Features Of Chinese Characters Research Articles

COSMOLOGICAL CONCEPT OF TIME IN “THE CHANGE OF ZHOU” (“THE BOOL OF CHANGE”)

This paper represents one chapter from my PhD dissertation, written and successfully defended at Beijing Normal University in 2012. The whole content is translated from my doctoral dissertation that was originally written in Chinese language. According to the existing references, the earliest Chinese philosopher who mentioned in his essays the word “cosmos – yuzhou” was Shi Jiao (390-330 BC). He said: “four directions of Heaven and Earth are called ′yu′, time since ancient times till today is called ′zhou′”. In other words, yu represents space, while zhou symbolizes time, and that is why Chinese ancient cosmology may be even called “concept of space and time”. Inscriptions on the oracle bones and tortoiseshells with the symbols called “ganzhi”, originating from the Yin dynasty, were used to record hours, days, months and years i.e. time, but at the same time, Chinese people used these symbols to mark the directions of space, too. All of the cosmological diagrams used by “The Change of Zhou” and yijingologists, like Fuxi’s transcendental bagua diagram, king Wen’s empirical bagua diagram, river map and Luo book simultaneously depict time and space. In the eyes of the Chinese people, since ancient times, time and space are one whole and inseparable unity. However, being a foreigner, the writer of these lines will separately analyze the meaning of time in “The Book of Change”. As a foreigner, I find that the greatest charm of Chinese characters is that every character holds the whole history and etymology of one word or term in it. In other words, the most prominent feature of Chinese characters is symbolism. Therefore, if we want to explore “the time” concept, we must start from the character 时- shi meaning “time”. Character shi is formed of one radical meaning “sun”, and another meaning “measure”. This shows that the meaning of character shi is “measuring the sun”, i.e. measuring the length of sun shadow. The shorter the sun shadow is, the weather is being warmer and the Yang energy gets more abundant. It explains that character shi depicts increasing and decreasing of Yin and Yang energy in the cosmos and on the Earth. This paper analyzes time concept in “The Change of Zhou” from the Chinese and Western point of view. It explains meanings of Heavenly dao, Heavenly time, chance and synchronicity. The paper uses time theories (like simplicity of time, presentism, etc.) of German scientist Leibniz（1646 – 1716）, and famous Swiss psychologist Carl Jung’s（1875 - 1961） concepts (like cosmical archetypes and synchronicities), and compare them with ideas of “the cognition of subtle”, “progress through change” and “capability to grasp the chance” in “The Change of Zhou”. Finally, it combines and summarises the concept of time from “The Change of Zhou” and Western point of view.

Chinese Named Entity Recognition method based on multi-feature fusion and biaffine

Chinese Named Entity Recognition (CNER) focuses on precisely identifying predefined structural categories in unstructured Chinese text. Most existing CNER models do not consider the unique glyph and pinyin features of Chinese characters, but the rich semantic features hidden behind these features have a good effect on enhancing the judgment ability of language models. At the same time, it is difficult to identify the boundaries of Chinese nested entities, and accurately identifying the boundaries of entities within nested entities is also a difficult problem to solve. We propose a CNER method based on multi-feature fusion technology and biaffine mechanism to address the above issues: In the input representation layer, integrate the glyph and pinyin features of Chinese characters together, intuitively capturing the semantics of Chinese characters. Furthermore, biaffine mechanism has been introduced to provide a comprehensive view of the input on a global scale. This mechanism effectively converts the task of entity recognition into a problem of assigning scores to spans, hence enhancing the precision of identifying entity borders. In order to evaluate the efficacy of the model, a series of experiments were done on three Chinese entity recognition datasets: Resume, MSRA, and People Daily. The experimental results show that the solid boundary can be identified more accurately, and the F1 values of 96.49%, 96.26% and 96.19% are obtained respectively, which has a better recognition effect than the baseline model.

DAT-MT Accelerated Graph Fusion Dependency Parsing Model for Small Samples in Professional Fields.

