Display omitted
•We propose a hybrid method based on RNN and CRF for de-identification.•Extend the LSTM-based model by adding some context features to neural network.•An ensemble classifier is ...deployed to combine results of different methods.•Our system achieves a strict F1-score of 91.43% on N-GRID corpus, which ranks first.
De-identification, identifying information from data, such as protected health information (PHI) present in clinical data, is a critical step to enable data to be shared or published. The 2016 Centers of Excellence in Genomic Science (CEGS) Neuropsychiatric Genome-scale and RDOC Individualized Domains (N-GRID) clinical natural language processing (NLP) challenge contains a de-identification track in de-identifying electronic medical records (EMRs) (i.e., track 1). The challenge organizers provide 1000 annotated mental health records for this track, 600 out of which are used as a training set and 400 as a test set. We develop a hybrid system for the de-identification task on the training set. Firstly, four individual subsystems, that is, a subsystem based on bidirectional LSTM (long-short term memory, a variant of recurrent neural network), a subsystem-based on bidirectional LSTM with features, a subsystem based on conditional random field (CRF) and a rule-based subsystem, are used to identify PHI instances. Then, an ensemble learning-based classifiers is deployed to combine all PHI instances predicted by above three machine learning-based subsystems. Finally, the results of the ensemble learning-based classifier and the rule-based subsystem are merged together. Experiments conducted on the official test set show that our system achieves the highest micro F1-scores of 93.07%, 91.43% and 95.23% under the “token”, “strict” and “binary token” criteria respectively, ranking first in the 2016 CEGS N-GRID NLP challenge. In addition, on the dataset of 2014 i2b2 NLP challenge, our system achieves the highest micro F1-scores of 96.98%, 95.11% and 98.28% under the “token”, “strict” and “binary token” criteria respectively, outperforming other state-of-the-art systems. All these experiments prove the effectiveness of our proposed method.
By embedding prior knowledge into the learning structure, this paper presents a two-step semi-supervised learning method called fuzzy deep belief networks (FDBN) for sentiment classification. First, ...we train the general deep belief networks (DBN) by the semi-supervised learning taken on training dataset. Then, we design a fuzzy membership function for each class of reviews based on the learned deep architecture. Since the training of DBN maps each review into the DBN output space, the distribution of all training samples in the space is treated as prior knowledge and is encoded by a series of fuzzy membership functions. Second, based on the fuzzy membership functions and the DBN obtained in the first step, a novel FDBN architecture is constructed and the supervised learning stage is applied to improve the classification performance of the FDBN. FDBN not only inherits the powerful abstraction ability of DBN, but also demonstrates the attractive fuzzy classification ability for handling sentiment data. To inherit the advantages of both active learning and FDBN, we also propose an active FDBN (AFD) semi-supervised learning method. The empirical validation on five sentiment classification datasets demonstrates the effectiveness of FDBN and AFD methods.
In natural language processing community, sentiment classification based on insufficient labeled data is a well-known challenging problem. In this paper, a novel semi-supervised learning algorithm ...called active deep network (ADN) is proposed to address this problem. First, we propose the semi-supervised learning framework of ADN. ADN is constructed by restricted Boltzmann machines (RBM) with unsupervised learning based on labeled reviews and abundant of unlabeled reviews. Then the constructed structure is fine-tuned by gradient-descent based supervised learning with an exponential loss function. Second, in the semi-supervised learning framework, we apply active learning to identify reviews that should be labeled as training data, then using the selected labeled reviews and all unlabeled reviews to train ADN architecture. Moreover, we combine the information density with ADN, and propose information ADN (IADN) method, which can apply the information density of all unlabeled reviews in choosing the manual labeled reviews. Experiments on five sentiment classification datasets show that ADN and IADN outperform classical semi-supervised learning algorithms, and deep learning techniques applied for sentiment classification.
The evolution of brown carbon (BrC) during atmospheric aging, including the changes in optical properties and chemical compositions, is still unclear. Light absorption and fluorescence of BrC ...fraction extracted from fresh and ozonized propane soot particles by methanol were systematically measured, which showed that (1) the mass absorption efficiencies (MAE) sharply decreased by ozone (O3) aging (e.g., 1.2 ± 0.3–0.8 ± 0.1 m2 g−1 for MAE365), but changed slowly with increased O3 concentration (e.g., from 0.7 ± 0.2–0.8 ± 0.1 m2 g−1 for MAE365); (2) the fluorescence emission peaks were blue shifted, implying a loss of conjugated structures; (3) excitation–emission matrix analysis suggested that humic-like substances, charge transfer complexes, and polycyclic aromatic hydrocarbon (PAH)-like substances were the main chromophores. The PAH loss, accompanied by the decline of surface C˭C content, contributed more to the change of optical properties than the oxygenated PAH formation, thereby leading to the decrease in light absorption and fluorescence with O3 aging. This research reveals the importance of identifying the components responsible for optical properties in investigating the evolution of BrC during atmospheric aging, and is benefit for improving the evaluation of BrC’s radiative forcing.
