Automatic pronunciation assessment (APA) seeks to quantify a second language (L2) learner’s pronunciation proficiency in a target language by offering timely and fine-grained diagnostic feedback. Most existing efforts on APA have predominantly concentrated on highly constrained reading-aloud tasks (where learners are prompted to read a reference text aloud); however, assessing pronunciation quality in unscripted speech (or free-speaking scenarios) remains relatively underexplored. In light of this, we first propose HiPPO, a hierarchical pronunciation assessment model tailored for spoken languages, which evaluates an L2 learner’s oral proficiency at multiple linguistic levels based solely on the speech uttered by the learner. To improve the overall accuracy of assessment, a contrastive ordinal regularizer and a curriculum learning strategy are introduced for model training. The former aims to generate score-discriminative features by exploiting the ordinal nature of regression targets, while the latter gradually ramps up the training complexity to facilitate the assessment task that takes unscripted speech as input. Experiments conducted on the Speechocean762 benchmark dataset validates the feasibility and superiority of our method in relation to several cutting-edge baselines.

Prior efforts in building computer-assisted pronunciation training (CAPT) systems often treat automatic pronunciation assessment (APA) and mispronunciation detection and diagnosis (MDD) as separate fronts: the former aims to provide multiple pronunciation aspect scores across diverse linguistic levels, while the latter focuses instead on pinpointing the precise phonetic pronunciation errors made by non-native language learners. However, it is generally expected that a full-fledged CAPT system should perform both functionalities simultaneously and efficiently. In response to this surging demand, we in this work first propose HMamba, a novel CAPT approach that seamlessly integrates APA and MDD tasks in parallel. In addition, we introduce a novel loss function, decoupled cross-entropy loss (deXent), specifically tailored for MDD to facilitate better-supervised learning for detecting mispronounced phones, thereby enhancing overall performance. A comprehensive set of empirical results on the speechocean762 benchmark dataset demonstrates the effectiveness of our approach on APA. Notably, our proposed approach also yields a considerable improvement in MDD performance over a strong baseline, achieving an F1-score of 63.85%. Our codes are made available at https://github.com/Fuann/hmamba

Retrieval-Augmented Generation (RAG) has proven effective for text-only question answering, yet expanding it to visually rich documents remains a challenge. Existing multimodal benchmarks, often derived from visual question answering (VQA) datasets, or large vision-language model (LVLM)-generated query-image pairs, which often contain underspecified questions that assume direct image access. To mitigate this issue, we propose a two-stage query rewriting framework that first generates OCR-based image descriptions and then reformulates queries into precise, retrieval-friendly forms under explicit constraints. Experiments show consistent improvements across dense, hybrid and multimodal retrieval paradigms, with the most pronounced gains in visual document retrieval – Hits@1 rises from 21.0% to 56.6% with VDocRetriever and further to 79.3% when OCR-based descriptions are incorporated. These results indicate that query rewriting, particularly when combined with multimodal fusion, provides a reliable and scalable solution to bridge underspecified queries and improve retrieval over visually rich documents.

End-to-End Neural Diarization (EEND) has undergone substantial development, particularly with powerset classification methods that enhance performance but can exacerbate speaker confusion. To address this, we propose a novel training strategy that complements the standard cross entropy loss with an auxiliary ordinal log loss, guided by a distance matrix of speaker combinations. Our experiments reveal that while this approach yields significant relative improvements of 15.8% in false alarm rate and 10.0% in confusion error rate, it also uncovers a critical trade-off with an increased missed error rate. The primary contribution of this work is the identification and analysis of this trade-off, which stems from the model adopting a more conservative prediction strategy. This insight is crucial for designing more balanced and effective loss functions in speaker diarization.

