Improving end-to-end speech recognition by incorporating external text data has been a longstanding research topic. There has been a recent focus on training E2E ASR models that get the performance benefits of external text data without incurring the extra cost of evaluating an external language model at inference time. In this work, we propose training ASR model jointly with a set of text-to-text auxiliary tasks with which it shares a decoder and parts of the encoder. When we jointly train ASR and masked language model with the 960-hour Librispeech and Opensubtitles data respectively, we observe WER reductions of 16% and 20% on test-other and test-clean respectively over an ASR-only baseline without any extra cost at inference time, and reductions of 6% and 8% compared to a stronger MUTE-L baseline which trains the decoder with the same text data as our model. We achieve further improvements when we train masked language model on Librispeech data or when we use machine translation as the auxiliary task, without significantly sacrificing performance on the task itself.
Midrash collections are complex rabbinic works that consist of text in multiple languages, which evolved through long processes of unstable oral and written transmission. Determining the origin of a given passage in such a compilation is not always straightforward and is often a matter of dispute among scholars, yet it is essential for scholars' understanding of the passage and its relationship to other texts in the rabbinic corpus. To help solve this problem, we propose a system for classification of rabbinic literature based on its style, leveraging recently released pretrained Transformer models for Hebrew. Additionally, we demonstrate how our method can be applied to uncover lost material from Midrash Tanhuma.
Latent text representations exhibit geometric regularities, such as the famous analogy: queen is to king what woman is to man. Such structured semantic relations were not demonstrated on image representations. Recent works aiming at bridging this semantic gap embed images and text into a multimodal space, enabling the transfer of text-defined transformations to the image modality. We introduce the SIMAT dataset to evaluate the task of text-driven image transformation. SIMAT contains 6k images and 18k "transformation queries" that aim at either replacing scene elements or changing their pairwise relationships. The goal is to retrieve an image consistent with the (source image, transformation) query. We use an image/text matching oracle (OSCAR) to assess whether the image transformation is successful. The SIMAT dataset will be publicly available. We use SIMAT to show that vanilla CLIP multimodal embeddings are not very well suited for text-driven image transformation, but that a simple finetuning on the COCO dataset can bring dramatic improvements. We also study whether it is beneficial to leverage the geometric properties of pretrained universal sentence encoders (FastText, LASER and LaBSE).
Embeddings play an important role in many recent end-to-end solutions for language processing problems involving more than one data modality. Although there has been some effort to understand the properties of single-modality embedding spaces, particularly that of text, their cross-modal counterparts are less understood. In this work, we study a joint speech-text embedding space trained for semantic matching by minimizing the distance between paired utterance and transcription inputs. This was done through dual encoders in a teacher-student model setup, with a pretrained language model acting as the teacher and a transformer-based speech encoder as the student. We extend our method to incorporate automatic speech recognition through both pretraining and multitask scenarios and found that both approaches improve semantic matching. Multiple techniques were utilized to analyze and evaluate cross-modal semantic alignment of the embeddings: a quantitative retrieval accuracy metric, zero-shot classification to investigate generalizability, and probing of the encoders to observe the extent of knowledge transfer from one modality to another.
Warning: this paper contains content that may be offensive or upsetting. In the current context where online platforms have been effectively weaponized in a variety of geo-political events and social issues, Internet memes make fair content moderation at scale even more difficult. Existing work on meme classification and tracking has focused on black-box methods that do not explicitly consider the semantics of the memes or the context of their creation. In this paper, we pursue a modular and explainable architecture for Internet meme understanding. We design and implement multimodal classification methods that perform example- and prototype-based reasoning over training cases, while leveraging both textual and visual SOTA models to represent the individual cases. We study the relevance of our modular and explainable models in detecting harmful memes on two existing tasks: Hate Speech Detection and Misogyny Classification. We compare the performance between example- and prototype-based methods, and between text, vision, and multimodal models, across different categories of harmfulness (e.g., stereotype and objectification). We devise a user-friendly interface that facilitates the comparative analysis of examples retrieved by all of our models for any given meme, informing the community about the strengths and limitations of these explainable methods.
