The performance of spoofing countermeasure systems depends fundamentally upon the use of sufficiently representative training data. With this usually being limited, current solutions typically lack generalisation to attacks encountered in the wild. Strategies to improve reliability in the face of uncontrolled, unpredictable attacks are hence needed. We report in this paper our efforts to use self-supervised learning in the form of a wav2vec 2.0 front-end with fine tuning. Despite initial base representations being learned using only bona fide data and no spoofed data, we obtain the lowest equal error rates reported in the literature for both the ASVspoof 2021 Logical Access and Deepfake databases. When combined with data augmentation,these results correspond to an improvement of almost 90% relative to our baseline system.

Large language models (LLMs) have been widely used for relevance assessment in information retrieval. However, our study demonstrates that combining two distinct small language models (SLMs) with different architectures can outperform LLMs in this task. Our approach -- QUPID -- integrates a generative SLM with an embedding-based SLM, achieving higher relevance judgment accuracy while reducing computational costs compared to state-of-the-art LLM solutions. This computational efficiency makes QUPID highly scalable for real-world search systems processing millions of queries daily. In experiments across diverse document types, our method demonstrated consistent performance improvements (Cohen's Kappa of 0.646 versus 0.387 for leading LLMs) while offering 60x faster inference times. Furthermore, when integrated into production search pipelines, QUPID improved nDCG@5 scores by 1.9%. These findings underscore how architectural diversity in model combinations can significantly enhance both search relevance and operational efficiency in information retrieval systems.

Vision-Language Pretrained (VLP) models have achieved impressive performance on multimodal tasks, including text-image retrieval, based on dense representations. Meanwhile, Learned Sparse Retrieval (LSR) has gained traction in text-only settings due to its interpretability and efficiency with fast term-based lookup via inverted indexes. Inspired by these advantages, recent work has extended LSR to the multimodal domain. However, these methods often rely on computationally expensive contrastive pre-training, or distillation from a frozen dense model, which limits the potential for mutual enhancement. To address these limitations, we propose a simple yet effective framework that enables bi-directional learning between dense and sparse representations through Self-Knowledge Distillation. This bi-directional learning is achieved using an integrated similarity score-a weighted sum of dense and sparse similarities-which serves as a shared teacher signal for both representations. To ensure efficiency, we fine-tune the final layer of the dense encoder and the sparse projection head, enabling easy adaptation of any existing VLP model. Experiments on MSCOCO and Flickr30k demonstrate that our sparse retriever not only outperforms existing sparse baselines, but also achieves performance comparable to-or even surpassing-its dense counterparts, while retaining the benefits of sparse models.

For successful scene text recognition (STR) models, synthetic text image generators have alleviated the lack of annotated text images from the real world. Specifically, they generate multiple text images with diverse backgrounds, font styles, and text shapes and enable STR models to learn visual patterns that might not be accessible from manually annotated data. In this paper, we introduce a new synthetic text image generator, SynthTIGER, by analyzing techniques used for text image synthesis and integrating effective ones under a single algorithm. Moreover, we propose two techniques that alleviate the long-tail problem in length and character distributions of training data. In our experiments, SynthTIGER achieves better STR performance than the combination of synthetic datasets, MJSynth (MJ) and SynthText (ST). Our ablation study demonstrates the benefits of using sub-components of SynthTIGER and the guideline on generating synthetic text images for STR models. Our implementation is publicly available at this https URL.

The objective of this paper is to learn representations of speaker identity without access to manually annotated data. To do so, we develop a self-supervised learning objective that exploits the natural cross-modal synchrony between faces and audio in video. The key idea behind our approach is to tease apart--without annotation--the representations of linguistic content and speaker identity. We construct a two-stream architecture which: (1) shares low-level features common to both representations; and (2) provides a natural mechanism for explicitly disentangling these factors, offering the potential for greater generalisation to novel combinations of content and identity and ultimately producing speaker identity representations that are more robust. We train our method on a large-scale audio-visual dataset of talking heads `in the wild', and demonstrate its efficacy by evaluating the learned speaker representations for standard speaker recognition performance.

