolution in a given language occurs via borrowing from other languages with which it is in contact. We use ideas and concepts from statistical physics to formulate a series of static and dynamical models which illustrate these issues in general terms. The static models emphasise the relative numbers of rules and exceptions, while the dynamical models focus on the emergence of exceptional forms. These unlikely survivors among various competing grammatical forms are winners against the odds. Our analysis suggests that they emerge when the influence of neighbouring languages exceeds the generic tendency towards regularisation within individual languages.

This chapter presents a practical guide for conducting Sentiment Analysis using Natural Language Processing (NLP) techniques in the domain of tick-borne disease text. The aim is to demonstrate the process of how the presence of bias in the discourse surrounding chronic manifestations of the disease can be evaluated. The goal is to use a dataset of 5643 abstracts collected from scientific journals on the topic of chronic Lyme disease to demonstrate using Python, the steps for conducting sentiment analysis using pre-trained language models and the process of validating the preliminary results using both interpretable machine learning tools, as well as a novel methodology of using emerging state-of-the-art large language models like ChatGPT. This serves as a useful resource for researchers and practitioners interested in using NLP techniques for sentiment analysis in the medical domain.

Recent works have shown that selecting an optimal model architecture suited to the differential privacy setting is necessary to achieve the best possible utility for a given privacy budget using differentially private stochastic gradient descent (DP-SGD)(Tramer and Boneh 2020; Cheng et al. 2022). In light of these findings, we empirically analyse how different fairness notions, belonging to distinct classes of statistical fairness criteria (independence, separation and sufficiency), are impacted when one selects a model architecture suitable for DP-SGD, optimized for utility. Using standard datasets from ML fairness literature, we show using a rigorous experimental protocol, that by selecting the optimal model architecture for DP-SGD, the differences across groups concerning the relevant fairness metrics (demographic parity, equalized odds and predictive parity) more often decrease or are negligibly impacted, compared to the non-private baseline, for which optimal model architecture has also been selected to maximize utility. These findings challenge the understanding that differential privacy will necessarily exacerbate unfairness in deep learning models trained on biased datasets.

Prior work in semantic parsing has shown that conventional seq2seq models fail at compositional generalization tasks. This limitation led to a resurgence of methods that model alignments between sentences and their corresponding meaning representations, either implicitly through latent variables or explicitly by taking advantage of alignment annotations. We take the second direction and propose TPOL, a two-step approach that first translates input sentences monotonically and then reorders them to obtain the correct output. This is achieved with a modular framework comprising a Translator and a Reorderer component. We test our approach on two popular semantic parsing datasets. Our experiments show that by means of the monotonic translations, TPOL can learn reliable lexico-logical patterns from aligned data, significantly improving compositional generalization both over conventional seq2seq models, as well as over other approaches that exploit gold alignments.

Speaker embeddings represent a means to extract representative vectorial representations from a speech signal such that the representation pertains to the speaker identity alone. The embeddings are commonly used to classify and discriminate between different speakers. However, there is no objective measure to evaluate the ability of a speaker embedding to disentangle the speaker identity from the other speech characteristics. This means that the embeddings are far from ideal, highly dependent on the training corpus and still include a degree of residual information pertaining to factors such as linguistic content, recording conditions or speaking style of the utterance. This paper introduces an analysis over six sets of speaker embeddings extracted with some of the most recent and high-performing DNN architectures, and in particular, the degree to which they are able to truly disentangle the speaker identity from the speech signal. To correctly evaluate the architectures, a large multi-speaker parallel speech dataset is used. The dataset includes 46 speakers uttering the same set of prompts, recorded in either a professional studio or their home environments. The analysis looks into the intra- and inter-speaker similarity measures computed over the different embedding sets, as well as if simple classification and regression methods are able to extract several residual information factors from the speaker embeddings. The results show that the discriminative power of the analyzed embeddings is very high, yet across all the analyzed architectures, residual information is still present in the representations in the form of a high correlation to the recording conditions, linguistic contents and utterance duration.

Density-based clustering is a commonly used tool in data science. Today many data science works are utilizing high-dimensional neural embeddings. However, traditional density-based clustering techniques like DBSCAN have a degraded performance on high-dimensional data. In this paper, we propose LAF, a generic learned accelerator framework to speed up the original DBSCAN and the sampling-based variants of DBSCAN on high-dimensional data with angular distance metric. This framework consists of a learned cardinality estimator and a post-processing module. The cardinality estimator can fast predict whether a data point is core or not to skip unnecessary range queries, while the post-processing module detects the false negative predictions and merges the falsely separated clusters. The evaluation shows our LAF-enhanced DBSCAN method outperforms the state-of-the-art efficient DBSCAN variants on both efficiency and quality.

The recent explosion of question answering (QA) datasets and models has increased the interest in the generalization of models across multiple domains and formats by either training on multiple datasets or by combining multiple models. Despite the promising results of multi-dataset models, some domains or QA formats may require specific architectures, and thus the adaptability of these models might be limited. In addition, current approaches for combining models disregard cues such as question-answer compatibility. In this work, we propose to combine expert agents with a novel, flexible, and training-efficient architecture that considers questions, answer predictions, and answer-prediction confidence scores to select the best answer among a list of answer candidates. Through quantitative and qualitative experiments we show that our model i) creates a collaboration between agents that outperforms previous multi-agent and multi-dataset approaches in both in-domain and out-of-domain scenarios, ii) is highly data-efficient to train, and iii) can be adapted to any QA format. We release our code and a dataset of answer predictions from expert agents for 16 QA datasets to foster future developments of multi-agent systems on https://github.com/UKPLab/MetaQA.