The frequencies at which individual words occur across languages follow power law distributions, a pattern of findings known as Zipf's law. A vast literature argues over whether this serves to optimize the efficiency of human communication, however this claim is necessarily post hoc, and it has been suggested that Zipf's law may in fact describe mixtures of other distributions. From this perspective, recent findings that Sinosphere first (family) names are geometrically distributed are notable, because this is actually consistent with information theoretic predictions regarding optimal coding. First names form natural communicative distributions in most languages, and I show that when analyzed in relation to the communities in which they are used, first name distributions across a diverse set of languages are both geometric and, historically, remarkably similar, with power law distributions only emerging when empirical distributions are aggregated. I then show this pattern of findings replicates in communicative distributions of English nouns and verbs. These results indicate that if lexical distributions support efficient communication, they do so because their functional structures directly satisfy the constraints described by information theory, and not because of Zipf's law. Understanding the function of these information structures is likely to be key to explaining humankind's remarkable communicative capacities.
Graduate School of Information Science and Technology（信息科学与技术研究生院）；University of Tokyo（东京大学）
Word frequency is assumed to correlate with word familiarity, but the strength of this correlation has not been thoroughly investigated. In this paper, we report on our analysis of the correlation between a word familiarity rating list obtained through a psycholinguistic experiment and the log-frequency obtained from various corpora of different kinds and sizes (up to the terabyte scale) for English and Japanese. Major findings are threefold: First, for a given corpus, familiarity is necessary for a word to achieve high frequency, but familiar words are not necessarily frequent. Second, correl…
Investigating the Frequency Distortion of Word Embeddings and Its Impact on Bias Metrics
Instituto de Investigación en Ciencias de la Computación, CONICET-UBA（CONICET-布宜诺斯艾利斯大学计算机科学研究所）；Universidad de Buenos Aires (UBA)（布宜诺斯艾利斯大学（UBA））；Departamento de Computación, FCEyN, UBA（布宜诺斯艾利斯大学理学院计算机系）
How does word frequency in pre-training data affect the behavior of similarity metrics in contextualized BERT embeddings? Are there systematic ways in which some word relationships are exaggerated or understated? In this work, we explore the geometric characteristics of contextualized word embeddings with two novel tools: (1) an identity probe that predicts the identity of a word using its embedding; (2) the minimal bounding sphere for a word's contextualized representations. Our results reveal that words of high and low frequency differ significantly with respect to their representational geo…
Word meaning change can be inferred from drifts of time-varying word embeddings. However, temporal data may be too sparse to build robust word embeddings and to discriminate significant drifts from noise. In this paper, we compare three models to learn diachronic word embeddings on scarce data: incremental updating of a Skip-Gram from Kim et al. (2014), dynamic filtering from Bamler and Mandt (2017), and dynamic Bernoulli embeddings from Rudolph and Blei (2018). In particular, we study the performance of different initialisation schemes and emphasise what characteristics of each model are more…
National Institute of Information and Communications Technology（国立信息与通信技术研究所）
Let $f (\cdot)$ be the absolute frequency of words and $r$ be the rank of words in decreasing order of frequency, then the following function can fit the rank-frequency relation \[ f (r;s,t) = \left(\frac{r_{\tt max}}{r}\right)^{1-s} \left(\frac{r_{\tt max}+t \cdot r_{\tt exp}}{r+t \cdot r_{\tt exp}}\right)^{1+(1+t)s} \] where $r_{\tt max}$ and $r_{\tt exp}$ are the maximum and the expectation of the rank, respectively; $s>0$ and $t>0$ are parameters estimated from data. On well-behaved data, there should be $s<1$ and $s \cdot t < 1$.
