Pre-trained contextual representations like BERT have achieved great success in natural language processing. However, the sentence embeddings from the pre-trained language models without fine-tuning have been found to poorly capture semantic meaning of sentences. In this paper, we argue that the semantic information in the BERT embeddings is not fully exploited. We first reveal the theoretical connection between the masked language model pre-training objective and the semantic similarity task theoretically, and then analyze the BERT sentence embeddings empirically. We find that BERT always induces a non-smooth anisotropic semantic space of sentences, which harms its performance of semantic similarity. To address this issue, we propose to transform the anisotropic sentence embedding distribution to a smooth and isotropic Gaussian distribution through normalizing flows that are learned with an unsupervised objective. Experimental results show that our proposed BERT-flow method obtains significant performance gains over the state-of-the-art sentence embeddings on a variety of semantic textual similarity tasks. The code is available at https://github.com/bohanli/BERT-flow.
Pre-trained language models have achieved huge success on a wide range of NLP tasks. However, contextual representations from pre-trained models contain entangled semantic and syntactic information, and therefore cannot be directly used to derive useful semantic sentence embeddings for some tasks. Paraphrase pairs offer an effective way of learning the distinction between semantics and syntax, as they naturally share semantics and often vary in syntax. In this work, we present ParaBART, a semantic sentence embedding model that learns to disentangle semantics and syntax in sentence embeddings o…
Refining Sentence Embedding Model through Ranking Sentences Generation with Large Language Models
State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China（中国科学技术大学认知智能国家重点实验室）；Institute of Artificial Intelligence, Hefei Comprehensive National Science Center（合肥综合性国家科学中心人工智能研究院）；Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际 Graduate School）；Zhejiang University（浙江大学）
The University of Tokyo（东京大学）；Microsoft Corporation（微软公司）
Learning sentence embeddings in an unsupervised manner is fundamental in natural language processing. Recent common practice is to couple pre-trained language models with unsupervised contrastive learning, whose success relies on augmenting a sentence with a semantically-close positive instance to construct contrastive pairs. Nonetheless, existing approaches usually depend on a mono-augmenting strategy, which causes learning shortcuts towards the augmenting biases and thus corrupts the quality of sentence embeddings. A straightforward solution is resorting to more diverse positives from a mult…
Following SimCSE, contrastive learning based methods have achieved the state-of-the-art (SOTA) performance in learning sentence embeddings. However, the unsupervised contrastive learning methods still lag far behind the supervised counterparts. We attribute this to the quality of positive and negative samples, and aim to improve both. Specifically, for positive samples, we propose switch-case augmentation to flip the case of the first letter of randomly selected words in a sentence. This is to counteract the intrinsic bias of pre-trained token embeddings to frequency, word cases and subwords. …
Decoder transformers have continued increasing in scale reaching hundreds of billions of parameters. Due to their scale the same decoder sets state-of-the-art results on various language tasks via prompting or fine-tuning. Yet, these large foundation models remain unusable for the related fields of semantic search and sentence embeddings. This prevents possibly new state-of-the-art results and forces organizations to train and maintain separate models. To this end, we propose SGPT to use decoders for sentence embeddings and semantic search via prompting or fine-tuning. At 5.8 billion parameter…
We present Emu, a system that semantically enhances multilingual sentence embeddings. Our framework fine-tunes pre-trained multilingual sentence embeddings using two main components: a semantic classifier and a language discriminator. The semantic classifier improves the semantic similarity of related sentences, whereas the language discriminator enhances the multilinguality of the embeddings via multilingual adversarial training. Our experimental results based on several language pairs show that our specialized embeddings outperform the state-of-the-art multilingual sentence embedding model o…
Sentence embeddings encode sentences in fixed dense vectors and have played an important role in various NLP tasks and systems. Methods for building sentence embeddings include unsupervised learning such as Quick-Thoughts and supervised learning such as InferSent. With the success of pretrained NLP models, recent research shows that fine-tuning pretrained BERT on SNLI and Multi-NLI data creates state-of-the-art sentence embeddings, outperforming previous sentence embeddings methods on various evaluation benchmarks. In this paper, we propose a new method to build sentence embeddings by doing su…
We propose a novel model architecture and training algorithm to learn bilingual sentence embeddings from a combination of parallel and monolingual data. Our method connects autoencoding and neural machine translation to force the source and target sentence embeddings to share the same space without the help of a pivot language or an additional transformation. We train a multilayer perceptron on top of the sentence embeddings to extract good bilingual sentence pairs from nonparallel or noisy parallel data. Our approach shows promising performance on sentence alignment recovery and the WMT 2018 …
Pair-based metric learning has been widely adopted to learn sentence embedding in many NLP tasks such as semantic text similarity due to its efficiency in computation. Most existing works employed a sequence encoder model and utilized limited sentence pairs with a pair-based loss to learn discriminating sentence representation. However, it is known that the sentence representation can be biased when the sampled sentence pairs deviate from the true distribution of all sentence pairs. In this paper, our theoretical analysis shows that existing works severely suffered from a good pair sampling an…
National Research University Higher School of Economics（国立研究型大学高等经济学院）
We investigate the performance of sentence embeddings models on several tasks for the Russian language. In our comparison, we include such tasks as multiple choice question answering, next sentence prediction, and paraphrase identification. We employ FastText embeddings as a baseline and compare it to ELMo and BERT embeddings. We conduct two series of experiments, using both unsupervised (i.e., based on similarity measure only) and supervised approaches for the tasks. Finally, we present datasets for multiple choice question answering and next sentence prediction in Russian.
