Significant improvements in end-to-end speech translation (ST) have been achieved through the application of multi-task learning. However, the extent to which auxiliary tasks are highly consistent with the ST task, and how much this approach truly helps, have not been thoroughly studied. In this paper, we investigate the consistency between different tasks, considering different times and modules. We find that the textual encoder primarily facilitates cross-modal conversion, but the presence of noise in speech impedes the consistency between text and speech representations. Furthermore, we propose an improved multi-task learning (IMTL) approach for the ST task, which bridges the modal gap by mitigating the difference in length and representation. We conduct experiments on the MuST-C dataset. The results demonstrate that our method attains state-of-the-art results. Moreover, when additional data is used, we achieve the new SOTA result on MuST-C English to Spanish task with 20.8% of the training time required by the current SOTA method.
Recent studies have made some progress in refining end-to-end (E2E) speech recognition encoders by applying Connectionist Temporal Classification (CTC) loss to enhance named entity recognition within transcriptions. However, these methods have been constrained by their exclusive use of the ASCII character set, allowing only a limited array of semantic labels. Our proposed solution extends the E2E automatic speech recognition (ASR) system's vocabulary by adding a set of unused placeholder symbols, conceptually akin to the <pad> tokens used in sequence modeling. These placeholders are then assigned to represent semantic tags and are integrated into the transcription process as distinct tokens. We demonstrate notable improvements in entity tagging, intent discernment, and transcription accuracy on the SLUE benchmark and yields results that are on par with those for the SLURP dataset. Additionally, we provide a visual analysis of the system's proficiency in accurately pinpointing meaningful tokens over time, illustrating the enhancement in transcription quality through the utilization of supplementary semantic tags.
Social media data is a valuable resource for research, yet it contains a wide range of non-standard words (NSW). These irregularities hinder the effective operation of NLP tools. Current state-of-the-art methods for the Vietnamese language address this issue as a problem of lexical normalization, involving the creation of manual rules or the implementation of multi-staged deep learning frameworks, which necessitate extensive efforts to craft intricate rules. In contrast, our approach is straightforward, employing solely a sequence-to-sequence (Seq2Seq) model. In this research, we provide a dataset for textual normalization, comprising 2,181 human-annotated comments with an inter-annotator agreement of 0.9014. By leveraging the Seq2Seq model for textual normalization, our results reveal that the accuracy achieved falls slightly short of 70%. Nevertheless, textual normalization enhances the accuracy of the Hate Speech Detection (HSD) task by approximately 2%, demonstrating its potential to improve the performance of complex NLP tasks. Our dataset is accessible for research purposes.
With the ever-growing presence of social media platforms comes the increased spread of harmful content and the need for robust hate speech detection systems. Such systems easily overfit to specific targets and keywords, and evaluating them without considering distribution shifts that might occur between train and test data overestimates their benefit. We challenge hate speech models via new train-test splits of existing datasets that rely on the clustering of models' hidden representations. We present two split variants (Subset-Sum-Split and Closest-Split) that, when applied to two datasets using four pretrained models, reveal how models catastrophically fail on blind spots in the latent space. This result generalises when developing a split with one model and evaluating it on another. Our analysis suggests that there is no clear surface-level property of the data split that correlates with the decreased performance, which underscores that task difficulty is not always humanly interpretable. We recommend incorporating latent feature-based splits in model development and release two splits via the GenBench benchmark.
Conventional speech-to-text translation (ST) systems are trained on single-speaker utterances, and they may not generalize to real-life scenarios where the audio contains conversations by multiple speakers. In this paper, we tackle single-channel multi-speaker conversational ST with an end-to-end and multi-task training model, named Speaker-Turn Aware Conversational Speech Translation, that combines automatic speech recognition, speech translation and speaker turn detection using special tokens in a serialized labeling format. We run experiments on the Fisher-CALLHOME corpus, which we adapted by merging the two single-speaker channels into one multi-speaker channel, thus representing the more realistic and challenging scenario with multi-speaker turns and cross-talk. Experimental results across single- and multi-speaker conditions and against conventional ST systems, show that our model outperforms the reference systems on the multi-speaker condition, while attaining comparable performance on the single-speaker condition. We release scripts for data processing and model training.
Code-switching (CS), i.e. mixing different languages in a single sentence, is a common phenomenon in communication and can be challenging in many Natural Language Processing (NLP) settings. Previous studies on CS speech have shown promising results for end-to-end speech translation (ST), but have been limited to offline scenarios and to translation to one of the languages present in the source (\textit{monolingual transcription}). In this paper, we focus on two essential yet unexplored areas for real-world CS speech translation: streaming settings, and translation to a third language (i.e., a language not included in the source). To this end, we extend the Fisher and Miami test and validation datasets to include new targets in Spanish and German. Using this data, we train a model for both offline and streaming ST and we establish baseline results for the two settings mentioned earlier.
One of the key issues in Mandarin Chinese text-to-speech (TTS) systems is polyphone disambiguation when doing grapheme-to-phoneme (G2P) conversion. In this paper, we introduce a novel method to solve the problem as a generation task. Following the trending research of large language models (LLM) and prompt learning, the proposed method consists of three modules. Retrieval module incorporates external knowledge which is a multi-level semantic dictionary of Chinese polyphonic characters to format the sentence into a prompt. Generation module adopts the decoder-only Transformer architecture to induce the target text. Postprocess module corrects the generated text into a valid result if needed. Experimental results show that our method outperforms the existing methods on a public dataset called CPP. We also empirically study the impacts of different templates of the prompt, different sizes of training data, and whether to incorporate external knowledge.
Target speech extraction aims to extract, based on a given conditioning cue, a target speech signal that is corrupted by interfering sources, such as noise or competing speakers. Building upon the achievements of the state-of-the-art (SOTA) time-frequency speaker separation model TF-GridNet, we propose AV-GridNet, a visual-grounded variant that incorporates the face recording of a target speaker as a conditioning factor during the extraction process. Recognizing the inherent dissimilarities between speech and noise signals as interfering sources, we also propose SAV-GridNet, a scenario-aware model that identifies the type of interfering scenario first and then applies a dedicated expert model trained specifically for that scenario. Our proposed model achieves SOTA results on the second COG-MHEAR Audio-Visual Speech Enhancement Challenge, outperforming other models by a significant margin, objectively and in a listening test. We also perform an extensive analysis of the results under the two scenarios.
The objective of this work is to extract target speaker's voice from a mixture of voices using visual cues. Existing works on audio-visual speech separation have demonstrated their performance with promising intelligibility, but maintaining naturalness remains a challenge. To address this issue, we propose AVDiffuSS, an audio-visual speech separation model based on a diffusion mechanism known for its capability in generating natural samples. For an effective fusion of the two modalities for diffusion, we also propose a cross-attention-based feature fusion mechanism. This mechanism is specifically tailored for the speech domain to integrate the phonetic information from audio-visual correspondence in speech generation. In this way, the fusion process maintains the high temporal resolution of the features, without excessive computational requirements. We demonstrate that the proposed framework achieves state-of-the-art results on two benchmarks, including VoxCeleb2 and LRS3, producing speech with notably better naturalness.
This paper compares different pre-trained and fine-tuned large language models (LLMs) for hate speech detection. Our research underscores challenges in LLMs' cross-domain validity and overfitting risks. Through evaluations, we highlight the need for fine-tuned models that grasp the nuances of hate speech through greater label heterogeneity. We conclude with a vision for the future of hate speech detection, emphasizing cross-domain generalizability and appropriate benchmarking practices.