Recent NLP literature has seen growing interest in improving model interpretability. Along this direction, we propose a trainable neural network layer that learns a global interaction graph between words and then selects more informative words using the learned word interactions. Our layer, we call WIGRAPH, can plug into any neural network-based NLP text classifiers right after its word embedding layer. Across multiple SOTA NLP models and various NLP datasets, we demonstrate that adding the WIGRAPH layer substantially improves NLP models' interpretability and enhances models' prediction performance at the same time.
National Research University Higher School of Economics（国家研究型高等经济大学）
Consider a continuous word embedding model. Usually, the cosines between word vectors are used as a measure of similarity of words. These cosines do not change under orthogonal transformations of the embedding space. We demonstrate that, using some canonical orthogonal transformations from SVD, it is possible both to increase the meaning of some components and to make the components more stable under re-learning. We study the interpretability of components for publicly available models for the Russian language (RusVectores, fastText, RDT).
ASELSAN Research Center（阿瑟兰研究中心）；Bilkent University（比尔肯特大学）；Ludwig Maximillian University（路德维希·马克西米利安大学）
As an ubiquitous method in natural language processing, word embeddings are extensively employed to map semantic properties of words into a dense vector representation. They capture semantic and syntactic relations among words but the vectors corresponding to the words are only meaningful relative to each other. Neither the vector nor its dimensions have any absolute, interpretable meaning. We introduce an additive modification to the objective function of the embedding learning algorithm that encourages the embedding vectors of words that are semantically related to a predefined concept to ta…
RWTH Aachen（亚琛工业大学）；L3S Research Center（L3S研究中心）；University of Mannheim（曼海姆大学）；GESIS（盖泽斯社会科学研究所）；CSH Vienna（维也纳复杂科学中心）
Adding interpretability to word embeddings represents an area of active research in text representation. Recent work has explored thepotential of embedding words via so-called polar dimensions (e.g. good vs. bad, correct vs. wrong). Examples of such recent approaches include SemAxis, POLAR, FrameAxis, and BiImp. Although these approaches provide interpretable dimensions for words, they have not been designed to deal with polysemy, i.e. they can not easily distinguish between different senses of words. To address this limitation, we present SensePOLAR, an extension of the original POLAR framewo…
To build an interpretable neural text classifier, most of the prior work has focused on designing inherently interpretable models or finding faithful explanations. A new line of work on improving model interpretability has just started, and many existing methods require either prior information or human annotations as additional inputs in training. To address this limitation, we propose the variational word mask (VMASK) method to automatically learn task-specific important words and reduce irrelevant information on classification, which ultimately improves the interpretability of model predict…
ASELSAN Research Center（ASELSAN研究中心）；Bilkent University（比尔肯大学）
Dense word embeddings, which encode semantic meanings of words to low dimensional vector spaces have become very popular in natural language processing (NLP) research due to their state-of-the-art performances in many NLP tasks. Word embeddings are substantially successful in capturing semantic relations among words, so a meaningful semantic structure must be present in the respective vector spaces. However, in many cases, this semantic structure is broadly and heterogeneously distributed across the embedding dimensions, which makes interpretation a big challenge. In this study, we propose a s…
Word embeddings are substantially successful in capturing semantic relations among words. However, these lexical semantics are difficult to be interpreted. Definition modeling provides a more intuitive way to evaluate embeddings by utilizing them to generate natural language definitions of corresponding words. This task is of great significance for practical application and in-depth understanding of word representations. We propose a novel framework for definition modeling, which can generate reasonable and understandable context-dependent definitions. Moreover, we introduce usage modeling and…
Isfahan University of Technology（伊斯法罕理工大学）；Isfahan University of Medical Sciences（伊斯法罕医科大学）
Word embeddings have found their way into a wide range of natural language processing tasks including those in the biomedical domain. While these vector representations successfully capture semantic and syntactic word relations, hidden patterns and trends in the data, they fail to offer interpretability. Interpretability is a key means to justification which is an integral part when it comes to biomedical applications. We present an inclusive study on interpretability of word embeddings in the medical domain, focusing on the role of sparse methods. Qualitative and quantitative measurements and…
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 -- hard). The core idea of our approach is to transform existing, pre-trained word embeddings via semantic differentials to a new "polar" space with interpretable dimensions defined by such polar opposites. Our framework also allows for selecting the most discriminative dimensions fro…
Duke University（杜克大学）；University of Washington（华盛顿大学）；Google Research（谷歌研究院）
This work shows how to improve and interpret the commonly used dual encoder model for response suggestion in dialogue. We present an attentive dual encoder model that includes an attention mechanism on top of the extracted word-level features from two encoders, one for context and one for label respectively. To improve the interpretability in the dual encoder models, we design a novel regularization loss to minimize the mutual information between unimportant words and desired labels, in addition to the original attention method, so that important words are emphasized while unimportant words ar…
