This paper is a technical report to share our experience and findings building a Korean and English bilingual multimodal model. While many of the multimodal datasets focus on English and multilingual multimodal research uses machine-translated texts, employing such machine-translated texts is limited to describing unique expressions, cultural information, and proper noun in languages other than English. In this work, we collect 1.1 billion image-text pairs (708 million Korean and 476 million English) and train a bilingual multimodal model named KELIP. We introduce simple yet effective training schemes, including MAE pre-training and multi-crop augmentation. Extensive experiments demonstrate that a model trained with such training schemes shows competitive performance in both languages. Moreover, we discuss multimodal-related research questions: 1) strong augmentation-based methods can distract the model from learning proper multimodal relations; 2) training multimodal model without cross-lingual relation can learn the relation via visual semantics; 3) our bilingual KELIP can capture cultural differences of visual semantics for the same meaning of words; 4) a large-scale multimodal model can be used for multimodal feature analogy. We hope that this work will provide helpful experience and findings for future research. We provide an open-source pre-trained KELIP.
This paper presents the slurk software, a lightweight interaction server for setting up dialog data collections and running experiments. Slurk enables a multitude of settings including text-based, speech and video interaction between two or more humans or humans and bots, and a multimodal display area for presenting shared or private interactive context. The software is implemented in Python with an HTML and JS frontend that can easily be adapted to individual needs. It also provides a setup for pairing participants on common crowdworking platforms such as Amazon Mechanical Turk and some example bot scripts for common interaction scenarios.
We address causal inference with text documents. For example, does adding a theorem to a paper affect its chance of acceptance? Does reporting the gender of a forum post author affect the popularity of the post? We estimate these effects from observational data, where they may be confounded by features of the text such as the subject or writing quality. Although the text suffices for causal adjustment, it is prohibitively high-dimensional. The challenge is to find a low-dimensional text representation that can be used in causal inference. A key insight is that causal adjustment requires only the aspects of text that are predictive of both the treatment and outcome. Our proposed method adapts deep language models to learn low-dimensional embeddings from text that predict these values well; these embeddings suffice for causal adjustment. We establish theoretical properties of this method. We study it empirically on semi-simulated and real data on paper acceptance and forum post popularity. Code is available at https://github.com/blei-lab/causal-text-embeddings.
The attention-based encoder-decoder framework has recently achieved impressive results for scene text recognition, and many variants have emerged with improvements in recognition quality. However, it performs poorly on contextless texts (e.g., random character sequences) which is unacceptable in most of real application scenarios. In this paper, we first deeply investigate the decoding process of the decoder. We empirically find that a representative character-level sequence decoder utilizes not only context information but also positional information. Contextual information, which the existing approaches heavily rely on, causes the problem of attention drift. To suppress such side-effect, we propose a novel position enhancement branch, and dynamically fuse its outputs with those of the decoder attention module for scene text recognition. Specifically, it contains a position aware module to enable the encoder to output feature vectors encoding their own spatial positions, and an attention module to estimate glimpses using the positional clue (i.e., the current decoding time step) only. The dynamic fusion is conducted for more robust feature via an element-wise gate mechanism. Theoretically, our proposed method, dubbed \emph{RobustScanner}, decodes individual characters with dynamic ratio between context and positional clues, and utilizes more positional ones when the decoding sequences with scarce context, and thus is robust and practical. Empirically, it has achieved new state-of-the-art results on popular regular and irregular text recognition benchmarks while without much performance drop on contextless benchmarks, validating its robustness in both contextual and contextless application scenarios.
Research on definition extraction has been conducted for well over a decade, largely with significant constraints on the type of definitions considered. In this work, we present DeftEval, a SemEval shared task in which participants must extract definitions from free text using a term-definition pair corpus that reflects the complex reality of definitions in natural language. Definitions and glosses in free text often appear without explicit indicators, across sentences boundaries, or in an otherwise complex linguistic manner. DeftEval involved 3 distinct subtasks: 1)Sentence classification, 2) sequence labeling, and 3) relation extraction.
Unlike classical lexical overlap metrics such as BLEU, most current evaluation metrics (such as BERTScore or MoverScore) are based on black-box language models such as BERT or XLM-R. They often achieve strong correlations with human judgments, but recent research indicates that the lower-quality classical metrics remain dominant, one of the potential reasons being that their decision processes are transparent. To foster more widespread acceptance of the novel high-quality metrics, explainability thus becomes crucial. In this concept paper, we identify key properties and propose key goals of explainable machine translation evaluation metrics. We also provide a synthesizing overview over recent approaches for explainable machine translation metrics and discuss how they relate to those goals and properties. Further, we conduct own novel experiments, which (among others) find that current adversarial NLP techniques are unsuitable for automatically identifying limitations of high-quality black-box evaluation metrics, as they are not meaning-preserving. Finally, we provide a vision of future approaches to explainable evaluation metrics and their evaluation. We hope that our work can help catalyze and guide future research on explainable evaluation metrics and, mediately, also contribute to better and more transparent text generation systems.
