e-book Literature and Science (Traditional Chinese Edition)

Free download. Book file PDF easily for everyone and every device. You can download and read online Literature and Science (Traditional Chinese Edition) file PDF Book only if you are registered here. And also you can download or read online all Book PDF file that related with Literature and Science (Traditional Chinese Edition) book. Happy reading Literature and Science (Traditional Chinese Edition) Bookeveryone. Download file Free Book PDF Literature and Science (Traditional Chinese Edition) at Complete PDF Library. This Book have some digital formats such us :paperbook, ebook, kindle, epub, fb2 and another formats. Here is The CompletePDF Book Library. It's free to register here to get Book file PDF Literature and Science (Traditional Chinese Edition) Pocket Guide.
Traditional Chinese Medicine and Constitutional Medicine in China, Japan and Korea: A and Hypertension Control: A Narrative Review of Chinese Literature.
Table of contents

We then generated the candidate relations between the co-occurring heterogeneous entity pairs and identified: Note that we did not extract herb-formula relations, because all the formulae in the dictionary are classical TCM formulae, and their relations with herbs have already been well-defined. We asked three TCM experts respectively denoted as Kang, Tang, and Zhan, according to their family names , who specialize in Chinese materia medica and clinical TCM, to annotate the data.

The statistics of our dataset are summarized in Table 1 , and the entire dataset is available online at http: We used Kappa statistics to measure inter-annotator agreements. After that, we employed the majority rule to decide the final label of each candidate relation. In this paper, we propose extracting relations in the context of heterogeneous TCM networks. So, we first give the definition of heterogeneous TCM networks, then present the problem formulation. Our objective is to then estimate the values of Y , and we can use a joint posterior probability P Y X , G to model its distribution.

Here, G denotes all forms of network information. This joint probability indicates that the labels of the edges depend not only on the local attributes associated with each edge, but also on the structure of the network.

Scientific Study of Chinese Medicine

To represent the dependencies between the labels Y and the attributes X and the correlations among the labels Y , we can define the following two categories of local factors:. Evidence factors , which are used to capture the dependencies between the labels of edges and their attributes. Compatibility factors , which are used to capture the compatibility among the labels of edges. We use triadic closures in heterogeneous TCM networks to construct compatibility factors. Triadic closure 38 is one of the fundamental processes of linking information in a network and has been applied in many aspects of social network mining, such as in inferring social ties 34 as well as social roles and statuses.

Sean T Bradley | Asian Languages & Literature | University of Washington

Figure 3 gives the graphical representation of an HFGM. The dotted ellipse at the bottom of the figure encloses the constructed heterogeneous TCM network, in which a node represents a TCM entity of a certain type. The dotted ellipse in the middle of the figure encloses the set of candidate relations, each of which corresponds to an edge in the input network. The dotted ellipse at the top of the figure encloses the factor graph generated from the input network, in which the colored ovals represent variables labels corresponding to the candidate relations, and the squares represent factors.

The green ovals represent the known labels that are taken as supervised information, while the red ovals represent the unknown labels to be predicted. The black squares represent the evidence factors between variables and their attributes, while the blue squares represent the compatibility factors of triadic closures. Graphical representation of a heterogeneous factor graph model HFGM. In general, the features of evidence factors should be able to reflect prior knowledge of the labels of edges.

In this study, we employed three categories of features for evidence factors:.

  • Working From Home: Legal, Tax, Insurance and Other Things That Matter!
  • Introduction.
  • Nothings Too Good For My Baby.
  • Traditional Chinese Medicine Zheng in the Era of Evidence-Based Medicine: A Literature Analysis?

This is the simplest feature, which represents the instances in which the two end-entities of a relation co-occur in the same document. Intuitively, the context surrounding the two end-entities is very important for identifying a relation. Six surrounding words were collected for each instance of an entity, three before and three after the instance. After removing the infrequent words, distinct words remained. We then defined a feature function for each of the words, to indicate the frequency that each word appeared around the two end-entities of a relation.

