Abstract
Semantic role labelling (SRL) is a task in natural language processing which detects and classifies the semantic arguments associated with the predicates of a sentence. It is an important step towards understanding the meaning of a natural language. There exists SRL systems for well-studied languages like English, Chinese or Japanese but there is not any such system for the Vietnamese language. In this paper, we present the first SRL system for Vietnamese with encouraging accuracy. We first demonstrate that a simple application of SRL techniques developed for English could not give a good accuracy for Vietnamese. We then introduce a new algorithm for extracting candidate syntactic constituents, which is much more accurate than the common node-mapping algorithm usually used in the identification step. Finally, in the classification step, in addition to the common linguistic features, we propose novel and useful features for use in SRL. Our SRL system achieves an F1 score of 73.53% on the Vietnamese PropBank corpus. This system, including software and corpus, is available as an open source project and we believe that it is a good baseline for the development of future Vietnamese SRL systems.
Introduction: The Challenge of Vietnamese SRL
Semantic Role Labelling (SRL) is the process of identifying the "Who did What to Whom, When, and Why" within a sentence. It identifies the predicates (usually verbs) and classifies their associated arguments—the actors, objects, and circumstances that provide context to the action.
While SRL is well-developed for languages like English and Chinese, Vietnamese—an isolating language with distinct syntactic structures—lacked a dedicated system. Researchers discovered that simply "porting" English SRL techniques to Vietnamese fails significantly. Specifically, the standard "node-mapping" algorithm used to identify potential arguments in a sentence tree only achieved a 35.84% $F_1$ score on Vietnamese data. This gap necessitated a system built specifically for Vietnamese linguistic rules.
Methodology
The authors followed the traditional two-step SRL pipeline—identification and classification—but overhauled the internal logic for the Vietnamese context.
Argument Identification
Instead of the standard 1-1 node-mapping, the researchers developed a new Constituent Extraction Algorithm. This algorithm traverses bracketed parse trees by starting at the predicate and moving upward to the root, intelligently collecting sibling nodes and their children based on phrase types and function tags.
As shown in the phrase structure tree above, the algorithm must navigate the hierarchy to correctly identify "Nam" and "bóng" as arguments for the predicate "đá".
Argument Classification and Feature Engineering
Once candidates are identified, the system classifies them into roles (e.g., Arg0 for agent, ArgM-TMP for time). The study compared Maximum Entropy (ME) and Support Vector Machines (SVM) models, finding that SVMs significantly outperformed ME.
Beyond standard features like phrase type or head words, the system introduced:
- Function Tags: Labels indicating grammatical roles (e.g., subject, object), which improved the $F_1$ score by approximately 4%.
- Word Clusters: Using Gaussian mixture analysis on word embeddings to group similar words, helping the model handle "unseen" words in small datasets.
- Predicate Type: Identifying if the predicate is a verb, noun, adjective, or preposition.
Data and Results
The study utilized the Vietnamese PropBank, containing roughly 5,000 annotated sentences. This is significantly smaller than the English PropBank (~50,000 sentences), leading to higher data sparseness.
The results highlight the success of the custom approach:
- Identification Step: The new extraction algorithm achieved an 83.63% $F_1$ score, a massive leap from the 35.84% achieved by the baseline English-style algorithm.
- Overall System: The final system reached a 73.53% $F_1$ score using the SVM classifier.
- Baseline Comparison: The baseline system (using standard English techniques) only managed a 6.20% $F_1$ score on the Vietnamese dataset, proving that the authors' custom approach was essential.
Key Insights and Future Directions
The failure of English-centric methods highlights that syntactic-semantic mappings in isolating languages like Vietnamese are fundamentally different from inflectional languages. Function tags and word clustering proved to be high-value features for mitigating the challenges of a smaller corpus.
Future Work: * Enlarging the Corpus: More data is needed to reach performance levels seen in English SRL. * Complex Sentences: Moving beyond "simple" sentences with single predicates to handle nested structures. * Advanced Inference: Exploring techniques like Integer Linear Programming (ILP) to ensure global consistency of predicted semantic roles.
The system is available as an open-source project, providing a robust baseline for future Vietnamese natural language understanding research.
Summary updated