In this repository you find data that has been gathered when conducting in-situ experiments in a conversational cooking setting. These data include transcripts,annotations and code to analyse these data.
When you are here while reading our paper, you might have following questions:
The raw transcripts are located in the transcripts folder. All experiments were conducted in Germany. Therefore, we provide the original german version as well as an english version transcribed with DeepL. Since we put much more effort in the transcription and annotation process compared to our workshop paper (=pilot study), we also provide the annotated data from that paper (see folder transcripts/workshop_paper).
A visual representation can be found in the annotation folder. InformationNeedTaxonomy.svg is the file you are looking for.
A description of all information needs as well as examples for them are located in the annotation folder. annotation_schema_cookversational_search_german.xlsx is the german version, annotation_schema_cookversational_search.xlsx is the english one. The codebook from our pilot study is located in this folder, too.
You find code for both the baseline models, the BERT based models and the results in the experiments folder.
Yes! Statistical Analysis.ipynb in the experimentsfolder is the relevant file.
The fully annotated dataset with all information needs, german and english turns (translated with DeepL) is here: annotation/corpus/cookversational_search_dataset
@article{frummet2022whatcani,
author = {Alexander Frummet and
David Elsweiler and
Bernd Ludwig},
title = {{"What Can {I} Cook with these Ingredients?" - Understanding Cooking-Related
Information Needs in Conversational Search}},
journal = {{ACM} Trans. Inf. Syst.},
volume = {40},
number = {4},
pages = {81:1--81:32},
year = {2022},
url = {https://doi.org/10.1145/3498330},
}