This is a repository for the development of a provenance summarization pipeline. The team will use this repository to share code, the dataset, and more.
We are building and testing the possibility of generating natural language summaries from interaction histories. Wether this is generated from a corpus of sentences for every event and summarized or other intermediary steps, we want to understand:
- Evaluate human preferences
- feasibility of generating sentences from interaction histories
- Comparison of features between human generated summaries of interaction histories and the automated approach
We implement a pipeline for conducting experiments using prompt engineering techniques. Please refer to the README in the prompt_engineering package