Determining the latent semantic relatedness of the two end-entities may also be helpful for identifying a relation. For calculating semantic distance, we first need to represent the semantic meanings of each entity. Distributed vector representations facilitate learning word meanings from large collections of text.

Alternative Medicine; Traditional Chinese Medicine

Each word is learned as a distinct pattern of continuous values over a single, large vector, with each dimension corresponding to a latent topic. We can then measure the semantic relatedness among words in terms of distances in the resulting vector space.

The quest for modernisation of traditional Chinese medicine

We used word2vec 40 https: We then defined a feature function on each dimension by using the absolute value of the difference between the values of the dimensions of the two end-entities. Except for the above three categories of features, syntactic structure ie, dependency relation is another type of information that is useful for relation extraction. However, being able to determine syntactic structure requires that the two end-entities co-occur within one single sentence. So, we did not take syntactic structure into account in this study. Illustration of the transitive property of triadic closures.

We propose using a semi-supervised learning algorithm to estimate the parameters of the model see Supplementary Appendix A , which enables us to predict the labels of unknown edges based on the estimated parameters see Supplementary Appendix B. We trained and evaluated our proposed HFGM on the dataset we collected and annotated. Five types of relations — herb-syndrome, herb-disease, formula-syndrome, formula-disease, and syndrome-disease — were extracted from the TCM literature we analyzed.

The entire semi-supervised learning and inference process took about 4. A traditional classification approach was employed as the baseline, in which the co-occurring frequency, lexical context, as well as the aforementioned semantic distance were taken as the classification features. Because the number of features involved in this approach was very large, we used an SVM 41 www. We designed an iterative SVM classifier in our experiments that takes the results of the basic SVM classifier as the initiate values of all the labels, then iteratively updates the label of each relation by simultaneously using its observed features and inferred neighboring label values, until the labels no longer change.

In this study, we also defined iterative features based on the transitive property of triadic closures. Specifically, we calculated the numbers of triadic closures with different sums ie, 0, 1, 2, and 3, as shown in Figure 4 that were formed by each relation with its neighboring relations. We performed a five-fold cross-validation to evaluate the performance of our model.

Table 3 shows the performance of the HFGM as compared with that of two other approaches. To further determine the effectiveness of our approach, we plotted receiver operating characteristic curves of the basic SVM and HFGM approaches as shown in Figure 5 , where the y -axis represents the rate of predicated positive labels in all the positive samples, and the x -axis represents the rate of predicted positive labels in all the negative samples.

Receiver operating characteristic curves of different approaches for extracting a herb-syndrome, b herb-disease, c formula-syndrome, d formula-disease, and e syndrome-disease relations. The y -axis represents the true positive rate and the x -axis represents the false positive rate. After performing an in-depth analysis of some specific instances, we found that our HFGM significantly improves the accuracy of relation extraction in the following cases in which traditional classifiers have difficult identify relations:. The context is very short. For a candidate relation, we took the co-occurring frequency of the two end-entities appearing in the same documents and the frequencies of their surrounding words as the classification features, so if a relation only appears once in a single, short document, then the context information will not be enough to identify the relation.

The context contains confusing information. In some TCM treatment experiments, one or more control groups are used to compare the effectiveness of different treatments, which often misleads traditional classifiers to extract some relations from the control groups that are not actually present. Several different studies are reported on in the same document.

Occasionally, several studies will be covered within a single article or even a single sentence. The name of an entity is a polysemous word. Some Chinese names for TCM entities have other meanings. The effectiveness of the HFGM demonstrates the existence of correlations among different types of relations. In addition, using the transitive property of triadic closures to model the dependencies among the labels of edges in the network is a reasonable and practicable approach. However, some of our results can be improved upon, and some of the approaches we employ can be expanded upon in the future.

Firstly, some other types of important TCM entities, such as symptoms, are not incorporated into our model. This is because there is not a standard or unified terminology glossary for TCM symptoms, so entity recognition techniques are needed to detect the instances of symptom entities in text. If we can bring such entities into our unified model in the future, then more types of relations can be extracted. Another challenge is that the computational complexity of learning the HFGM is very high, because multiple rounds of approximate inferences are required over the entire dataset see Algorithm 1 of the online supplementary data.