The rapid development of information technology has made the amount of information in massive texts far exceed human intuitive cognition, and dependency parsing can effectively deal with information overload. In the background of domain specialization, the migration and application of syntactic treebanks and the speed improvement in syntactic analysis models become the key to the efficiency of syntactic analysis. To realize domain migration of syntactic tree library and improve the speed of text parsing, this paper proposes a novel approach-the Double-Array Trie and Multi-threading (DAT-MT) accelerated graph fusion dependency parsing model. It effectively combines the specialized syntactic features from small-scale professional field corpus with the generalized syntactic features from large-scale news corpus, which improves the accuracy of syntactic relation recognition. Aiming at the problem of high space and time complexity brought by the graph fusion model, the DAT-MT method is proposed. It realizes the rapid mapping of massive Chinese character features to the model's prior parameters and the parallel processing of calculation, thereby improving the parsing speed. The experimental results show that the unlabeled attachment score (UAS) and the labeled attachment score (LAS) of the model are improved by 13.34% and 14.82% compared with the model with only the professional field corpus and improved by 3.14% and 3.40% compared with the model only with news corpus; both indicators are better than DDParser and LTP 4 methods based on deep learning. Additionally, the method in this paper achieves a speedup of about 3.7 times compared to the method with a red-black tree index and a single thread. Efficient and accurate syntactic analysis methods will benefit the real-time processing of massive texts in professional fields, such as multi-dimensional semantic correlation, professional feature extraction, and domain knowledge graph construction.

Age-related differences in associative memory recognition of Chinese characters and hippocampal subfield volumes

Associative memory is a type of hippocampal-dependent episodic memory that declines with age. Studies have examined the neural substrates underlying associative memory and considered the hippocampus holistically; however, the association between associative memory decline and volumetric change in hippocampal subfields in the context of normal aging remains uncharacterized. Leveraging the distinct linguistic features of Chinese characters to evaluate distinct types of false recognition, we investigated age-related differences in associative recognition and hippocampal subfield volumes, as well as the relationship between behavioral performance and hippocampal morphometry in 25 younger adults and 32 older adults. The results showed an age-related associative memory deficit, which was exacerbated after a 30-min delay. Older adults showed higher susceptibility to false alarm errors with recombined and orthographically related foils compared to phonologically or semantically related ones. Moreover, we detected a disproportionately age-related, time-dependent increase in orthographic errors. Older adults exhibited smaller volumes in all hippocampal subfields when compared to younger adults, with a less pronounced effect observed in the CA2/3 subfield. Group-collapsed correlational analyses revealed associations between specific hippocampal subfields and associative memory but not item memory. Additionally, multi-subfield regions had prominent associations with delayed recognition. These findings underscore the significance of multiple hippocampal subfields in various hippocampal-dependent processes including associative memory, recollection-based retrieval, and pattern separation ability. Moreover, our observations of age-related difficulty in differentiating perceptually similar foils from targets provide a unique opportunity for examining the essential contribution of individual hippocampal subfields to the pattern separation process in mnemonic recognition.

Visual analysis of inscriptions in the Tang Dynasty: a case study on the calligraphy style of Wang Xizhi

The art of inscriptions is one of the most important components of the glorious ancient culture in China. It is an essential method not only for ancients to record history and disseminate culture but also to provide original written and pictorial records for subsequent generations to learn about ancient Chinese culture. However, the art of inscriptions and calligraphy is currently not widely available because of its professional chirography, style and composition. Moreover, traditional calligraphy exhibitions cannot provide the public with an interactive and intuitive way to present the features of Chinese characters, such as stroke thickness. Therefore, in this paper, we present an interactive visual system to support the public in understanding and appreciating the calligraphic style of the inscriptions of the Tang Dynasty and the evolutionary path of the calligraphic style of Wang Xizhi. We first employ image processing technology to extract calligraphy features. Then, we help users explore the development of calligraphy from the spatial-temporal dimension and analyze the similarity between the works at two granularity levels: Chinese character structure and the works’ style. Furthermore, the system also provides a metaphorical visualization method to enhance the concretization of calligraphic appreciation. Case studies and comprehensive evaluation demonstrate the usability of our proposed visual analysis system of the calligraphic style of the inscriptions in the Tang Dynasty.

