Mirella Lapata · page 3
Papers
- Whodunnit? Crime Drama as a Case for Natural Language Understanding (2018 · 22 citations)
Frermann, Cohen and Lapata propose crime drama as a testbed for natural language understanding. Unlike most comprehension benchmarks, it combines several modalities, spoken dialogue and reasoning that unfolds over time. Every CSI episode asks who committed the crime and later reveals the answer, so the supervision… - What Is That Talk About? A Video-to-Text Summarization Dataset for Scientific Presentations (2025 · 13 citations)
Large multimodal models handle general videos fairly well, but they struggle with scientific talks, which contain specialized vocabulary, figures and tables. - Text Generation from Discourse Representation Structures (2021 · 13 citations)
The paper generates document text from DRSs on the Groningen Meaning Bank. Two problems make DRSs hard to generate from. First, the conditions of a box form a set, so one meaning can arrive in many orders; the paper learns a canonical order with a graph model (a graph neural network over conditions followed by a… - Dolomites: Domain-Specific Long-Form Methodical Tasks (2025 · 12 citations)
Professionals such as clinicians, lawyers and teachers often produce structured documents (care plans, petitions, lesson plans) by following a fairly fixed routine. - Explanatory Summarization with Discourse-Driven Planning (2025 · 8 citations)
Lay summaries of scientific papers rely heavily on explanations such as analogies, background and causal links. - Think Before you Write: QA-Guided Reasoning for Character Descriptions in Books (2026)
Writing descriptions of characters from full-length books is difficult. A model must follow how a character's relationships and circumstances change, combine clues spread across more than 100k tokens, and infer details the text only implies. The authors find that reasoning-capable LLMs perform worse on this task when…
This page in Markdown → https://md.research.elamur.ai/free/en/authors/mirella-lapata/page/3/