The Role of Verb Semantics in Hungarian Verb-Object Order

Dorottya Demszky, László Kálmán, Dan Jurafsky, Beth LevinarXiv 2020 · arXiv

Abstract

Hungarian is traditionally classified as a "discourse-configurational" language, where word order is dictated by information structure—topic and focus—rather than rigid grammatical rules. However, this study challenges the exclusivity of that model by demonstrating that the lexical semantics of a verb are a powerful and systematic driver of sentence structure. Through a large-scale analysis of 1.3 million verb-object pairs, the research reveals that a verb's semantic class is a more reliable predictor of word order than traditional factors such as object definiteness or constituent weight.

Introduction: Beyond Discourse Configuration

In linguistic theory, Hungarian is often cited for its "free word order." Because grammatical roles are clearly marked by suffixes, speakers can rearrange sentences to emphasize different elements. The prevailing consensus has been that this flexibility is governed almost entirely by discourse context: the "topic" (known information) precedes the "focus" (new or emphasized information).

Despite this, linguists have long observed that certain verbs appear to favor specific positions regardless of context. This paper quantifies this "lexical influence," asking whether the inherent meaning of a transitive verb—whether it denotes an action, a state, or a perception—systematically biases the choice between Object-Verb (OV) and Verb-Object (VO) ordering.

Methodology

The authors utilized a data-driven approach centered on a logistic regression framework to isolate the impact of semantics from other variables.

  1. Data Extraction: The study utilized the Hungarian Gigaword Corpus, extracting verb-object pairs from a 1.5-billion-token collection. To ensure a clean dataset, the authors filtered out confounding factors such as verbs with preverbs (which have fixed positions), bare nouns, and ditransitive verbs.
  2. Semantic Classification: 380 unique transitive verbs were manually categorized into 11 semantic classes (e.g., Activity, Affect, Change, Creation, Perception) based on the Levin (1993) framework.
  3. Modeling: The researchers compared three types of features: lexical (verb lemma and class), information structure proxies (object definiteness), and syntactic weight (length of the noun phrase). These were used to predict the binary outcome of OV vs. VO word order.

Key Findings

The Primacy of Semantics

One of the most striking results is that the semantic class of a verb has a significantly larger effect size on word order than object definiteness. In Hungarian, the "textbook rule" suggests that definite objects follow the verb while indefinite objects precede it. However, the data shows that the verb's meaning often overrides this rule.

The Stative vs. Non-Stative Divide

The research identified a clear split based on the nature of the verb:

  • Stative Verbs: Verbs describing states or spatial configurations (e.g., "cover," "contain") show a strong preference for OV order.
  • Non-Stative Verbs: Verbs describing actions or changes (e.g., "reduce," "kiss") strongly prefer VO order.

Notable Exceptions: Psych and Creation Verbs

Two categories provided unique insights into the intersection of meaning and structure:

  • Psych Verbs: While verbs of emotion (e.g., "admire," "hate") are technically stative, they behave like action verbs, showing a strong preference for VO.
  • Creation Verbs: Verbs of creation (e.g., "create," "breed") prefer OV. The authors hypothesize that because the object does not exist until the action is complete, it is inherently "new" information, making it a natural fit for the preverbal focus position.

Statistical Results

The predictive power of the semantic models was substantial:

  • Verb-Only Model: Predicting order based solely on the specific verb achieved 68% accuracy, compared to a 53% majority baseline.
  • Semantic Class Model: Using only 11 broad categories achieved 65.04% accuracy. The small gap between individual lemmas and broad classes suggests that word order preferences are not random lexical quirks but are tied to categorical meanings.
  • Comparison: Models based solely on definiteness (57.29%) or object weight (54.90%) were significantly less accurate than those based on verb semantics.

Conclusion

This study demonstrates that Hungarian word order is not merely a product of discourse-level emphasis, but is deeply rooted in the lexical semantics of the verb. While VO remains the more common "neutral" state (62% of significant cases), the OV preference is a systematic feature of specific semantic classes.

Future research aims to incorporate actual discourse context labels and expand the analysis to include the animacy of subjects and objects. Understanding these patterns provides a clearer picture of how cognitive constraints and meaning-making influence the structural choices speakers make in real-time communication.

Summary updated

Abstract

Hungarian is often referred to as a discourse-configurational language, since the structural position of constituents is determined by their logical function (topic or comment) rather than their grammatical function (e.g., subject or object). We build on work by Komlósy (1989) and argue that in addition to discourse context, the lexical semantics of the verb also plays a significant role in determining Hungarian word order. In order to investigate the role of lexical semantics in determining Hungarian word order, we conduct a large-scale, data-driven analysis on the ordering of 380 transitive verbs and their objects, as observed in hundreds of thousands of examples extracted from the Hungarian Gigaword Corpus. We test the effect of lexical semantics on the ordering of verbs and their objects by grouping verbs into 11 semantic classes. In addition to the semantic class of the verb, we also include two control features related to information structure, object definiteness and object NP weight, chosen to allow a comparison of their effect size to that of verb semantics. Our results suggest that all three features have a significant effect on verb-object ordering in Hungarian and among these features, the semantic class of the verb has the largest effect. Specifically, we find that stative verbs, such as fed "cover", jelent "mean" and övez "surround", tend to be OV-preferring (with the exception of psych verbs which are strongly VO-preferring) and non-stative verbs, such as bírál "judge", csökkent "reduce" and csókol "kiss", verbs tend to be VO-preferring. These findings support our hypothesis that lexical semantic factors influence word order in Hungarian.
BibTeX
@misc{demszky2020role,
  author = {Dorottya Demszky and László Kálmán and Dan Jurafsky and Beth Levin},
  title = {{The Role of Verb Semantics in Hungarian Verb-Object Order}},
  year = {2020},
  eprint = {2006.09432},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2006.09432},
}
Figure