What can Neural Referential Form Selectors Learn?

2021/08/15 by Guanyi Chen, Chen, Guanyi, Fahime Same +3
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Speech and dialogue systems #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2108.06806

openalex publication_date 2021/08/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Despite achieving encouraging results, neural Referring Expression Generation models are often thought to lack transparency. We probed neural Referential Form Selection (RFS) models to find out to what extent the linguistic features influencing the RE form are learnt and captured by state-of-the-art RFS models. The results of 8 probing tasks show that all the defined features were learnt to some extent. The probing tasks pertaining to referential status and syntactic position exhibited the highest performance. The lowest performance was achieved by the probing models designed to predict discourse structure properties beyond the sentence level.

Citations

• Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View
• Building Natural-Language Generation Systems
• Sequence to Sequence Learning with Neural Networks

Related

• Assessing Neural Referential Form Selectors on a Realistic Multilingual Dataset
• NeuralREG: An end-to-end approach to referring expression generation
• Does referent predictability affect the choice of referential form? A\n computational approach using masked coreference resolution
• Intrinsic Task-based Evaluation for Referring Expression Generation
• Non-neural Models Matter: A Re-evaluation of Neural Referring Expression\n Generation Systems