University College London（伦敦大学学院）；Tilburg University（蒂尔堡大学）
Recent efforts have shown that neural text processing models are vulnerable to adversarial examples, but the nature of these examples is poorly understood. In this work, we show that adversarial attacks against CNN, LSTM and Transformer-based classification models perform word substitutions that are identifiable through frequency differences between replaced words and their corresponding substitutions. Based on these findings, we propose frequency-guided word substitutions (FGWS), a simple algorithm exploiting the frequency properties of adversarial word substitutions for the detection of adve…
Instituto de Física Enrique Gaviola, Consejo Nacional de Investigaciones Científicas y Técnicas（恩里克·加维奥拉物理研究所，国家科学技术研究委员会）；Universidad Nacional de Córdoba（科尔多瓦国立大学）；Centro Atómico Bariloche（巴里洛切原子能中心）；Instituto Balseiro（巴尔塞iro研究所）；Comisión Nacional de Energía Atómica（国家原子能委员会）；Universidad Nacional de Cuyo（库约国立大学）；Consejo Nacional de Investigaciones Científicas y Técnicas（国家科学技术研究委员会）
We analyze the frequency-rank relationship in sub-vocabularies corresponding to three different grammatical classes (nouns, verbs, and others) in a collection of literary works in English, whose words have been automatically tagged according to their grammatical role. Comparing with a null hypothesis which assumes that words belonging to each class are uniformly distributed across the frequency-ranked vocabulary of the whole work, we disclose statistically significant differences between the three classes. This results point to the fact that frequency-rank relationships may reflect linguistic …
National Engineering Research Center for Software Engineering, Peking University（北京大学国家软件工程研究中心）；Alibaba Group（阿里巴巴集团）
Low-frequency word prediction remains a challenge in modern neural machine translation (NMT) systems. Recent adaptive training methods promote the output of infrequent words by emphasizing their weights in the overall training objectives. Despite the improved recall of low-frequency words, their prediction precision is unexpectedly hindered by the adaptive objectives. Inspired by the observation that low-frequency words form a more compact embedding space, we tackle this challenge from a representation learning perspective. Specifically, we propose a frequency-aware token-level contrastive lea…
In this paper we study how different ways of combining character and word-level representations affect the quality of both final word and sentence representations. We provide strong empirical evidence that modeling characters improves the learned representations at the word and sentence levels, and that doing so is particularly useful when representing less frequent words. We further show that a feature-wise sigmoid gating mechanism is a robust method for creating representations that encode semantic similarity, as it performed reasonably well in several word similarity datasets. Finally, our …
Laboratoire d'Informatique de Grenoble, Univ. Grenoble Alpes (UGA)（格勒诺布尔阿尔卑斯大学格勒诺布尔信息学实验室）；School of Computer Science and Electronic Engineering, University of Essex（埃塞克斯大学计算机科学与电子工程学院）；Institute of Informatics, Federal University of Rio Grande do Sul（南里奥格兰德联邦大学信息学研究所）
Since Bahdanau et al. [1] first introduced attention for neural machine translation, most sequence-to-sequence models made use of attention mechanisms [2, 3, 4]. While they produce soft-alignment matrices that could be interpreted as alignment between target and source languages, we lack metrics to quantify their quality, being unclear which approach produces the best alignments. This paper presents an empirical evaluation of 3 main sequence-to-sequence models (CNN, RNN and Transformer-based) for word discovery from unsegmented phoneme sequences. This task consists in aligning word sequences i…
Institute of Informatics, Federal University of Rio Grande do Sul（联邦南里奥格兰德大学信息学院）；School of Computer Science and Electronic Engineering, University of Essex（埃塞克斯大学计算机科学与电子工程学院）
Increasing the capacity of recurrent neural networks (RNN) usually involves augmenting the size of the hidden layer, with significant increase of computational cost. Recurrent neural tensor networks (RNTN) increase capacity using distinct hidden layer weights for each word, but with greater costs in memory usage. In this paper, we introduce restricted recurrent neural tensor networks (r-RNTN) which reserve distinct hidden layer weights for frequent vocabulary words while sharing a single set of weights for infrequent words. Perplexity evaluations show that for fixed hidden layer sizes, r-RNTNs…
Kata Research Team（Kata研究团队）；Fondazione Bruno Kessler（布鲁诺·凯斯勒基金会）；University of Trento（特伦托大学）
