Understanding the quality of a performance evaluation metric is crucial for ensuring that model outputs align with human preferences. However, it remains unclear how well each metric captures the diverse aspects of these preferences, as metrics often excel in one particular area but not across all dimensions. To address this, it is essential to systematically calibrate metrics to specific aspects of human preference, catering to the unique characteristics of each aspect. We introduce MetaMetrics, a calibrated meta-metric designed to evaluate generation tasks across different modalities in a supervised manner. MetaMetrics optimizes the combination of existing metrics to enhance their alignment with human preferences. Our metric demonstrates flexibility and effectiveness in both language and vision downstream tasks, showing significant benefits across various multilingual and multi-domain scenarios. MetaMetrics aligns closely with human preferences and is highly extendable and easily integrable into any application. This makes MetaMetrics a powerful tool for improving the evaluation of generation tasks, ensuring that metrics are more representative of human judgment across diverse contexts.

This paper investigates a novel approach to end-to-end speech translation (ST) based on aligning frozen pre-trained automatic speech recognition (ASR) and machine translation (MT) models via a small connector module (Q-Former, our Subsampler-Transformer Encoder). This connector bridges the gap between the speech and text modalities, transforming ASR encoder embeddings into the latent representation space of the MT encoder while being the only part of the system optimized during training. Experiments are conducted on the How2 English-Portuguese dataset as we investigate the alignment approach in a small-scale scenario focusing on ST. While keeping the size of the connector module constant and small in comparison ( < 5% of the size of the larger aligned models), increasing the size and capability of the foundation ASR and MT models universally improves translation results. We also find that the connectors can serve as domain adapters for the foundation MT models, significantly improving translation performance in the aligned ST setting. We conclude that this approach represents a viable and scalable approach to training end-to-end ST systems.

Sign language is a visual language that encompasses all linguistic features of natural languages and serves as the primary communication method for the deaf and hard-of-hearing communities. While many studies have successfully adapted pretrained language models (LMs) for sign language translation (sign-to-text), drawing inspiration from its linguistic characteristics, the reverse task of sign language generation (SLG, text-to-sign) remains largely unexplored. Most existing approaches treat SLG as a visual content generation task, employing techniques such as diffusion models to produce sign videos, 2D keypoints, or 3D avatars based on text inputs, overlooking the linguistic properties of sign languages. In this work, we introduce a multilingual sign language model, Signs as Tokens (SOKE), which can generate 3D sign avatars autoregressively from text inputs using a pretrained LM. To align sign language with the LM, we develop a decoupled tokenizer that discretizes continuous signs into token sequences representing various body parts. These sign tokens are integrated into the raw text vocabulary of the LM, allowing for supervised fine-tuning on sign language datasets. To facilitate multilingual SLG research, we further curate a large-scale Chinese sign language dataset, CSL-Daily, with high-quality 3D pose annotations. Extensive qualitative and quantitative evaluations demonstrate the effectiveness of SOKE. The project page is available at https://2000zrl.github.io/soke/.

Tokenization is a crucial step in processing protein sequences for machine learning models, as proteins are complex sequences of amino acids that require meaningful segmentation to capture their functional and structural properties. However, existing subword tokenization methods, developed primarily for human language, may be inadequate for protein sequences, which have unique patterns and constraints. This study evaluates three prominent tokenization approaches, Byte-Pair Encoding (BPE), WordPiece, and SentencePiece, across varying vocabulary sizes (400-6400), analyzing their effectiveness in protein sequence representation, domain boundary preservation, and adherence to established linguistic laws. Our comprehensive analysis reveals distinct behavioral patterns among these tokenizers, with vocabulary size significantly influencing their performance. BPE demonstrates better contextual specialization and marginally better domain boundary preservation at smaller vocabularies, while SentencePiece achieves better encoding efficiency, leading to lower fertility scores. WordPiece offers a balanced compromise between these characteristics. However, all tokenizers show limitations in maintaining protein domain integrity, particularly as vocabulary size increases. Analysis of linguistic law adherence shows partial compliance with Zipf's and Brevity laws but notable deviations from Menzerath's law, suggesting that protein sequences may follow distinct organizational principles from natural languages. These findings highlight the limitations of applying traditional NLP tokenization methods to protein sequences and emphasize the need for developing specialized tokenization strategies that better account for the unique characteristics of proteins.

Streaming generation models are increasingly utilized across various fields, with the Transducer architecture being particularly popular in industrial applications. However, its input-synchronous decoding mechanism presents challenges in tasks requiring non-monotonic alignments, such as simultaneous translation, leading to suboptimal performance in these contexts. In this research, we address this issue by tightly integrating Transducer's decoding with the history of input stream via a learnable monotonic attention mechanism. Our approach leverages the forwardbackward algorithm to infer the posterior probability of alignments between the predictor states and input timestamps, which is then used to estimate the context representations of monotonic attention in training. This allows Transducer models to adaptively adjust the scope of attention based on their predictions, avoiding the need to enumerate the exponentially large alignment space. Extensive experiments demonstrate that our MonoAttn-Transducer significantly enhances the handling of non-monotonic alignments in streaming generation, offering a robust solution for Transducer-based frameworks to tackle more complex streaming generation tasks. Unlike modern turn-based large language models, streaming models need to start generating the output before the input is completely read. Popular streaming generation methods can be broadly divided into two categories: Attentionbased Encoder-Decoder (AED; Bahdanau et al., 2015) and Transducer (Graves, 2012). Streaming AED models adapt the conventional sequence-to-sequence framework (Bahdanau, 2014) to support streaming generation. They often rely on an external policy module to determine the READ/WRITE actions in inference and to direct the scope of cross-attention in training. Examples include Wait-k policy (Ma et al., 2019) and monotonic attention-based methods (Raffel et al., 2017; Arivazhagan et al., 2019; Ma et al., 2020d; 2023a).

