arXiv research

Text corpora are widely used resources for measuring societal biases and stereotypes. The common approach to measuring such biases using a corpus is by calculating the similarities between the embedding vector of a word (like nurse) and the vectors of the representative words of the concepts of interest (such as gender…

We combine concepts from random matrix theory and free probability together with ideas from the theory of commutator length in groups and maps from surfaces, and establish new connections between the two. More particularly, we study measures induced by free words on the unitary groups $U(n)$. Every word $w$ in the free…

PCA whitening weighted by Zipfian word frequencies improves task performance.

New measure shows how LSTM models compose hierarchical representations.

The paper explores how word embeddings affect the stability of downstream NLP models.

We adapt the Higher Criticism (HC) goodness-of-fit test to measure the closeness between word-frequency tables. We apply this measure to authorship attribution challenges, where the goal is to identify the author of a document using other documents whose authorship is known. The method is simple yet performs well witho…

Word embedding is a powerful tool in natural language processing. In this paper we consider the problem of word embedding composition \--- given vector representations of two words, compute a vector for the entire phrase. We give a generative model that can capture specific syntactic relations between words. Under our …

Given a measure on the Thurston boundary of Teichmueller space, one can pick a geodesic ray joining some basepoint to a randomly chosen point on the boundary. Different choices of measures may yield typical geodesics with different geometric properties. In particular, we consider two families of measures: the ones whic…

Taxonomies are semantic hierarchies of concepts. One limitation of current taxonomy learning systems is that they define concepts as single words. This position paper argues that contextualized word representations, which recently achieved state-of-the-art results on many competitive NLP tasks, are a promising method t…

Concerns about interpretability, computational resources, and principled inductive priors have motivated efforts to engineer sparse neural models for NLP tasks. If sparsity is important for NLP, might well-trained neural models naturally become roughly sparse? Using the Taxi-Euclidean norm to measure sparsity, we find …

We introduce Probabilistic FastText, a new model for word embeddings that can capture multiple word senses, sub-word structure, and uncertainty information. In particular, we represent each word with a Gaussian mixture density, where the mean of a mixture component is given by the sum of n-grams. This representation al…

Improves document summarization by combining word embeddings and n-grams.

End-to-end automatic speech recognition (ASR) commonly transcribes audio signals into sequences of characters while its performance is evaluated by measuring the word-error rate (WER). This suggests that predicting sequences of words directly may be helpful instead. However, training with word-level supervision can be …

We describe our language-independent unsupervised word sense induction system. This system only uses topic features to cluster different word senses in their global context topic space. Using unlabeled data, this system trains a latent Dirichlet allocation (LDA) topic model then uses it to infer the topics distribution…

The depth of a link measures the minimum height of a resolving tree for the link whose leaves are all unlinks. We show that the depth of the closure of a strictly positive braid word is the length of the word minus the number of distinct letters.

Text classification has become indispensable due to the rapid increase of text in digital form. Over the past three decades, efforts have been made to approach this task using various learning algorithms and statistical models based on bag-of-words (BOW) features. Despite its simple implementation, BOW features lack se…

While the celebrated Word2Vec technique yields semantically rich representations for individual words, there has been relatively less success in extending to generate unsupervised sentences or documents embeddings. Recent work has demonstrated that a distance measure between documents called \emph{Word Mover's Distance…

We investigate the average-case complexity of decision problems for finitely generated groups, in particular the word and membership problems. Using our recent results on ``generic-case complexity'' we show that if a finitely generated group $G$ has the word problem solvable in subexponential time and has a subgroup of…

Embedding words in a vector space has gained a lot of attention in recent years. While state-of-the-art methods provide efficient computation of word similarities via a low-dimensional matrix embedding, their motivation is often left unclear. In this paper, we argue that word embedding can be naturally viewed as a rank…

Word embeddings are commonly obtained as optimizers of a criterion function $f$ of a text corpus, but assessed on word-task performance using a different evaluation function $g$ of the test data. We contend that a possible source of disparity in performance on tasks is the incompatibility between classes of transformat…

The following paper presents a method of comparing two sets of vectors. The method can be applied in all tasks, where it is necessary to measure the closeness of two objects presented as sets of vectors. It may be applicable when we compare the meanings of two sentences as part of the problem of paraphrasing. This is t…

We introduce POLAR - a framework that adds interpretability to pre-trained word embeddings via the adoption of semantic differentials. Semantic differentials are a psychometric construct for measuring the semantics of a word by analysing its position on a scale between two polar opposites (e.g., cold -- hot, soft -- ha…

Improves topic modeling using LLM embeddings and Poisson process.