Ionian University（伊奥尼亚大学）；University of the Aegean（爱琴海大学）
The domain of natural language processing (NLP), which has greatly evolved over the last years, has highly benefited from the recent developments in word and sentence embeddings. Such embeddings enable the transformation of complex NLP tasks, like semantic similarity or Question and Answering (Q&A), into much simpler to perform vector comparisons. However, such a problem transformation raises new challenges like the efficient comparison of embeddings and their manipulation. In this work, we will discuss about various word and sentence embeddings algorithms, we will select a sentence embedding …
College of Computing and Informatics Drexel University（德雷塞尔大学计算机与信息学院）；Drexel University（德雷塞尔大学）
Sentence embeddings encode natural language sentences as low-dimensional dense vectors. A great deal of effort has been put into using sentence embeddings to improve several important natural language processing tasks. Relation extraction is such an NLP task that aims at identifying structured relations defined in a knowledge base from unstructured text. A promising and more efficient approach would be to embed both the text and structured knowledge in low-dimensional spaces and discover semantic alignments or mappings between them. Although a number of techniques have been proposed in the lit…
CICS, University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校信息与计算机科学学院）
Most unsupervised NLP models represent each word with a single point or single region in semantic space, while the existing multi-sense word embeddings cannot represent longer word sequences like phrases or sentences. We propose a novel embedding method for a text sequence (a phrase or a sentence) where each sequence is represented by a distinct set of multi-mode codebook embeddings to capture different semantic facets of its meaning. The codebook embeddings can be viewed as the cluster centers which summarize the distribution of possibly co-occurring words in a pre-trained word embedding spac…
Carnegie Mellon University（卡内基梅隆大学）；University of California San Diego（加州大学圣地亚哥分校）
Semantic sentence embedding models encode natural language sentences into vectors, such that closeness in embedding space indicates closeness in the semantics between the sentences. Bilingual data offers a useful signal for learning such embeddings: properties shared by both sentences in a translation pair are likely semantic, while divergent properties are likely stylistic or language-specific. We propose a deep latent variable model that attempts to perform source separation on parallel sentences, isolating what they have in common in a latent semantic vector, and explaining what is left ove…
Continuous sentence embeddings using recurrent neural networks (RNNs), where variable-length sentences are encoded into fixed-dimensional vectors, are often the main building blocks of architectures applied to language tasks such as dialogue generation. While it is known that those embeddings are able to learn some structures of language (e.g. grammar) in a purely data-driven manner, there is very little work on the objective evaluation of their ability to cover the whole language space and to generalize to sentences outside the language bias of the training data. Using a manually designed con…
Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning t…
User language data can contain highly sensitive personal content. As such, it is imperative to offer users a strong and interpretable privacy guarantee when learning from their data. In this work, we propose SentDP: pure local differential privacy at the sentence level for a single user document. We propose a novel technique, DeepCandidate, that combines concepts from robust statistics and language modeling to produce high-dimensional, general-purpose $ε$-SentDP document embeddings. This guarantees that any single sentence in a document can be substituted with any other sentence while keeping …
This paper have two parts. In the first part we discuss word embeddings. We discuss the need for them, some of the methods to create them, and some of their interesting properties. We also compare them to image embeddings and see how word embedding and image embedding can be combined to perform different tasks. In the second part we implement a convolutional neural network trained on top of pre-trained word vectors. The network is used for several sentence-level classification tasks, and achieves state-of-art (or comparable) results, demonstrating the great power of pre-trainted word embedding…
Answer selection aims at identifying the correct answer for a given question from a set of potentially correct answers. Contrary to previous works, which typically focus on the semantic similarity between a question and its answer, our hypothesis is that question-answer pairs are often in analogical relation to each other. Using analogical inference as our use case, we propose a framework and a neural network architecture for learning dedicated sentence embeddings that preserve analogical properties in the semantic space. We evaluate the proposed method on benchmark datasets for answer selecti…