Sign language translation (SLT) is a challenging task that involves translating sign language images into spoken language. For SLT models to perform this task successfully, they must bridge the modality gap and identify subtle variations in sign language components to understand their meanings accurately. To address these challenges, we propose a novel gloss-free SLT framework called Multimodal Sign Language Translation (MMSLT), which leverages the representational capabilities of off-the-shelf multimodal large language models (MLLMs). Specifically, we generate detailed textual descriptions of sign language components using MLLMs. Then, through our proposed multimodal-language pre-training module, we integrate these description features with sign video features to align them within the spoken sentence space. Our approach achieves state-of-the-art performance on benchmark datasets PHOENIX14T and CSL-Daily, highlighting the potential of MLLMs to be effectively utilized in SLT.

Linguists have long held that a key aspect of natural language syntax is the recursive organization of language units into constituent structures, and research has suggested that current state-of-the-art language models lack an inherent bias towards this feature. A number of alternative models have been proposed to provide inductive biases towards constituency, including the Tree Transformer, which utilizes a modified attention mechanism to organize tokens into constituents. We investigate Tree Transformers to study whether they utilize meaningful and/or useful constituent structures. We pretrain a large Tree Transformer on language modeling in order to investigate the learned constituent tree representations of sentences, finding little evidence for meaningful structures. Next, we evaluate Tree Transformers with similar transformer models on error detection tasks requiring constituent structure. We find that while the Tree Transformer models may slightly outperform at these tasks, there is little evidence to suggest a meaningful improvement. In general, we conclude that there is little evidence to support Tree Transformer as an effective model of syntactic constituency.

Sign language processing has traditionally relied on task-specific models,limiting the potential for transfer learning across tasks. We introduce SHuBERT (Sign Hidden-Unit BERT), a self-supervised transformer encoder that learns strong representations from approximately 1,000 hours of American Sign Language (ASL) video content. Inspired by the success of the HuBERT speech representation model, SHuBERT adapts masked prediction for multi-stream visual sign language input, learning to predict multiple targets for corresponding to clustered hand, face, and body pose streams. SHuBERT achieves state-of-the-art performance across multiple benchmarks. On sign language translation, it outperforms prior methods trained on publicly available data on the How2Sign (+0.7 BLEU), OpenASL (+10.0 BLEU), and FLEURS-ASL (+0.3 BLEU) benchmarks. Similarly for isolated sign language recognition, SHuBERT's accuracy surpasses that of specialized models on ASL-Citizen (+5\%) and SEM-LEX (+20.6\%), while coming close to them on WLASL2000 (-3\%). Ablation studies confirm the contribution of each component of the approach.

Audio description (AD) is a crucial accessibility service provided to blind persons and persons with visual impairment, designed to convey visual information in acoustic form. Despite recent advancements in multilingual machine translation research, the lack of well-crafted and time-synchronized AD data impedes the development of audio description translation (ADT) systems that address the needs of multilingual countries such as Switzerland. Furthermore, since the majority of ADT systems rely solely on text, uncertainty exists as to whether incorporating visual information from the corresponding video clips can enhance the quality of ADT outputs. In this work, we present SwissADT, the first ADT system implemented for three main Swiss languages and English. By collecting well-crafted AD data augmented with video clips in German, French, Italian, and English, and leveraging the power of Large Language Models (LLMs), we aim to enhance information accessibility for diverse language populations in Switzerland by automatically translating AD scripts to the desired Swiss language. Our extensive experimental ADT results, composed of both automatic and human evaluations of ADT quality, demonstrate the promising capability of SwissADT for the ADT task. We believe that combining human expertise with the generation power of LLMs can further enhance the performance of ADT systems, ultimately benefiting a larger multilingual target population.

This study presents a comprehensive evaluation of GPT-4's translation capabilities compared to human translators of varying expertise levels. Through systematic human evaluation using the MQM schema, we assess translations across three language pairs (Chinese$\longleftrightarrow$English, Russian$\longleftrightarrow$English, and Chinese$\longleftrightarrow$Hindi) and three domains (News, Technology, and Biomedical). Our findings reveal that GPT-4 achieves performance comparable to junior-level translators in terms of total errors, while still lagging behind senior translators. Unlike traditional Neural Machine Translation systems, which show significant performance degradation in resource-poor language directions, GPT-4 maintains consistent translation quality across all evaluated language pairs. Through qualitative analysis, we identify distinctive patterns in translation approaches: GPT-4 tends toward overly literal translations and exhibits lexical inconsistency, while human translators sometimes over-interpret context and introduce hallucinations. This study represents the first systematic comparison between LLM and human translators across different proficiency levels, providing valuable insights into the current capabilities and limitations of LLM-based translation systems.