Personalizing a speech synthesis system is a highly desired application, where the system can generate speech with the user's voice with rare enrolled recordings. There are two main approaches to build such a system in recent works: speaker adaptation and speaker encoding. On the one hand, speaker adaptation methods fine-tune a trained multi-speaker text-to-speech (TTS) model with few enrolled samples. However, they require at least thousands of fine-tuning steps for high-quality adaptation, making it hard to apply on devices. On the other hand, speaker encoding methods encode enrollment utterances into a speaker embedding. The trained TTS model can synthesize the user's speech conditioned on the corresponding speaker embedding. Nevertheless, the speaker encoder suffers from the generalization gap between the seen and unseen speakers. In this paper, we propose applying a meta-learning algorithm to the speaker adaptation method. More specifically, we use Model Agnostic Meta-Learning (MAML) as the training algorithm of a multi-speaker TTS model, which aims to find a great meta-initialization to adapt the model to any few-shot speaker adaptation tasks quickly. Therefore, we can also adapt the meta-trained TTS model to unseen speakers efficiently. Our experiments compare the proposed method (Meta-TTS) with two baselines: a speaker adaptation method baseline and a speaker encoding method baseline. The evaluation results show that Meta-TTS can synthesize high speaker-similarity speech from few enrollment samples with fewer adaptation steps than the speaker adaptation baseline and outperforms the speaker encoding baseline under the same training scheme. When the speaker encoder of the baseline is pre-trained with extra 8371 speakers of data, Meta-TTS can still outperform the baseline on LibriTTS dataset and achieve comparable results on VCTK dataset.
Hateful meme detection is a new multimodal task that has gained significant traction in academic and industry research communities. Recently, researchers have applied pre-trained visual-linguistic models to perform the multimodal classification task, and some of these solutions have yielded promising results. However, what these visual-linguistic models learn for the hateful meme classification task remains unclear. For instance, it is unclear if these models are able to capture the derogatory or slurs references in multimodality (i.e., image and text) of the hateful memes. To fill this research gap, this paper propose three research questions to improve our understanding of these visual-linguistic models performing the hateful meme classification task. We found that the image modality contributes more to the hateful meme classification task, and the visual-linguistic models are able to perform visual-text slurs grounding to a certain extent. Our error analysis also shows that the visual-linguistic models have acquired biases, which resulted in false-positive predictions.
Segmentation has emerged as a fundamental field of computer vision and natural language processing, which assigns a label to every pixel/feature to extract regions of interest from an image/text. To evaluate the performance of segmentation, the Dice and IoU metrics are used to measure the degree of overlap between the ground truth and the predicted segmentation. In this paper, we establish a theoretical foundation of segmentation with respect to the Dice/IoU metrics, including the Bayes rule and Dice/IoU-calibration, analogous to classification-calibration or Fisher consistency in classification. We prove that the existing thresholding-based framework with most operating losses are not consistent with respect to the Dice/IoU metrics, and thus may lead to a suboptimal solution. To address this pitfall, we propose a novel consistent ranking-based framework, namely RankDice/RankIoU, inspired by plug-in rules of the Bayes segmentation rule. Three numerical algorithms with GPU parallel execution are developed to implement the proposed framework in large-scale and high-dimensional segmentation. We study statistical properties of the proposed framework. We show it is Dice-/IoU-calibrated, and its excess risk bounds and the rate of convergence are also provided. The numerical effectiveness of RankDice/mRankDice is demonstrated in various simulated examples and Fine-annotated CityScapes and Pascal VOC datasets with state-of-the-art deep learning architectures.
Active learning is a state-of-art machine learning approach to deal with an abundance of unlabeled data. In the field of Natural Language Processing, typically it is costly and time-consuming to have all the data annotated. This inefficiency inspires out our application of active learning in text classification. Traditional unsupervised k-means clustering is first modified into a semi-supervised version in this research. Then, a novel attempt is applied to further extend the algorithm into active learning scenario with Penalized Min-Max-selection, so as to make limited queries that yield more stable initial centroids. This method utilizes both the interactive query results from users and the underlying distance representation. After tested on a Chinese news dataset, it shows a consistent increase in accuracy while lowering the cost in training.
Understanding the semantic of a collection of texts is a challenging task. Topic models are probabilistic models that aims at extracting "topics" from a corpus of documents. This task is particularly difficult when the corpus is composed of short texts, such as posts on social networks. Following several previous research papers, we explore in this paper a set of collected tweets about bitcoin. In this work, we train three topic models and evaluate their output with several scores. We also propose a concrete application of the extracted topics.
