Multilingual pre-trained language models (MPLMs) not only can handle tasks in different languages but also exhibit surprising zero-shot cross-lingual transferability. However, MPLMs usually are not able to achieve comparable supervised performance on rich-resource languages compared to the state-of-the-art monolingual pre-trained models. In this paper, we aim to improve the multilingual model's supervised and zero-shot performance simultaneously only with the resources from supervised languages. Our approach is based on transferring knowledge from high-performance monolingual models with a teacher-student framework. We let the multilingual model learn from multiple monolingual models simultaneously. To exploit the model's cross-lingual transferability, we propose MBLM (multi-branch multilingual language model), a model built on the MPLMs with multiple language branches. Each branch is a stack of transformers. MBLM is trained with the zero-shot-aware training strategy that encourages the model to learn from the mixture of zero-shot representations from all the branches. The results on two cross-lingual classification tasks show that, with only the task's supervised data used, our method improves both the supervised and zero-shot performance of MPLMs.
Decision forest algorithms model data by learning a binary tree structure recursively where every node splits the feature space into two regions, sending examples into the left or right branches. This "decision" is the result of the evaluation of a condition. For example, a node may split input data by applying a threshold to a numerical feature value. Such decisions are learned using (often greedy) algorithms that attempt to optimize a local loss function. Crucially, whether an algorithm exists to find and evaluate splits for a feature type (e.g., text) determines whether a decision forest algorithm can model that feature type at all. In this work, we set out to devise such an algorithm for textual features, thereby equipping decision forests with the ability to directly model text without the need for feature transformation. Our algorithm is efficient during training and the resulting splits are fast to evaluate with our extension of the QuickScorer inference algorithm. Experiments on benchmark text classification datasets demonstrate the utility and effectiveness of our proposal.
This paper presents TableQuery, a novel tool for querying tabular data using deep learning models pre-trained to answer questions on free text. Existing deep learning methods for question answering on tabular data have various limitations, such as having to feed the entire table as input into a neural network model, making them unsuitable for most real-world applications. Since real-world data might contain millions of rows, it may not entirely fit into the memory. Moreover, data could be stored in live databases, which are updated in real-time, and it is impractical to serialize an entire database to a neural network-friendly format each time it is updated. In TableQuery, we use deep learning models pre-trained for question answering on free text to convert natural language queries to structured queries, which can be run against a database or a spreadsheet. This method eliminates the need for fitting the entire data into memory as well as serializing databases. Furthermore, deep learning models pre-trained for question answering on free text are readily available on platforms such as HuggingFace Model Hub (7). TableQuery does not require re-training; when a newly trained model for question answering with better performance is available, it can replace the existing model in TableQuery.
The recent surge of complex attention-based deep learning architectures has led to extraordinary results in various downstream NLP tasks in the English language. However, such research for resource-constrained and morphologically rich Indian vernacular languages has been relatively limited. This paper proffers team SPPU\_AKAH's solution for the TechDOfication 2020 subtask-1f: which focuses on the coarse-grained technical domain identification of short text documents in Marathi, a Devanagari script-based Indian language. Availing the large dataset at hand, a hybrid CNN-BiLSTM attention ensemble model is proposed that competently combines the intermediate sentence representations generated by the convolutional neural network and the bidirectional long short-term memory, leading to efficient text classification. Experimental results show that the proposed model outperforms various baseline machine learning and deep learning models in the given task, giving the best validation accuracy of 89.57\% and f1-score of 0.8875. Furthermore, the solution resulted in the best system submission for this subtask, giving a test accuracy of 64.26\% and f1-score of 0.6157, transcending the performances of other teams as well as the baseline system given by the organizers of the shared task.
In natural scenes and documents, we can find the correlation between a text and its color. For instance, the word, "hot", is often printed in red, while "cold" is often in blue. This correlation can be thought of as a feature that represents the semantic difference between the words. Based on this observation, we propose the idea of using text color for word embeddings. While text-only word embeddings (e.g. word2vec) have been extremely successful, they often represent antonyms as similar since they are often interchangeable in sentences. In this paper, we try two tasks to verify the usefulness of text color in understanding the meanings of words, especially in identifying synonyms and antonyms. First, we quantify the color distribution of words from the book cover images and analyze the correlation between the color and meaning of the word. Second, we try to retrain word embeddings with the color distribution of words as a constraint. By observing the changes in the word embeddings of synonyms and antonyms before and after re-training, we aim to understand the kind of words that have positive or negative effects in their word embeddings when incorporating text color information.
