WORK EXPERIENCE:

Senior Research Scientist & Senior Systems Designer   7/01 - present
CLAIRVOYANCE Corporation/JSERI, Pittsburgh PA.

• Text Classification and Machine Learning
  • Project and technical lead. Developed an adaptive filtering system in Java integrating several types of filters (IR and SVM-based). Designed and implemented a GUI application using the toolkit as an adaptive news/email filtering system (Java/Swing).
  • Designed and implemented a Java class library supporting SVM-based classification (Adatron and SMO variants; support for pluggable non-linear kernels; IO infrastructure etc.). This toolkit was integrated into a product sold in Japan as a basis of its text filtering functionality.
  • Conducted research in (as well as designed and implemented toolkit components for) using tuned thresholds for SVM-generated scores as applied to text filtering.
  • Conducted research on kernel compression and efficiency of generation of kernels for SVMs by restricting (thresholding) number of entries in the kernel matrix. Designed and implemented a toolkit supporting kernel thresholding.
  • Developed, maintained and expanded (for instance, to expose functionality to Java through JNI) various elements of a large C++ class library for text retrieval and filtering. Maintained portability of that library to Linux.
• Text Search
  • Designed and implemented a desktop Java application (including a Swing-based GUI) for query/topic visualization and analysis.
  • Conducted research on the problem of searching XML document repositories, especially in the context of utilization of unstructured text content present in these documents. (Leveraging IR and text classification approaches.)
  • Designed and implemented browser-based interface to support research on search results presentation and organization. Conducted related research.
  • Analyzed an RDB-based commercial grade retrieval engine for disk footprint optimization deficiencies. Proposed and implemented a solution by inserting a compression layer in the RDB access code.
• Social Networks
  • Designed and implemented algorithms for graph analysis (related to causal modeling). Designed and implemented components of a toolkit supporting research in this area. Conducted the research.
  • Designed and developed a prototype interface (in Java) to represent (as a directed graph), and allow research into, the temporal evolution of a social network of email users defined by their email communications.

• Designed and developed elements of a distributed (CORBA, both server and client) network- and host-intrusion detection and response system written in Java and C++.
• Participated in architecture design for statistical and machine learning based anomaly detection in network traffic and host logging data.

Research Scientist     11/97 - 3/01
CLARITECH Corporation, Pittsburgh PA.

• Project and technical lead. Lead a team designing and developing a new module of a large C++ class library for text filtering using a graph of filters.
• Created and maintained a port of a large C++ class library to Linux.
• Conducted research in applications of machine learning methods to routing, filtering and classification of text documents based on unstructured text analysis.
• Developed and implemented a score normalization technique for building multiple classifiers out of ranking engines.
• Evaluated effectiveness of in-house summarization algorithms for news reports (general and domain-specific), e-mail and other types of messages.
• Designed and implemented C++ applications for running retrieval and text classification experiments (to evaluate algorithm performance) and prototyping, including COM server-based GUIs.

• Administered the departmental LAN jointly with the IS department.
• Set up and administered the departmental (experimental) web server (Apache on Linux).
• Designed and implemented several relational databases using Microsoft Access and deployed them on the LAN.
• Developed Perl script library driving a Unix-based custom database engine with a boolean query interface.
• Designed and implemented reports in FOCUS to access IBM mainframe-based database.

• Managed a Unix/PC LAN including user support.
• Administered e-mail server (sendmail) for a University department; administered a primary DNS server.
• Administered systems running SunOS, Solaris, AIX, Ultrix, and Linux.
• Monitored the LAN for security breaches. Designed solutions to recurring sniffer attacks.
• Wrote a booklet introducing new users to the local Unix network environment.

Programming Languages/Technologies: C++, Java, C, XML, Perl, client-server/tiered (various types), multi-threading/concurrency.

Operating Systems Administration: Linux, Solaris, SunOS, AIX, FreeBSD, Digital Unix, Ultrix, IRIX, Windows, Mac OS.

EDUCATION:

Doctor of Philosophy      8/98
Department of Physics and Astronomy, University of Pittsburgh.

Master of Science      8/11/91
Department of Physics and Astronomy, University of Pittsburgh.