The conventional success of textual classification relies on annotated data, and the new paradigm of pre-trained language models (PLMs) still requires a few labeled data for downstream tasks. However, in real-world applications, label noise inevitably exists in training data, damaging the effectiveness, robustness, and generalization of the models constructed on such data. Recently, remarkable achievements have been made to mitigate this dilemma in visual data, while only a few explore textual data. To fill this gap, we present SelfMix, a simple yet effective method, to handle label noise in text classification tasks. SelfMix uses the Gaussian Mixture Model to separate samples and leverages semi-supervised learning. Unlike previous works requiring multiple models, our method utilizes the dropout mechanism on a single model to reduce the confirmation bias in self-training and introduces a textual-level mixup training strategy. Experimental results on three text classification benchmarks with different types of text show that the performance of our proposed method outperforms these strong baselines designed for both textual and visual data under different noise ratios and noise types. Our code is available at \url{https://github.com/noise-learning/SelfMix}.
The advent of personalized reality has arrived. Rapid development in AR/MR/VR enables users to augment or diminish their perception of the physical world. Robust tooling for digital interface modification enables users to change how their software operates. As digital realities become an increasingly-impactful aspect of human lives, we investigate the design of a system that enables users to manipulate the perception of both their physical realities and digital realities. Users can inspect their view history from either reality, and generate interventions that can be interoperably rendered cross-reality in real-time. Personalized interventions can be generated with mask, text, and model hooks. Collaboration between users scales the availability of interventions. We verify our implementation against our design requirements with cognitive walkthroughs, personas, and scalability tests.
DNN-based models achieve high performance in the speaker verification (SV) task with substantial computation costs. The model size is an essential concern in applying models on resource-constrained devices, while model compression for SV models has not been studied extensively in previous works. Weight quantization is exploited to compress DNN-based speaker embedding extraction models in this paper. Uniform and Powers-of-Two quantization are utilized in the experiments. The results on VoxCeleb show that the weight quantization can decrease the size of ECAPA-TDNN and ResNet by 4 times with insignificant performance decline. The quantized 4-bit ResNet achieves similar performance to the original model with an 8 times smaller size. We empirically show that the performance of ECAPA-TDNN is more sensitive than ResNet to quantization due to the difference in weight distribution. The experiments on CN-Celeb also demonstrate that quantized models are robust for SV in the language mismatch scenario.
The paper explores the feasibility of using machine learning techniques, in particular neural networks, for classification of the experimental data from the joint $^\text{nat}$C(n,p) and $^\text{nat}$C(n,d) reaction cross section measurement from the neutron time of flight facility n_TOF at CERN. Each relevant $\Delta E$-$E$ pair of strips from two segmented silicon telescopes is treated separately and afforded its own dedicated neural network. An important part of the procedure is a careful preparation of training datasets, based on the raw data from Geant4 simulations. Instead of using these raw data for the training of neural networks, we divide a relevant 3-parameter space into discrete voxels, classify each voxel according to a particle/reaction type and submit these voxels to a training procedure. The classification capabilities of the structurally optimized and trained neural networks are found to be superior to those of the manually selected cuts.
Diffusion models have emerged as a powerful tool for point cloud generation. A key component that drives the impressive performance for generating high-quality samples from noise is iteratively denoise for thousands of steps. While beneficial, the complexity of learning steps has limited its applications to many 3D real-world. To address this limitation, we propose Point Straight Flow (PSF), a model that exhibits impressive performance using one step. Our idea is based on the reformulation of the standard diffusion model, which optimizes the curvy learning trajectory into a straight path. Further, we develop a distillation strategy to shorten the straight path into one step without a performance loss, enabling applications to 3D real-world with latency constraints. We perform evaluations on multiple 3D tasks and find that our PSF performs comparably to the standard diffusion model, outperforming other efficient 3D point cloud generation methods. On real-world applications such as point cloud completion and training-free text-guided generation in a low-latency setup, PSF performs favorably.