China 
Africa 
Explore chapters and articles related to this topic
Extraction of Medical Entities Using a Matrix-Based Pattern-Matching Method
Published in Himansu Das, Jitendra Kumar Rout, Suresh Chandra Moharana, Nilanjan Dey, Applied Intelligent Decision Making in Machine Learning, 2020
In previous works of clinical entity identification, different methods have been used such as the dictionary look-up method, and rules-based and machine learning. The dictionary look-up method used in [3], in which the authors identified clinical entities using dictionaries compiled from the corpus, performed experimentation on I2b2 2010 dataset and obtained the average F score of 48% for the Beth dataset and 50% for the Partners dataset. The rule-based method is also used in [13, 14], in which some rules are created based on corpus words and word occurrences, and words are then found in the corpus and mapped to a corresponding category and provided a 42% F score. Machine-learning-based approaches like SVM (support vector machine) and CRF (conditional random field) have been used for entity boundary identification and entity classification [1, 15], and which is based on the beginning, inner, and outside (BOI) model for sequence labeling. An unsupervised approach has also been used to extract named entities from biomedical texts [16], in which authors have developed a noun phrase chunker followed by a filter based on inverse document frequency. The classification of multiword entities is carried out by using the concept of distributional semantics.
Weakly supervised co-segmentation by neural attention
Published in Yigang He, Xue Qing, Automatic Control, Mechatronics and Industrial Engineering, 2019
Y. Zhao, F. Zhang, Z.L. Zhang, X.H. Liang
As a classical probabilistic graphical model, Fully Connected Conditional Random Field (FC-CRF) (Krahenbuhl & Koltun, 2011) considers both node priors and consistency between nodes. Since the attention map obtained in the previous subsection is vague, we apply an FC-CRF model to both enhance spatial coherence and refine boundary accuracy. Every pixel is regarded as a node, and every node is connected to every other node. The energy function is showed as follows: () E(x)=∑iθi(xi)+∑i,jθi,j(xi,xj)
Conditional random fields, syntactic parsing, and more
Published in Jun Wu, Rachel Wu, Yuxi Candice Wang, The Beauty of Mathematics in Computer Science, 2018
This feature demonstrates that it only depends on x1 but no other variables. If the values of some variables corresponding to a certain feature function are zero, then this feature function has no effect on these variables. Applying these features in the model, we obtain the following, P(x1,x2,…,xn,y1,y2,…,ym)=ef1+f1+⋯+fkZ, where k is the total number of features and Z is the normalization constant that makes P a proper probability. Using shallow parsing as an example, we can demonstrate how a conditional random field model is trained and how it works.
Structured prediction models for argumentative claim parsing from text
Published in Automatika, 2020
The tasks of argumentation mining involves the transformation of text into structured representations. These are typically solved using structured prediction, a supervised machine learning paradigm that predicts structured objects such as sequences, trees, and graphs [43]. Conditional random fields (CRF) is a very powerful class of probabilistic modelling methods used for structured prediction [20]. Whereas a classifier predicts a label for an instance independently of other instances, a CRF can account for context. CRFs, particularly linear-chain CRFs, have been widely applied in NLP. Recent approaches to structured prediction rely on deep learning models. Long short-term memory network (LSTM) [21] is a recurrent neural network architecture with feedback connections that models sequences of data. LSTM networks modelling data in both forward and backward directions (BiLSTM) are often used to solve text classification problems [22] or sequence labelling problems [23, 24]. Distributed word representations [25] are often used as input features to solve such problems [26]. A popular alternative to probabilistic and deep learning models for structured prediction is chain classification [27]. Since the ordering of classifiers may significantly impact performance, ensembling of chain classifiers is often employed [28].
A Levenshtein distance-based method for word segmentation in corpus augmentation of geoscience texts
Published in Annals of GIS, 2023
Jinqu Zhang, Lang Qian, Shu Wang, Yunqiang Zhu, Zhenji Gao, Hailong Yu, Weirong Li
For Chinese word segmentation, taking into account the relationship between adjacent labels is necessary. For instance, the latter of a B (Begin) label should be an E (End) label or an M (Middle) label, and cannot be an S (Single) label. Conditional random fields (CRFs) use a single exponential model to represent the joint probability of a whole label sequence, and therefore can effectively solve the problem of label deviation and ensure the validity of predicted labels (Haruechaiyasak, Kongyoung, and Dailey 2008). Thus, we use CRFs to model the label sequence jointly rather than independently.