Despite the long history of named-entity recognition (NER) task in the natural language processing community, previous work rarely studied the task on conversational texts. Such texts are challenging because they contain a lot of word variations which increase the number of out-of-vocabulary (OOV) words. The high number of OOV words poses a difficulty for word-based neural models. Meanwhile, there is plenty of evidence to the effectiveness of character-based neural models in mitigating this OOV problem. We report an empirical evaluation of neural sequence labeling models with character embeddi…
ICC (UBA-CONICET)（ICC（布宜诺斯艾利斯大学-国家科学技术研究委员会））；Maestría en Data Mining (UBA)（数据挖掘硕士项目（布宜诺斯艾利斯大学））；Escuela IDAES (UNSAM)（IDAES学院（圣马丁将军大学））
Numerous works use word embedding-based metrics to quantify societal biases and stereotypes in texts. Recent studies have found that word embeddings can capture semantic similarity but may be affected by word frequency. In this work we study the effect of frequency when measuring female vs. male gender bias with word embedding-based bias quantification methods. We find that Skip-gram with negative sampling and GloVe tend to detect male bias in high frequency words, while GloVe tends to return female bias in low frequency words. We show these behaviors still exist when words are randomly shuffl…
This work investigates the role of factors like training method, training corpus size and thematic relevance of texts in the performance of word embedding features on sentiment analysis of tweets, song lyrics, movie reviews and item reviews. We also explore specific training or post-processing methods that can be used to enhance the performance of word embeddings in certain tasks or domains. Our empirical observations indicate that models trained with multithematic texts that are large and rich in vocabulary are the best in answering syntactic and semantic word analogy questions. We further ob…
Stanford University（斯坦福大学）；University of Michigan（密歇根大学）
Cosine similarity of contextual embeddings is used in many NLP tasks (e.g., QA, IR, MT) and metrics (e.g., BERTScore). Here, we uncover systematic ways in which word similarities estimated by cosine over BERT embeddings are understated and trace this effect to training data frequency. We find that relative to human judgements, cosine similarity underestimates the similarity of frequent words with other instances of the same word or other words across contexts, even after controlling for polysemy and other factors. We conjecture that this underestimation of similarity for high frequency words i…
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）
Chinese input recommendation plays an important role in alleviating human cost in typing Chinese words, especially in the scenario of mobile applications. The fundamental problem is to predict the conditional probability of the next word given the sequence of previous words. Therefore, statistical language models, i.e.~n-grams based models, have been extensively used on this task in real application. However, the characteristics of extremely different typing behaviors usually lead to serious sparsity problem, even n-gram with smoothing will fail. A reasonable approach to tackle this problem is…
Universidad de Buenos Aires（布宜诺斯艾利斯大学）；CONICET（阿根廷国家科学与技术研究理事会）；Academia Argentina de Letras（阿根廷文学院）
The task of detecting regionalisms (expressions or words used in certain regions) has traditionally relied on the use of questionnaires and surveys, and has also heavily depended on the expertise and intuition of the surveyor. The irruption of Social Media and its microblogging services has produced an unprecedented wealth of content, mainly informal text generated by users, opening new opportunities for linguists to extend their studies of language variation. Previous work on automatic detection of regionalisms depended mostly on word frequencies. In this work, we present a novel metric based…
Neural language models learn, to varying degrees of accuracy, the grammatical properties of natural languages. In this work, we investigate whether there are systematic sources of variation in the language models' accuracy. Focusing on subject-verb agreement and reflexive anaphora, we find that certain nouns are systematically understood better than others, an effect which is robust across grammatical tasks and different language models. Surprisingly, we find that across four orders of magnitude, corpus frequency is unrelated to a noun's performance on grammatical tasks. Finally, we find that …
Word, Subword or Character? An Empirical Study of Granularity in Chinese-English NMT
National Laboratory of Pattern Recognition, CASIA（中国科学院自动化研究所模式识别国家重点实验室）；University of Chinese Academy of Sciences（中国科学院大学）；CAS Center for Excellence in Brain Science and Intelligence Technology（中国科学院脑科学与智能技术卓越创新中心）