University of Sheffield（谢菲尔德大学）；King Saud University（沙特国王大学）
Topics models, such as LDA, are widely used in Natural Language Processing. Making their output interpretable is an important area of research with applications to areas such as the enhancement of exploratory search interfaces and the development of interpretable machine learning models. Conventionally, topics are represented by their n most probable words, however, these representations are often difficult for humans to interpret. This paper explores the re-ranking of topic words to generate more interpretable topic representations. A range of approaches are compared and evaluated in two expe…
Research and Technology Center, Robert Bosch（罗伯特·博世研究与技术中心）
Word embeddings are a basic building block of modern NLP pipelines. Efforts have been made to learn rich, efficient, and interpretable embeddings for large generic datasets available in the public domain. However, these embeddings have limited applicability for small corpora from specific domains such as automotive, manufacturing, maintenance and support, etc. In this work, we present a comprehensive notion of interpretability for word embeddings and propose a novel method to generate highly interpretable and efficient embeddings for a domain-specific small corpus. We report the evaluation res…
On the Interpretability and Significance of Bias Metrics in Texts: a PMI-based Approach
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（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）；Universidad Nacional de San Martín（圣马丁国立大学）；Harvard University（哈佛大学）
MIT Media Lab（麻省理工学院媒体实验室）；Boston University（波士顿大学）；MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）；Basis Technologies
Retrofitting is a technique used to move word vectors closer together or further apart in their space to reflect their relationships in a Knowledge Base (KB). However, retrofitting only works on concepts that are present in that KB. RetroGAN uses a pair of Generative Adversarial Networks (GANs) to learn a one-to-one mapping between concepts and their retrofitted counterparts. It applies that mapping (post-specializes) to handle concepts that do not appear in the original KB in a manner similar to how some natural language systems handle out-of-vocabulary entries. We test our system on three wo…
Centre for Artificial Intelligence Research（人工智能研究中心）；University of Agder（阿格德尔大学）
Tsetlin Machine (TM) is an interpretable pattern recognition algorithm based on propositional logic, which has demonstrated competitive performance in many Natural Language Processing (NLP) tasks, including sentiment analysis, text classification, and Word Sense Disambiguation. To obtain human-level interpretability, legacy TM employs Boolean input features such as bag-of-words (BOW). However, the BOW representation makes it difficult to use any pre-trained information, for instance, word2vec and GloVe word representations. This restriction has constrained the performance of TM compared to dee…
University of Maryland, Baltimore County（马里兰大学巴尔的摩县分校）
Word embeddings improve the performance of NLP systems by revealing the hidden structural relationships between words. Despite their success in many applications, word embeddings have seen very little use in computational social science NLP tasks, presumably due to their reliance on big data, and to a lack of interpretability. I propose a probabilistic model-based word embedding method which can recover interpretable embeddings, without big data. The key insight is to leverage mixed membership modeling, in which global representations are shared, but individual entities (i.e. dictionary words)…
Social scientists have recently turned to analyzing text using tools from natural language processing like word embeddings to measure concepts like ideology, bias, and affinity. However, word embeddings are difficult to use in the regression framework familiar to social scientists: embeddings are are neither identified, nor directly interpretable. I offer two advances on standard embedding models to remedy these problems. First, I develop Bayesian Word Embeddings with Automatic Relevance Determination priors, relaxing the assumption that all embedding dimensions have equal weight. Second, I ap…
The human language can be expressed through multiple sources of information known as modalities, including tones of voice, facial gestures, and spoken language. Recent multimodal learning with strong performances on human-centric tasks such as sentiment analysis and emotion recognition are often black-box, with very limited interpretability. In this paper we propose Multimodal Routing, which dynamically adjusts weights between input modalities and output representations differently for each input sample. Multimodal routing can identify relative importance of both individual modalities and cros…
Most of existing work learn sentiment-specific word representation for improving Twitter sentiment classification, which encoded both n-gram and distant supervised tweet sentiment information in learning process. They assume all words within a tweet have the same sentiment polarity as the whole tweet, which ignores the word its own sentiment polarity. To address this problem, we propose to learn sentiment-specific word embedding by exploiting both lexicon resource and distant supervised information. We develop a multi-level sentiment-enriched word embedding learning method, which uses parallel…