The objective of this paper is 'open-set' speaker recognition of unseen speakers, where ideal embeddings should be able to condense information into a compact utterance-level representation that has small intra-speaker and large inter-speaker distance. A popular belief in speaker recognition is that networks trained with classification objectives outperform metric learning methods. In this paper, we present an extensive evaluation of most popular loss functions for speaker recognition on the VoxCeleb dataset. We demonstrate that the vanilla triplet loss shows competitive performance compared to classification-based losses, and those trained with our proposed metric learning objective outperform state-of-the-art methods.

This paper summarises the findings from the VoxCeleb Speaker Recognition Challenge 2022 (VoxSRC-22), which was held in conjunction with INTERSPEECH 2022. The goal of this challenge was to evaluate how well state-of-the-art speaker recognition systems can diarise and recognise speakers from speech obtained "in the wild". The challenge consisted of: (i) the provision of publicly available speaker recognition and diarisation data from YouTube videos together with ground truth annotation and standardised evaluation software; and (ii) a public challenge and hybrid workshop held at INTERSPEECH 2022. We describe the four tracks of our challenge along with the baselines, methods, and results. We conclude with a discussion on the new domain-transfer focus of VoxSRC-22, and on the progression of the challenge from the previous three editions.

Artefacts that serve to distinguish bona fide speech from spoofed or deepfake speech are known to reside in specific subbands and temporal segments. Various approaches can be used to capture and model such artefacts, however, none works well across a spectrum of diverse spoofing attacks. Reliable detection then often depends upon the fusion of multiple detection systems, each tuned to detect different forms of attack. In this paper we show that better performance can be achieved when the fusion is performed within the model itself and when the representation is learned automatically from raw waveform inputs. The principal contribution is a spectro-temporal graph attention network (GAT) which learns the relationship between cues spanning different sub-bands and temporal intervals. Using a model-level graph fusion of spectral (S) and temporal (T) sub-graphs and a graph pooling strategy to improve discrimination, the proposed RawGAT-ST model achieves an equal error rate of 1.06 % for the ASVspoof 2019 logical access database. This is one of the best results reported to date and is reproducible using an open source implementation.

Large Language Models (LLMs) excel at understanding the semantic relationships between queries and documents, even with lengthy and complex long-tail queries. These queries are challenging for feedback-based rankings due to sparse user engagement and limited feedback, making LLMs' ranking ability highly valuable. However, the large size and slow inference of LLMs necessitate the development of smaller, more efficient models (sLLMs). Recently, integrating ranking label generation into distillation techniques has become crucial, but existing methods underutilize LLMs' capabilities and are cumbersome. Our research, RRADistill: Re-Ranking Ability Distillation, propose an efficient label generation pipeline and novel sLLM training methods for both encoder and decoder models. We introduce an encoder-based method using a Term Control Layer to capture term matching signals and a decoder-based model with a ranking layer for enhanced understanding. A/B testing on a Korean-based search platform, validates the effectiveness of our approach in improving re-ranking for long-tail queries.

Recent progress in fine-grained gesture and action classification, and machine translation, point to the possibility of automated sign language recognition becoming a reality. A key stumbling block in making progress towards this goal is a lack of appropriate training data, stemming from the high complexity of sign annotation and a limited supply of qualified annotators. In this work, we introduce a new scalable approach to data collection for sign recognition in continuous videos. We make use of weakly-aligned subtitles for broadcast footage together with a keyword spotting method to automatically localise sign-instances for a vocabulary of 1,000 signs in 1,000 hours of video. We make the following contributions: (1) We show how to use mouthing cues from signers to obtain high-quality annotations from video data - the result is the BSL-1K dataset, a collection of British Sign Language (BSL) signs of unprecedented scale; (2) We show that we can use BSL-1K to train strong sign recognition models for co-articulated signs in BSL and that these models additionally form excellent pretraining for other sign languages and benchmarks - we exceed the state of the art on both the MSASL and WLASL benchmarks. Finally, (3) we propose new large-scale evaluation sets for the tasks of sign recognition and sign spotting and provide baselines which we hope will serve to stimulate research in this area.