Stemmatology is a subfield of philology where one approach to understand the copy-history of textual variants of a text (witnesses of a tradition) is to generate an evolutionary tree. Computational methods are partly shared between the sister discipline of phylogenetics and stemmatology. In 2022, a surveypaper in nature communications found that Deep Learning (DL), which otherwise has brought about major improvements in many fields (Krohn et al 2020) has had only minor successes in phylogenetics and that "it is difficult to conceive of an end-to-end DL model to directly estimate phylogenetic trees from raw data in the near future"(Sapoval et al. 2022, p.8). In stemmatology, there is to date no known DL approach at all. In this paper, we present a new DL approach to placement of manuscripts on a stemma and demonstrate its potential. This could be extended to phylogenetics where the universal code of DNA might be an even better prerequisite for the method using sequence to sequence based neural networks in order to retrieve tree distances.
Content is created for a well-defined purpose, often described by a metric or a signal represented in the form of structured information. The relationship between the metrics or the goal of a target content and the content itself are non-trivial. While large scale language models show promising text generation capabilities, guiding and informing the generated text with external metrics is challenging. These metrics and the content tend to have inherent relationships and not all of them may directly impact the content. We introduce a CaM-Gen: Causally-aware Generative Networks guided by user-defined input metrics incorporating the causal relationships between the metric and the content features. We leverage causal inference techniques to identify the causally significant aspects of text that leads to the target metric and then explicitly guide the generative model towards these by a feedback mechanism. We propose this mechanism for variational autoencoder-based and transformer-based generative models. The proposed models beat baselines in terms of the target metric accuracy while maintaining the fluency and the language quality of the generated text. To the best of our knowledge, this is one of the early attempts at incorporating a metric-guide using causal inference towards controlled generation.
In cross-lingual text classification, it is required that task-specific training data in high-resource source languages are available, where the task is identical to that of a low-resource target language. However, collecting such training data can be infeasible because of the labeling cost, task characteristics, and privacy concerns. This paper proposes an alternative solution that uses only task-independent word embeddings of high-resource languages and bilingual dictionaries. First, we construct a dictionary-based heterogeneous graph (DHG) from bilingual dictionaries. This opens the possibility to use graph neural networks for cross-lingual transfer. The remaining challenge is the heterogeneity of DHG because multiple languages are considered. To address this challenge, we propose dictionary-based heterogeneous graph neural network (DHGNet) that effectively handles the heterogeneity of DHG by two-step aggregations, which are word-level and language-level aggregations. Experimental results demonstrate that our method outperforms pretrained models even though it does not access to large corpora. Furthermore, it can perform well even though dictionaries contain many incorrect translations. Its robustness allows the usage of a wider range of dictionaries such as an automatically constructed dictionary and crowdsourced dictionary, which are convenient for real-world applications.
It is presented here a machine learning-based (ML) natural language processing (NLP) approach capable to automatically recognize and extract categorical and numerical parameters from a corpus of articles. The approach (named a.RIX) operates with a concomitant/interchangeable use of ML models such as neuron networks (NNs), latent semantic analysis (LSA) and naive-Bayes classifiers (NBC), and a pattern recognition model using regular expression (REGEX). To demonstrate the efficiency of the a.RIX engine, it was processed a corpus of 7,873 scientific articles dealing with natural products (NPs). The engine automatically extracts categorical and numerical parameters such as (i) the plant species from which active molecules are extracted, (ii) the microorganisms species for which active molecules can act against, and (iii) the values of minimum inhibitory concentration (MIC) against these microorganisms. The parameters are extracted without part-of-speech tagging (POS) and named entity recognition (NER) approaches (i.e. without the need of text annotation), and the models training is performed with unsupervised approaches. In this way, a.RIX can be essentially used on articles from any scientific field. Finally, it has a potential to make obsolete the currently used articles reviewing process in some areas, specially those in which texts structure, text semantics and latent knowledge is captured by machine learning models.
We consider a multi-agent multi-armed bandit setting in which $n$ honest agents collaborate over a network to minimize regret but $m$ malicious agents can disrupt learning arbitrarily. Assuming the network is the complete graph, existing algorithms incur $O( (m + K/n) \log (T) / \Delta )$ regret in this setting, where $K$ is the number of arms and $\Delta$ is the arm gap. For $m \ll K$, this improves over the single-agent baseline regret of $O(K\log(T)/\Delta)$. In this work, we show the situation is murkier beyond the case of a complete graph. In particular, we prove that if the state-of-the-art algorithm is used on the undirected line graph, honest agents can suffer (nearly) linear regret until time is doubly exponential in $K$ and $n$. In light of this negative result, we propose a new algorithm for which the $i$-th agent has regret $O( ( d_{\text{mal}}(i) + K/n) \log(T)/\Delta)$ on any connected and undirected graph, where $d_{\text{mal}}(i)$ is the number of $i$'s neighbors who are malicious. Thus, we generalize existing regret bounds beyond the complete graph (where $d_{\text{mal}}(i) = m$), and show the effect of malicious agents is entirely local (in the sense that only the $d_{\text{mal}}(i)$ malicious agents directly connected to $i$ affect its long-term regret).