Automated speaking assessment (ASA) has become a crucial component in computer-assisted language learning, providing scalable, objective, and timely feedback to second-language learners. While early ASA systems relied on hand-crafted features and shallow classifiers, recent advances in self-supervised learning (SSL) have enabled richer representations for both text and speech, improving assessment accuracy. Despite these advances, challenges remain in evaluating long speech responses, due to limited labeled data, class imbalance, and the importance of pronunciation clarity and fluency, especially for read-aloud tasks. In this work, we propose a segment-based ASA framework leveraging WhisperX to split long responses into shorter fragments, generate weak labels from holistic scores, and aggregate segment-level predictions to obtain final proficiency scores. Experiments on the GEPT corpus demonstrate that our framework outperforms baseline holistic models, generalizes robustly to unseen prompts and speakers, and provides diagnostic insights at both segment and response levels.

Automatic speech recognition (ASR) for low-resource languages such as Taiwanese Hokkien is difficult due to the scarcity of annotated data. However, direct fine-tuning on Han-character transcriptions often fails to capture detailed phonetic and tonal cues, while training only on romanization lacks lexical and syntactic coverage. In addition, prior studies have rarely explored staged strategies that integrate both annotation types. To address this gap, we present CLiFT-ASR, a cross-lingual fine-tuning framework that builds on Mandarin HuBERT models and progressively adapts them to Taiwanese Hokkien. The framework employs a two-stage process in which it first learns acoustic and tonal representations from phonetic Tai-lo annotations and then captures vocabulary and syntax from Han-character transcriptions. This progressive adaptation enables effective alignment between speech sounds and orthographic structures. Experiments on the TAT-MOE corpus demonstrate that CLiFT-ASR achieves a 24.88% relative reduction in character error rate (CER) compared with strong baselines. The results indicate that CLiFT-ASR provides an effective and parameter-efficient solution for Taiwanese Hokkien ASR and that it has potential to benefit other low-resource language scenarios.

Sentence stress reflects the relative prominence of words within a sentence. It is fundamental to speech intelligibility and naturalness, and is particularly important in second language (L2) learning. Accurate stress production facilitates effective communication and reduces misinterpretation. In this work, we investigate sentence stress detection (SSD) using Whisper-based transformer speech models under diverse settings, including model scaling, backbone–decoder interactions, architectural and regularization enhancements, and embedding visualization for interpretability. Results show that smaller Whisper variants achieve stronger performance under limited data, while architectural and regularization enhancements improves stability and generalization. Embedding analysis reveal clear separation between stressed and unstressed words. These findings offer practical insights into model selection, architecture design, and interpretability for SSD applications, with implications for L2 learning support tools.

反洗錢(Anti-Money Laundering, AML)是金融科技領域的重要研究課題,其目標在於識別潛在的可疑帳戶與交易。然而隨著跨境支付與新型態交易的興起,洗錢行為往往具有高度隱匿性與複雜的網路結構,傳統規則式方法在偵測效能與泛化能力上皆表現不足。近年來,雖然有研究嘗試將機器學習或深度學習方法應用於 AML,但仍存在許多挑戰。為了解決這些問題,本研究提出一個基於序列圖融合的 AML 帳戶風險預測框架。該方法的核心在於同時建模帳戶的個體時序行為與其在交易網路中的結構特徵。首先,將每個帳戶的交易歷史分解為入邊和出邊序列,使用雙分支GRU架構分別編碼,捕捉帳戶的時序交易模式,接著使用雙向注意力圖卷積層,通過差異感知的消息傳遞機制同時處理正向和反向鄰居關係,學習帳戶間的行為差異,並通過注意力機制自適應融合節點自身特徵與雙向鄰居聚合特徵。此外,針對 AML 資料集的極度不平衡特性,引入類別重加權與平衡採樣策略。我們在公開的反洗錢資料集上驗證所提方法,實驗結果顯示該框架在極度不平衡的情境下能取得穩定的 F1 表現,相較於傳統基線方法具有顯著優勢。