We provide a comprehensive analysis of the interactions between pre-trained word embeddings, character models and POS tags in a transition-based dependency parser. While previous studies have shown POS information to be less important in the presence of character models, we show that in fact there are complex interactions between all three techniques. In isolation each produces large improvements over a baseline system using randomly initialised word embeddings only, but combining them quickly leads to diminishing returns. We categorise words by frequency, POS tag and language in order to syst…
University of Sheffield（谢菲尔德大学）；Beihang University（北京航空航天大学）
Target-oriented opinion words extraction (TOWE) (Fan et al., 2019b) is a new subtask of target-oriented sentiment analysis that aims to extract opinion words for a given aspect in text. Current state-of-the-art methods leverage position embeddings to capture the relative position of a word to the target. However, the performance of these methods depends on the ability to incorporate this information into word representations. In this paper, we explore a variety of text encoders based on pretrained word embeddings or language models that leverage part-of-speech and position embeddings, aiming t…
The effect of amplifiers, downtoners, and negations has been studied in general and particularly in the context of sentiment analysis. However, there is only limited work which aims at transferring the results and methods to discrete classes of emotions, e. g., joy, anger, fear, sadness, surprise, and disgust. For instance, it is not straight-forward to interpret which emotion the phrase "not happy" expresses. With this paper, we aim at obtaining a better understanding of such modifiers in the context of emotion-bearing words and their impact on document-level emotion classification, namely, m…
Department of Computer Science, CUNY Graduate Center, USA（纽约市立大学研究生中心计算机科学系，美国）；Department of Mathematics and Computer Science, York College (CUNY), USA（纽约市立大学约克学院数学与计算机科学系，美国）；Department of Electrical Engineering, Columbia University, USA（哥伦比亚大学电气工程系，美国）；OATH Inc, USA（OATH公司，美国）
To build a satisfying chatbot that has the ability of managing a goal-oriented multi-turn dialogue, accurate modeling of human conversation is crucial. In this paper we concentrate on the task of response selection for multi-turn human-computer conversation with a given context. Previous approaches show weakness in capturing information of rare keywords that appear in either or both context and correct response, and struggle with long input sequences. We propose Cross Convolution Network (CCN) and Multi Frequency word embedding to address both problems. We train several models using the Ubuntu…
The University of Sydney（悉尼大学）；Tencent AI Lab（腾讯人工智能实验室）；University of Macau（澳门大学）；JD Explore Academy, JD.com（京东探索研究院）
Knowledge distillation (KD) is commonly used to construct synthetic data for training non-autoregressive translation (NAT) models. However, there exists a discrepancy on low-frequency words between the distilled and the original data, leading to more errors on predicting low-frequency words. To alleviate the problem, we directly expose the raw data into NAT by leveraging pretraining. By analyzing directed alignments, we found that KD makes low-frequency source words aligned with targets more deterministically but fails to align sufficient low-frequency words from target to source. Accordingly,…
Shanghai University of Finance and Economics（上海财经大学）
Text augmentation techniques are widely used in text classification problems to improve the performance of classifiers, especially in low-resource scenarios. Whilst lots of creative text augmentation methods have been designed, they augment the text in a non-selective manner, which means the less important or noisy words have the same chances to be augmented as the informative words, and thereby limits the performance of augmentation. In this work, we systematically summarize three kinds of role keywords, which have different functions for text classification, and design effective methods to e…
ETH Zürich（苏黎世联邦理工学院）；Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）
Languages are continuously undergoing changes, and the mechanisms that underlie these changes are still a matter of debate. In this work, we approach language evolution through the lens of causality in order to model not only how various distributional factors associate with language change, but how they causally affect it. In particular, we study slang, which is an informal language that is typically restricted to a specific group or social setting. We analyze the semantic change and frequency shift of slang words and compare them to those of standard, nonslang words. With causal discovery an…
Peking University（北京大学）；Key Laboratory of Machine Perception, MOE, School of EECS, Peking University（北京大学信息科学技术学院机器感知与智能教育部重点实验室）；Microsoft Research Asia（微软亚洲研究院）；Center for Data Science, Peking University（北京大学数据科学中心）；Beijing Institute of Big Data Research（北京大数据研究院）