Most end-to-end speech recognition systems model text directly as a sequence of characters or sub-words. Current approaches to sub-word extraction only consider character sequence frequencies, which at times produce inferior sub-word segmentation that might lead to erroneous speech recognition output. We propose pronunciation-assisted sub-word modeling (PASM), a sub-word extraction method that leverages the pronunciation information of a word. Experiments show that the proposed method can greatly improve upon the character-based baseline, and also outperform commonly used byte-pair encoding me…
Gradient-based explanation methods play an important role in the field of interpreting complex deep neural networks for NLP models. However, the existing work has shown that the gradients of a model are unstable and easily manipulable, which impacts the model's reliability largely. According to our preliminary analyses, we also find the interpretability of gradient-based methods is limited for complex tasks, such as aspect-based sentiment classification (ABSC). In this paper, we propose an \textbf{I}nterpretation-\textbf{E}nhanced \textbf{G}radient-based framework for \textbf{A}BSC via a small…
This work presents an unsupervised approach for improving WordNet that builds upon recent advances in document and sense representation via distributional semantics. We apply our methods to construct Wordnets in French and Russian, languages which both lack good manual constructions.1 These are evaluated on two new 600-word test sets for word-to-synset matching and found to improve greatly upon synset recall, outperforming the best automated Wordnets in F-score. Our methods require very few linguistic resources, thus being applicable for Wordnet construction in low-resources languages, and may…
Technion – Israel Institute of Technology（技术离子-以色列理工学院）
One of the main challenges in mechanistic interpretability is circuit discovery, determining which parts of a model perform a given task. We build on the Mechanistic Interpretability Benchmark (MIB) and propose three key improvements to circuit discovery. First, we use bootstrapping to identify edges with consistent attribution scores. Second, we introduce a simple ratio-based selection strategy to prioritize strong positive-scoring edges, balancing performance and faithfulness. Third, we replace the standard greedy selection with an integer linear programming formulation. Our methods yield mo…
Data and Web Science Group（数据与网络科学组）；LATTICE Lab (ENS, CNRS, Paris 3)（LATTICE实验室（巴黎高等师范学院、法国国家科学研究中心、巴黎第三大学））；University of Mannheim（曼海姆大学）；PSL Research Univ, Sorbonne Paris Cité（巴黎文理研究大学，索邦巴黎西岱大学）
In order to create a corpus exploration method providing topics that are easier to interpret than standard LDA topic models, here we propose combining two techniques called Entity linking and Labeled LDA. Our method identifies in an ontology a series of descriptive labels for each document in a corpus. Then it generates a specific topic for each label. Having a direct relation between topics and labels makes interpretation easier; using an ontology as background knowledge limits label ambiguity. As our topics are described with a limited number of clear-cut labels, they promote interpretabilit…
The inclusion of semantic information in any similarity measures improves the efficiency of the similarity measure and provides human interpretable results for further analysis. The similarity calculation method that focuses on features related to the text's words only, will give less accurate results. This paper presents three different methods that not only focus on the text's words but also incorporates semantic information of texts in their feature vector and computes semantic similarities. These methods are based on corpus-based and knowledge-based methods, which are: cosine similarity us…
We suggest a new method for creating and using gold-standard datasets for word similarity evaluation. Our goal is to improve the reliability of the evaluation, and we do this by redesigning the annotation task to achieve higher inter-rater agreement, and by defining a performance measure which takes the reliability of each annotation decision in the dataset into account.
Cross-domain Chinese Word Segmentation (CWS) remains a challenge despite recent progress in neural-based CWS. The limited amount of annotated data in the target domain has been the key obstacle to a satisfactory performance. In this paper, we propose a semi-supervised word-based approach to improving cross-domain CWS given a baseline segmenter. Particularly, our model only deploys word embeddings trained on raw text in the target domain, discarding complex hand-crafted features and domain-specific dictionaries. Innovative subsampling and negative sampling methods are proposed to derive word em…
With the dramatic advances in deep learning technology, machine learning research is focusing on improving the interpretability of model predictions as well as prediction performance in both basic and applied research. While deep learning models have much higher prediction performance than traditional machine learning models, the specific prediction process is still difficult to interpret and/or explain. This is known as the black-boxing of machine learning models and is recognized as a particularly important problem in a wide range of research fields, including manufacturing, commerce, roboti…
Kohli Center On Intelligent Systems (KCIS)（科利智能系统中心）；International Institute of Information Technology Hyderabad (IIIT-H)（海得拉巴国际信息技术学院）
We present a language independent, unsupervised method for building word embeddings using morphological expansion of text. Our model handles the problem of data sparsity and yields improved word embeddings by relying on training word embeddings on artificially generated sentences. We evaluate our method using small sized training sets on eleven test sets for the word similarity task across seven languages. Further, for English, we evaluated the impacts of our approach using a large training set on three standard test sets. Our method improved results across all languages.
