Papers with neural systems struggle
Content Type Profiling of Data-to-Text Generation Datasets (2022.coling-1)
Copied to clipboard
| Challenge: | Data-to-Text Generation (D2T) problems can be seen as a stream of time-stamped events with a textual summary of each event presenting the insights. |
| Approach: | They propose a typology of content types to classify the contents of event summaries using a dataset as the distribution of the aggregated content types. |
| Outcome: | The proposed typology shows that neural systems struggle in generating complex types on different datasets. |