The Goodness of Pronunciation (GOP) score for pronunciation quality assessment is a key technology in computer-assisted language learning. Recent studies have shown that computing GOP scores directly from the acoustic model’s raw output logits outperforms traditional softmax-probability-based methods, because logits avoid probability saturation issues and retain richer discriminative information. However, existing logit-based methods mostly rely on basic statistics such as maxima, means, or variances, which neglect the more complex dynamic distributions and temporal characteristics of logit sequences over phoneme durations. To more comprehensively capture pronunciation details embedded in logit sequences, this study proposes a multi-faceted statistical analysis method. We explore five higher-order statistical indicators that describe different characteristics of logit sequences: (1) moment-generating functions to compute distribution skewness and kurtosis; (2) information theory, using entropy to quantify model uncertainty; (3) Gaussian mixture models (GMMs) to fit multimodal distributions of logits; (4) time-series analysis, computing autocorrelation coefficients to measure logit stability; and (5) extreme value theory, using top-k averaging to obtain more robust peak-confidence estimates. We conduct experiments on the public L2 English speech corpus SpeechOcean762, comparing these newly proposed statistical indicators with baseline methods from the literature (GOP_MaxLogit, GOP_margin). Preliminary results show that some higher-order statistical indicators—particularly those that describe logit-sequence stability and distribution shape—achieve higher accuracy on pronunciation-error detection classification tasks and exhibit stronger correlation with human expert ratings. This study demonstrates that deeper statistical modeling of logit sequences is an effective approach to improving the performance of automated pronunciation assessment systems.

The Formosa Speech Recognition Challenge 2025 (FSR-2025) focuses on Taiwanese Hakka, a low-resource language with limited data diversity and channel coverage. To address this challenge, we propose a channel-aware, data-centric framework that leverages multilingual foundation models to mitigate mismatches between field recordings and training data. Our method integrates unsupervised anomaly detection and channel-conditioned augmentation to enhance data representativeness before ASR fine-tuning, aiming to explore the potential for improving robustness in low-resource Hakka speech recognition.

Automatic pronunciation assessment (APA) manages to quantify a second language (L2) learner’s pronunciation proficiency in a target language by providing fine-grained feedback with multiple pronunciation aspect scores at various linguistic levels. Most existing efforts on APA typically parallelize the modeling process, namely predicting multiple aspect scores across various linguistic levels simultaneously. This inevitably makes both the hierarchy of linguistic units and the relatedness among the pronunciation aspects sidelined. Recognizing such a limitation, we in this paper first introduce HierTFR, a hierarchal APA method that jointly models the intrinsic structures of an utterance while considering the relatedness among the pronunciation aspects. We also propose a correlation-aware regularizer to strengthen the connection between the estimated scores and the human annotations. Furthermore, novel pre-training strategies tailored for different linguistic levels are put forward so as to facilitate better model initialization. An extensive set of empirical experiments conducted on the speechocean762 benchmark dataset suggest the feasibility and effectiveness of our approach in relation to several competitive baselines.

Automated speaking assessment (ASA) typically involves automatic speech recognition (ASR) and hand-crafted feature extraction from the ASR transcript of a learner’s speech. Recently, self-supervised learning (SSL) has shown stellar performance compared to traditional methods. However, SSL-based ASA systems are faced with at least three data-related challenges: limited annotated data, uneven distribution of learner proficiency levels and non-uniform score intervals between different CEFR proficiency levels. To address these challenges, we explore the use of two novel modeling strategies: metric-based classification and loss re-weighting, leveraging distinct SSL-based embedding features. Extensive experimental results on the ICNALE benchmark dataset suggest that our approach can outperform existing strong baselines by a sizable margin, achieving a significant improvement of more than 10% in CEFR prediction accuracy.