The goal of this paper is to learn robust speaker representation for bilingual speaking scenario. The majority of the world's population speak at least two languages; however, most speaker recognition systems fail to recognise the same speaker when speaking in different languages. Popular speaker recognition evaluation sets do not consider the bilingual scenario, making it difficult to analyse the effect of bilingual speakers on speaker recognition performance. In this paper, we publish a large-scale evaluation set named VoxCeleb1-B derived from VoxCeleb that considers bilingual scenarios. We introduce an effective disentanglement learning strategy that combines adversarial and metric learning-based methods. This approach addresses the bilingual situation by disentangling language-related information from speaker representation while ensuring stable speaker representation learning. Our language-disentangled learning method only uses language pseudo-labels without manual information.

The goal of this work is to determine 'who spoke when' in real-world meetings. The method takes surround-view video and single or multi-channel audio as inputs, and generates robust diarisation outputs. To achieve this, we propose a novel iterative approach that first enrolls speaker models using audio-visual correspondence, then uses the enrolled models together with the visual information to determine the active speaker. We show strong quantitative and qualitative performance on a dataset of real-world meetings. The method is also evaluated on the public AMI meeting corpus, on which we demonstrate results that exceed all comparable methods. We also show that beamforming can be used together with the video to further improve the performance when multi-channel audio is available.

As the number of video content has mushroomed in recent years, automatic video summarization has come useful when we want to just peek at the content of the video. However, there are two underlying limitations in generic video summarization task. First, most previous approaches read in just visual features as input, leaving other modality features behind. Second, existing datasets for generic video summarization are relatively insufficient to train a caption generator used for extracting text information from a video and to train the multimodal feature extractors. To address these two problems, this paper proposes the Multimodal Frame-Scoring Transformer (MFST), a framework exploiting visual, text, and audio features and scoring a video with respect to frames. Our MFST framework first extracts each modality features (audio-visual-text) using pretrained encoders. Then, MFST trains the multimodal frame-scoring transformer that uses multimodal representation based on extracted features as inputs and predicts frame-level scores. Our extensive experiments with previous models and ablation studies on TVSum and SumMe datasets demonstrate the effectiveness and superiority of our proposed method by a large margin in both F1 score and Rank-based evaluation.

Audio anti-spoofing for automatic speaker verification aims to safeguard users' identities from spoofing attacks. Although state-of-the-art spoofing countermeasure(CM) models perform well on specific datasets, they lack generalization when evaluated with different datasets. To address this limitation, previous studies have explored large pre-trained models, which require significant resources and time. We aim to develop a compact but well-generalizing CM model that can compete with large pre-trained models. Our approach involves multi-dataset co-training and sharpness-aware minimization, which has not been investigated in this domain. Extensive experiments reveal that proposed method yield competitive results across various datasets while utilizing 4,000 times less parameters than the large pre-trained models.

Feature selection, which is a technique to select key features in recommender systems, has received increasing research attention. Recently, Adaptive Feature Selection (AdaFS) has shown remarkable performance by adaptively selecting features for each data instance, considering that the importance of a given feature field can vary significantly across data. However, this method still has limitations in that its selection process could be easily biased to major features that frequently occur. To address these problems, we propose Multi-view Feature Selection (MvFS), which selects informative features for each instance more effectively. Most importantly, MvFS employs a multi-view network consisting of multiple sub-networks, each of which learns to measure the feature importance of a part of data with different feature patterns. By doing so, MvFS mitigates the bias problem towards dominant patterns and promotes a more balanced feature selection process. Moreover, MvFS adopts an effective importance score modeling strategy which is applied independently to each field without incurring dependency among features. Experimental results on real-world datasets demonstrate the effectiveness of MvFS compared to state-of-the-art baselines.

In this paper, we propose a novel end-to-end user-defined keyword spotting method that utilizes linguistically corresponding patterns between speech and text sequences. Unlike previous approaches requiring speech keyword enrollment, our method compares input queries with an enrolled text keyword sequence. To place the audio and text representations within a common latent space, we adopt an attention-based cross-modal matching approach that is trained in an end-to-end manner with monotonic matching loss and keyword classification loss. We also utilize a de-noising loss for the acoustic embedding network to improve robustness in noisy environments. Additionally, we introduce the LibriPhrase dataset, a new short-phrase dataset based on LibriSpeech for efficiently training keyword spotting models. Our proposed method achieves competitive results on various evaluation sets compared to other single-modal and cross-modal baselines.