Polyphonic Poetics: Reading Yang Shaoping’s Pidgin Bamboo Branch Lyrics

Pidgin English appeared in China along the Pearl River Delta as a contact language composed of a mixture of English, Cantonese, Portuguese, Hindi, and Malay as early as the eighteenth century. It spread to the Yangzi River delta in the mid-nineteenth century when an increasing number of Cantonese merchants and workers traveled to other treaty ports after the First Opium War. Focusing on a series of poems called Pidgin Bamboo Branch Lyrics (Bieqin zhuzhici) published in the newspaper Shen Bao in 1873, this paper explores how pidgin was appropriated in Chinese poetry and altered the literary soundscape in late Qing China. I argue that these pidgin English terms created a heteroglot poetic space wherein Sinographs and European words, literary language and local tongues, and classical images and contemporary anecdotes intertwine to generate diverse and contradictory meanings. Through an innovative collocation of the graphic, literal, and phonetic features of Chinese characters, the pidgin words and expressions in these poems produce a multilayered structure of meanings with various possibilities for interpretation. Moreover, unlike traditional bamboo branch lyrics in which non-Chinese words are incorporated to domesticate strangeness and consolidate the imperial order of center and periphery, these poems have an inverse effect: their use of pidgin English instead alienates the native, the conventional, and the familiar landscape/soundscape. This poetry made of language crossings depicts modernity as an uncanny sonic experience and makes foreignness appear not in a faraway land but in one’s own poetic language and everyday life.

Chinese Clinical Named Entity Recognition From Electronic Medical Records Based on Multisemantic Features by Using Robustly Optimized Bidirectional Encoder Representation From Transformers Pretraining Approach Whole Word Masking and Convolutional Neural Networks: Model Development and Validation

Clinical electronic medical records (EMRs) contain important information on patients' anatomy, symptoms, examinations, diagnoses, and medications. Large-scale mining of rich medical information from EMRs will provide notable reference value for medical research. With the complexity of Chinese grammar and blurred boundaries of Chinese words, Chinese clinical named entity recognition (CNER) remains a notable challenge. Follow-up tasks such as medical entity structuring, medical entity standardization, medical entity relationship extraction, and medical knowledge graph construction largely depend on medical named entity recognition effects. A promising CNER result would provide reliable support for building domain knowledge graphs, knowledge bases, and knowledge retrieval systems. Furthermore, it would provide research ideas for scientists and medical decision-making references for doctors and even guide patients on disease and health management. Therefore, obtaining excellent CNER results is essential. We aimed to propose a Chinese CNER method to learn semantics-enriched representations for comprehensively enhancing machines to understand deep semantic information of EMRs by using multisemantic features, which makes medical information more readable and understandable. First, we used Robustly Optimized Bidirectional Encoder Representation from Transformers Pretraining Approach Whole Word Masking (RoBERTa-wwm) with dynamic fusion and Chinese character features, including 5-stroke code, Zheng code, phonological code, and stroke code, extracted by 1-dimensional convolutional neural networks (CNNs) to obtain fine-grained semantic features of Chinese characters. Subsequently, we converted Chinese characters into square images to obtain Chinese character image features from another modality by using a 2-dimensional CNN. Finally, we input multisemantic features into Bidirectional Long Short-Term Memory with Conditional Random Fields to achieve Chinese CNER. The effectiveness of our model was compared with that of the baseline and existing research models, and the features involved in the model were ablated and analyzed to verify the model's effectiveness. We collected 1379 Yidu-S4K EMRs containing 23,655 entities in 6 categories and 2007 self-annotated EMRs containing 118,643 entities in 7 categories. The experiments showed that our model outperformed the comparison experiments, with F1-scores of 89.28% and 84.61% on the Yidu-S4K and self-annotated data sets, respectively. The results of the ablation analysis demonstrated that each feature and method we used could improve the entity recognition ability. Our proposed CNER method would mine the richer deep semantic information in EMRs by multisemantic embedding using RoBERTa-wwm and CNNs, enhancing the semantic recognition of characters at different granularity levels and improving the generalization capability of the method by achieving information complementarity among different semantic features, thus making the machine semantically understand EMRs and improving the CNER task accuracy.