Continuous word representation (aka word embedding) is a basic building block in many neural network-based models used in natural language processing tasks. Although it is widely accepted that words with similar semantics should be close to each other in the embedding space, we find that word embeddings learned in several tasks are biased towards word frequency: the embeddings of high-frequency and low-frequency words lie in different subregions of the embedding space, and the embedding of a rare word and a popular word can be far from each other even if they are semantically similar. This mak…
Understanding how words change their meanings over time is key to models of language and cultural evolution, but historical data on meaning is scarce, making theories hard to develop and test. Word embeddings show promise as a diachronic tool, but have not been carefully evaluated. We develop a robust methodology for quantifying semantic change by evaluating word embeddings (PPMI, SVD, word2vec) against known historical changes. We then use this methodology to reveal statistical laws of semantic evolution. Using six historical corpora spanning four languages and two centuries, we propose two q…
The Institute of Statistical Mathematics（统计数理研究所）；The Research Organization of Information and Systems（信息与系统研究机构）
Ultraslow diffusion (i.e. logarithmic diffusion) has been extensively studied theoretically, but has hardly been observed empirically. In this paper, firstly, we find the ultraslow-like diffusion of the time-series of word counts of already popular words by analysing three different nationwide language databases: (i) newspaper articles (Japanese), (ii) blog articles (Japanese), and (iii) page views of Wikipedia (English, French, Chinese, and Japanese). Secondly, we use theoretical analysis to show that this diffusion is basically explained by the random walk model with the power-law forgetting…
Several variants of deep neural networks have been successfully employed for building parametric models that project variable-duration spoken word segments onto fixed-size vector representations, or acoustic word embeddings (AWEs). However, it remains unclear to what degree we can rely on the distance in the emerging AWE space as an estimate of word-form similarity. In this paper, we ask: does the distance in the acoustic embedding space correlate with phonological dissimilarity? To answer this question, we empirically investigate the performance of supervised approaches for AWEs with differen…
Word forms - not just their lengths- are optimized for efficient communication
University of California, Berkeley（加州大学伯克利分校）
一个词的长度和它的使用频率之间的反比关系，最早由G. K. Zipf在1935年提出，是一个经典的经验法则，适用于各种人类语言。我们证明，长度是一个更普遍的属性的一个方面的话：他们是如何独特的相对于其他词在一种语言。区别性在识别流利语音中的单词中起着关键作用，因为它反映了在为模糊信号选择最佳候选时潜在竞争者的实力。语音信息量是在一种语言的声音或字符序列的统计模型下对单词的字符串概率的测量，它简明地捕捉了独特性。通过对13种语言的大规模语料库的分析，我们发现独特性在预测频率方面明显优于字长。这一发现提供了证据表明，听众的处理限制形状细粒度方面的词的形式跨越语言。
New York University（纽约大学）；Johns Hopkins University（约翰斯·霍普金斯大学）
Spoken word recognition involves at least two basic computations. First is matching acoustic input to phonological categories (e.g. /b/, /p/, /d/). Second is activating words consistent with those phonological categories. Here we test the hypothesis that the listener's probability distribution over lexical items is weighted by the outcome of both computations: uncertainty about phonological discretisation and the frequency of the selected word(s). To test this, we record neural responses in auditory cortex using magnetoencephalography, and model this activity as a function of the size and rela…
IR models using a pretrained language model significantly outperform lexical approaches like BM25. In particular, SPLADE, which encodes texts to sparse vectors, is an effective model for practical use because it shows robustness to out-of-domain datasets. However, SPLADE still struggles with exact matching of low-frequency words in training data. In addition, domain shifts in vocabulary and word frequencies deteriorate the IR performance of SPLADE. Because supervision data are scarce in the target domain, addressing the domain shifts without supervision data is necessary. This paper proposes a…
Word embeddings learnt from large corpora have been adopted in various applications in natural language processing and served as the general input representations to learning systems. Recently, a series of post-processing methods have been proposed to boost the performance of word embeddings on similarity comparison and analogy retrieval tasks, and some have been adapted to compose sentence representations. The general hypothesis behind these methods is that by enforcing the embedding space to be more isotropic, the similarity between words can be better expressed. We view these methods as an …