Undergraduate Education      6/30/89
Major: Physics.
University of Gdańsk (Gdańsk, Poland)
Honors: Ministry of Education Academic Achievement Stipend.

PATENTS:

1. Patent: Roma, Norbert "Method and Apparatus for Comparing Scores in a Vector Spaces Retrieval Process", U.S. Patent No. 7,356,604

2. Patent: Shanahan, James G.; Roma, Norbert; Evans, David A. "Method and Apparatus for Adjusting the Model Threshold for a Support Vector Machine for Text Classification and Filtering", U.S. Patent No. 7,356,187

3. Patent Application: Shanahan, James G.; Roma, Norbert; Evans, David A. "Method and Apparatus for Adjusting the Model Threshold of a Support Vector Machine for Text Classification and Filtering", U.S. Patent Application No. 20050228783

4. Patent Application: Shanahan, James G.; Roma, Norbert; Evans, David A. "Method and Apparatus for Constructing a Compact Similarity Structure and for Using the Same in Analyzing Document Relevance", U.S. Patent Application No. 20070136336

5. Patent Application: Hull, David A.; Roma, Norbert "Methods and Apparatus for Identifying Workflow Graphs Using an Iterative Analysis of Empirical Data", U.S. Patent Application No. 20080065448

PUBLICATIONS:

1. Roma, Norbert; Hagiwara, Shigeki; Matsumoto, Noriyoshi "xfy Technology: Fusing XML and Java Technology to Open New Dimensions of Application and Document Creation", JavaOne 2005 Conference, Technical Sessions, Web Tier,TS-3094 (link)

2. Shanahan, James G.; Roma, Norbert "Boosting Support Vector Machines for Text Classification through Parameter-Free Threshold Relaxation", Conference on Information and Knowledge Management, CIKM-2003

3. Shanahan, James G.; Roma, Norbert "Improving SVM Text Classification Performance through Threshold Adjustment", European Conference on Machine Learning, ECML-2003

4. Evans, David A.; Shanahan, James; Roma, Norbert; Bennett, Jeffrey; Sheftel, Victor; Stoica, Emilia; Montgomery, Jesse; Hull, David A.; Tembe, Waibhav "CLARIT Experiments in Batch Filtering: Term Selection and Threshold Optimization in IR and SVM Filters." In E.M. Voorhees and Lori P. Buckland (Editors), The Eleventh Text Retrieval Conference, TREC-2002. NIST Special Publication 500-251. Washington, DC: U.S. Government Printing Office, 2003, 166-175.

5. Evans, David A.; Shanahan, James; Tong, Xiang; Roma, Norbert; Stoica, Emilia; Sheftel, Victor; Montgomery, Jesse; Bennett, Jeffrey; Fujita, Sumio; Grefenstette, Gregory "Topic-Specific Optimization and Structuring." In E.M. Voorhees and D.K. Harman (Editors), The Tenth Text Retrieval Conference (TREC-2001). NIST Special Publication 500-250 Washington DC: U.S. Government Printing Office, 2002, 123-141.

6. Zhai, Chengxiang; Jansen, Peter; Roma, Norbert; Stoica, Emilia; Evans, David A. "Optimization in CLARIT TREC-8 Adaptive Filtering", The Eighth Text Retrieval Conference (TREC-8), NIST SP 500-246, p. 253-258 (2000).

7. Zhai, Chengxiang; Jansen, Peter; Stoica, Emilia; Grot, Norbert; Evans, David A. "Threshold Calibration in CLARIT Adaptive Filtering", The Seventh Text Retrieval Conference (TREC-7), NIST SP 500-242, p. 149 (1999).

8. Grot, Norbert; Rovelli, Carlo "Weave states in Loop Quantum Gravity", General Relativity and Gravitation, Vol. 29, No 8, p. 1039 (1997).

9. Grot, Norbert; Rovelli, Carlo; Tate, Ranjeet "Time of arrival in quantum mechanics", Physical Review A, Vol. 54, No 6, p. 4676 (1996).

10. Grot, Norbert; Rovelli, Carlo "Moduli-space structure of knots with intersections", Journal of Mathematical Physics, Vol 37, p. 3014 (1996).