Machine learning methodologies can be adopted in cultural applications and propose new ways to distribute or even present the cultural content to the public. For instance, speech analytics can be adopted to automatically generate subtitles in theatrical plays, in order to (among other purposes) help people with hearing loss. Apart from a typical speech-to-text transcription with Automatic Speech Recognition (ASR), Speech Emotion Recognition (SER) can be used to automatically predict the underlying emotional content of speech dialogues in theatrical plays, and thus to provide a deeper understanding how the actors utter their lines. However, real-world datasets from theatrical plays are not available in the literature. In this work we present GreThE, the Greek Theatrical Emotion dataset, a new publicly available data collection for speech emotion recognition in Greek theatrical plays. The dataset contains utterances from various actors and plays, along with respective valence and arousal annotations. Towards this end, multiple annotators have been asked to provide their input for each speech recording and inter-annotator agreement is taken into account in the final ground truth generation. In addition, we discuss the results of some indicative experiments that have been conducted with machine and deep learning frameworks, using the dataset, along with some widely used databases in the field of speech emotion recognition.
Similarity query is the family of queries based on some similarity metrics. Unlike the traditional database queries which are mostly based on value equality, similarity queries aim to find targets "similar enough to" the given data objects, depending on some similarity metric, e.g., Euclidean distance, cosine similarity and so on. To measure the similarity between data objects, traditional methods normally work on low level or syntax features(e.g., basic visual features on images or bag-of-word features of text), which makes them weak to compute the semantic similarities between objects. So for measuring data similarities semantically, neural embedding is applied. Embedding techniques work by representing the raw data objects as vectors (so called "embeddings" or "neural embeddings" since they are mostly generated by neural network models) that expose the hidden semantics of the raw data, based on which embeddings do show outstanding effectiveness on capturing data similarities, making it one of the most widely used and studied techniques in the state-of-the-art similarity query processing research. But there are still many open challenges on the efficiency of embedding based similarity query processing, which are not so well-studied as the effectiveness. In this survey, we first provide an overview of the "similarity query" and "similarity query processing" problems. Then we talk about recent approaches on designing the indexes and operators for highly efficient similarity query processing on top of embeddings (or more generally, high dimensional data). Finally, we investigate the specific solutions with and without using embeddings in selected application domains of similarity queries, including entity resolution and information retrieval. By comparing the solutions, we show how neural embeddings benefit those applications.
Fake news and misinformation are a matter of concern for people around the globe. Users of the internet and social media sites encounter content with false information much frequently. Fake news detection is one of the most analyzed and prominent areas of research. These detection techniques apply popular machine learning and deep learning algorithms. Previous work in this domain covers fake news detection vastly among text circulating online. Platforms that have extensively been observed and analyzed include news websites and Twitter. Facebook, Reddit, WhatsApp, YouTube, and other social applications are gradually gaining attention in this emerging field. Researchers are analyzing online data based on multiple modalities composed of text, image, video, speech, and other contributing factors. The combination of various modalities has resulted in efficient fake news detection. At present, there is an abundance of surveys consolidating textual fake news detection algorithms. This review primarily deals with multi-modal fake news detection techniques that include images, videos, and their combinations with text. We provide a comprehensive literature survey of eighty articles presenting state-of-the-art detection techniques, thereby identifying research gaps and building a pathway for researchers to further advance this domain.
Topic models are popular models for analyzing a collection of text documents. The models assert that documents are distributions over latent topics and latent topics are distributions over words. A nested document collection is where documents are nested inside a higher order structure such as stories in a book, articles in a journal, or web pages in a web site. In a single collection of documents, topics are global, or shared across all documents. For web pages nested in web sites, topic frequencies likely vary between web sites. Within a web site, topic frequencies almost certainly vary between web pages. A hierarchical prior for topic frequencies models this hierarchical structure and specifies a global topic distribution. Web site topic distributions vary around the global topic distribution and web page topic distributions vary around the web site topic distribution. In a nested collection of web pages, some topics are likely unique to a single web site. Local topics in a nested collection of web pages are topics unique to one web site. For US local health department web sites, brief inspection of the text shows local geographic and news topics specific to each department that are not present in others. Topic models that ignore the nesting may identify local topics, but do not label topics as local nor do they explicitly identify the web site owner of the local topic. For web pages nested inside web sites, local topic models explicitly label local topics and identifies the owning web site. This identification can be used to adjust inferences about global topics. In the US public health web site data, topic coverage is defined at the web site level after removing local topic words from pages. Hierarchical local topic models can be used to identify local topics, adjust inferences about if web sites cover particular health topics, and study how well health topics are covered.