Tagging Scheme Research Articles

Fine Grained Named Entity Recognition via Seq2seq Framework

Fine-grained Named entity recognition (NER) is crucial to natural language processing (NLP) applications like relation extraction and knowledge graph construction. Most existing fine-grained NER systems suffer from inefficiency problem as they use manually annotated training datasets. To address such issue, our NER system could automatically generate datasets from Wikipedia in distant supervision paradigm through mapping hyperlinks in Wikipedia documents to Freebase. In addition, previous NER models can not effectively process fine-grained labels with more than 100 types. So we introduce a ‘BIO’ tagging strategy which can identify the position and type attributes simultaneously. Such tagging scheme transfers NER problem into a sequence-to-sequence (seq2seq) based issue. We propose a seq2seq framework to comprehend the input sentence in a comprehensive way. Specifically, we adopt a Bi-LSTM as the encoder to equally process the past and future information of the input. Then we add a self-attention mechanism to handle the long-term dependency problem in a long sequence. When classifying the entity tags, we choose CRF model as it adds more constraints to avoid position logical problem. Experiments are performed on large-scale datasets for fine-grained NER tasks. Experimental results verify the effectiveness of FSeqC, and it outperforms other state-of-the-art alternatives consistently and significantly.

A step towards information extraction: Named entity recognition in Bangla using deep learning

Information Extraction allows machines to decipher natural language through using two tasks: Named Entity Recognition and Relation Extraction. In order to build such a system for Bangla Language, in this work a Named Entity Recognition (NER) System is proposed, which requires a minimum information to deliver a decent performance having less dependency on handcrafted features. The proposed model is based on Deep Learning, which is accomplished through the use of a Densely Connected Network (DCN) in collaboration with a Bidirectional-LSTM (BiLSTM) and word embedding, i.e., DCN-BiLSTM. Such a system, specific to the Bangla language, has never been done before. Furthermore, a unique dataset was made since no Named Entity Recognition dataset exists for Bangla language till date. In the dataset, over 71 thousand Bangla sentences have been collected, annotated, and classified into four different groups using IOB tagging scheme. Those groups are person, location, organization, and object. Due to Bangla’s morphological structure, character level feature extraction is also applied so that we can access more features to determine relational structure between different words. This is initially done with the use of a Convolutional Neural Network but is later outperformed by our second approach which is through the use of a Densely Connected Network (DCN). As for the training portion, it has been done for two variations of word embedding which are word2vec and glove, the outcome being the largest vocabulary size known to both models. A detailed discussion in regard to the methodology of the NER system is explained in a comprehensive manner followed by an examination of the various evaluation scores achieved. The proposed model in this work resulted in having a F1 score of 63.37, which is evaluated at Named Entity Level.

An input information enhanced model for relation extraction

We present a novel end-to-end model to jointly extract semantic relations and argument entities from sentence texts. This model does not require any handcrafted feature set or auxiliary toolkit, and hence it could be easily extended to a wide range of sequence tagging tasks. A new method of using the word morphology feature for relation extraction is studied in this paper. We combine the word morphology feature and the semantic feature to enrich the representing capacity of input vectors. Then, an input information enhanced unit is developed for the bidirectional long short-term memory network (Bi-LSTM) to overcome the information loss caused by the gate operations and the concatenation operations in the LSTM memory unit. A new tagging scheme using uncertain labels and a corresponding objective function are exploited to reduce the interference information from non-entity words. Experiments are performed on three datasets: The New York Times (NYT) and ACE2005 datasets for relation extraction and the SemEval 2010 task 8 dataset for relation classification. The results demonstrate that our model achieves a significant improvement over the state-of-the-art model for relation extraction on the NYT dataset and achieves a competitive performance on the ACE2005 dataset.

Joint Extraction of Entities and Overlapping Relations Using Position-Attentive Sequence Labeling

Joint entity and relation extraction is to detect entity and relation using a single model. In this paper, we present a novel unified joint extraction model which directly tags entity and relation labels according to a query word position p, i.e., detecting entity at p, and identifying entities at other positions that have relationship with the former. To this end, we first design a tagging scheme to generate n tag sequences for an n-word sentence. Then a position-attention mechanism is introduced to produce different sentence representations for every query position to model these n tag sequences. In this way, our method can simultaneously extract all entities and their type, as well as all overlapping relations. Experiment results show that our framework performances significantly better on extracting overlapping relations as well as detecting long-range relation, and thus we achieve state-of-the-art performance on two public datasets.

