Data augmentation is a widely used technique to address the problem of text classification when there is a limited amount of training data.Recent work often tackles this problem using large language models (LLMs) like GPT3 that can generate new examples given already available ones.In this work, we propose a method to generate more helpful augmented data by utilizing the LLM's abilities to follow instructions and perform few-shot classifications.Our specific PromptMix method consists of two steps: 1) generate challenging text augmentations near class boundaries; however, generating borderline examples increases the risk of false positives in the dataset, so we 2) relabel the text augmentations using a prompting-based LLM classifier to enhance the correctness of labels in the generated data.We evaluate the proposed method in challenging 2-shot and zero-shot settings on four text classification datasets: Bank-ing77, TREC6, Subjectivity (SUBJ), and Twitter Complaints.Our experiments show that generating and, crucially, relabeling borderline examples facilitates the transfer of knowledge of a massive LLM like GPT3.5-turbo into smaller and cheaper classifiers like DistilBERT base and BERT base .Furthermore, 2-shot Prompt-Mix outperforms multiple 5-shot data augmentation methods on the four datasets.

Abstract Text style transfer is an important task in natural language generation, which aims to control certain attributes in the generated text, such as politeness, emotion, humor, and many others. It has a long history in the field of natural language processing, and recently has re-gained significant attention thanks to the promising performance brought by deep neural models. In this article, we present a systematic survey of the research on neural text style transfer, spanning over 100 representative articles since the first neural text style transfer work in 2017. We discuss the task formulation, existing datasets and subtasks, evaluation, as well as the rich methodologies in the presence of parallel and non-parallel data. We also provide discussions on a variety of important topics regarding the future development of this task.1

Vikash Balasubramanian, Ivan Kobyzev, Hareesh Bahuleyan, Ilya Shapiro, Olga Vechtomova. Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. 2021.

Effective fusion of data from multiple modalities, such as video, speech, and text, is challenging due to the heterogeneous nature of multimodal data.In this paper, we propose adaptive fusion techniques that aim to model context from different modalities effectively.Instead of defining a deterministic fusion operation, such as concatenation, for the network, we let the network decide "how" to combine a given set of multimodal features more effectively.We propose two networks: 1) Auto-Fusion, which learns to compress information from different modalities while preserving the context, and 2) GAN-Fusion, which regularizes the learned latent space given context from complementing modalities.A quantitative evaluation on the tasks of multimodal machine translation and emotion recognition suggests that our lightweight, adaptive networks can better model context from other modalities than existing methods, many of which employ massive transformer-based networks.1

We present a novel iterative, edit-based approach to unsupervised sentence simplification.Our model is guided by a scoring function involving fluency, simplicity, and meaning preservation.Then, we iteratively perform word and phrase-level edits on the complex sentence.Compared with previous approaches, our model does not require a parallel training set, but is more controllable and interpretable.Experiments on Newsela and WikiLarge datasets show that our approach is nearly as effective as state-of-the-art supervised approaches. 1

Automatic sentence summarization produces a shorter version of a sentence, while preserving its most important information.A good summary is characterized by language fluency and high information overlap with the source sentence.We model these two aspects in an unsupervised objective function, consisting of language modeling and semantic similarity metrics.We search for a high-scoring summary by discrete optimization.Our proposed method achieves a new state-of-the art for unsupervised sentence summarization according to ROUGE scores.Additionally, we demonstrate that the commonly reported ROUGE F1 metric is sensitive to summary length.Since this is unwillingly exploited in recent work, we emphasize that future evaluation should explicitly group summarization systems by output length brackets.1

Generating relevant responses in a dialog is challenging, and requires not only proper modeling of context in the conversation, but also being able to generate fluent sentences during inference.In this paper, we propose a two-step framework based on generative adversarial nets for generating conditioned responses.Our model first learns a meaningful representation of sentences by autoencoding, and then learns to map an input query to the response representation, which is in turn decoded as a response sentence.Both quantitative and qualitative evaluations show that our model generates more fluent, relevant, and diverse responses than existing state-of-the-art methods. 1

Variational auto-encoders (VAEs) are widely used in natural language generation due to the regularization of the latent space.However, generating sentences from the continuous latent space does not explicitly model the syntactic information.In this paper, we propose to generate sentences from disentangled syntactic and semantic spaces.Our proposed method explicitly models syntactic information in the VAE's latent space by using the linearized tree sequence, leading to better performance of language generation.Additionally, the advantage of sampling in the disentangled syntactic and semantic latent spaces enables us to perform novel applications, such as the unsupervised paraphrase generation and syntaxtransfer generation.Experimental results show that our proposed model achieves similar or better performance in various tasks, compared with state-of-the-art related work.

This paper tackles the problem of disentangling the latent representations of style and content in language models.We propose a simple yet effective approach, which incorporates auxiliary multi-task and adversarial objectives, for style prediction and bag-of-words prediction, respectively.We show, both qualitatively and quantitatively, that the style and content are indeed disentangled in the latent space.This disentangled latent representation learning can be applied to style transfer on non-parallel corpora.We achieve high performance in terms of transfer accuracy, content preservation, and language fluency, in comparison to various previous approaches. 1

The variational encoder-decoder (VED) encodes source information as a set of random variables using a neural network, which in turn is decoded into target data using another neural network. In natural language processing, sequence-to-sequence (Seq2Seq) models typically serve as encoder-decoder networks. When combined with a traditional (deterministic) attention mechanism, the variational latent space may be bypassed by the attention model, and thus becomes ineffective. In this paper, we propose a variational attention mechanism for VED, where the attention vector is also modeled as Gaussian distributed random variables. Results on two experiments show that, without loss of quality, our proposed method alleviates the bypassing phenomenon as it increases the diversity of generated sentences.