How can prompting a large language model like GPT-3 with explanations improve in-context learning? We focus specifically on two NLP tasks that involve reasoning over text, namely question answering and natural language inference. Including explanations in the prompt and having the model generate them does not consistently improve performance in the settings we study, contrary to recent results on symbolic reasoning tasks (Nye et al., 2021; Wei et al., 2022). Despite careful prompting, explanations generated by GPT-3 may not even be factually grounded in the input, even on simple tasks with straightforward extractive explanations. However, these flawed explanations can still be useful as a way to verify GPT-3's predictions post-hoc. Through analysis in three settings, we show that explanations judged as good by humans--those that are logically consistent with the input and the prediction--usually indicate more accurate predictions. Following these observations, we present a framework for calibrating model predictions based on the reliability of the explanations. Our framework trains calibrators using automatically extracted scores that approximately assess the reliability of explanations, which helps improve performance across three different datasets.
The recently developed pitch-controllable text-to-speech (TTS) model, i.e. FastPitch, was conditioned for the pitch contours. However, the quality of the synthesized speech degraded considerably for pitch values that deviated significantly from the average pitch; i.e. the ability to control pitch was limited. To address this issue, we propose two algorithms to improve the robustness of FastPitch. First, we propose a novel timbre-preserving pitch-shifting algorithm for natural pitch augmentation. Pitch-shifted speech samples sound more natural when using the proposed algorithm because the speaker's vocal timbre is maintained. Moreover, we propose a training algorithm that defines FastPitch using pitch-augmented speech datasets with different pitch ranges for the same sentence. The experimental results demonstrate that the proposed algorithms improve the pitch controllability of FastPitch.
Transformer-based language models are able to generate fluent text and be efficiently adapted across various natural language generation tasks. However, language models that are pretrained on large unlabeled web text corpora have been shown to suffer from degenerating toxic content and social bias behaviors, consequently hindering their safe deployment. Various detoxification methods were proposed to mitigate the language model's toxicity; however, these methods struggled to detoxify language models when conditioned on prompts that contain specific social identities related to gender, race, or religion. In this study, we propose Reinforce-Detoxify; A reinforcement learning-based method for mitigating toxicity in language models. We address the challenge of safety in language models and propose a new reward model that is able to detect toxic content and mitigate unintended bias towards social identities in toxicity prediction. The experiments demonstrate that the Reinforce-Detoxify method for language model detoxification outperforms existing detoxification approaches in automatic evaluation metrics, indicating the ability of our approach in language model detoxification and less prone to unintended bias toward social identities in generated content.
Current deep learning methods for anomaly detection in text rely on supervisory signals in inliers that may be unobtainable or bespoke architectures that are difficult to tune. We study a simpler alternative: fine-tuning Transformers on the inlier data with self-supervised objectives and using the losses as an anomaly score. Overall, the self-supervision approach outperforms other methods under various anomaly detection scenarios, improving the AUROC score on semantic anomalies by 11.6% and on syntactic anomalies by 22.8% on average. Additionally, the optimal objective and resultant learnt representation depend on the type of downstream anomaly. The separability of anomalies and inliers signals that a representation is more effective for detecting semantic anomalies, whilst the presence of narrow feature directions signals a representation that is effective for detecting syntactic anomalies.
This paper aims to perform an emotion analysis of social media comments in Tamil. Emotion analysis is the process of identifying the emotional context of the text. In this paper, we present the findings obtained by Team Optimize_Prime in the ACL 2022 shared task "Emotion Analysis in Tamil." The task aimed to classify social media comments into categories of emotion like Joy, Anger, Trust, Disgust, etc. The task was further divided into two subtasks, one with 11 broad categories of emotions and the other with 31 specific categories of emotion. We implemented three different approaches to tackle this problem: transformer-based models, Recurrent Neural Networks (RNNs), and Ensemble models. XLM-RoBERTa performed the best on the first task with a macro-averaged f1 score of 0.27, while MuRIL provided the best results on the second task with a macro-averaged f1 score of 0.13.