End-to-end automatic speech recognition (E2E ASR) systems often suffer from mistranscription of domain-specific phrases, such as named entities, sometimes leading to catastrophic failures in downstream tasks. A family of fast and lightweight named entity correction (NEC) models for ASR have recently been proposed, which normally build on pho-netic-level edit distance algorithms and have shown impressive NEC performance. However, as the named entity (NE) list grows, the problems of phonetic confusion in the NE list are exacerbated; for example, homophone ambiguities increase substantially. In view of this, we proposed a novel Description Augmented Named entity CorrEctoR (dubbed DANCER), which leverages entity descriptions to provide additional information to facilitate mitigation of phonetic con-fusion for NEC on ASR transcription. To this end, an efficient entity description augmented masked language model (EDA-MLM) comprised of a dense retrieval model is introduced, enabling MLM to adapt swiftly to domain-specific entities for the NEC task. A series of experiments conducted on the AISHELL-1 and Homophone datasets confirm the effectiveness of our modeling approach. DANCER outperforms a strong baseline, the phonetic edit-distance-based NEC model (PED-NEC), by a character error rate (CER) reduction of about 7% relatively on AISHELL-1 for named entities. More notably, when tested on Homophone that contain named entities of high phonetic confusion, DANCER offers a more pronounced CER reduction of 46% relatively over PED-NEC for named entities. The code is available at https://github.com/Amiannn/Dancer.

Due to the surge in global demand for English as a second language (ESL), developments of automated methods for grading speaking proficiency have gained considerable attention. This paper aims to present a computerized regime of grading the spontaneous spoken language for ESL learners. Based on the speech corpus of ESL learners recently collected in Taiwan, we first extract multi-view features (e.g., pronunciation, fluency, and prosody features) from either automatic speech recognition (ASR) transcription or audio signals. These extracted features are, in turn, fed into a tree-based classifier to produce a new set of indicative features as the input of the automated assessment system, viz. the grader. Finally, we use different machine learning models to predict ESL learners’ respective speaking proficiency and map the result into the corresponding CEFR level. The experimental results and analysis conducted on the speech corpus of ESL learners in Taiwan show that our approach holds great potential for use in automated speaking assessment, meanwhile offering more reliable predictive results than the human experts.

The goal of an information retrieval system is to retrieve documents that are most relevant to a given user query from a huge collection of documents, which usually requires time-consuming multiple comparisons between the query and candidate documents so as to find the most relevant ones. Recently, a novel retrieval modeling approach, dubbed Differentiable Search Index (DSI), has been proposed. DSI dramatically simplifies the whole retrieval process by encoding all information about the document collection into the parameter space of a single Transformer model, on top of which DSI can in turn generate the relevant document identities (IDs) in an autoregressive manner in response to a user query. Although DSI addresses the shortcomings of traditional retrieval systems, previous studies have pointed out that DSI might fail to retrieve relevant documents because DSI uses the document IDs as the pivotal mechanism to establish the relationship between queries and documents, whereas not every document in the document collection has its corresponding relevant and irrelevant queries for the training purpose. In view of this, we put forward to leveraging supervised contrastive learning to better render the relationship between queries and documents in the latent semantic space. Furthermore, an approximate nearest neighbor search strategy is employed at retrieval time to further assist the Transformer model in generating document IDs relevant to a posed query more efficiently. A series of experiments conducted on the Nature Question benchmark dataset confirm the effectiveness and practical feasibility of our approach in relation to some strong baseline systems.

With the recent breakthrough of deep learning technologies, research on machine reading comprehension (MRC) has attracted much attention and found its versatile applications in many use cases. MRC is an important natural language processing (NLP) task aiming to assess the ability of a machine to understand natural language expressions, which is typically operationalized by first asking questions based on a given text paragraph and then receiving machine-generated answers in accordance with the given context paragraph and questions. In this paper, we leverage two novel pretrained language models built on top of Bidirectional Encoder Representations from Transformers (BERT), namely BERT-wwm and MacBERT, to develop effective MRC methods. In addition, we also seek to investigate whether additional incorporation of the categorical information about a context paragraph can benefit MRC or not, which is achieved based on performing context paragraph clustering on the training dataset. On the other hand, an ensemble learning approach is proposed to harness the synergistic power of the aforementioned two BERT-based models so as to further promote MRC performance.