Selective Impairments in Fine Neural Tuning for Print in Chinese Children with Developmental Dyslexia.

Neural tuning for print refers to differential neural responses (e.g., the N1 component of event-related potentials) to different orthographic forms and other visual stimuli. While impaired neural tuning for print has been well established in dyslexic children who read alphabetic scripts, it remains unclear whether such effects exist in dyslexic children who read Chinese, which dramatically differs in visual and linguistic characteristics from alphabetic words. To fill this gap, we examined two levels of the neural tuning for print: coarse tuning (i.e., false character vs. stroke combination), and fine tuning (i.e., sub-lexical tuning: pseudo character vs. false character; and lexical tuning: real character vs. pseudo character). Using the event-related potential technique, we examined 14 typically developing children and 16 dyslexic children who were screened from 216 nine-year-old children in the third grade. For typically developing children, we observed both coarse and sub-lexical tuning. Critically, for dyslexic children, we found stronger N1 for false character than for stroke combination, suggesting intact coarse tuning, but a reduced N1 difference between false character and pseudo character, suggesting impaired sub-lexical tuning. These results clearly show selective impairments in fine neural tuning at the sub-lexical level in Chinese dyslexic children. Our findings may be associated with unique features of Chinese characters.

Chinese Named Entity Recognition Based on Multi-Feature Fusion FLAT Model in the Medical Field

The complexity of syntax and the specialized nature of Chinese electronic medical record data make it challenging to accurately identify medical entities using named entity recognition models. In order to precisely extract complex medical vocabulary from electronic medical records, this paper proposes a multi-feature-based named entity recognition model that addresses the issue of insufficient internal feature extraction in the FLAT model. Firstly, the radical and pinyin features of Chinese characters are extracted for enrichment and perfection of their semantic information. These features are then combined with the word embeddings extracted by the FLAT-lattice method. Finally, the fused features and position encoding are input to the Transformer encoder for encoding, followed by decoding using the CRF method. Experimental results demonstrate that the proposed model outperforms many existing algorithms on the CCL2021 dataset, with an F1 score of up to 91.75%.

Chinese named entity recognition method for the finance domain based on enhanced features and pretrained language models

For some named entities in the Chinese finance domain that are long, with difficult to delineate boundaries and diverse forms of expression, we propose a method based on pretrained language models for named entity recognition with enhanced features. First, the method considers entity boundary delineation and entity classification as two separate tasks and learns enhanced Chinese character features by introducing a gating-based multi-channel attention mechanism to delineate financial entity boundaries on the basis of a pretrained language model. Then, the boundary demarcation results are input into the pretrained language model in the form of mask units for data enhancement. Subsequently, document-level entity-based enhancement features are introduced to construct a finance entity classification model. We experimentally identified the best-performing Chinese pretrained language model from several state-of-the-art models and then embedded it into our method to compare against other benchmark models. The experimental results showed that our model is superior to other benchmark models on the named entity recognition task in the finance domain.