Display omitted
•A “propane soot + O3” system is used to study the evolution of brown carbon (BrC).•Decreasing absorption and blue-shifted fluorescence peaks of BrC were observed.•Results of the optical properties correspond to composition characterization.•Polycyclic aromatic hydrocarbon loss appears to match optical property changes.
The chemical characteristics of complex organic matter in atmospheric aerosols remain poorly understood. Water-insoluble organic matter (WISOM) and water-soluble organic matter (WSOM) in the total ...suspended particulates collected in the city of Nagoya in summer/early autumn and winter were extracted using multiple solvents. Two fractions of humic-like substances, showing neutral and acidic behavior (HULIS-n and HULIS-a, respectively), and the remaining highly polar part (HP-WSOM) were fractionated from WSOM using solid phase extraction. The chemical structural characteristics and concentrations of the organic matter were investigated using mass spectrometry and Fourier transform infrared (FT-IR) spectroscopy. WISOM and HULIS-n had low O/C ratios (0.1 and 0.4, respectively) and accounted for a large fraction of the organics in aerosols (70%). HULIS-a and HP-WSOM had higher O/C ratios (0.7 and 1.0, respectively), and their concentrations in summer and early autumn were on average ∼2 times higher than those in winter. The mass spectrum and FT-IR analyses suggest the following: (1) WISOM were high-molecular-weight aliphatics (primarily C27–C32) with small proportions of −CH3, −OH, and CO groups; (2) HULIS-n was abundant in aliphatic structures and hydroxyl groups (primarily C9–C18) and by branched structures; (3) HULIS-a and HP-WSOM contained relatively large amounts of low-molecular-weight carboxylic acids and alcohols (primarily C4–C10); and (4) WISOM and HULIS-n were relatively abundant in amines and organic nitrates.
It is urgent to control the emission of volatile organic compounds (VOCs) due to their harmful effects on the environment and human health. A hybrid system integrating non-thermal-plasma and ...catalysis is regarded as one of the most promising technologies for VOCs removal due to their high VOCs removal efficiency, product selectivity and energy efficiency. This review systematically documents the main findings and improvements of VOCs removal using plasma-catalysis technology in recent 10 years. To better understand the fundamental relation between different aspects of this research field, this review mainly addresses the catalyst development, key influential factors, generation of by-products and reaction mechanism of VOCs decomposition in the plasma-catalysis process. Also, a comparison of the performance in various VOCs removal processes is provided. Particular emphasis is given to the importance of the selected catalyst and the synergy of plasma and catalyst in the VOCs removal in the hybrid system, which can be used as a reference point for future studies in this field.
Display omitted
•Recent process achieved in VOCs removal in plasma-catalysis system are summarized.•The main influential factors affecting the VOCs removal are discussed.•The synergetic interactions between plasma and catalyst are provided.•The properties of catalyst suitable for plasma-catalysis system are summarized.•The comparison of the performance of various VOCs removal process is presented.
Understanding the characteristics and sources of atmospheric chromophores is essential to assess their impact on climate change and the quality of atmospheric environment. In this work, the fine ...particulate matter (PM2.5) samples of Xi'an, China in 2017 were analyzed by excitation-emission matrices and parallel factor analysis (EEM-PARAFAC) method to obtain the species, content, sources and seasonal variation characteristics of atmospheric chromophores. The results showed that humic-like (HULIS) chromophores and polycyclic aromatic hydrocarbons-like (PAHs-like) chromophores were the most abundant chromophores in the samples, accounting for 42% and 33%, respectively. With the aggravation of air pollution, the relative content of low-polarity chromophores increased markedly, while the relative content of polar chromophores decreased. The concentrations of atmospheric chromophores exhibited obvious seasonal variation characteristics: high in winter and low in summer. Similarly, the relative contributions of atmospheric chromophores from each source varied with the season. In addition, special weather and human activities had a significant influence on the source of atmospheric chromophores. Dust source was an important source of atmospheric chromophores, which was susceptible to long-range incoming air masses from northwestern regions in spring. However, the chromophores from the dust source were easily removed by wet precipitation, which was the same as the chromophores from the combustion source. The chromophores from the combustion source were susceptible to human activities. The contribution of combustion source to atmospheric chromophores was reduced due to the implementation of air pollution control policies during the Chinese Spring Festival. In summer, the formation of photochemical secondary chromophores was more significant than in other seasons, and the photochemical secondary chromophores increased due to the formation of liquid phase reactions under high relative humidity conditions.