A large body of research into semantic textual similarity has focused on constructing state-of-the-art embeddings using sophisticated modelling, careful choice of learning signals and many clever tricks. By contrast, little attention has been devoted to similarity measures between these embeddings, with cosine similari…

We bound the value of the Casson invariant of any integral homology 3-sphere $M$ by a constant times the distance-squared to the identity, measured in any word metric on the Torelli group $\T$, of the element of $\T$ associated to any Heegaard splitting of $M$. We construct examples which show this bound is asymptotica…

Detects out-of-distribution sentences in Neural Machine Translation.

Representing a word by its co-occurrences with other words in context is an effective way to capture the meaning of the word. However, the theory behind remains a challenge. In this work, taking the example of a word classification task, we give a theoretical analysis of the approaches that represent a word X by a func…

Improved bounds on moments of word measures on unitary groups.

This paper evaluates debiasing methods on word embeddings to reduce religious bias.

Quadratic-time algorithm computes stretch factors and foliations for pseudo-Anosov mapping classes.

Accelerometer measurements are the prime type of sensor information most think of when seeking to measure physical activity. On the market, there are many fitness measuring devices which aim to track calories burned and steps counted through the use of accelerometers. These measurements, though good enough for the aver…

The power of machine learning systems not only promises great technical progress, but risks societal harm. As a recent example, researchers have shown that popular word embedding algorithms exhibit stereotypical biases, such as gender bias. The widespread use of these algorithms in machine learning systems, from automa…

This paper improves topic modeling by embedding words and topics together.

The paper introduces a model to measure ASR fairness, addressing key issues.

Since the 1970's, physicists and mathematicians who study random matrices in the GUE or GOE models are aware of intriguing connections between integrals of such random matrices and enumeration of graphs on surfaces. We establish a new aspect of this theory: for random matrices sampled from the group $\mathcal{U}\left(n…

We consider the relations between different measures of complexity for free homotopy classes of curves on a surface $Σ$, including the minimum number of self-intersections, the minimum length of the words representing them in a geometric presentation of $π_1(Σ)$, and the minimum degree of the coverings of $Σ$ to which …

The ability to measure similarity between documents enables intelligent summarization and analysis of large corpora. Past distances between documents suffer from either an inability to incorporate semantic similarities between words or from scalability issues. As an alternative, we introduce hierarchical optimal transp…

We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate encoder outputs) and eliminating the redundant vectors. b) determining which word-vectors to eliminate …

Text documents can be described by a number of abstract concepts such as semantic category, writing style, or sentiment. Machine learning (ML) models have been trained to automatically map documents to these abstract concepts, allowing to annotate very large text collections, more than could be processed by a human in …

Compressing word embeddings is important for deploying NLP models in memory-constrained settings. However, understanding what makes compressed embeddings perform well on downstream tasks is challenging---existing measures of compression quality often fail to distinguish between embeddings that perform well and those th…

In topic modeling, many algorithms that guarantee identifiability of the topics have been developed under the premise that there exist anchor words -- i.e., words that only appear (with positive probability) in one topic. Follow-up work has resorted to three or higher-order statistics of the data corpus to relax the an…

Sequence-to-sequence models, such as attention-based models in automatic speech recognition (ASR), are typically trained to optimize the cross-entropy criterion which corresponds to improving the log-likelihood of the data. However, system performance is usually measured in terms of word error rate (WER), not log-likel…

Following the financial crisis of the late 2000s, policy makers have shown considerable interest in monitoring financial stability. Several central banks now publish indices of financial stress, which are essentially based upon market related data. In this paper, we examine the potential for improving the indices by de…

The Manhattan curve connects metrics of hyperbolic groups, showing rigidity.

LIDS assesses LLM summaries with interpretable key words.

Transformer architectures show significant promise for natural language processing. Given that a single pretrained model can be fine-tuned to perform well on many different tasks, these networks appear to extract generally useful linguistic features. A natural question is how such networks represent this information in…

In this paper, we propose a novel approach for measuring the degree of similarity between categories of two pieces of Persian text, which were published as descriptions of two separate advertisements. We built an appropriate dataset for this work using a dataset which consists of advertisements posted on an e-commerce …

We address the problem of speech enhancement generalisation to unseen environments by performing two manipulations. First, we embed an additional recording from the environment alone, and use this embedding to alter activations in the main enhancement subnetwork. Second, we scale the number of noise environments presen…