A Unified Model for Opinion Target Extraction and Target Sentiment Prediction

Target-based sentiment analysis involves opinion target extraction and target sentiment classification. However, most of the existing works usually studied one of these two sub-tasks alone, which hinders their practical use. This paper aims to solve the complete task of target-based sentiment analysis in an end-to-end fashion, and presents a novel unified model which applies a unified tagging scheme. Our framework involves two stacked recurrent neural networks: The upper one predicts the unified tags to produce the final output results of the primary target-based sentiment analysis; The lower one performs an auxiliary target boundary prediction aiming at guiding the upper network to improve the performance of the primary task. To explore the inter-task dependency, we propose to explicitly model the constrained transitions from target boundaries to target sentiment polarities. We also propose to maintain the sentiment consistency within an opinion target via a gate mechanism which models the relation between the features for the current word and the previous word. We conduct extensive experiments on three benchmark datasets and our framework achieves consistently superior results.

LayerCode

With the advance of personal and customized fabrication techniques, the capability to embed information in physical objects becomes evermore crucial. We present LayerCode , a tagging scheme that embeds a carefully designed barcode pattern in 3D printed objects as a deliberate byproduct of the 3D printing process. The LayerCode concept is inspired by the structural resemblance between the parallel black and white bars of the standard barcode and the universal layer-by-layer approach of 3D printing. We introduce an encoding algorithm that enables the 3D printing layers to carry information without altering the object geometry. We also introduce a decoding algorithm that reads the LayerCode tag of a physical object by just taking a photo. The physical deployment of LayerCode tags is realized on various types of 3D printers, including Fused Deposition Modeling printers as well as Stereolithography based printers. Each offers its own advantages and tradeoffs. We show that LayerCode tags can work on complex, nontrivial shapes, on which all previous tagging mechanisms may fail. To evaluate LayerCode thoroughly, we further stress test it with a large dataset of complex shapes using virtual rendering. Among 4,835 tested shapes, we successfully encode and decode on more than 99% of the shapes.

Toward an Effective Igbo Part-of-Speech Tagger

Part-of-speech (POS) tagging is a well-established technology for most Western European languages and a few other world languages, but it has not been evaluated on Igbo, an agglutinative African language. This article presents POS tagging experiments conducted using an Igbo corpus as a test bed for identifying the POS taggers and the Machine Learning (ML) methods that can achieve a good performance with the small dataset available for the language. Experiments have been conducted using different well-known POS taggers developed for English or European languages, and different training data styles and sizes. Igbo has a number of language-specific characteristics that present a challenge for effective POS tagging. One interesting case is the wide use of verbs (and nominalizations thereof) that have an inherent noun complement , which form “linked pairs” in the POS tagging scheme, but which may appear discontinuously. Another issue is Igbo’s highly productive agglutinative morphology, which can produce many variant word forms from a given root. This productivity is a key cause of the out-of-vocabulary (OOV) words observed during Igbo tagging. We report results of experiments on a promising direction for improving tagging performance on such morphologically-inflected OOV words.

Dependentzia Unibertsalen eredura egokitutako euskarazko zuhaitz-bankua

Hizkuntzaren Prozesamenduan kokatzen den Dependentzia Unibertsalen proiektuaren helburua da hainbat hizkuntzatan sortu diren dependentzia-ereduan oinarritutako zuhaitz-bankuak etiketatze-eskema estandar berera egokitzea. Artikulu honetan, eredu horretara automatikoki egokitu den euskarazko zuhaitz-bankua aurkezten da; halaber, egokitzapen-lan hori nola gauzatu den deskribatzen da eta, azkenik, horretan oinarrituta, azaltzen da zer antzekotasun eta zer desberdintasun diren jatorrizko zuhaitza-bankuaren eta Dependentzia Unibertsalen eredura egokitutako zuhaitz-bankuaren artean.

Imaging individual barium atoms in solid xenon for barium tagging in nEXO.

