My research has developed over more than three decades around a central question: how can linguistic structure, variation, and relatedness across languages be formally represented, computationally modeled, and empirically tested, and how do these relationships shape multilingual language processing? My scientific work integrates formal linguistics, computational grammar engineering, multilingual language resources, and experimental research on intercomprehension. These are successive stages of a coherent research program: from analyzing the structure of individual languages, through representing shared grammatical knowledge across related languages, to investigating how speakers exploit cross-linguistic regularities in comprehension and multilingual communication.

Foundational is my work in Head-Driven Phrase Structure Grammar (HPSG) and formal Slavic linguistics. Early research examined Bulgarian verbal and nominal clitics, word order, phrase structure, relative clauses, and the syntax of free-word-order languages. A dissertation, Word Order and Clitics in Bulgarian (1997), brought these interests together and established an agenda focused on how highly inflected languages can be described within a precise, constraint-based framework while accounting for substantial syntactic flexibility. Subsequent work addressed clitic placement and replication, argument structure, grammatical relations, valence, agreement, and diathesis (Avgustinova 1994, 1997, 1998, 2001, 2002). On such a basis, a central methodological principle emerged: formal linguistic hypotheses should be explicit enough to be implemented, compared, and empirically evaluated. My work on argument structure and valence, for example, examined how grammatical relations and diathetic alternations can be represented within a formally explicit architecture. Further investigations focused the interaction of syntax, prosody, and discourse-related properties in Slavic languages. Rather than treating Slavic-specific phenomena as collections of unrelated exceptions, these studies sought to identify general grammatical principles while preserving structural distinctions among languages. Such a perspective led naturally to a broader question: not only how can an individual language be represented, but also what can be shared across related languages and how can systematic variation be encoded within a common grammatical architecture?

My habilitation, Shared Grammatical Resources for Slavic Languages: Selected Topics in Multilingual Grammar Design (2002), and the subsequent monograph, Language Family Oriented Perspective in Multilingual Grammar Design (2007), developed this question explicitly. The actual proposal was that linguistic relatedness should itself have a principled representation in formal grammar and grammar engineering. Rather than constructing grammars for related languages independently, a multilingual architecture can distinguish a shared grammatical core from language-specific extensions. This approach was developed in collaboration with Hans Uszkoreit on shared and non-shared grammar and on an ontology of systematic relations (Avgustinova & Uszkoreit 2000, 2006). Its significance is both theoretical and methodological. Cross-linguistic generalizations become computationally explicit and therefore testable, while computational implementation reveals whether linguistic generalizations are sufficiently precise to support actual grammar development. The same principle guided my subsequent work on the parallel construction of Slavic grammatical resources. I proposed a common Slavic core combined with language-specific extensions and corpus-based elaboration (Avgustinova 2011). Collaborations within the DELPH-IN consortium resulted in resource grammars for Russian and Bulgarian, while work with Yi Zhang involved exploiting the Russian National Corpus for grammar development, constructing a Russian HPSG resource grammar, and converting dependency representations into HPSG derivations (Avgustinova & Zhang 2009, 2010). Later collaboration on CLIMB extended this approach through meta-grammar engineering and code generation, aiming to make multilingual grammar development more systematic, reusable, and scalable. Grammar engineering thus became a research methodology rather than merely an application of linguistic theory. By requiring linguistic hypotheses to be represented in executable and comparable form, it provided a means of testing claims about linguistic relatedness and identifying both shared structures and systematic differences across languages.

The question of shared grammatical structure led to a broader empirical question: if related languages share systematic properties, to what extent can speakers exploit these properties when understanding a language they have not explicitly learned? My research on receptive multilingualism and Slavic intercomprehension, particularly through the INCOMSLAV program, represents a direct continuation of this earlier work. The focus consequently shifted from the representation of linguistic relatedness to its cognitive and communicative consequences. Quantitative analyses and psycholinguistic experiments made it possible to investigate how multilingual speakers recognize and exploit cross-linguistic correspondences, how linguistic similarity contributes to comprehension, and how multilingual competence emerges from the interaction of language-specific and shared knowledge. More recent joint research on microsyntax (Avgustinova & Iomdin 2019), multilingual competence, and language contact extends this perspective to finer-grained linguistic phenomena and more complex multilingual environments.