Search

Automated analysis of full-text life science research articles and technical documents is becoming increasingly important. In contrast to abstracts, accessing and processing full-text is considerably more complex. GetItFull is a tool for downloading and pre-processing full-text journal articles. GetItFull automatically connects to a journal's Web site, downloads the journal content and performs various commonly used pre-processing steps. The output comprises a structured XML document for each article with tags identifying the various sections and journal information. The output may then be used as the basis for text mining applications or exported to a database for further processing. [ABSTRACT FROM AUTHOR]

Because of the importance of proteins in life sciences, biologists have put great effort to elucidate their structures, functions and expression profiles to help us understand their roles in living cells in the past few decades. Currently, protein databases are widely used by biologists. Hence it is critical that the information that researcher work with should be as accurate as possible. However, the sizes of these databases are increasing rapidly, and existing protein databases are already known to contain annotation errors. In this paper, we investigate the reason why protein databases possess mis-annotated sequence data. Then, by using some statistical approaches, we derive a method to automatically filter and assess the reliability of the data from databases. This is important to provide accurate information to researchers and will help reduce further errors in annotation resulting from existed mis-annotated sequence data. Our initial experiments proved our theoretical findings, and show that our methods can effectively detect the mis-annotated sequence data. [ABSTRACT FROM AUTHOR]

In text mining, to calculate precise keyword frequency distributions in a particular document collection, we need to map different keywords that denote the same entity to a canonical form. In the life science domain, we can construct a large dictionary that contains the canonical forms and their variants based on the information from external resources and use this dictionary for the term aggregation. However, in this automatically generated dictionary, there are many invalid entries that have negative effects on the calculations of keyword frequencies. In this paper, we propose and test methods to detect invalid entries in the dictionary. [ABSTRACT FROM AUTHOR]

Most existing text classification work assumes that training data are completely labeled. In real life, some information retrieval problems can only be described as learning a binary classifier from a set of incompletely labeled examples, where a small set of labeled positive examples and a very large set of unlabeled ones are provided. In this case, all of the traditional text classification methods can't work properly. In this paper, we propose a method called Weighted Voting Classifier, which is an improved 1-DNF algorithm. Experimental results on the Reuters-21578 set show that our algorithm Weighting Voting Classifier outperforms PEBL and one-class SVM in terms of F measure. Weighting Voting Classifier can achieve high F score when comparing with PEBL and one-class SVM. Furthermore, the reduction of iterations is 2.26 when comparing the method of PEBL with ours. [ABSTRACT FROM AUTHOR]

Identifying proper names, like gene names, DNAs, or proteins is useful to help researchers to mining the text information. Learning to extract proper names in natural language text is a named entity recognition (NER) task. Previous studies focus on combining abundant human made rules, trigger words, to enhance the system performance. However these methods require domain experts to build up these rules and word set which relies on lots of human efforts. In this paper, we present a robust named entity recognition system based on support vector machines (SVM). By integrating with rich feature set and the proposed mask method, the system performance is satisfactory on the MUC-7 and biology named entity recognition tasks which outperforms famous machine learning-based method, such as hidden markov model (HMM), and maximum entropy model (MEM). We compare our method to previous systems that were performed on the same data set. The experiments show that when training with the MUC-7 data set, our system achieves 86.4 in F(β = 1) rate and 81.57 for the biology corpus. Besides, our named entity system is able to handle real time processing applications, the turn around time on a 63 K words document set is less than 30 seconds. [ABSTRACT FROM AUTHOR]

We are developing an information extraction system for life science literature. We are currently focusing on PubMed abstracts and trying to extract named entities and their relationships, especially protein names and protein-protein interactions. We are adopting methods including natural language processing, machine learning, and text processing. But we are not developing a new tagging or parsing technique. Developing a new tagger or a new parser specialized in life science literature is a very complex job. And it is not easy to get a good result by tuning an existing parser or by training it without a sufficient corpus. These all are another research topics and we are trying to extract information, not to develop something to help the extracting job or else. In this paper, we introduce our method to use an existing full parser without training or tuning. After tagging sentences and extracting proteins, we make sentences simple by substituting some words like named entities, nouns into one word. Then parsing errors are reduced and parsing precision is increased by this sentence simplification. We parse the simplified sentences syntactically with an existing syntactic parser and extract protein-protein interactions from its results. We show the effects of sentence simplification and syntactic parsing. [ABSTRACT FROM AUTHOR]

We represent research themes by the temporal topic profiles based on MeSH terms from MEDLINE citations. By comparing the differences of the temporal profiles for the same topic at the two different periods, we find that the temporal profiles for a topic at the new period may result from three kinds of concepts replacements of the temporal profiles at the old period, namely broad replacement, parallel replacement and narrow replacement. Such findings provide new ways to generate potential relationships from the current biomedical literature. [ABSTRACT FROM AUTHOR]

In this paper, Support Vector Machines (SVMs) are applied to the identification and automatic annotation of biomedical named entities in the domain of molecular biology, as an extension of the traditional named entity recognition task to special domains. The effect of the use of well-known features such as word formation patterns, lexical, morphological, and surface words on recognition performance is investigated. Experiments have been conducted using the train and test data made public at the Bio-Entity Recognition Task at JNLPBA 2004. An F-score of 69.87% was obtained by using a carefully selected combination of a minimal set of features, which can be easily computed from training data without any use of post-processing or external resources. [ABSTRACT FROM AUTHOR]