School of software Engineering, Beijing University of Posts and Telecommunications（北京邮电大学软件学院）；Department of statistics, Harvard University（哈佛大学统计系）
In this paper, we develop a low than character feature embedding called radical embedding, and apply it on LSTM model for sentence segmentation of pre modern Chinese texts. The datasets includes over 150 classical Chinese books from 3 different dynasties and contains different literary styles. LSTM CRF model is a state of art method for the sequence labeling problem. Our new model adds a component of radical embedding, which leads to improved performances. Experimental results based on the aforementioned Chinese books demonstrates a better accuracy than earlier methods on sentence segmentation…
This paper contributes a new State Of The Art (SOTA) for Semantic Textual Similarity (STS). We compare and combine a number of recently proposed sentence embedding methods for STS, and propose a novel and simple ensemble knowledge distillation scheme that improves on previous approaches. Our experiments demonstrate that a model trained to learn the average embedding space from multiple ensemble students outperforms all the other individual models with high robustness. Utilizing our distillation method in combination with previous methods, we significantly improve on the SOTA unsupervised STS, …
Drexel University（德雷塞尔大学）；Metadata Research Center（元数据研究中心）；College of Computing and Informatics（计算与信息学院）
Sentence embedding methods offer a powerful approach for working with short textual constructs or sequences of words. By representing sentences as dense numerical vectors, many natural language processing (NLP) applications have improved their performance. However, relatively little is understood about the latent structure of sentence embeddings. Specifically, research has not addressed whether the length and structure of sentences impact the sentence embedding space and topology. This paper reports research on a set of comprehensive clustering and network analyses targeting sentence and sub-s…
Recent state-of-the-art natural language understanding models, such as BERT and XLNet, score a pair of sentences (A and B) using multiple cross-attention operations - a process in which each word in sentence A attends to all words in sentence B and vice versa. As a result, computing the similarity between a query sentence and a set of candidate sentences, requires the propagation of all query-candidate sentence-pairs throughout a stack of cross-attention layers. This exhaustive process becomes computationally prohibitive when the number of candidate sentences is large. In contrast, sentence em…
Singapore University of Technology and Design（新加坡科技设计大学）；DAMO Academy, Alibaba Group（阿里巴巴集团达摩院）；ZJU-UIUC Institue（浙江大学-伊利诺伊大学厄巴纳-香槟分校联合学院）
BERT is inefficient for sentence-pair tasks such as clustering or semantic search as it needs to evaluate combinatorially many sentence pairs which is very time-consuming. Sentence BERT (SBERT) attempted to solve this challenge by learning semantically meaningful representations of single sentences, such that similarity comparison can be easily accessed. However, SBERT is trained on corpus with high-quality labeled sentence pairs, which limits its application to tasks where labeled data is extremely scarce. In this paper, we propose a lightweight extension on top of BERT and a novel self-super…
Pre-trained sentence embeddings have been shown to be very useful for a variety of NLP tasks. Due to the fact that training such embeddings requires a large amount of data, they are commonly trained on a variety of text data. An adaptation to specific domains could improve results in many cases, but such a finetuning is usually problem-dependent and poses the risk of over-adapting to the data used for adaptation. In this paper, we present a simple universal method for finetuning Google's Universal Sentence Encoder (USE) using a Siamese architecture. We demonstrate how to use this approach for …
We study methods for learning sentence embeddings with syntactic structure. We focus on methods of learning syntactic sentence-embeddings by using a multilingual parallel-corpus augmented by Universal Parts-of-Speech tags. We evaluate the quality of the learned embeddings by examining sentence-level nearest neighbours and functional dissimilarity in the embedding space. We also evaluate the ability of the method to learn syntactic sentence-embeddings for low-resource languages and demonstrate strong evidence for transfer learning. Our results show that syntactic sentence-embeddings can be lear…
We introspect black-box sentence embeddings by conditionally generating from them with the objective to retrieve the underlying discrete sentence. We perceive of this as a new unsupervised probing task and show that it correlates well with downstream task performance. We also illustrate how the language generated from different encoders differs. We apply our approach to generate sentence analogies from sentence embeddings.