Double-β-decay involves the simultaneous conversion of two neutrons into two protons, and the emission of two electrons and two neutrinos; the neutrinoless process, although not yet observed, is thought to involve the emission of the two electrons but no neutrinos. The search for neutrinoless-double-β-decay probes fundamental properties of neutrinos, including whether or not the neutrino and antineutrino are distinct particles. Double-β-decay detectors are large and expensive, so it is essential to achieve the highest possible sensitivity with each study, and removing spurious contributions ('background') from detected signals is crucial. In the nEXO neutrinoless-double-β-decay experiment, the identification, or 'tagging', of the 136Ba daughter atom resulting from the double-β decay of 136Xeprovides a technique for discriminating background. The tagging scheme studied here uses a cryogenic probe to trap the barium atom in a solid xenon matrix, where the barium atom is tagged through fluorescence imaging. Here we demonstrate the imaging and counting of individual barium atoms in solid xenon by scanning a focused laser across a solid xenon matrix deposited on a sapphire window. When the laser irradiates an individual atom, the fluorescence persists for about 30 seconds before dropping abruptly to the background level-a clear confirmation of one-atom imaging. Following evaporation of a barium deposit, the residual barium fluorescence is 0.16 per cent or less. Our technique achieves the imaging of single atoms in a solid noble element, establishing the basic principle of barium tagging for nEXO.

Tagger: Practical PFC Deadlock Prevention in Data Center Networks

Remote direct memory access over converged Ethernet deployments is vulnerable to deadlocks induced by priority flow control. Prior solutions for deadlock prevention either require significant changes to routing protocols or require excessive buffers in the switches. In this paper, we propose Tagger, a scheme for deadlock prevention. It does not require any changes to the routing protocol and needs only modest buffers. Tagger is based on the insight that given a set of expected lossless routes, a simple tagging scheme can be developed to ensure that no deadlock will occur under any failure conditions. Packets that do not travel on these lossless routes may be dropped under extreme conditions. We design such a scheme, prove that it prevents deadlock, and implement it efficiently on commodity hardware.

Joint Entity and Relation Extraction Based on Reinforcement Learning

Information extraction is a crucial technology to construct a knowledge base. In this paper, a novel model was proposed to extract entities and relations from plain text. This model consists of two components: a joint network and a reinforcement learning agent. The joint network is designed end-to-end, which can extract entities and relations simultaneously. In the joint model, a new tagging scheme was adopted, then the entity and relation extraction can be modeled as a joint sequence tagging problem. To enhance the robustness of the model, we also introduced a reinforcement learning (RL) agent to remove the noisy data from the training dataset. The RL agent aims at determining whether a candidate instance should be removed from the training dataset or not. When the agent completes a selection process, the training dataset will be divided into two parts: clean data and noisy data. Then the joint network can be trained again on the clean dataset to generate a better model. To assess the validity of the model we proposed, extensive experiments were conducted on the New York Times dataset (NYT10 and NYT 11). The experimental results showed that the model we proposed is superior compared with the baselines, achieving the F1 value on NYT10 and NYT11 with 0.612 and 0.549, respectively.

Semantic Interpretation of the map with Diabetes-Related Websites

Diabetes as a chronic disease requires continuous medical care and constant patient self-management which involve several stakeholders to improve health outcome and patient quality of life. In our prior work, we used the networks of World Wide Web to highlight how stakeholders of diabetes link to each other online. The aim of this study is to use a semantic approach focusing on the diabetes-related websites to better understand the common interest shared by the same clusters which were detected in our previous study of stakeholders on diabetes. To achieve this, we employed the data annotation and machine learning to study which combinations of tags can predict or explain the clusters. In the end, a total of 430 websites which are detected into 5 clusters have been tagged with 38 different tags from 6 different dimensions in this study. Although the result shows a very low prediction performance using tags to determine the clusters of diabetes-related websites, except for cluster 1 and cluster 2, this reflects the community reality: a mix of websites of different types that create a mixed but localized space. It proves the community can have a tagging scheme occasionally but it is still hard to use semantical approach to predict accurately the clusters.

Supporting global initiatives on climate change: The EPO's “Y02-Y04S” tagging scheme

The EPO has developed a dedicated classification scheme for climate change mitigation technologies (CCMTs), where relevant patent publications are “tagged” and classified into a separated scheme which is fully integrated within the Cooperative Patent Classification (CPC). This allows for non-patent experts to search for climate change-related technologies in a more user-friendly fashion. The scheme has been continually expanded in recent years and is now considered complete for CCMTs, covering seven main categories, namely energy, greenhouse gases (GHG) capture, buildings, industry (including agriculture), transport and waste and wastewater management. A classification scheme for smart grids, Y04S (a spin-off of Y02), has been also developed. Both classifications ease the access to patent documentation in sustainable technologies. An overview of the structure of the Y02/Y04S scheme is presented together with examples of how this can be used (i) to find relevant technologies and their owners and (ii) to provide technical, legal and business information to support strategic decision-making in the field of climate change.