AIP: A Named Entity Recognition Method Combining Glyphs and Sounds

In recent years, a large number of Chinese electronic texts have been produced in the process of information construction in various fields. Identifying specific entities in these electronic texts has become a major research focus. Most existing research methods use radicals to extract the glyph features of Chinese characters but have seen its limitation. This paper extracts the features of Chinese characters from three aspects: glyph features, phonetic features, and character features, and improves conventional feature extraction methods for each kind of feature. A new named entity recognition method (AIP) is proposed by transforming Chinese characters into corresponding images for glyph feature extraction, dividing pinyin into initials, vowels, and tones for phonetic feature extraction, and fine-tuning the A Lite Bert model for character feature extraction to improve the performance of the model. This paper compares the performance of the AIP model and mainstream neural network models on Chinese named entity recognition tasks on commonly used data sets and the data sets in specific domains. The results showed that AIP achieved better results than the related work. The F1 values on the two data sets are 94.4% and 80.5%, respectively, which validates the model's versatility.

Hong Kong Chinese character psycholinguistic norms: ratings of 4376 single Chinese characters on semantic radical transparency, age-of-acquisition, familiarity, imageability, and concreteness

Several norms of psycholinguistic features of Chinese characters exist in Mandarin Chinese, but only a few are available in Cantonese or in the traditional script, and none includes semantic radical transparency ratings. This study presents subjective ratings of age-of-acquisition (AoA), familiarity, imageability, concreteness, and semantic radical transparency in 4376 Chinese characters. The single Chinese characters were rated individually on the five dimensions by 20 native Cantonese speakers in Hong Kong to form the Hong Kong Chinese Character Psycholinguistic Norms (HKCCPN). The split-half reliability and intra-class correlations testified to the high internal reliability of the ratings. Their convergent and discriminant patterns in relations to other psycholinguistic measures echoed previous findings reported on Chinese. There were high correlations for semantic radical transparency, imageability and concreteness, and moderate-to-high correlations for AoA and familiarity among subsets of items that had been collected in previous studies. Concurrent validity analyses showed convergence in predicting behavioral response times in various tasks (lexical decision, naming, and writing-to-dictation) when compared with other Chinese character databases. High predictive validity was shown in writing-to-dictation data from an independent sample of 20 native Cantonese speakers. Several objective psycholinguistic measures (character frequency, stroke number, number of words formed, number of homophones and number of meanings) were included in this database to facilitate its use. These new ratings extend the currently available norms in language and reading research in Cantonese Chinese for researchers, clinicians, and educators, as well as provide them with a wider choice of stimuli.

Pronounce differently, mean differently: A multi-tagging-scheme learning method for Chinese NER integrated with lexicon and phonetic features

Named Entity Recognition (NER) aims to automatically extract specific entities from the unstructured text. Compared with performing NER in English, Chinese NER is more challenging in recognizing entity boundaries because there are no explicit delimiters between Chinese characters. However, most previous researches focused on the semantic information of the Chinese language on the character level but ignored the importance of the phonetic characteristics. To address these issues, we integrated phonetic features of Chinese characters with the lexicon information to help disambiguate the entity boundary recognition by fully exploring the potential of Chinese as a pictophonetic language. In addition, a novel multi-tagging-scheme learning method was proposed, based on the multi-task learning paradigm, to alleviate the data sparsity and error propagation problems that occurred in the previous tagging schemes, by separately annotating the segmentation information of entities and their corresponding entity types. Extensive experiments performed on four Chinese NER benchmark datasets: OntoNotes4.0, MSRA, Resume, and Weibo, show that our proposed method consistently outperforms the existing state-of-the-art baseline models. The ablation experiments further demonstrated that the introduction of the phonetic feature and the multi-tagging-scheme has a significant positive effect on the improvement of the Chinese NER task.

The LightGBM-based classification algorithm for Chinese characters speech imagery BCI system.

Brain-computer interface (BCI) can obtain text information by decoding language induced electroencephalogram (EEG) signals, so as to restore communication ability for patients with language impairment. At present, the BCI system based on speech imagery of Chinese characters has the problem of low accuracy of features classification. In this paper, the light gradient boosting machine (LightGBM) is adopted to recognize Chinese characters and solve the above problems. Firstly, the Db4 wavelet basis function is selected to decompose the EEG signals in six-layer of full frequency band, and the correlation features of Chinese characters speech imagery with high time resolution and high frequency resolution are extracted. Secondly, the two core algorithms of LightGBM, gradient-based one-side sampling and exclusive feature bundling, are used to classify the extracted features. Finally, we verify that classification performance of LightGBM is more accurate and applicable than the traditional classifiers according to the statistical analysis methods. We evaluate the proposed method through contrast experiment. The experimental results show that the average classification accuracy of the subjects' silent reading of Chinese characters "(left)", "(one)" and simultaneous silent reading is improved by 5.24%, 4.90% and 12.44% respectively.