As genomic research advances, the knowledge discovery from a large collection of scientific papers becomes more important for efficient biological and biomedical research. Even though current databases continue to update new protein-protein interactions, valuable information still remains in biomedical literature. Thus data mining techniques are required to extract the information. In this paper, we present a tree kernel-based method to mine protein-protein interactions from biomedical literature. The tree kernel is designed to consider grammatical structures for given sentences. A support vector machine classifier is combined with the tree kernel and trained on predefined interaction corpus and set of interaction patterns. Experimental results show that the proposed method gives promising results by utilizing the structure patterns. [ABSTRACT FROM AUTHOR]

Natural Language Processing (NLP) techniques have been used for the task of extracting and mining knowledge from biomedical literature. One of the critical steps of such a task is biomedical named entity tagging (BNER) which usually contains two steps: the first step is the identification of biomedical names in text and the second is the assignment of semantic classes predefined to names identified by the first step. Headwords and suffixes have been used frequently by BNER systems as features for the assignment of semantic classes to names in text. However, there are few studies to evaluate the performance of headwords and suffixes in predicting semantic classes of biomedical terms utilizing knowledge sources in an unsupervised way. We conducted a study to evaluate the performance of headwords and suffixes using names in the Unified Medical Language System (UMLS) where the semantic classes associated with these names were obtained by modifying an existing UMLS semantic group system and incorporating the GENIA ontology. We define headwords and suffixes that are significantly associated with a specific semantic class as semantic suffixes. The performance of semantic assignment using semantic suffixes achieved an F-measure of 86.4% with a precision of 91.6% and a recall of 81.7%. When applying these semantic suffixes obtained using the UMLS to names extracted from the GENIA corpus, the system achieved an F-measure of 73.4% with a precision of 84.2% and a recall of 65.1% where these performance measures could be improved dramatically when limited to names associated with classes that have the corresponding GENIA types. [ABSTRACT FROM AUTHOR]

In this paper we present a picture search engine for life science literature and show how it can be used to improve literature preselection. This preselection is needed as a way to compensate for the vast amounts of literature that are available. While searching for DNA binding sites for example, we wanted to add the results of specific experiments (DNAse I footprint and EMSA) to our database. The preselection via abstract search was very unspecific (150 000 hits), but by looking for paper with images concerning the experiments, we could improve precision immensely. They are displayed like hits in a search engine, allowing easy and quick quality assessment without having to read through the whole paper. The images are found by their annotation in the paper: the figure caption. To identify that, we analyse the layout of the paper: the position of the image and the surrounding text. [ABSTRACT FROM AUTHOR]

In recent years many biomedical ontologies have been developed and many of these ontologies contain overlapping information. To be able to use multiple ontologies they have to be aligned. In this paper we propose strategies for aligning ontologies based on life science literature. We propose a basic algorithm as well as extensions that take the structure of the ontologies into account. We evaluate the strategies and compare them with strategies implemented in the alignment system SAMBO. We also evaluate the combination of the proposed strategies and the SAMBO strategies. [ABSTRACT FROM AUTHOR]

We demonstrate the application of a grid infrastructure for conducting text mining over distributed data and computational resources. The approach is based on using LexiQuest Mine, a text mining workbench, in a grid computing environment. We describe our architecture and approach and provide an illustrative example of mining full-text journal articles to create a knowledge base of gene relations. The number of patterns found increased from 0.74 per full-text articles from a corpus of 1000 articles to 0.83 when the corpus contained 5000 articles. However, it was also shown that mining a corpus of 5000 full-text articles took 26 hours on a single computer, whilst the process was completed in less than 2.5 hours on a grid comprising of 20 computers. Thus whilst increasing the size of the corpus improved the efficiency of the text-mining process, a grid infrastructure was required to complete the task in a timely manner. [ABSTRACT FROM AUTHOR]

The challenge of knowledge management in the pharmaceutical industry is twofold. First it has to address the integration of sequence data with the vast and growing body of data from functional analysis of genes with the information in huge historical archival databases. Second, as the number of biomedical publications exponentially increases (Medline now contains more than 13 million records), researchers require assistance in order to broaden their vision and comprehension of scientific domains. Analogous to data mining in the sense that it uncovers relationships in information, text mining uncovers relationships in a text collection and leverages the creativity of the knowledge worker in the exploration of these relationships and in the discovery of new knowledge. We describe herein a text mining method to automatically detect protein interactions which are described across a large amount of scientific publications. This method relies on natural language processing to identify protein names, their synonyms and the various interactions they can bear with other proteins. We have then compared text mining analysis on abstracts to the same kind of analysis on full text articles to assess how much information is lost when only abstracts are processed. Our results show that: 1)LexiQuest Mine is a very versatile and accurate tool when mining biomedical literature to analyze interactions between proteins. 2)Mining only abstracts can be sufficient and time saving for applications that do not require a high level of detail on a large scale whereas mining full text articles is to be chosen for more exhaustive applications designed to address a specific issue. Availability: LexiQuest Mine is available for commercial licensing from SPSS, Inc. [ABSTRACT FROM AUTHOR]

Automated extraction of information from biological literature promises to play an increasingly important role in text-based knowledge discovery processes. This is particularly true in regards to high throughput approaches such as microarrays and combining data from different sources in a systems biology approach. We have developed an integrated system that combines protein/gene name dictionaries, synonymy dictionaries, natural language processing, and pattern matching rules to extract and organize gene relationships from full text articles. In the first phase full text articles were collected from 20 peer-reviewed journals in the field of molecular biology and biomedicine over the last 5 years (1999-2003). The extracted relationships were organized in a database that included the unique PubMed ID and section id (abstract, introduction, materials and method, and results and discussion) to identify the source article and section from which concepts were extracted. The system architecture, its uniqueness and advantages are presented in this paper. It is hoped that the resulting knowledge base will assist in the understanding of gene lists generated from microarray experiments. [ABSTRACT FROM AUTHOR]