Despite the great success of word embedding, sentence embedding remains a not-well-solved problem. In this paper, we present a supervised learning framework to exploit sentence embedding for the medical question answering task. The learning framework consists of two main parts: 1) a sentence embedding producing module, and 2) a scoring module. The former is developed with contextual self-attention and multi-scale techniques to encode a sentence into an embedding tensor. This module is shortly called Contextual self-Attention Multi-scale Sentence Embedding (CAMSE). The latter employs two scorin…
We present the Siamese Continuous Bag of Words (Siamese CBOW) model, a neural network for efficient estimation of high-quality sentence embeddings. Averaging the embeddings of words in a sentence has proven to be a surprisingly successful and efficient way of obtaining sentence embeddings. However, word embeddings trained with the methods currently available are not optimized for the task of sentence representation, and, thus, likely to be suboptimal. Siamese CBOW handles this problem by training word embeddings directly for the purpose of being averaged. The underlying neural network learns w…
Sentence embedding is an important research topic in natural language processing. It is essential to generate a good embedding vector that fully reflects the semantic meaning of a sentence in order to achieve an enhanced performance for various natural language processing tasks, such as machine translation and document classification. Thus far, various sentence embedding models have been proposed, and their feasibility has been demonstrated through good performances on tasks following embedding, such as sentiment analysis and sentence classification. However, because the performances of senten…
National Center for Biotechnology Information (NCBI), National Library of Medicine (NLM), National Institutes of Health (NIH)（美国国立卫生研究院国家医学图书馆国家生物技术信息中心）
Sentence embeddings have become an essential part of today's natural language processing (NLP) systems, especially together advanced deep learning methods. Although pre-trained sentence encoders are available in the general domain, none exists for biomedical texts to date. In this work, we introduce BioSentVec: the first open set of sentence embeddings trained with over 30 million documents from both scholarly articles in PubMed and clinical notes in the MIMIC-III Clinical Database. We evaluate BioSentVec embeddings in two sentence pair similarity tasks in different text genres. Our benchmarki…
We introduce a model for bidirectional retrieval of images and sentences through a multi-modal embedding of visual and natural language data. Unlike previous models that directly map images or sentences into a common embedding space, our model works on a finer level and embeds fragments of images (objects) and fragments of sentences (typed dependency tree relations) into a common space. In addition to a ranking objective seen in previous work, this allows us to add a new fragment alignment objective that learns to directly associate these fragments across modalities. Extensive experimental eva…
A variety of contextualised language models have been proposed in the NLP community, which are trained on diverse corpora to produce numerous Neural Language Models (NLMs). However, different NLMs have reported different levels of performances in downstream NLP applications when used as text representations. We propose a sentence-level meta-embedding learning method that takes independently trained contextualised word embedding models and learns a sentence embedding that preserves the complementary strengths of the input source NLMs. Our proposed method is unsupervised and is not tied to a par…
Recently, BERT realized significant progress for sentence matching via word-level cross sentence attention. However, the performance significantly drops when using siamese BERT-networks to derive two sentence embeddings, which fall short in capturing the global semantic since the word-level attention between two sentences is absent. In this paper, we propose a Dual-view distilled BERT~(DvBERT) for sentence matching with sentence embeddings. Our method deals with a sentence pair from two distinct views, i.e., Siamese View and Interaction View. Siamese View is the backbone where we generate sent…
While important properties of word vector representations have been studied extensively, far less is known about the properties of sentence vector representations. Word vectors are often evaluated by assessing to what degree they exhibit regularities with regard to relationships of the sort considered in word analogies. In this paper, we investigate to what extent commonly used sentence vector representation spaces as well reflect certain kinds of regularities. We propose a number of schemes to induce evaluation data, based on lexical analogy data as well as semantic relationships between sent…
