The cloning of a speaker's voice using an untranscribed reference sample is one of the great advances of modern neural text-to-speech (TTS) methods. Approaches for mimicking the prosody of a transcribed reference audio have also been proposed recently. In this work, we bring these two tasks together for the first time through utterance level normalization in conjunction with an utterance level speaker embedding. We further introduce a lightweight aligner for extracting fine-grained prosodic features, that can be finetuned on individual samples within seconds. We show that it is possible to clone the voice of a speaker as well as the prosody of a spoken reference independently without any degradation in quality and high similarity to both original voice and prosody, as our objective evaluation and human study show. All of our code and trained models are available, alongside static and interactive demos.
Existing Natural Language Inference (NLI) datasets, while being instrumental in the advancement of Natural Language Understanding (NLU) research, are not related to scientific text. In this paper, we introduce SciNLI, a large dataset for NLI that captures the formality in scientific text and contains 107,412 sentence pairs extracted from scholarly papers on NLP and computational linguistics. Given that the text used in scientific literature differs vastly from the text used in everyday language both in terms of vocabulary and sentence structure, our dataset is well suited to serve as a benchmark for the evaluation of scientific NLU models. Our experiments show that SciNLI is harder to classify than the existing NLI datasets. Our best performing model with XLNet achieves a Macro F1 score of only 78.18% and an accuracy of 78.23% showing that there is substantial room for improvement.
Localizing objects in 3D scenes based on natural language requires understanding and reasoning about spatial relations. In particular, it is often crucial to distinguish similar objects referred by the text, such as "the left most chair" and "a chair next to the window". In this work we propose a language-conditioned transformer model for grounding 3D objects and their spatial relations. To this end, we design a spatial self-attention layer that accounts for relative distances and orientations between objects in input 3D point clouds. Training such a layer with visual and language inputs enables to disambiguate spatial relations and to localize objects referred by the text. To facilitate the cross-modal learning of relations, we further propose a teacher-student approach where the teacher model is first trained using ground-truth object labels, and then helps to train a student model using point cloud inputs. We perform ablation studies showing advantages of our approach. We also demonstrate our model to significantly outperform the state of the art on the challenging Nr3D, Sr3D and ScanRefer 3D object grounding datasets.
Existing text-to-SQL semantic parsers are typically designed for particular settings such as handling queries that span multiple tables, domains or turns which makes them ineffective when applied to different settings. We present UniSAr (Unified Structure-Aware Autoregressive Language Model), which benefits from directly using an off-the-shelf language model architecture and demonstrates consistently high performance under different settings. Specifically, UniSAr extends existing autoregressive language models to incorporate three non-invasive extensions to make them structure-aware: (1) adding structure mark to encode database schema, conversation context, and their relationships; (2) constrained decoding to decode well structured SQL for a given database schema; and (3) SQL completion to complete potential missing JOIN relationships in SQL based on database schema. On seven well-known text-to-SQL datasets covering multi-domain, multi-table and multi-turn, UniSAr demonstrates highly comparable or better performance to the most advanced specifically-designed text-to-SQL models. Importantly, our UniSAr is non-invasive, such that other core model advances in text-to-SQL can also adopt our extensions to further enhance performance.
The problem of directivity enhancement, leading to the increase in the directivity gain over a certain desired angle of arrival/departure (AoA/AoD), is considered in this work. A new formulation of the volumetric array directivity problem is proposed using the rectangular coordinates to describe each antenna element and the desired azimuth and elevation angles with a general element pattern. Such a directivity problem is formulated to find the optimal minimum distance between the antenna elements $d_\text{min}$ aiming to achieve as high directivity gains as possible. {An expedited implementation method is developed to place the antenna elements in a distinctive plane dependent on ($\theta_0$; $\phi_0$). A novel concept on optimizing directivity for the uniform planar array (OUPA) is introduced to find a quasi-optimal solution for the non-convex optimization problem with low complexity. This solution is reached by deploying the proposed successive evaluation and validation (SEV) method. {Moreover, the genetic} algorithm (GA) method was deployed to find the directivity optimization solution expeditiously. For a small number of antenna elements {, typically $N\in [4,\dots, 9]$,} the achievable directivity by GA optimization demonstrates gains of $\sim 3$ dBi compared with the traditional beamforming technique, using steering vector for uniform linear arrays (ULA) and uniform circular arrays (UCA), while gains of $\sim1.5$ dBi are attained when compared with an improved UCA directivity method. For a larger number of antenna elements {, two improved GA procedures, namely GA-{\it marginal} and GA-{\it stall}, were} proposed and compared with the OUPA method. OUPA also indicates promising directivity gains surpassing $30$ dBi for massive MIMO scenarios.