In exploratory search, how information is presented to the user and how the user interacts with the presented information heavily influence the user's success. In this paper, we examine two different spatial representations of search results: knowledge graphs and hierarchical trees. Through interaction logs we show that knowledge graphs can effectively reduce the need to read source content with no reduction in the quality of the information gathered by the user. Through qualitative interviews and thinkalounds we explore factors that influence user perception of different search results representations including biases, task, perceived structure of the data, and problem-solving approach. Overall, these results enhance our understanding of the role each of these representations can play in information seeking.

Nugget-based evaluations, such as those deployed in the TREC Temporal Summarization and Question Answering tracks, require human assessors to determine whether a nugget is present in a given piece of text. This process, known as nugget annotation, is labor-intensive. In this paper, we present two active learning techniques that prioritize the sequence in which candidate nugget/sentence pairs are presented to an assessor, based on the likelihood that the sentence contains a nugget. Our approach builds on the recognition that nugget annotation is similar to high-recall retrieval, and we adapt proven existing solutions. Simulation experiments with four existing TREC test collections show that our techniques yield far more matches for a given level of effort than baselines that are typically deployed in previous nugget-based evaluations.

The article describes a semi‐supervised approach to extracting multiword aspects of user‐written reviews that belong to a given category. The method starts with a small set of seed words, representing the target category, and calculates distributional similarity between the candidate and seed words. We compare 3 distributional similarity measures (Lin's,Weeds's, and balAPinc), and a document retrieval function,BM25, adapted as a word similarity measure. We then introduce a method for identifying multiword aspects by using a combination of syntactic rules and a co‐occurrence association measure. Finally, we describe a method for ranking multiword aspects by the likelihood of belonging to the target aspect category. The task used for evaluation is extraction of restaurant dish names from a corpus of restaurant reviews.

Introduction to Information Retrieval by Manning, Raghavan, and Sch ütze is an up-todate, thorough, and systematic introduction to information retrieval (IR) from a computer science perspective.Written as a textbook, its main audience is graduate and senior undergraduate students taking IR courses.The book will also be valuable to researchers in other computer science fields, such as computational linguistics, as well as to professional practitioners wishing to delve into the IR field.The book is structured into 21 chapters, which gradually unfold the subject of information retrieval, starting with the fundamentals (such as Boolean retrieval, document indexing, vector-space model, and evaluation in IR) and moving on to more advanced topics (such as probabilistic models, XML retrieval, text classification, machine learning for IR, document clustering, and Web retrieval).Pedagogical features of the book include short exercises at the end of each section and brief overviews of related research literature at the end of each chapter.Some of the major strengths of the book are its accessibility, clarity, and good balance between theory and practice.There are many concrete examples throughout the book that facilitate understanding of complex topics.Although the book covers a broad selection of the major established and emerging topics in IR, it largely bypasses two important subjects, in my opinion: natural language processing techniques in IR and interactive information retrieval.Although the authors refer to some research done in these areas in various chapters, they do not give them the same thorough treatment given to other topics in the book.To compensate, in the preface the authors provide references to the detailed coverage of these and some other topics in other textbooks.It also might have been useful if the authors introduced some specialized IR tasks, such as opinion retrieval or enterprise search, which might benefit from more advanced NLP techniques.Chapter 1 gives a succinct and focused introduction to the main concepts in IR, such as term, index, document, query, recall, precision, and so on.It outlines the main principles of Boolean retrieval, briefly criticizes it, and compares it to ranked retrieval.The authors also present a good real-world example of a commercial Boolean retrieval system.Chapter 2 provides a detailed discussion of the initial stages of the document indexing process that include tokenization, stemming and lemmatization, stopwords removal, and approaches to dealing with phrases at the indexing stage, namely bigram indexing and the use of positional indexes.In this chapter the authors discuss some linguistic aspects of these processes.For example, when examining tokenization, they discuss various morphological and other aspects of languages that complicate this process (e.g., hyphenation in English, compound nouns in German, and word

Evaluation measures act as objective functions to be optimized by information retrieval systems. Such objective functions must accurately reflect user requirements, particularly when tuning IR systems and learning ranking functions. Ambiguity in queries and redundancy in retrieved documents are poorly reflected by current evaluation measures. In this paper, we present a framework for evaluation that systematically rewards novelty and diversity. We develop this framework into a specific evaluation measure, based on cumulative gain. We demonstrate the feasibility of our approach using a test collection based on the TREC question answering track.

Mailing list archives in an enterprise are a valuable source for employees to dig into the past proceedings of the organization that could be relevant to their present task. Going through the proceedings of discussions about certain topics might be cumbersome and regular search techniques might not work in this context due to the genre that the documents belong to. In this paper, we propose methods, based on theory of subjectivity, to retrieve email messages that could contain argumentative discussions about the topic that the user is interested in.