Recently multi-lingual pre-trained language models (PLM) such as mBERT and XLM-R have achieved impressive strides in cross-lingual dense retrieval. Despite its successes, they are general-purpose PLM while the multilingual PLM tailored for cross-lingual retrieval is still unexplored. Motivated by an observation that the sentences in parallel documents are approximately in the same order, which is universal across languages, we propose to model this sequential sentence relation to facilitate cross-lingual representation learning. Specifically, we propose a multilingual PLM called masked sentence model (MSM), which consists of a sentence encoder to generate the sentence representations, and a document encoder applied to a sequence of sentence vectors from a document. The document encoder is shared for all languages to model the universal sequential sentence relation across languages. To train the model, we propose a masked sentence prediction task, which masks and predicts the sentence vector via a hierarchical contrastive loss with sampled negatives. Comprehensive experiments on four cross-lingual retrieval tasks show MSM significantly outperforms existing advanced pre-training models, demonstrating the effectiveness and stronger cross-lingual retrieval capabilities of our approach. Code and model will be available. Cross-lingual retrieval (also including multi-lingual retrieval) is becoming increasingly important as new texts in different languages are being generated every day, and people query and search for the relevant documents in different languages (Zhang et al., 2021b; Asai et al., 2021a). This is a fundamental and challenging task and plays an essential part in real-world search engines, for example, Google and Bing search which serve hundreds of countries across diverse languages. In addition, it's also a vital component to solve many cross-lingual downstream problems, such as open-domain question answering (Asai et al., 2021a) or fact checking (Huang et al., 2022). With the rapid development of deep neural models, cross-lingual retrieval has progressed from translation-based methods (Nie, 2010), cross-lingual word embeddings (Sun & Duh, 2020), and now to dense retrieval built on the top of multi-lingual pre-trained models (Devlin et al., 2019; Conneau et al., 2019).

In this work, we provide a literature review of active learning (AL) for its applications in natural language processing (NLP). In addition to a fine-grained categorization of query strategies, we also investigate several other important aspects of applying AL to NLP problems. These include AL for structured prediction tasks, annotation cost, model learning (especially Figure 1: Counts of AL (left) and "neural" (right) papers with deep neural models), and starting in the ACL Anthology over the past twenty years.

Multimodality Representation Learning, as a technique of learning to embed information from different modalities and their correlations, has achieved remarkable success on a variety of applications, such as Visual Question Answering (VQA), Natural Language for Visual Reasoning (NLVR), and Vision Language Retrieval (VLR). Among these applications, cross-modal interaction and complementary information from different modalities are crucial for advanced models to perform any multimodal task, e.g., understand, recognize, retrieve, or generate optimally. Researchers have proposed diverse methods to address these tasks. The different variants of transformer-based architectures performed extraordinarily on multiple modalities. This survey presents the comprehensive literature on the evolution and enhancement of deep learning multimodal architectures to deal with textual, visual and audio features for diverse cross-modal and modern multimodal tasks. This study summarizes the (i) recent task-specific deep learning methodologies, (ii) the pretraining types and multimodal pretraining objectives, (iii) from state-of-the-art pretrained multimodal approaches to unifying architectures, and (iv) multimodal task categories and possible future improvements that can be devised for better multimodal learning. Moreover, we prepare a dataset section for new researchers that covers most of the benchmarks for pretraining and finetuning. Finally, major challenges, gaps, and potential research topics are explored. A constantly-updated paperlist related to our survey is maintained at https://github.com/marslanm/multimodality-representation-learning.

This study investigates the use of Visually Grounded Speech (VGS) models for keyword localisation in speech. The study focusses on two main research questions: (1) Is keyword localisation possible with VGS models and (2) Can keyword localisation be done cross-lingually in a real low-resource setting? Four methods for localisation are proposed and evaluated on an English dataset, with the best-performing method achieving an accuracy of 57%. A new dataset containing spoken captions in Yoruba language is also collected and released for cross-lingual keyword localisation. The cross-lingual model obtains a precision of 16% in actual keyword localisation and this performance can be improved by initialising from a model pretrained on English data. The study presents a detailed analysis of the model's success and failure modes and highlights the challenges of using VGS models for keyword localisation in low-resource settings.

