We investigate a lattice-structured LSTM model for Chinese NER, which encodes a sequence of input characters as well as all potential words that match a lexicon. Compared with character-based methods, our model explicitly leverages word and word sequence information. Compared with word-based methods, lattice LSTM does not suffer from segmentation errors. Gated recurrent cells allow our model to choose the most relevant characters and words from a sentence for better NER results. Experiments on various datasets show that lattice LSTM outperforms both word-based and character-based LSTM baselines, achieving the best results.

Neural network models have recently been applied to the task of automatic essay scoring, giving promising results. Existing work used recurrent neural networks and convolutional neural networks to model input essays, giving grades based on a single vector representation of the essay. On the other hand, the relative advantages of RNNs and CNNs have not been compared. In addition, different parts of the essay can contribute differently for scoring, which is not captured by existing models. We address these issues by building a hierarchical sentence-document model to represent essays, using the attention mechanism to automatically decide the relative weights of words and sentences. Results show that our model outperforms the previous stateof-the-art methods, demonstrating the effectiveness of the attention mechanism.

This paper describes NCRF++, a toolkit for neural sequence labeling. NCRF++ is designed for quick implementation of different neural sequence labeling models with a CRF inference layer. It provides users with an inference for building the custom model structure through configuration file with flexible neural feature design and utilization. Built on PyTorch 1 , the core operations are calculated in batch, making the toolkit efficient with the acceleration of GPU. It also includes the implementations of most state-of-the-art neural sequence labeling models such as LSTM-CRF, facilitating reproducing and refinement on those methods.

Gene therapyhas immense potential as atherapeutic approach to serious diseases.H owever,e fficient delivery and real-time tracking of gene therapeutic agents have not been solved well for successful gene-based therapeutics.H erein we present av ersatile gene-delivery strategy for efficient and visualizedd elivery of therapeutic genes into the targeted nucleus.W ed eveloped an integrin-targeted, cell-permeable, and nucleocytoplasmic trafficking peptide-conjugated AIEgen named T D NCP for the efficient and sequential targeted delivery of an antisense single-stranded DNAo ligonucleotide (ASO) and tracking of the delivery process into the nucleus.A s compared with T D NCP/siRNA-NPs (siRNAfunctions mainly in the cytoplasm), T D NCP/ASO-NPs (ASO functions mainly in the nucleus) exhibited ab etter interference effect, which further indicates that T D NCP is an ucleus-targeting vector. Moreover,T D NCP/ASO-NPs showed af avorable tumorsuppressive effect in vivo.

Neural word segmentation research has benefited from large-scale raw texts by leveraging them for pretraining character and word embeddings. On the other hand, statistical segmentation research has exploited richer sources of external information, such as punctuation, automatic segmentation and POS. We investigate the effectiveness of a range of external training sources for neural word segmentation by building a modular segmentation model, pretraining the most important submodule using rich external sources. Results show that such pretraining significantly improves the model, leading to accuracies competitive to the best methods on six benchmarks.

Objective To examine changes in outpatient visits for mental health and/or substance use disorders (MH/SUD) in an integrated healthcare organization during the initial Massachusetts COVID-19 surge and partial state reopening. Methods Observational study of outpatient MH/SUD visits January 1st-June 30th, 2018–2020 by: 1) visit diagnosis group, 2) provider type, 3) patient race/ethnicity, 4) insurance, and 5) visit method (telemedicine vs. in-person). Results Each year, January–June 52,907–73,184 patients were seen for a MH/SUD visit. While non-MH/SUD visits declined during the surge relative to 2020 pre-pandemic (−38.2%), MH/SUD visits increased (9.1%)—concentrated in primary care (35.3%) and non-Hispanic Whites (10.5%). During the surge, MH visit volume increased 11.7% while SUD decreased 12.7%. During partial reopening, while MH visits returned to 2020 pre-pandemic levels, SUD visits declined 31.1%; MH/SUD visits decreased by Hispanics (−33.0%) and non-Hispanic Blacks (−24.6%), and among Medicaid (−19.4%) and Medicare enrollees (−20.9%). Telemedicine accounted for ~5% of MH/SUD visits pre-pandemic and 83.3%–83.5% since the surge. Conclusions MH/SUD visit volume increased during the COVID surge and was supported by rapidly-scaled telemedicine. Despite this, widening diagnostic and racial/ethnic disparities in MH/SUD visit volume during the surge and reopening suggest additional barriers for these vulnerable populations, and warrant continued monitoring and research.

Integration of multiple agent therapy (MAT) into one probe is promising for improving therapeutic efficiency for cancer treatment. However, MAT probe, if entering the cell as a whole, may not be optimal for each therapeutic agent (with different physicochemical properties), to achieve their best performance, hindering strategy optimization. A peptideconjugated-AIEgen (FC-PyTPA) is presented: upon loading with siRNA, it self-assembles into FC siRNA-PyTPA. When approaching the region near tumor cells, FC siRNA-PyTPA responds to extracellular MMP-2 and is cleaved into FC siRNA and PyTPA. The former enters cells mainly by macropinocytosis and the latter is internalized into cells mainly through caveolae-mediated endocytosis. This two-part strategy greatly improves the internalization efficiency of each individual therapeutic agent. Inside the cell, self-assembly of nanofiber precursor F, gene interference of C siRNA , and ROS production of PyTPA are activated to inhibit tumor growth.

In this paper, we introduce YEDDA, a lightweight but efficient and comprehensive open-source tool for text span annotation. YEDDA provides a systematic solution for text span annotation, ranging from collaborative user annotation to administrator evaluation and analysis. It overcomes the low efficiency of traditional text annotation tools by annotating entities through both command line and shortcut keys, which are configurable with custom labels. YEDDA also gives intelligent recommendations by learning the up-to-date annotated text. An administrator client is developed to evaluate annotation quality of multiple annotators and generate detailed comparison report for each annotator pair. Experiments show that the proposed system can reduce the annotation time by half compared with existing annotation tools. And the annotation time can be further compressed by 16.47% through intelligent recommendation.