Display omitted
•The low-polarity chromophores increased with the aggravation of air pollution.•Dust source was an important source of atmospheric chromophores in spring.•The formation of photochemical secondary chromophores was observed in summer.•The combustion source was reduced during the Chinese Spring Festival.
In biomedical research, chemical and disease relation extraction from unstructured biomedical literature is an essential task. Effective context understanding and knowledge integration are two main ...research problems in this task. Most work of relation extraction focuses on classification for entity mention pairs. Inspired by the effectiveness of machine reading comprehension (RC) in the respect of context understanding, solving biomedical relation extraction with the RC framework at both intra-sentential and inter-sentential levels is a new topic worthy to be explored. Except for the unstructured biomedical text, many structured knowledge bases (KBs) provide valuable guidance for biomedical relation extraction. Utilizing knowledge in the RC framework is also worthy to be investigated. We propose a knowledge-enhanced reading comprehension (KRC) framework to leverage reading comprehension and prior knowledge for biomedical relation extraction. First, we generate questions for each relation, which reformulates the relation extraction task to a question answering task. Second, based on the RC framework, we integrate knowledge representation through an efficient knowledge-enhanced attention interaction mechanism to guide the biomedical relation extraction.
The proposed model was evaluated on the BioCreative V CDR dataset and CHR dataset. Experiments show that our model achieved a competitive document-level F1 of 71.18% and 93.3%, respectively, compared with other methods.
Result analysis reveals that open-domain reading comprehension data and knowledge representation can help improve biomedical relation extraction in our proposed KRC framework. Our work can encourage more research on bridging reading comprehension and biomedical relation extraction and promote the biomedical relation extraction.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Entity recognition is one of the most primary steps for text analysis and has long attracted considerable attention from researchers. In the clinical domain, various types of entities, such as ...clinical entities and protected health information (PHI), widely exist in clinical texts. Recognizing these entities has become a hot topic in clinical natural language processing (NLP), and a large number of traditional machine learning methods, such as support vector machine and conditional random field, have been deployed to recognize entities from clinical texts in the past few years. In recent years, recurrent neural network (RNN), one of deep learning methods that has shown great potential on many problems including named entity recognition, also has been gradually used for entity recognition from clinical texts.
In this paper, we comprehensively investigate the performance of LSTM (long-short term memory), a representative variant of RNN, on clinical entity recognition and protected health information recognition. The LSTM model consists of three layers: input layer - generates representation of each word of a sentence; LSTM layer - outputs another word representation sequence that captures the context information of each word in this sentence; Inference layer - makes tagging decisions according to the output of LSTM layer, that is, outputting a label sequence.
Experiments conducted on corpora of the 2010, 2012 and 2014 i2b2 NLP challenges show that LSTM achieves highest micro-average F1-scores of 85.81% on the 2010 i2b2 medical concept extraction, 92.29% on the 2012 i2b2 clinical event detection, and 94.37% on the 2014 i2b2 de-identification, which is considerably competitive with other state-of-the-art systems.
LSTM that requires no hand-crafted feature has great potential on entity recognition from clinical texts. It outperforms traditional machine learning methods that suffer from fussy feature engineering. A possible future direction is how to integrate knowledge bases widely existing in the clinical domain into LSTM, which is a case of our future work. Moreover, how to use LSTM to recognize entities in specific formats is also another possible future direction.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
•E1 and BPA were the dominant compounds detected among the target compounds.•Estrogenic compounds in Yangtze River Estuary were mainly influenced by Shanghai.•TOC was the factor primarily influencing ...the E1 and BPA distributions in sediments.•The BPA and E1 distributions in sediments did not obviously vary with the seasons.
Surface water and sediments in the Yangtze River Estuary (YRE) and the adjacent East China Sea (ECS) were sampled to investigate the spatial and seasonal distributions of Estrone (E1), 17β-estradiol (E2), 17α-ethinylestradiol (EE2) and bisphenol A (BPA). E1 and BPA were the dominant compounds detected. The zones of the highest E1 and BPA concentrations in water were located at the mouth of the Huangpu River and the outfalls of wastewater treatment plants. The zones of the highest concentrations in sediments were mainly located in the down reach of the estuary and the adjacent sea. The relationship between E1 and BPA in sediments with those in water was not significant (BPA: r=0.16, p=0.21; E1: r=0.18, p=0.24), but positive correlations with the total organic carbon (TOC) contents of sediments (BPA: r=0.57, p<0.01; E1: r=0.33, p=0.04) and negative correlations with the sand contents of sediments (BPA: r=−0.52, p<0.01; E1: r=–0.16, p=0.23) were found. The TOC contents were the major factor influencing the E1 and BPA distributions in sediments. The hierarchical cluster analysis further indicated the spatial distributions of BPA and E1 in sediments were obviously affected by TOC and sediment particle sizes, whilst the distributions varied slightly with seasons.