Knowledge Graphs (KGs) often have two characteristics: heterogeneous graph structure and text-rich entity/relation information. KG representation models should consider graph structures and text semantics, but no comprehensive open-sourced framework is mainly designed for KG regarding informative text description. In this paper, we present PromptKG, a prompt learning framework for KG representation learning and application that equips the cutting-edge text-based methods, integrates a new prompt learning model and supports various tasks (e.g., knowledge graph completion, question answering, recommendation, and knowledge probing). PromptKG is publicly open-sourced at https://github.com/zjunlp/PromptKG with long-term technical support.
Automated completion of open knowledge bases (KBs), which are constructed from triples of the form (subject phrase, relation phrase, object phrase) obtained via open information extraction (IE) from text, is useful for discovering novel facts that may not directly be present in the text. However, research in open knowledge base completion (KBC) has so far been limited to resource-rich languages like English. Using the latest advances in multilingual open IE, we construct the first multilingual open KBC dataset, called mOKB6, that contains facts from Wikipedia in six languages (including English). Improving the previous open KB construction pipeline by doing multilingual coreference resolution and keeping only entity-linked triples, we create a dense open KB. We experiment with several baseline models that have been proposed for both open and closed KBs and observe a consistent benefit of using knowledge gained from other languages. The dataset and accompanying code will be made publicly available.
Predicting emotions expressed in text is a well-studied problem in the NLP community. Recently there has been active research in extracting the cause of an emotion expressed in text. Most of the previous work has done causal emotion entailment in documents. In this work, we propose neural models to extract emotion cause span and entailment in conversations. For learning such models, we use RECCON dataset, which is annotated with cause spans at the utterance level. In particular, we propose MuTEC, an end-to-end Multi-Task learning framework for extracting emotions, emotion cause, and entailment in conversations. This is in contrast to existing baseline models that use ground truth emotions to extract the cause. MuTEC performs better than the baselines for most of the data folds provided in the dataset.
Knowledge Distillation (KD) is a commonly used technique for improving the generalization of compact Pre-trained Language Models (PLMs) on downstream tasks. However, such methods impose the additional burden of training a separate teacher model for every new dataset. Alternatively, one may directly work on the improvement of the optimization procedure of the compact model toward better generalization. Recent works observe that the flatness of the local minimum correlates well with better generalization. In this work, we adapt Stochastic Weight Averaging (SWA), a method encouraging convergence to a flatter minimum, to fine-tuning PLMs. We conduct extensive experiments on various NLP tasks (text classification, question answering, and generation) and different model architectures and demonstrate that our adaptation improves the generalization without extra computation cost. Moreover, we observe that this simple optimization technique is able to outperform the state-of-the-art KD methods for compact models.
We study speech-to-speech translation (S2ST) that translates speech from one language into another language and focuses on building systems to support languages without standard text writing systems. We use English-Taiwanese Hokkien as a case study, and present an end-to-end solution from training data collection, modeling choices to benchmark dataset release. First, we present efforts on creating human annotated data, automatically mining data from large unlabeled speech datasets, and adopting pseudo-labeling to produce weakly supervised data. On the modeling, we take advantage of recent advances in applying self-supervised discrete representations as target for prediction in S2ST and show the effectiveness of leveraging additional text supervision from Mandarin, a language similar to Hokkien, in model training. Finally, we release an S2ST benchmark set to facilitate future research in this field. The demo can be found at https://huggingface.co/spaces/facebook/Hokkien_Translation .