Request Expert

He is an expert in the use of computers to identify document authorship by linguistic analysis. This includes such tasks as determining if a document has been plagiarized, identifying whether a given individual was the author of an anonymous document (authorship attribution), and identifying characteristics of a document's author (their sex, age, likely educational background, personality, etc.). The methods he has developed are applicable to civil investigations in areas such as copyright and intellectual property, as well as for criminal investigations where author authentication or attribution is needed to find or incriminate a suspect.

Expert's published research includes several influential results in this area. His work on determining author gender from writing style was the first to show clear differences between male and female writing in both formal and informal genres, and across multiple languages and time-periods. He has shown how certain aspects of personality are reliably indicated by certain patterns of word usage, obtaining more accurate results than competing approaches. A primary contribution of his research on stylistic text analysis, is a novel methodology for constructing meaningful features for authorship analysis (based on functional/semantic principles) which give better results than previous methods in many cases.

His research has also been instrumental in working to solidify the methodological bases of the fields of automated authorship attribution and computational linguistics, which heretofore were plagued by methodological inconsistencies and low standards of reproducibility. He was the first to give a clear theoretical explanation of the popular "Delta" method for authorship attribution, thus enabling better analysis of its reliability in different circumstances. He was the principal investigator of an NSF-funded project to solidify the methodological bases of the field of automated authorship attribution, and is currently co-PI of the NSF-funded followup effort to create a standardized corpus for research in the area.

Expert has over 20 years of research experience in diverse areas of artificial intelligence and machine learning. His published research includes work on active learning, ensemble learning, learning with prior knowledge, robotic navigation, visual place recognition, shallow parsing, sentiment analysis, computational stylistics, and authorship attribution.

Expert has published numerous papers in many areas of natural language processing and computational linguistics, including shallow parsing, sentiment analysis, computational stylistics, genre analysis, text classification, part-of-speech tagging, information extraction, information retrieval, morphological analysis and rhetorical analysis. He has consulted to industry on projects in legal document processing, information retrieval, and business intelligence.

Expert consulted for AOL Research (2006), providing new algorithms and technology for development of a prototype system to predict emerging trends in web search. The system could predict in advance many of the most popular web searches for the following day, as well as give meaningful explanations for why people suddenly became interested in them.
Expert consulted for Radcom, Ltd. (1998-99) in Tel-Aviv, Israel, helping develop novel proprietary technology more more efficient analysis of Internet protocol traffic.
From 1995 to 1998, Expert had a long-term consulting arrangement with Expert-Ease Development, Inc., now part of Thomson Reuters. His primary contribution was the development of the patented text mining algorithms that form the core of the company's flagship product, Deal Proof (available from Thomson West). He also helped with the design and development of many other aspects of that product. Expert consulted for Compassware Development, Inc. (1995-1997), an information retrieval software company in Jerusalem, Israel. He developed new methods for effective information retrieval, document clustering, and text mining, as part of the company's InfoMagnet enterprise content management system.

Research interests:
Natural language processing and computational linguistics, particularly metaphor
analysis, systemic functional linguistics, shallow parsing, and discourse structure.
Computational stylistics, including text categorization by writing style, gender issues in
linguistic expression, forensic linguistics, and sociolinguistics, with applications to
humanities scholarship, cybersecurity, and counterterrorism.
Data mining and machine learning from very large textual corpora, including intelligent
text retrieval, categorization, summarization, document search, and pattern
recognition.
Scientic methodology, epistemology, and ethics; the relationship between the logical,
statistical, cognitive, and social structures in science and reproducibility and
reliability of scientic research and societal implications.
Digital humanities, particularly application of machine learning and natural language
processing to literary textual analysis and visualization.

Expert has been consulting since 1995.
Forensic linguistics:
2018: Corporate civil arbitration authorship analysis case. Case ongoing.
2018: Public Defender Service for District of Columbia.
Murder case, involving validation of authorship analysis of short online texts. Provided evaluation and validation of text analytic methodologies.
2014: Patent infringement case dealing with network security technology, specically
scanning for byte signatures in a data stream. Case dismissed before nal report
completed.
2012: Evaluation of whether a company's e-discovery procedures comply with discovery
requirements. Case settled before nal report completed.
2011: Analysis of the authorship of text messages and emails as evidence of conspiracy
and murder. Exploratory analysis performed, but no reports or testimony required.
2010: Authorship analysis of text messages containing death threats. Preliminary report
submitted; full report and testimony not required.