With the widespread commercialization of smart devices, research on environmental sound classification has gained more and more attention in recent years. In this paper, we set out to make effective use of large-scale audio pretrained model and semi-supervised model training paradigm for environmental sound classification. To this end, an environmental sound classification method is first put forward, whose component model is built on top a large-scale audio pretrained model. Further, to simulate a low-resource sound classification setting where only limited supervised examples are made available, we instantiate the notion of transfer learning with a recently proposed training algorithm (namely, FixMatch) and a data augmentation method (namely, SpecAugment) to achieve the goal of semi-supervised model training. Experiments conducted on bench-mark dataset UrbanSound8K reveal that our classification method can lead to an accuracy improvement of 2.4% in relation to a current baseline method.

There has been increasing demand to develop effective computer-assisted language training (CAPT) systems, which can provide feedback on mispronunciations and facilitate second-language (L2) learners to improve their speaking proficiency through repeated practice. Due to the shortage of non-native speech for training the automatic speech recognition (ASR) module of a CAPT system, the corresponding mispronunciation detection performance is often affected by imperfect ASR. Recognizing this importance, we in this paper put forward a two-stage mispronunciation detection method. In the first stage, the speech uttered by an L2 learner is processed by an end-to-end ASR module to produce N-best phone sequence hypotheses. In the second stage, these hypotheses are fed into a pronunciation model which seeks to faithfully predict the phone sequence hypothesis that is most likely pronounced by the learner, so as to improve the performance of mispronunciation detection. Empirical experiments conducted a English benchmark dataset seem to confirm the utility of our method.

Natural language understanding has recently seen a surge of progress with the use of sentence encoders like ELMo (Peters et al., 2018a) and BERT (Devlin et al., 2019) which are pretrained on variants of language modeling. We conduct the first large-scale systematic study of candidate pretraining tasks, comparing 19 different tasks both as alternatives and complements to language modeling. Our primary results support the use language modeling, especially when combined with pretraining on additional labeled-data tasks. However, our results are mixed across pretraining tasks and show some concerning trends: In ELMo’s pretrain-then-freeze paradigm, random baselines are worryingly strong and results vary strikingly across target tasks. In addition, fine-tuning BERT on an intermediate task often negatively impacts downstream transfer. In a more positive trend, we see modest gains from multitask training, suggesting the development of more sophisticated multitask and transfer learning techniques as an avenue for further research.

In the context of natural language processing, representation learning has emerged as a newly active research subject because of its excellent performance in many applications. Learning representations of words is a pioneering study in this school of research. However, paragraph (or sentence and document) embedding learning is more suitable/reasonable for some tasks, such as sentiment classification and document summarization. Nevertheless, as far as we are aware, there is only a dearth of research focusing on launching unsupervised paragraph embedding methods. Classic paragraph embedding methods infer the representation of a given paragraph by considering all of the words occurring in the paragraph. Consequently, those stop or function words that occur frequently may mislead the embedding learning process to produce a misty paragraph representation. Motivated by these observations, our major contributions are twofold. First, we propose a novel unsupervised paragraph embedding method, named the essence vector (EV) model, which aims at not only distilling the most representative information from a paragraph but also excluding the general background information to produce a more informative low-dimensional vector representation for the paragraph. We evaluate the proposed EV model on benchmark sentiment classification and multi-document summarization tasks. The experimental results demonstrate the effectiveness and applicability of the proposed embedding method. Second, in view of the increasing importance of spoken content processing, an extension of the EV model, named the denoising essence vector (D-EV) model, is proposed. The D-EV model not only inherits the advantages of the EV model but also can infer a more robust representation for a given spoken paragraph against imperfect speech recognition. The utility of the D-EV model is evaluated on a spoken document summarization task, confirming the effectiveness of the proposed embedding method in relation to several well-practiced and state-of-the-art summarization methods.