More Like this

1. Algebraic Systems Motivated by DNA Origami

   Garrett, J.; Jonoska, N.; Saito, M (January 2019, Algebraic Informatics)

   Ćirić, M.; Droste, M.; Pin, JÉ. (Ed.)

   We initiate an algebraic approach to study DNA origami structures. We identify two types of basic building blocks and describe a DNA origami structure by their composition. These building blocks are taken as generators of a monoid, called the origami monoid, and motivated by the well studied Temperley-Lieb algebras, we identify a set of relations that characterize the origami monoid. We present several observations about Green’s relations for the origami monoid and study the relations to a direct product of Jones monoids, which is a morphic image of an origami monoid. 

   more » « less
2. Parsimonious Morpheme Segmentation with an Application to Enriching Word Embeddings

   https://doi.org/10.1109/BigData47090.2019.9005957  

   El-Kishky, Ahmed; Xu, Frank; Zhang, Aston; Han, Jiawei (December 2019, 2019 {IEEE} International Conference on Big Data (Big Data))

   Traditionally, many text-mining tasks treat individual word-tokens as the finest meaningful semantic granularity. However, in many languages and specialized corpora, words are composed by concatenating semantically meaningful subword structures. Word-level analysis cannot leverage the semantic information present in such subword structures. With regard to word embedding techniques, this leads to not only poor embeddings for infrequent words in long-tailed text corpora but also weak capabilities for handling out-of-vocabulary words. In this paper we propose MorphMine for unsupervised morpheme segmentation. MorphMine applies a parsimony criterion to hierarchically segment words into the fewest number of morphemes at each level of the hierarchy. This leads to longer shared morphemes at each level of segmentation. Experiments show that MorphMine segments words in a variety of languages into human-verified morphemes. Additionally, we experimentally demonstrate that utilizing MorphMine morphemes to enrich word embeddings consistently improves embedding quality on a variety of of embedding evaluations and a downstream language modeling task. 

   more » « less
3. RNA origami design tools enable cotranscriptional folding of kilobase-sized nanoscaffolds

   https://doi.org/10.1038/s41557-021-00679-1  

   Cody Geary, Guido Grossi (May 2021, Nature chemistry)

   null (Ed.)

   RNA origami is a framework for the modular design of nanoscaffolds that can be folded from a single strand of RNA and used to organize molecular components with nanoscale precision. The design of genetically expressible RNA origami, which must fold cotranscriptionally, requires modelling and design tools that simultaneously consider thermodynamics, the folding pathway, sequence constraints and pseudoknot optimization. Here, we describe RNA Origami Automated Design software (ROAD), which builds origami models from a library of structural modules, identifies potential folding barriers and designs optimized sequences. Using ROAD, we extend the scale and functional diversity of RNA scaffolds, creating 32 designs of up to 2,360 nucleotides, five that scaffold two proteins, and seven that scaffold two small molecules at precise distances. Micrographic and chromatographic comparisons of optimized and non-optimized structures validate that our principles for strand routing and sequence design substantially improve yield. By providing efficient design of RNA origami, ROAD may simplify the construction of custom RNA scaffolds for nanomedicine and synthetic biology. 

   more » « less
4. Learning to Scale Multilingual Representations for Vision-Language Tasks

   Andrea Burns, Donghyun Kim (January 2020, European Conference on Computer Vision)

   null (Ed.)

   Current multilingual vision-language models either require a large number of additional parameters for each supported language, or suffer performance degradation as languages are added. In this paper, we-9*6 propose a Scalable Multilingual Aligned Language Representation (SMALR) that supports many languages with few model parameters without sacrificing downstream task performance. SMALR learns a fixed size language-agnostic representation for most words in a multilingual vocabulary, keeping language-specific features for just a few. We use a masked cross-language modeling loss to align features with context from other languages. Additionally, we propose a cross-lingual consistency module that ensures predictions made for a query and its machine translation are comparable. The effectiveness of SMALR is demonstrated with ten diverse languages, over twice the number supported in vision-language tasks to date. We evaluate on multilingual image-sentence retrieval and outperform prior work by 3–4% with less than 1/5th the training parameters compared to other word embedding methods. 

   more » « less
5. Learning to Scale Multilingual Representations for Vision-Language Tasks

   https://doi.org/10.1007/978-3-030-58548-8  

   Andrea Burns, Donghyun Kim (January 2020, European Conference on Computer Vision)

   null (Ed.)

   Current multilingual vision-language models either require a large number of additional parameters for each supported language, or suffer performance degradation as languages are added. In this paper, we-9*6 propose a Scalable Multilingual Aligned Language Representation (SMALR) that supports many languages with few model parameters without sacrificing downstream task performance. SMALR learns a fixed size language-agnostic representation for most words in a multilingual vocabulary, keeping language-specific features for just a few. We use a masked cross-language modeling loss to align features with context from other languages. Additionally, we propose a cross-lingual consistency module that ensures predictions made for a query and its machine translation are comparable. The effectiveness of SMALR is demonstrated with ten diverse languages, over twice the number supported in vision-language tasks to date. We evaluate on multilingual image-sentence retrieval and outperform prior work by 3–4% with less than 1/5th the training parameters compared to other word embedding methods. 

   more » « less