|
# Enriched Topical-Chat: A Dialogue Act and Knowledge Sentence annotated version of Topical-Chat |
|
|
|
This README describes Enriched Topical-Chat, an augmentation of Topical-Chat that contains dialogue act and knowledge sentence annotations for each turn in the dataset. Each annotation is automatically annotated using off-the-shelf models. |
|
|
|
## Knowledge Sentence Annoations |
|
|
|
Each conversation in Topical-Chat has a pair of reading sets which consists of a set of knowledge sentences. For every turn and knowledge sentence in the Topical-Chat dataset, we computed a TFIDF vector. We then computed the cosine similarity between a turn in the conversation and selected the top-1 knowledge sentence. In the new data release, for each conversation turn, we will present the knowledge sentences selected along with the similarity score. |
|
|
|
## Dialogue Act Annoations |
|
|
|
We obtain the dialogue acts for each turn by running an off-the-shelf SVM dialogue act tagger released by (https: |
|
This tagger was trained on five datasets (Switchboard, Oasis BT, Maptask, VerbMobil2, AMI). |
|
|
|
## Prerequisites |
|
|
|
After cloning the repo you must first follow the instructions in https: |
|
|
|
### Conversations: |
|
The data is hosted on s3. To pull the data run these commands: |
|
``` |
|
wget https: |
|
|
|
wget https: |
|
|
|
wget https: |
|
|
|
wget https: |
|
|
|
wget https: |
|
``` |
|
|
|
Each .json file has the specified format: |
|
|
|
``` |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
, |
|
}, |
|
|
|
|
|
|
|
, |
|
} |
|
], |
|
"gt_turn_ks": |
|
|
|
}, |
|
``` |
|
``` |
|
The additional fields are: |
|
|
|
message: a list containing the segments of each turn |
|
segmented_annotations: a list of annotations for each segment within a turn each. |
|
da: ground truth dialog act associated with segmented response |
|
gt_ks: ground truth knowledge sentence associated with segmented response |
|
ds: data source knowledge retrieved from. wiki, fun_facts or article |
|
fun_facts: |
|
section: which section containing the fun facts i.e. FS1 |
|
index: which element in the list of fun facts |
|
wiki: |
|
section: which section containing the wikipedia sentence i.e. FS2 |
|
start_index: index of beginning character of sentence in article |
|
end_index: index of end character |
|
article: |
|
section: which section of article. i.e. AS4 |
|
start_index: index of beginning character of sentence in article |
|
end_index: index of end character |
|
gt_turn_ks: ground truth knowlege sentence associated with turn |
|
``` |
|
|
|
## Citation |
|
If you use this dataset, please cite the following two papers: |
|
### Enriched Topical-Chat |
|
``` |
|
@article |
|
, |
|
author=, |
|
journal=, |
|
year= |
|
} |
|
|
|
``` |
|
### Topical-Chat |
|
``` |
|
@inproceedings |
|
, |
|
title=}, |
|
year=, |
|
booktitle= |
|
} |
|
|
|
``` |
|
|