arXiv research

Scene text magnifier aims to magnify text in natural scene images without recognition. It could help the special groups, who have myopia or dyslexia to better understand the scene. In this paper, we design the scene text magnifier through interacted four CNN-based networks: character erasing, character extraction, char…

The paper classifies actions of a specific group on certain manifolds.

The paper solves the Nielsen realization problem for high degree del Pezzo surfaces.

We establish a gluing construction for Higgs bundles over a connected sum of Riemann surfaces in terms of solutions to the $\text{Sp(4}\text{,}\mathbb{R}\text{)}$-Hitchin equations using the linearization of a relevant elliptic operator. The construction can be used to provide model Higgs bundles in all the $2g-3$ exce…

A new text representation model combines CNN and VAE for better semantic extraction.

We propose two algorithms that can find local minima faster than the state-of-the-art algorithms in both finite-sum and general stochastic nonconvex optimization. At the core of the proposed algorithms is $\text{One-epoch-SNVRG}^+$ using stochastic nested variance reduction (Zhou et al., 2018a), which outperforms the s…

New families of Lie groups with special foliations discovered.

Improved text summarization using belief propagation on weighted bipartite graphs.

New TC variant dTC better fits motion planning for some systems.

Let $\text{Man}_{*}$ denote the category of closed, connected, oriented and based $3$-manifolds, with basepoint preserving diffeomorphisms between them. Juhász, Thurston and Zemke showed that the Heegaard Floer invariants are natural with respect to diffeomorphisms, in the sense that there are functors $HF^{\circ}: \te…

Paper generates diverse, readable adversarial texts from scratch.

SCROLLS benchmarks long text NLP tasks, improving existing models.

New method detects text changes under dependencies, outperforming baselines.

Contextual bandits study how reward variance affects regret bounds.

The paper studies symplectic structures on character varieties of Sasakian threefolds.

Most of the information is stored as text, so text mining is regarded as having high commercial potential. Aiming at the semantic constraint problem of classification methods based on sparse representation, we propose a weighted recurrent neural network (W-RNN), which can fully extract text serialization semantic infor…

Recent years have seen remarkable progress of text generation in different contexts, such as the most common setting of generating text from scratch, and the emerging paradigm of retrieval-and-rewriting. Text infilling, which fills missing text portions of a sentence or paragraph, is also of numerous use in real life, …

We study the behavior of the second order Renormalization Group flow on locally homogeneous metrics on closed three-manifolds. In the cases $\mathbb R^3$ and $\text{SO}(3)\times \R$, the flow is qualitatively the same as the Ricci flow. In the cases $\text{H}(3)$ and $\text{H}(2)\times \R$, if the curvature is small, t…

Joyce showed that for a classical knot $K$, the involutory medial quandle $\text{IMQ}(K)$ is isomorphic to the core quandle of the homology group $H_1(X_2)$, where $X_2$ is the cyclic double cover of $\mathbb S ^3$, branched over $K$. It follows that $|\text{IMQ}(K)| = | \det K |$. In the present paper, the extension o…

The paper classifies nilmanifolds with specific SL(3,C) structures.

Convolutional neural network improves assertion detection in multi-label clinical text.

Proposes a flexible neural recommendation framework for better prediction performance.

Let $\text{G}(n)$ be equal either to $\text{PO}(n,1),\text{PU}(n,1)$ or $\text{PSp}(n,1)$ and let $Γ\leq \text{G}(n)$ be a uniform lattice. Denote by $\mathbb{H}^n_K$ the hyperbolic space associated to $\text{G}(n)$, where $K$ is a division algebra over the reals of dimension $d=\dim_{\mathbb{R}} K$. Assume $d(n-1) \ge…

Traditionally, text generation models take in a sequence of text as input, and iteratively generate the next most probable word using pre-trained parameters. In this work, we propose the architecture to use images instead of text as the input of the text generation model, called StoryGen. In the architecture, we design…

A new framework uses text descriptions to improve protein design.

MoleculeSTM learns from molecule structures and texts for better drug design.

Unified causal model improves controllable text generation without bias.

Image captioning has demonstrated models that are capable of generating plausible text given input images or videos. Further, recent work in image generation has shown significant improvements in image quality when text is used as a prior. Our work ties these concepts together by creating an architecture that can enabl…

Estimates watermarked content proportions in mixed-source texts.

TBIP uses texts to quantify lawmakers' political positions.

Bayesian Topic Regression models causal inference with text and numerical data.

New coarse LS-category introduced for groups and spaces.

SpeakerStew verifies 46 languages with reduced training and inference costs.

In this work, we study abstractive text summarization by exploring different models such as LSTM-encoder-decoder with attention, pointer-generator networks, coverage mechanisms, and transformers. Upon extensive and careful hyperparameter tuning we compare the proposed architectures against each other for the abstractiv…

For a number ring $\mathcal{O}$, Borel and Serre proved that $\text{SL}_n(\mathcal{O})$ is a virtual duality group whose dualizing module is the Steinberg module. They also proved that $\text{GL}_n(\mathcal{O})$ is a virtual duality group. In contrast to $\text{SL}_n(\mathcal{O})$, we prove that the dualizing module of…

Single-stage neural architecture improves text-to-image synthesis.

Multimodal language analysis often considers relationships between features based on text and those based on acoustical and visual properties. Text features typically outperform non-text features in sentiment analysis or emotion recognition tasks in part because the text features are derived from advanced language mode…

Many challenges in natural language processing require generating text, including language translation, dialogue generation, and speech recognition. For all of these problems, text generation becomes more difficult as the text becomes longer. Current language models often struggle to keep track of coherence for long pi…

We simplify and prove a splitting theorem for mapping class groups of connect sums of $S^2 imes S^1$.

Combines BERT and graph CNN for improved text classification.

Study of isometry groups in skewed Γ-complexes.

Let $X$ be a compact connected Riemann surface of genus $g$, with $g \geq 2$. For each $d <η(X)$, where $η(X)$ is the gonality of $X$, the symmetric product $\text{Sym}^d(X)$ embeds into $\text{Pic}^d(X)$ by sending an effective divisor of degree $d$ to the corresponding holomorphic line bundle. Therefore, the restrict…

Let $M$ and $N$ be two closed $C^{\infty}$ manifolds and let $\text{Diff}_c(M)$ denote the group of $C^{\infty}$ diffeomorphisms isotopic to the identity. We prove that any (discrete) group homomorphism between $\text{Diff}_c(M)$ and $\text{Diff}_c(N)$ is continuous. We also show that a non-trivial group homomorphism $…

Unsupervised text embedding has shown great power in a wide range of NLP tasks. While text embeddings are typically learned in the Euclidean space, directional similarity is often more effective in tasks such as word similarity and document clustering, which creates a gap between the training stage and usage stage of t…

Benchmarking AutoML for tables with text fields, achieving top performance.

The paper studies $ext{Pin}^{\pm}$-structures on non-oriented 4-manifolds via Lefschetz fibrations.

Deep learning models outperform classical methods in text classification.

Study stabilizers of isotropic classes in rational 4-manifolds, finding diffeomorphisms that almost preserve Lefschetz fibrations.