Experimental Measurement and Markov Chain Modelling of Stroke Order Intuition

As a feature of Chinese characters, stroke order plays an important role in Chinese character education, Chinese character recognition and handwriting identification. However, in recent years, the research on stroke order is limited to elaboration and supplementation of existing stroke order and stroke order rules. There is no special research that focuses on the principle behind stroke order. To quantify the internal regularity of stroke order, we start with the stroke order intuition in the actual writing. Because stroke order has individual differences, we select a group of students with small differences as testers. In this paper, we design an algorithm to generate random strokes and collect the timing data during testers copying strokes. After pre-processing the collected data, a Markov chain is introduced to model stroke order intuition, which can divide the stroke sequences into three states. And then we classify them into three situations according to the stroke distribution. Next, we compare them with some well-known stroke order rules. The results show that the probability distribution of strokes in different situations is not always consistent with the empirical rules, as well as the relations between them.

Chinese Character Culture Communication and Comparative Analysis Based on Big Data Visualization

With the promotion of China’s international status, the study of Chinese characters has become the trend of scholars all over the world. Nowadays, the international promotion of Chinese has become an important part of the national diplomatic strategy. With the rapid development of big data visualization technology, Chinese culture can be spread quickly, and it is easy to achieve the above requirements to achieve visual accessibility. In view of the problems existing in the current dissemination of Chinese culture, such as too scattered dissemination content, unsystematic dissemination channels, and poor dissemination effect, this paper puts forward a method of Chinese character culture dissemination based on big data visualization. Firstly, the visual features of Chinese characters are recognized, and the rules of Chinese characters are detected; secondly, the large data visualization analysis of Chinese character imaging is carried out, and the performance comparison under different iterations is compared. Chinese visual technology has obvious advantages in the effect and performance of Chinese character transmission. Finally, the development trend of Chinese communication is analyzed, and the acceptance, speed, coverage, understanding, learning degree, and practical degree of Chinese communication abroad are compared. The big data visualization model has certain advantages in functional comparison.

An Entity Relationship Extraction Model Based on BERT-BLSTM-CRF for Food Safety Domain.

Dealing with food safety issues in time through online public opinion incidents can reduce the impact of incidents and protect human health effectively. Therefore, by the smart technology of extracting the entity relationship of public opinion events in the food field, the knowledge graph of the food safety field is constructed to discover the relationship between food safety issues. To solve the problem of multi-entity relationships in food safety incident sentences for few-shot learning, this paper adopts the pipeline-type extraction method. Entity relationship is extracted from Bidirectional Encoder Representation from Transformers (BERTs) joined Bidirectional Long Short-Term Memory (BLSTM), namely, the BERT-BLSTM network model. Based on the entity relationship types extracted from the BERT-BLSTM model and the introduction of Chinese character features, an entity pair extraction model based on the BERT-BLSTM-conditional random field (CRF) is established. In this paper, several common deep neural network models are compared with the BERT-BLSTM-CRF model with a food public opinion events dataset. Experimental results show that the precision of the entity relationship extraction model based on BERT-BLSTM-CRF is 3.29%∼23.25% higher than that of other models in the food public opinion events dataset, which verifies the validity and rationality of the model proposed in this paper.