In recent years, there has been a heightened consensus - both within academia and in the public discourse - that Social Media Platforms (SMPs), amplify the spread of hateful and negative sentiment content. Researchers have identified how hateful content, political propaganda, and targeted messaging contributed to real-world harms including insurrections against democratically elected governments, genocide, and breakdown of social cohesion due to heightened negative discourse towards certain communities in parts of the world. To counter these issues, SMPs have created semi-automated systems that can help identify toxic speech. In this paper we analyse the statistical distribution of hateful and negative sentiment contents within a representative Facebook dataset (n= 604,703) scrapped through 648 public Facebook pages which identify themselves as proponents (and followers) of far-right Hindutva actors. These pages were identified manually using keyword searches on Facebook and on CrowdTangleand classified as far-right Hindutva pages based on page names, page descriptions, and discourses shared on these pages. We employ state-of-the-art, open-source XLM-T multilingual transformer-based language models to perform sentiment and hate speech analysis of the textual contents shared on these pages over a period of 5.5 years. The result shows the statistical distributions of the predicted sentiment and the hate speech labels; top actors, and top page categories. We further discuss the benchmark performances and limitations of these pre-trained language models.

Archive collections are nowadays mostly available through search engines interfaces, which allow a user to retrieve documents by issuing queries. The study of these collections may be, however, impaired by some aspects of search engines, such as the overwhelming number of documents returned or the lack of contextual knowledge provided. New methods that could work independently or in combination with search engines are then required to access these collections. In this position paper, we propose to extend TimeLine Summarization (TLS) methods on archive collections to assist in their studies. We provide an overview of existing TLS methods and we describe a conceptual framework for an Archive TimeLine Summarization (ATLS) system, which aims to generate informative, readable and interpretable timelines.

Transfomer-based approaches advance the recent development of multi-camera 3D detection both in academia and industry. In a vanilla transformer architecture, queries are randomly initialised and optimised for the whole dataset, without considering the differences among input frames. In this work, we propose to leverage the predictions from an image backbone, which is often highly optimised for 2D tasks, as priors to the transformer part of a 3D detection network. The method works by (1). augmenting image feature maps with 2D priors, (2). sampling query locations via ray-casting along 2D box centroids, as well as (3). initialising query features with object-level image features. Experimental results shows that 2D priors not only help the model converge faster, but also largely improve the baseline approach by up to 12% in terms of average precision.

Modern data search is a complex domain. Vector similarity search, or VSS, represents data with contextual depth and returns more relevant information to the consumers in response to a search query. Let's take a simple example. Search queries like "data science" and "science fiction" refer to different types of content despite both having a common word ("science"). A traditional search technique would match common phrases to return relevant results, which would be inaccurate in this case.

Although the great success of open-domain dialogue generation, unseen entities can have a large impact on the dialogue generation task. It leads to performance degradation of the model in the dialog generation. Previous researches used retrieved knowledge of seen entities as the auxiliary data to enhance the representation of the model. Nevertheless, logical explanation of unseen entities remains unexplored, such as possible co-occurrence or semantically similar words of them and their entity category. In this work, we propose an approach to address the challenge above. We construct a graph by extracting entity nodes in them, enhancing the representation of the context of the unseen entity with the entity's 1-hop surrounding nodes. Furthermore, We added the named entity tag prediction task to apply the problem that the unseen entity does not exist in the graph. We conduct our experiments on an open dataset Wizard of Wikipedia and the empirical results indicate that our approach outperforms the state-of-the-art approaches on Wizard of Wikipedia.

Real-time social media data can provide useful information on evolving hazards. Alongside traditional methods of disaster detection, the integration of social media data can considerably enhance disaster management. In this paper, we investigate the problem of detecting geolocation-content communities on Twitter and propose a novel distributed system that provides in near real-time information on hazard-related events and their evolution. We show that content-based community analysis leads to better and faster dissemination of reports on hazards. Our distributed disaster reporting system analyzes the social relationship among worldwide geolocated tweets, and applies topic modeling to group tweets by topics. Considering for each tweet the following information: user, timestamp, geolocation, retweets, and replies, we create a publisher-subscriber distribution model for topics. We use content similarity and the proximity of nodes to create a new model for geolocation-content based communities. Users can subscribe to different topics in specific geographical areas or worldwide and receive real-time reports regarding these topics. As misinformation can lead to increase damage if propagated in hazards related tweets, we propose a new deep learning model to detect fake news. The misinformed tweets are then removed from display. We also show empirically the scalability capabilities of the proposed system.