Papers by Ioana Manolescu
Structured Discourse Representation for Factual Consistency Verification (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to verify factual consistency of text capture a performance gap compared with sentence-level entailment. |
| Approach: | They propose a method that combines structured discourse information extraction with a classifier, FDSpotter, for factual consistency verification. |
| Outcome: | The proposed method achieves competitive performance on two tasks: data-to-text generation and text summarisation. |
FactSpotter: Evaluating the Factual Faithfulness of Graph-to-Text Generation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Graph-to-text (G2T) generation is an important task in natural language generation as it renders graphs accessible to non-technical users in downstream applications such as question answering. |
| Approach: | They propose a metric that correctly identifies factual faithfulness and uses it to determine if a triple is present in a generated text. |
| Outcome: | The proposed metric achieves highest correlation with human annotations on data correctness, data coverage, and relevance. |
The Search for Conflicts of Interest: Open Information Extraction in Scientific Publications (2025.findings-emnlp)
Copied to clipboard
| Challenge: | public repositories of relations of interest are increasingly recommended or mandated in various domains to avoid COIs. |
| Approach: | They propose an LLM-based open information extraction framework for extracting financial or other types of interesting relations from scientific text. |
| Outcome: | The proposed framework extracts financial or other types of interesting relations from scientific text. |
Open Information Extraction with Entity Focused Constraints (2023.findings-eacl)
Copied to clipboard
| Challenge: | Open Information Extraction (OIE) is the task of extracting tuples from unstructured corpora without any knowledge of the type and lexical form of the subject, the object, or the subject. |
| Approach: | They exploit domain knowledge to inject constraints into the extraction through constrained inference and constraint-aware training. |
| Outcome: | The proposed approach improves the CaRB and WIRe57 metric and achieves a 29.17% improvement in the CARB and 24.37% improvement on the WIRe56 metric. |