USC College Department of Biology

Faculty

Biological Sciences has more than 60 full-time faculty members, as well as more than 20 faculty with joint appointments and 15 visiting or adjunct professors. Among its many distinctions and honors, the Ph.D. program faculty includes two members of the National Academy of Sciences, four members of the American Academy of Arts and Sciences, and 18 holders of endowed chairs and professorships.

Alexander Burns

Assistant Professor (research) of Biological Sciences

Contact Information
E-mail: gully@usc.edu
Phone: (213) 821-1837
Office: HNB 428

LINKS
Curriculum Vitae
Personal Website

Education

Ph.D. Neuroinformatics, Oxford University, 7/1997

Description of Research

Summary Statement of Research Interests

My research is concerned with Biomedical Knowledge Engineering, that is (A) understanding the best computational representations for biomedical knowledge (B) devising methods to extract that knowledge from large scale literature collections and (C) building working informatics systems that can be used by biologists and doctors.

Research Keywords

Biomedical Knowledge Engineering, Neuroinformatics, Natural Language Processing, Biomedical Informatics, Computer Science

Research Specialties

Natural Language Processing - Information Extraction - Knowledge Engineering - Neuroanatomy - Neuroscience - Biology

Detailed Statement of Research Interests

Biomedicine is a complex discipline with a great many sub-domains. Workers in different fields do not necessarily use the same theoretical structures, the same techniques or even the same terminology. There is a lack of ‘unifying theories’ within biological subjects. The discipline of biomedical informatics is concerned with the information infrastructure of biomedicine and so is concerned with standardizing concepts. Importantly, the logic used to plan experiments, theorize about the subject and understand scientific phenomena is based mainly on qualitative argumentation. Thus, the task of making the conceptual links between studies can only be performed by a human agent. Top-level biologists possess an encyclopedic knowledge of the literature and it can literally take a lifetime to master the subject.

From an informatics perspective, this is unsatisfactory. It should not be necessary to have to read through dozens of papers to answer a single straightforward question. It is necessary to delve into all the experimental minutiae of a specific phenomenon in order to say that a reported fact is indeed true or false. But, having done so once, a researcher should not have to repeat the same process after a few months have past and they have forgotten the intricate details of the paper in question.

My research interests address several key aspects of this general situation

1. Knowledge representations for biology based on experimental design. A formalized representation of experimental information would make the subject more transparent and permit increased communication between workers in different fields. Importantly, it would permit researchers to scale up the number of facts they incorporate in their thinking so that they can properly address the complex problems mandated by the subject. Most existing knowledge-based solutions to biomedical challenges do not use basic principles of experimental design (such as independent and dependent variables) used by scientists to formulate their internal representations of their knowledge. We seek to model these principles directly within our solutions to provide scientists with intelligent systems that they can immediately relate to and understand.

2. Practical intuitive tool construction. Unfortunately, researchers rarely perceive that anything might be wrong with the current paradigm. They are able to understand enough of their subject to perform experiments, obtain answers, write grants and push the subject forward. Importantly, they are already overworked and have only a limited amount of energy to invest into learning new approaches for their data. I am concerned with providing knowledge management tools that make biological and clinical researchers’ work easier and therefore introduce them implicitly to formal knowledge representations.

3. High throughput knowledge acquisition using Natural Language Processing. Another interesting (and timely) source of these semantic representations could be made from full-text articles in digital libraries as more textual data is coming online. Naturally, expertise in documents handling, information extraction and natural language processing is required to address these questions. Our preliminary research in this area involves the construction of a private corpus of textual data from the Journal of Comparative Neurology from 35 years of published articles. We have also obtained a license with Elsevier to develop systems that can mine the text of their ScienceDirect digital library.

Publications

Book Chapter

Burns, G. A., Feng, D., Hovy, E. H. (2007). Intelligent Approaches to Mining the Primary Research Literature: Techniques, Systems, and Examples. Springer.

Conference Proceeding

Burns, G. A., Herr, B., Newman, D., Ingulfsen, T., Pantel, P., Smyth, P. (2007). A snapshot of neuroscience: unsupervised natural language processing of abstracts from the Society for Neuroscience 2006 annual meeting. ". Society for Neuroscience Annual Meeting.
Feng, D., Burns, G. A., Hovy, E. (2007). Extracting Data Records from Unstructured Biomedical Full Text. The Joint Meeting of Conference on Empirical Methods in Natural Language Processing and Conference on Computational Natural Language Learning (EMNLP-CoNLL 2007).
Burns, G. A., Feng, D., Ingulfsen, T., Hovy, E. (2007). Infrastructure for Annotation-Driven Information Extraction from the Primary Scientific Literature: Principles and Practice. 1st IEEE International Workshop on Service Oriented Technologies for Biological Databases and Tools (SOBDAT 2007).

Journal Article

Burns, G. A., Cheng, W., Thompson, R. F., Swanson, L. W. (2006). The NeuARt II system: a viewing tool for neuroanatomical data based on published neuroanatomical atlases. BMC Bioinformatics. Vol. n/a
Khan, A. M., Cheng, W., Watts, A., Burns, G. A. (2006). NeuroScholar's Electronic Laboratory Notebook and its Application to Neuroendocrinology. Neuroinformatics/Humana. Vol. 4(2), pp. p. 139-160.
Burns, G. A., Cheng, W. (2006). Tools for Knowledge Acquisition within the NeuroScholar system and their application to anatomical tract-tracing data. Journal of Biomedical Discovery and Collaboration. Vol. 1(1), pp. p. 10..
Goto, M., Canteras, N., Burns, G. A., Swanson, L. W. (2005). Projections from the subfornical region of the lateral hypothalamic area. Journal of Comparative Neurology/Wiley. Vol. 493(3), pp. 412-38.

Proceedings

Burns, G. A., Khan, A. M. (2005). A computational knowledge representation for physiological experiments with direct links to the primary literature and raw data. Histochemical studies of stress-activated paraventricular hypothalamic neuroendocrine neurons: a neuroinformatics-based digital lab notebook to relate the primary literature to raw, unpublished data.
Burns, G. A. (2005). A 'Stress Management' System: Computational knowledge representation of experiments that delineate neural circuits and histochemical expression patterns based on the primary literature and raw data. A 'Stress Management' System: Computational knowledge representation of experiments that delineate neural circuits and histochemical expression patterns based on the primary literature and raw data.
Burns, G. A. (2005). Extracting and managing model parameters from the literature. World Assocation of Modellers, Biologically Accurate Modeling.
Cheng, W., Burns, G. A. (2005). Tools for knowledge representation, acquisition and retrieval of neuroanatomical data mapped onto a standard atlas. Society for Neuroscience.