Optimized super-selective Arterial Spin Labeling for quantitative flow territory mapping

Flow territory mapping using Arterial Spin Labeling (ASL) methods allows for deeper insights into the physiology of the brain. However, in most approaches assumptions of the labeling efficiency are used for calculation of brain perfusion which might result in inaccuracies. This becomes more important in super-selective ASL as the labeling focus could be positioned not exactly at the artery of interest. Therefore, measuring the labeling efficiency of the current scan seems important to obtain reliable results and to correct for potential errors. In this study, an optimized super-selective ASL tagging scheme is presented and the labeling efficiency is measured using quantitative phase-contrast angiography of the tagged artery and considering the volume of the supplied flow territory. The aim then is to evaluate the labeling efficiency of super-selective ASL and considering these values when shifting the labeling spot away from the artery of interest. The measured efficiency is compared to simulations performed for the optimized super-selective ASL approach. Considering the values of labeling efficiency after shifting the labeling focus, the values of cerebral blood flow still show accurate results despite the expected lower SNR up to an offset of 3 mm. Following this, to obtain accurate results for quantifying super-selective ASL perfusion images, measuring the labeling efficiency seems necessary to prevent false results.

Learning to recognize opinion targets using recurrent neural networks

Opinion target recognition can be deemed a problem of token-level sequence labelling, and each word in the sentence is assigned to a label with the standard BIO tagging scheme. Among a variety of methods, recently emerged recurrent neural network (RNN) is considered more effective to deal with such kinds of sequence annotation problem. Whereas existing RNN models mainly focus on learning the dependency relationships in the input sequence while ignoring the ones in the output sequence. To this end, we proposed an augmented RNN model called OLSRNN. The OLSRNN model adds self-connections to the output layer on the basics of conventional RNN models to further capture output temporal dependencies. Over the benchmark customer review datasets, experiment results demonstrate the effectiveness of the proposed approach in opinion target recognition in comparison with other baseline methods.

Contribution of emissions to concentrations: the TAGGING 1.0 submodel based on the Modular Earth Submodel System (MESSy 2.52)

Abstract. Questions such as what is the contribution of road traffic emissions to climate change? or what is the impact of shipping emissions on local air quality? require a quantification of the contribution of specific emissions sectors to the concentration of radiatively active species and air-quality-related species, respectively. Here, we present a diagnostics package, implemented in the Modular Earth Submodel System (MESSy), which keeps track of the contribution of source categories (mainly emission sectors) to various concentrations. The diagnostics package is implemented as a submodel (TAGGING) of EMAC (European Centre for Medium-Range Weather Forecasts – Hamburg (ECHAM)/MESSy Atmospheric Chemistry). It determines the contributions of 10 different source categories to the concentration of ozone, nitrogen oxides, peroxyacytyl nitrate, carbon monoxide, non-methane hydrocarbons, hydroxyl, and hydroperoxyl radicals ( = tagged tracers). The source categories are mainly emission sectors and some other sources for completeness. As emission sectors, road traffic, shipping, air traffic, anthropogenic non-traffic, biogenic, biomass burning, and lightning are considered. The submodel obtains information on the chemical reaction rates, online emissions, such as lightning, and wash-out rates. It then solves differential equations for the contribution of a source category to each of the seven tracers. This diagnostics package does not feed back to any other part of the model. For the first time, it takes into account chemically competing effects: for example, the competition between NOx, CO, and non-methane hydrocarbons (NMHCs) in the production and destruction of ozone. We show that the results are in-line with results from other tagging schemes and provide plausibility checks for concentrations of trace gases, such as OH and HO2, which have not previously been tagged. The budgets of the tagged tracers, i.e. the contribution from individual source categories (mainly emission sectors) to, e.g., ozone, are only marginally sensitive to changes in model resolution, though the level of detail increases. A reduction in road traffic emissions by 5 % shows that road traffic global tropospheric ozone is reduced by 4 % only, because the net ozone productivity increases. This 4 % reduction in road traffic tropospheric ozone corresponds to a reduction in total tropospheric ozone by ≈ 0.3 %, which is compensated by an increase in tropospheric ozone from other sources by 0.1 %, resulting in a reduction in total tropospheric ozone of ≈ 0.2 %. This compensating effect compares well with previous findings. The computational costs of the TAGGING submodel are low with respect to computing time, but a large number of additional tracers are required. The advantage of the tagging scheme is that in one simulation and at every time step and grid point, information is available on the contribution of different emission sectors to the ozone budget, which then can be further used in upcoming studies to calculate the respective radiative forcing simultaneously.