Recognition of the Agricultural Named Entities With Multifeature Fusion Based on ALBERT

High quality agricultural named entity recognition (NER) model can provide effective support for agricultural information extraction, semantic retrieval and other tasks. However, the existing models ignore the potential characteristics of Chinese characters, resulting in the lack of internal semantics. Moreover, the agricultural text sequence is long, which leads to the lack of long-distance dependence of model capture. In order to solve the above problems, a self-attention mechanism RSA-CANER agricultural named entity recognition model is proposed which incorporating the potential characteristics of Chinese characters. First of all, the model takes character features and potential features of Chinese characters as input to enrich semantic information. Among them, character features are obtained based on ALBERT pre training tool, radical features are extracted based on convolutional neural network (CNN), and stroke features are extracted based on bidirectional long short-term memory model (BiLSTM). Then, based on the BiLSTM, the sequence characteristic matrix is obtained, and the self-attention mechanism is used to further enhance the ability of the model to capture long-distance dependence. Finally, the global optimal sequence is generated based on conditional random field (CRF) model. It obtains an F1-score of 95.56. The experimental results show that the model learns semantic information at multiple fine-grained levels of radicals and strokes, enriches the vector expression of target words, and its recognition accuracy is better than other models, improving the generalization ability of the model.

Chinese-Named Entity Recognition From Adverse Drug Event Records: Radical Embedding-Combined Dynamic Embedding-Based BERT in a Bidirectional Long Short-term Conditional Random Field (Bi-LSTM-CRF) Model.

BackgroundWith the increasing variety of drugs, the incidence of adverse drug events (ADEs) is increasing year by year. Massive numbers of ADEs are recorded in electronic medical records and adverse drug reaction (ADR) reports, which are important sources of potential ADR information. Meanwhile, it is essential to make latent ADR information automatically available for better postmarketing drug safety reevaluation and pharmacovigilance.ObjectiveThis study describes how to identify ADR-related information from Chinese ADE reports.MethodsOur study established an efficient automated tool, named BBC-Radical. BBC-Radical is a model that consists of 3 components: Bidirectional Encoder Representations from Transformers (BERT), bidirectional long short-term memory (bi-LSTM), and conditional random field (CRF). The model identifies ADR-related information from Chinese ADR reports. Token features and radical features of Chinese characters were used to represent the common meaning of a group of words. BERT and Bi-LSTM-CRF were novel models that combined these features to conduct named entity recognition (NER) tasks in the free-text section of 24,890 ADR reports from the Jiangsu Province Adverse Drug Reaction Monitoring Center from 2010 to 2016. Moreover, the man-machine comparison experiment on the ADE records from Drum Tower Hospital was designed to compare the NER performance between the BBC-Radical model and a manual method.ResultsThe NER model achieved relatively high performance, with a precision of 96.4%, recall of 96.0%, and F1 score of 96.2%. This indicates that the performance of the BBC-Radical model (precision 87.2%, recall 85.7%, and F1 score 86.4%) is much better than that of the manual method (precision 86.1%, recall 73.8%, and F1 score 79.5%) in the recognition task of each kind of entity.ConclusionsThe proposed model was competitive in extracting ADR-related information from ADE reports, and the results suggest that the application of our method to extract ADR-related information is of great significance in improving the quality of ADR reports and postmarketing drug safety evaluation.

CHARM: An Improved Method for Chinese Precoding and Character-Level Embedding

The numerical transformation of text is a key step in natural language processing tasks, among which the word embedding model is the most representative. However, the word embedding model is insufficient in representing unregistered and low-frequency words, and the character-level embedding model makes up for it. Most Chinese character-level models focus on the independent use of Chinese character features such as strokes, radicals, and pinyin, or the shallow correlations between some features, while the inherent correlations among different features such as pronunciation, glyph, stroke order, and word frequency are not fully utilized. Through the statistical analyses of various features of Chinese characters, this paper proposes a precoding method based on Character Helix Alternative Representation Model (CHARM), which can realize the reversible mapping of Chinese characters or words to English-like sequences, and the advantage of this method is verified in the text classification test based on TextCNN. Experimental results on three test sets show that the model performs well, and can be a replacement character-level corpus for the original Chinese text.