Consequently, we need to develop efficient learning approaches. In addition, other approximation techniques, such as the pseudolikelihood measure, 42 , 43 may also be used in our collective inference methods. Our approach also be directly applied to relation extraction in the field of biomedical text mining. We can construct heterogeneous networks between biomedical entities eg, proteins, genes, phenotypes, biological targets, diseases, drugs, treatments gathered from biomedical literature or clinical records, then employ the HFGM to extract biomedical relations in the context of these heterogeneous biomedical networks.

The HFGM model proposed in this article is only suitable for heterogeneous entity networks that contain at least three kinds of entities, because we use triadic closures formed by three kinds of entities to construct compatibility factors in the model. Another limitation of the current version of our model is that it can only extract one class of relations between the same two types of entities at the same time, because we treat the relation extraction problem as a binary classification problem in this study.

Traditional Chinese Medicine Zheng in the Era of Evidence-Based Medicine: A Literature Analysis

In this article, we examine the problem of automatically extracting meaningful entity relations from TCM literature and propose an HFGM that exploits the power of collective inference in the context of heterogeneous entity networks to simultaneously and globally extract all types of relations eg, herb-syndrome, herb-disease, formula-syndrome, formula-disease, and syndrome-disease relations from the entire corpus of TCM data. We propose using a semi-supervised learning algorithm to estimate the parameters of the model.

The results of our analysis of a professionally annotated dataset show that our approach is superior to traditional classification methods in extracting multiple types of relations from TCM literature.

Skip straight to:

This work was mainly done while the first author was visiting KU Leuven, Belgium. The work was a collaboration between all the authors. All the authors have made valuable contributions to revising and approving the manuscript. Supplementary material is available online at Supplementary Data. Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide. Sign In or Create an Account. Close mobile search navigation Article navigation.

Extracting relations from traditional Chinese medicine literature via heterogeneous entity networks Huaiyu Wan. The result is shown in Figure 1. Many studies on TCM Zheng involved biomedical diseases, so the statistics included studies that involved diseases and those that did not involve diseases. In analyzing the Chinese literature, we filtered by title, keyword, and abstract. The count of independent Zheng studies increases by one if no disease name occurred; otherwise, the count of disease-related Zheng increases by one.

The result was shown in Figure 2. Similar to Section 2. The statistical method was similar to that of Section 2. The only difference was that the two methods were focused on different languages Chinese versus English, resp. By limiting the TCM Zheng literature to clinical studies, we obtained the frequencies of studies related to different diseases. The 10 most commonly associated diseases were listed in Table 1.

According to the results of Section 2. By analyzing these datasets with respect to their dates of publication, we obtained their annual distributions, shown in Figure 4. Based on the cooccurrence of TCM Zheng and applying the data slicing algorithm [ 9 ], we obtained the Zheng-Zheng network, shown in Figure 5. Because there was a strong connection between TCM Zheng and disease in both clinical practice and research studies, it was necessary to obtain the frequencies of different disease-Zheng association items that commonly existed in the Chinese literature.

By analyzing the literature associated with both TCM Zheng and disease names in a framework of Western medicine, we obtained a list of associated items of disease-Zheng and their frequencies. For simplicity, we list the 20 most common in Table 2. Because one disease could be involved with several Zhengs, it is necessary to explore the major Zhengs that are associated with each particular disease. These statistics were focused on the cooccurrence of disease names and Zheng terms. By analyzing the Chinese literature, we obtained a disease-Zheng network.

In Figure 6 , we listed the 5 most common diseases and their associated Zhengs. All analyses were performed based on these studies. Figure 1 showed an annual increase in the number of publications in the SinoMed database. The number of articles has increased rapidly in the past 2 decades. In addition, the portion of clinical studies has increased substantially, especially after Animal experimental studies remained insignificant, and the numbers of related articles remained a small proportion of the total, indicating that animal experimentation has not been a major part of Zheng-related studies.