Particle tagging and its implications for stellar population dynamics

We establish a controlled comparison between the properties of galactic stellar halos obtained with hydrodynamical simulations and with `particle tagging'. Tagging is a fast way to obtain stellar population dynamics: instead of tracking gas and star formation, it `paints' stars directly onto a suitably defined subset of dark matter particles in a collisionless, dark-matter-only simulation.Our study shows that there are conditions under which particle tagging generates good fits to the hydrodynamical stellar density profiles of a central Milky-Way-like galaxy and its most prominent substructure. Phase-space diffusion processes are crucial to reshaping the distribution of stars in infalling spheroidal systems and hence the final stellar halo. We conclude that the success of any particular tagging scheme hinges on this diffusion being taken into account, at a minimum by making use of `live' tagging schemes, in which particles are regularly tagged throughout the evolution of a galaxy.

A Proposed Adaptive Scheme for Arabic Part-of Speech Tagging

This paper presents an Arabic-compliant part-of-speech (POS) tagging scheme based on using atomic tag markers that are grouped together using brackets. This scheme promotes the speedy production of annotations while preserving the richness of resultant annotations. The proposed scheme is comprised of two main elements, a new tokenization approach and a custom tool that enables the semi-automatic implementation of this scheme. The proposed model can serve in many scenarios where the user is in a need for better Arabic support and more control over the Part-of-Speech tagging process. This scheme was used to annotate sample narratives and it demonstrated capability and adaptability while addressing the various distinguishing features of Arabic language including its unique declension system. It also sets new baselines that are prospect for further exploration by future efforts.

Rapid labeling of amino acid neurotransmitters with a fluorescent thiol in the presence of o-phthalaldehyde.

LIF detection often requires labeling of analytes with fluorophores; and fast fluorescent derivatization is valuable for high-throughput analysis with flow-gated CE. Here, we report a fast fluorescein-labeling scheme for amino acid neurotransmitters, which were then rapidly separated and detected in flow-gated CE. This scheme was based on the reaction between primary amines and o-phthalaldehyde in the presence of a fluorescent thiol, 2-((5-fluoresceinyl)aminocarbonyl)ethyl mercaptan (FACE-SH). The short reaction time (<30 s) was suited for on-line mixing and derivatization that was directly coupled with flow-gated CE for rapid electrophoretic separation and sensitive LIF detection. To maintain the effective concentration of reactive FACE-SH, Tris(2-carboxyethyl)phosphine was added to the derivatization reagents to prevent thiol loss due to oxidation. This labeling scheme was applied to the detection of neurotransmitters by coupling in vitro microdialysis with online derivatization and flow-gated CE. It is also anticipated that this fluorophore tagging scheme would be valuable for on-chip labeling of proteins retained on support in SPE.

A method for high precision sequencing of near full-length 16S rRNA genes on an Illumina MiSeq.

BackgroundThe bacterial 16S rRNA gene has historically been used in defining bacterial taxonomy and phylogeny. However, there are currently no high-throughput methods to sequence full-length 16S rRNA genes present in a sample with precision.ResultsWe describe a method for sequencing near full-length 16S rRNA gene amplicons using the high throughput Illumina MiSeq platform and test it using DNA from human skin swab samples. Proof of principle of the approach is demonstrated, with the generation of 1,604 sequences greater than 1,300 nt from a single Nano MiSeq run, with accuracy estimated to be 100-fold higher than standard Illumina reads. The reads were chimera filtered using information from a single molecule dual tagging scheme that boosts the signal available for chimera detection.ConclusionsThis method could be scaled up to generate many thousands of sequences per MiSeq run and could be applied to other sequencing platforms. This has great potential for populating databases with high quality, near full-length 16S rRNA gene sequences from under-represented taxa and environments and facilitates analyses of microbial communities at higher resolution.