Unsupervised learned representations of polysemous words generate a large of pseudo multi senses since unsupervised methods are overly sensitive to contextual variations. In this paper, we address the pseudo multi-sense detection for word embeddings by dimensionality reduction of sense pairs. We propose a novel principal analysis method, termed Ex-RPCA, designed to detect both pseudo multi senses and real multi senses. With Ex-RPCA, we empirically show that pseudo multi senses are generated systematically in unsupervised method. Moreover, the multi-sense word embeddings can by improved by a si…
Laboratoire d’analyse statistique des textes - LAST（文本统计分析实验室）；Institut de recherche en sciences psychologiques（心理科学研究所）；Université catholique de Louvain（鲁汶天主教大学）
This paper describes the system developed by the Laboratoire d'analyse statistique des textes (LAST) for the Lexical Complexity Prediction shared task at SemEval-2021. The proposed system is made up of a LightGBM model fed with features obtained from many word frequency lists, published lexical norms and psychometric data. For tackling the specificity of the multi-word task, it uses bigram association measures. Despite that the only contextual feature used was sentence length, the system achieved an honorable performance in the multi-word task, but poorer in the single word task. The bigram as…
In Word Sense Disambiguation (WSD), the predominant approach generally involves a supervised system trained on sense annotated corpora. The limited quantity of such corpora however restricts the coverage and the performance of these systems. In this article, we propose a new method that solves these issues by taking advantage of the knowledge present in WordNet, and especially the hypernymy and hyponymy relationships between synsets, in order to reduce the number of different sense tags that are necessary to disambiguate all words of the lexical database. Our method leads to state of the art r…
Despite the increasing popularity of NLP in the humanities and social sciences, advances in model performance and complexity have been accompanied by concerns about interpretability and explanatory power for sociocultural analysis. One popular model that balances complexity and legibility is Word Mover's Distance (WMD). Ostensibly adapted for its interpretability, WMD has nonetheless been used and further developed in ways which frequently discard its most interpretable aspect: namely, the word-level distances required for translating a set of words into another set of words. To address this a…
Peking University（北京大学）；Beijing Key Laboratory of Internet Culture and Digital Dissemination Research（北京网络文化与数字传播重点实验室）
Using low dimensional vector space to represent words has been very effective in many NLP tasks. However, it doesn't work well when faced with the problem of rare and unseen words. In this paper, we propose to leverage the knowledge in semantic dictionary in combination with some morphological information to build an enhanced vector space. We get an improvement of 2.3% over the state-of-the-art Heidel Time system in temporal expression recognition, and obtain a large gain in other name entity recognition (NER) tasks. The semantic dictionary Hownet alone also shows promising results in computin…
RWTH Aachen University（亚琛工业大学）；Human Language Technology and Pattern Recognition Group（人类语言技术与模式识别小组）
Unsupervised learning of cross-lingual word embedding offers elegant matching of words across languages, but has fundamental limitations in translating sentences. In this paper, we propose simple yet effective methods to improve word-by-word translation of cross-lingual embeddings, using only monolingual corpora but without any back-translation. We integrate a language model for context-aware search, and use a novel denoising autoencoder to handle reordering. Our system surpasses state-of-the-art unsupervised neural translation systems without costly iterative training. We also analyze the eff…
EPFL（洛桑联邦理工学院）；AXA（安盛集团）；Inria（法国国家信息与自动化研究所）；Sorbonne Université（索邦大学）；Polish Academy of Science（波兰科学院）
Current methods for Black-Box NLP interpretability, like LIME or SHAP, are based on altering the text to interpret by removing words and modeling the Black-Box response. In this paper, we outline limitations of this approach when using complex BERT-based classifiers: The word-based sampling produces texts that are out-of-distribution for the classifier and further gives rise to a high-dimensional search space, which can't be sufficiently explored when time or computation power is limited. Both of these challenges can be addressed by using segments as elementary building blocks for NLP interpre…
Distributional models are derived from co-occurrences in a corpus, where only a small proportion of all possible plausible co-occurrences will be observed. This results in a very sparse vector space, requiring a mechanism for inferring missing knowledge. Most methods face this challenge in ways that render the resulting word representations uninterpretable, with the consequence that semantic composition becomes hard to model. In this paper we explore an alternative which involves explicitly inferring unobserved co-occurrences using the distributional neighbourhood. We show that distributional …
Chemical patents are an important resource for chemical information. However, few chemical Named Entity Recognition (NER) systems have been evaluated on patent documents, due in part to their structural and linguistic complexity. In this paper, we explore the NER performance of a BiLSTM-CRF model utilising pre-trained word embeddings, character-level word representations and contextualized ELMo word representations for chemical patents. We compare word embeddings pre-trained on biomedical and chemical patent corpora. The effect of tokenizers optimized for the chemical domain on NER performance…
National University（国立大学）；De La Salle University（德拉萨大学）
Assessing the proper difficulty levels of reading materials or texts in general is the first step towards effective comprehension and learning. In this study, we improve the conventional methodology of automatic readability assessment by incorporating the Word Mover's Distance (WMD) of ranked texts as an additional post-processing technique to further ground the difficulty level given by a model. Results of our experiments on three multilingual datasets in Filipino, German, and English show that the post-processing technique outperforms previous vanilla and ranking-based models using SVM.
