Two-phase biomedical NE recognition based on SVMs

Abstract

Using SVMs for named entity recognition, we are often confronted with the multi-class problem. Larger as the number of classes is, more severe the multiclass problem is. Especially, one-vs-rest method is apt to drop the performance by generating severe unbalanced class distribution. In this study, to tackle the problem, we take a two-phase named entity recognition method based on SVMs and dictionary; at the first phase, we try to identify each entity by a SVM classifier and post-process the identified entities by a simple dictionary look-up; at the second phase, we try to classify the semantic class of the identified entity by SVMs. By dividing the task into two subtasks, i.e. the entity identification and the semantic classification, the unbalanced class distribution problem can be alleviated. Furthermore, we can select the features relevant to each task and take an alternative classification method according to the task. The experimental results on the GENIA corpus show that the proposed method is effective not only in the reduction of training cost but also in performance improvement: the identification performance is about 79.9(F β = 1), the semantic classification accuracy is about 66.5(F β = 1).

Various Features with Integrated Strategies for Protein Name Classification

“…In-context features dictionary is similar to the work of Lee et al [6]. The most right 3 words from name in the training data are collected as candidates.…”

“…Nowadays, Named Entity Recognition (NER) is proved to be fundamental in information extraction and understanding in biomedical domain. Based on the method, the NER system can be roughly split into three categorizes: rule-based methods [1][2], dictionary-based methods [3], and statistical-based methods [4][5][6][7], although there are also combination of dictionary-based and rule-based method [8].…”

PubMiner: Machine Learning-Based Text Mining System for Biomedical Information Mining

“…General medical term was trained with UMLS meta-thesaurus [12] and the biological entity and its interaction was trained with GENIA [13] corpus. The underlying NLP approaches for named entity recognition are based on the system of Hwang et al [14] and Lee et al [15] with collaborations. More detailed descriptions of language processing are elucidated in [16].…”

Named Entity Recognition in Greek Texts With an Ensemble of SVMS and Active Learning

“…); and in machine translation or cross-language information retrieval, special transliteration processes can be applied to entity names across languages with different alphabets, 4 provided that the names have been identified. Although ners that employ mostly hand-crafted rules 5,6 may perform very well, ners that use statistical and machine learning techniques, including Hidden Markov or Maximum Entropy Models, 7,8,9,10 decision tree learning and/or boosting, 11,12,13 and Support Vector Machines, 14,15 usually outperform them and they are easier to port to new text genres (e.g., biomedical, instead of news articles), where new name categories (e.g., protein names) may also need to be supported. However, supervised statistical and machine learning-based ners still require a tedious manual annotation phase, during which humans must tag occurrences of entity names in a training corpus.…”

“…Our two passes are also different from the approach whereby a first phase identifies all entity names and a second one categorizes them. 15 Furthermore, unlike the system of Shen et al, our ensemble acts as a single classifier with non-overlapping categories. In active learning, we select training examples for each pass by considering the distances from the hyperplanes of both svms of that pass, much as in Vlachos.…”