As a diagnostic method, TCM Zheng diagnosis can be integrated with a biomedical diagnosis in clinical practice, thus we can classify the whole studies into two categories, independent Zheng and Zheng in disease. The former indicates those studies considering only TCM Zheng classification without any biomedical disease information; the Zheng in disease studies refers to those studies aiming at the TCM Zheng research based on one or more biomedical diseases, or the integrative study on TCM Zheng and biomedical diseases. The majority of studies are independent of biomedical disease, as shown in Figure 2 , confirming that TCM Zheng classification can be discussed as a different classification system independent of disease diagnosis, although the integration of Zheng and disease diagnosis is common in clinical practice.

The proportion of studies that were correlated with biomedical diseases is increasing over time, especially after the year The advantage of integrating TCM Zheng with biomedical disease diagnoses has been emphasized in recent years, and a number of novel achievements have been acquired in this field. After , the annual number of articles in English-language journals on TCM Zheng in PubMed increased dramatically, but the total number was far less than the number of Chinese-language articles, as shown in Figure 3.

Among these studies, the percentage of clinical studies grew rapidly, a trend that was consistent with that of Chinese-language studies. A higher proportion of animal experimental studies was reported in PubMed than in SinoMed. The 10 most common diseases in Chinese-language TCM Zheng-related studies are summarized in Table 1 , and the annual numbers are shown in Figure 4.

From Table 1 and Figure 4 , it can be concluded that most of the TCM Zheng-related diseases are complex chronic diseases, which implies that researchers tend to focus on these chronic diseases in TCM Zheng-related studies due to the superior efficacy of herbal prescriptions in treating these diseases. There are thousands of studies per year focusing on TCM Zheng studies of diabetes mellitus and gastritis, and both of these diseases manifest with multiple symptoms with an increasing incidence in China and can be treated with herbal medicines. As a basic unit in a TCM diagnosis, Zheng can be shown in combination two or more Zhengs in a patient, and Zheng can change during the development of an illness.

During the data analysis, it can be found that most disease-Zhengs studies are published in Chinese. Although there are a small amount of English publications concerning the disease-Zheng research, most of them were published in English abstract, which actually were published in Chinese, and can be collected in SinoMed database. Thus we abandoned the English data in this analysis, for the data is too few, and also it is not appropriate in this study to combine both data together.

Figure 5 illustrates those Zhengs and the Zheng-Zheng association network. Clockwise from the largest node, the first is the liver-kidney yin deficiency pattern connecting five nodes: Six nodes of the network are connected to the second largest node kidney yin deficiency pattern. The yang deficiency pattern and pattern of dual deficiency of yin and yang are two patterns with relatively low frequencies.

The upper left corner is the dampness-heat pattern and connecting node spleen-stomach dampness-heat pattern. The upper right is the qi deficiency pattern, connecting with the spleen qi deficiency pattern and lung qi deficiency pattern. The lower right is the blood stasis pattern, connecting with the pattern of qi deficiency with blood stasis and pattern of qi stagnation with blood stasis.

The integration of disease diagnosis and TCM Zheng classification is a common model in clinical practice, and many studies have focused on this integration. According to Zheng-Zheng association analysis in Section 3. Details of the top 20 frequent disease-Zheng Zheng in a specific disease are provided in Table 2.

  • Sean T Bradley;
  • Evidence-Based Complementary and Alternative Medicine.
  • Ein Austauschjahr: Vom Zauber des Dazugehörens (German Edition).

In the pattern distribution, the patterns with yin deficiency were the most frequent 1,; To further confirm the disease-Zheng associations, 20 disease-Zheng were selected for more comprehensive analyses. Figure 6 reveals insights into the disease-Zheng association; it was built by analyzing 5 kinds of popular diseases.

  • Murder with a French Accent: An Alex Kertész Mystery (Alex Kertesz Mysteries Book 2)!
  • ;
  • The quest for modernisation of traditional Chinese medicine.
  • Why I Buy Real Estate Instead of Golfing: A memoir of what Not to Do in real estate!