Word2Vec is the most popular model for word representation and has been widely investigated in literature. However, its noise distribution for negative sampling is decided by empirical trials and the optimality has always been ignored. We suggest that the distribution is a sub-optimal choice, and propose to use a sub-sampled unigram distribution for better negative sampling. Our contributions include: (1) proposing the concept of semantics quantification and deriving a suitable sub-sampling rate for the proposed distribution adaptive to different training corpora; (2) demonstrating the advanta…
Recent advances in automatic evaluation metrics for text have shown that deep contextualized word representations, such as those generated by BERT encoders, are helpful for designing metrics that correlate well with human judgements. At the same time, it has been argued that contextualized word representations exhibit sub-optimal statistical properties for encoding the true similarity between words or sentences. In this paper, we present two techniques for improving encoding representations for similarity metrics: a batch-mean centering strategy that improves statistical properties; and a comp…
Word embeddings are a fixed, distributional representation of the context of words in a corpus learned from word co-occurrences. While word embeddings have proven to have many practical uses in natural language processing tasks, they reflect the attributes of the corpus upon which they are trained. Recent work has demonstrated that post-processing of word embeddings to apply information found in lexical dictionaries can improve their quality. We build on this post-processing technique by making it interactive. Our approach makes it possible for humans to adjust portions of a word embedding spa…
Though achieving impressive results on many NLP tasks, the BERT-like masked language models (MLM) encounter the discrepancy between pre-training and inference. In light of this gap, we investigate the contextual representation of pre-training and inference from the perspective of word probability distribution. We discover that BERT risks neglecting the contextual word similarity in pre-training. To tackle this issue, we propose an auxiliary gloss regularizer module to BERT pre-training (GR-BERT), to enhance word semantic similarity. By predicting masked words and aligning contextual embeddings…
Graph-based text representation focuses on how text documents are represented as graphs for exploiting dependency information between tokens and documents within a corpus. Despite the increasing interest in graph representation learning, there is limited research in exploring new ways for graph-based text representation, which is important in downstream natural language processing tasks. In this paper, we first propose a new heterogeneous word-character text graph that combines word and character n-gram nodes together with document nodes, allowing us to better learn dependencies among these en…
Fine-grained sentiment analysis is receiving increasing attention in recent years. Extracting opinion target expressions (OTE) in reviews is often an important step in fine-grained, aspect-based sentiment analysis. Retrieving this information from user-generated text, however, can be difficult. Customer reviews, for instance, are prone to contain misspelled words and are difficult to process due to their domain-specific language. In this work, we investigate whether character-level models can improve the performance for the identification of opinion target expressions. We integrate information…
ABBYY（ABBYY公司）；Moscow Institute of Physics and Technology（莫斯科物理技术学院）
In this paper, we explore the ways to improve POS-tagging using various types of auxiliary losses and different word representations. As a baseline, we utilized a BiLSTM tagger, which is able to achieve state-of-the-art results on the sequence labelling tasks. We developed a new method for character-level word representation using feedforward neural network. Such representation gave us better results in terms of speed and performance of the model. We also applied a novel technique of pretraining such word representations with existing word vectors. Finally, we designed a new variant of auxilia…
