Six studies on Persian complex predicates, Eastern Armenian causatives, and predicate composition, bringing together my previously published and circulated work with a new introduction.

Presents a data-driven approach to Persian pedagogy combining linguistic theory, acquisition research, computational methods, and AI. Uses compound verbs to illustrate how authentic materials and language patterns can support student-centered classroom learning.

Explores the use of prompted large language models to extract narrative events and participants from Persian and Sorani Kurdish texts. Finds promise in recovering implicit information, alongside inconsistencies and difficulties with complex tasks such as coreference resolution.

Introduces TaarofBench, a culturally grounded benchmark for evaluating Persian taarof in large language models. Role-play evaluations reveal gaps in cultural understanding, while experiments examine the effects of prompt language, gender, fine-tuning, and preference optimization.

Surveys responsible AI approaches, including organizational practices and methods for addressing bias, transparency, and trustworthy system development. Discusses how differences among AI methodologies and the lack of shared standards complicate implementation and evaluation.

Proposes a computational architecture for narrative-driven legal case elicitation and introduces six legal narrative corpora. Compares language models with event sequence alignment for building narrative schemas, finding the latter promising for small, sparse, heterogeneous datasets.

Describes ANAnSI, a distributed narrative analysis system combining linguistic resources, probabilistic parsing, knowledge-based reasoning, and machine learning. Extracts event structures and domain-specific information from large text collections for applications in health, news, and law enforcement.

Describes an NLP system for extracting substance use, co-occurring conditions, and treatment histories from probation records. PDF extraction, linguistic analysis, and temporal reasoning convert large collections of narrative documents into structured information and timelines.

Describes a global science and technology knowledge graph designed to map research activity, identify emerging topics and leading researchers and institutions, and reveal opportunities for collaboration across an increasingly interconnected research landscape.

Identifies and corrects annotation problems in the TimeBank temporal corpus, including guideline violations, inconsistent relations, disconnected graphs, and missing events and time expressions. Provides checking methods, code, and patches to support more reliable temporal analysis.

Presents a proof of concept for extracting case information from narrative records and automatically populating a risk assessment survey for online child pornography offenders. The work evaluates information extraction accuracy and its potential to reduce manual review.

Reviews differences in the linguistic competence of Persian heritage and second-language speakers. Examines phonology, morphology, and syntax, highlighting uneven strengths, difficulties with less salient forms, and the influence of the dominant language.