The constructed view shows three attributes. The second upper right attributes represent the 6 most influential Zheng in gastritis research. The third attribute represents the total number of shared Zheng among diabetes mellitus DM , hepatocirrhosis, and HF. Compared to a previous literature review [ 10 , 11 ], we report a new quantitative route for the synthesis of related literature and provide new quantitative evidence on TCM Zheng studies. A central problem is how to capture information from literature in a form that is suitable for analysis [ 12 ]. We address the information on Zheng and show that the frequencies of words in abstracts can be used to determine whether or not a given article discusses Zheng.

For those articles that have been determined to discuss this topic, relevant information can be obtained. Furthermore, suitable annotations can be obtained. These evaluations are based on limited but increasing evidence from animal studies and clinical studies. Among other limitations, the lack of quantitative assessment has consistently been cited as a fundamental problem in existing studies, and mining exploration has been used in a recent review [ 1 ].

The purpose of this study was to provide a comprehensive overview of quantitative levels. Over the past 30 years, an increasing number of Chinese researchers have focused their attention on developing evidence for Zheng and identifying the mechanism of Zheng. Recently, more studies were published in SCI indexed journals to introduce and evaluate the effectiveness of Zheng. However, relatively low numbers were reported for animal studies and RCTs. It is difficult to develop an animal model that perfectly reproduces the symptoms of Zheng in patients [ 13 ].

Researchers attempt to overcome this limitation by combining the disease and Zheng [ 14 ]. Chinese authors are becoming more aggressive about submitting animal experimental studies for Zheng. However, it is important to note that many Chinese scientists in international institutes bring innovation to worldwide TCM Zheng research. We believe that there is a growing trend of collaboration in combining a disease and Zheng between TCM researchers and biomedical scientists in animal experimental Zheng studies.

RCTs were not developed until the s. Recently, more advanced trial designs are being developed and will provide explicit Zheng theories based on long-term experience [ 15 , 16 ]. Nonetheless, there is a relatively small amount of evidence regarding RCTs with disease and Zheng designs for data mining.

The yin deficiency pattern is currently the preferred pattern for Zheng research compared to any other pattern because it is relatively major component of modern life. A yin deficiency may be due to excessive fluid loss or to the consumption of yin due to aging. If, due to overthinking, anxiety and worry, underexercise, faulty diet or erroneous medical treatments in modern life, the qi is damaged and becomes vacuous and weak, then the spleen will not be able to perform its various functions. As mentioned above, if yin does not nourish and enrich the liver and kidney, then the liver and kidney will not be able to governing coursing and discharging.

Hence, the liver and kidney will become depressed. Thus, it is clear that liver and kidney deficiencies are mutually engendering in the mining results. For the yin deficiency pattern, more research is needed to investigate its contribution to preventing and reversing chronic diseases that are consequences of a modern lifestyle. Similarly, damp heat typically complicates the diseases of many patients. In addition, dampness can be engendered internally, often due to spicy foods, alcohol, sugars, and sweets. Blood stasis is also a mechanism that is involved in most chronic disorders, especially when there is chronic severe pain at fixed locations.

In addition, there is less information available on the yang deficiency pattern compared to the yin deficiency pattern. For disease and Zheng correlation research, the results of all selected studies showed that the number of DM studies was the highest, followed by the number of studies on gastritis and HF. The 5 most common diseases in the mining results are chronic diseases. These chronic diseases are a likely explanation for the report that the yin deficiency pattern is substantially higher than the yang deficiency pattern in Zheng studies, and CM is able to provide a worldwide contribution for patients who suffer from chronic diseases [ 19 ].

Similar to DM, more detailed patterns of gastritis were generally consistent with patterns found in clinical practice. However, relatively few mean concentrations for some of patterns were reported for primary hypertension, cirrhosis, and HF. The results of this study suggest that DM and two diseases, cirrhosis and HF, share one common Zheng. For disease and Zheng correlation research such as TDDST, explorations of the existing biomedical networks between diseases are challenging.