Update README.md
Browse files
README.md
CHANGED
@@ -23,26 +23,41 @@ task_ids:
|
|
23 |
- text-scoring
|
24 |
- dialogue-modeling
|
25 |
---
|
26 |
-
|
|
|
27 |
[![CC BY-NC 4.0][cc-by-nc-shield]][cc-by-nc]
|
28 |
|
29 |
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
-
|
35 |
-
* `spont`: shorthand for Spontaneanation
|
36 |
-
* `cornell`: shorthand for Cornell Movie-Dialogs Corpus
|
37 |
-
* `subtle`: (not present in `spolin-valid.json`) shorthand for SubTle Corpus
|
38 |
|
39 |
-
|
40 |
-
*
|
41 |
-
*
|
42 |
-
*
|
43 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
|
45 |
-
##### `spolin-train.
|
46 |
|| yesands| non-yesands|
|
47 |
|--|---:|---:|
|
48 |
|Spontaneanation|10,459|5,587*|
|
@@ -51,7 +66,7 @@ Each second level item is a list of _yes-ands_ or non-_yes-ands_. Each item cont
|
|
51 |
|Total|67,188|43,409|
|
52 |
|
53 |
|
54 |
-
##### `spolin-train-acl.
|
55 |
|
56 |
|| yesands| non-yesands|
|
57 |
|--|---:|---:|
|
@@ -59,7 +74,7 @@ Each second level item is a list of _yes-ands_ or non-_yes-ands_. Each item cont
|
|
59 |
|Cornell|14,976|17,851|
|
60 |
|Total|25,435|23,438|
|
61 |
|
62 |
-
##### `spolin-valid.
|
63 |
|
64 |
|| yesands| non-yesands|
|
65 |
|--|---:|---:|
|
@@ -69,6 +84,25 @@ Each second level item is a list of _yes-ands_ or non-_yes-ands_. Each item cont
|
|
69 |
|
70 |
\*Artificially collected by mix & matching positive Spontaneanation samples to balance dataset for training classifier
|
71 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
|
73 |
### License
|
74 |
|
|
|
23 |
- text-scoring
|
24 |
- dialogue-modeling
|
25 |
---
|
26 |
+
|
27 |
+
# SPOLIN
|
28 |
[![CC BY-NC 4.0][cc-by-nc-shield]][cc-by-nc]
|
29 |
|
30 |
|
31 |
+
This is the repo for the paper ["Grounding Conversations with Improvised Dialogues"](https://aclanthology.org/2020.acl-main.218/) (ACL2020).
|
32 |
+
The _Selected Pairs of Learnable ImprovisatioN_ (SPOLIN) corpus is a collection of more than 68,000 "Yes, and" type dialogue pairs extracted from the Spontaneanation podcast by Paul F. Tompkins, the Cornell Movie-Dialogs Corpus, and the SubTle corpus. For more information, refer to our [paper](https://arxiv.org/abs/2004.09544) or our [project page](https://justin-cho.com/spolin).
|
33 |
+
|
34 |
+
|
35 |
+
### Available SPOLIN versions:
|
36 |
+
The core dataset that was used for the experiments in the paper only includes _yes-ands_ and non-_yes-ands_ from Spontaneanation and most of what is provided in those extracted from the Cornell Movie-Dialogs Corpus. After the submitting the paper, we continued our iterative data augmentation process, repeating another iteration with the Cornell Movie-Dialogs Corpus and extracting from the SubTle corpus. This expanded version is also included in this repository [here](/data). This latest version of SPOLIN was used to train the model used in our [demo](https://spolin.isi.edu).
|
37 |
+
|
38 |
+
|
39 |
+
In the `data` folder, we provide two versions of the SPOLIN training set:
|
40 |
+
1. Version used for experiments in the ACL paper: `data/spolin-train-acl.csv`
|
41 |
+
2. Expanded version: `data/spolin-train.csv`
|
42 |
|
43 |
+
### Relevant links:
|
|
|
|
|
|
|
44 |
|
45 |
+
* Project page: https://justin-cho.com/spolin
|
46 |
+
* Github repo: https://github.com/wise-east/spolin
|
47 |
+
* SpolinBot Demo: https://spolin.isi.edu
|
48 |
+
* ACL2020 Paper: https://aclanthology.org/2020.acl-main.218/
|
49 |
|
50 |
+
**Fiels**
|
51 |
+
* `id`: unique identifier
|
52 |
+
* `prompt`: first utterance in utterance pair
|
53 |
+
* `response`: second utterance in utterance pair
|
54 |
+
* `label`: yesand = 1, non-yesand = 0
|
55 |
+
* `source`: the source for the sample
|
56 |
+
* `split`: whether the sample belongs to the training set or the validation set
|
57 |
+
|
58 |
+
### Dataset Statistics
|
59 |
|
60 |
+
##### `spolin-train.csv`:
|
61 |
|| yesands| non-yesands|
|
62 |
|--|---:|---:|
|
63 |
|Spontaneanation|10,459|5,587*|
|
|
|
66 |
|Total|67,188|43,409|
|
67 |
|
68 |
|
69 |
+
##### `spolin-train-acl.csv`:
|
70 |
|
71 |
|| yesands| non-yesands|
|
72 |
|--|---:|---:|
|
|
|
74 |
|Cornell|14,976|17,851|
|
75 |
|Total|25,435|23,438|
|
76 |
|
77 |
+
##### `spolin-valid.csv`:
|
78 |
|
79 |
|| yesands| non-yesands|
|
80 |
|--|---:|---:|
|
|
|
84 |
|
85 |
\*Artificially collected by mix & matching positive Spontaneanation samples to balance dataset for training classifier
|
86 |
|
87 |
+
### ACL Presentation
|
88 |
+
|
89 |
+
[Video recording](https://slideslive.com/38928948/grounding-conversations-with-improvised-dialogues)
|
90 |
+
|
91 |
+
|
92 |
+
### Citation
|
93 |
+
|
94 |
+
If you use our data for your work, please cite our ACL2020 paper:
|
95 |
+
```
|
96 |
+
@inproceedings{cho2020spolin,
|
97 |
+
title={Grounding Conversations with Improvised Dialogues},
|
98 |
+
author={Cho, Hyundong and May, Jonathan},
|
99 |
+
booktitle ={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
|
100 |
+
publisher = {Association for Computational Linguistics},
|
101 |
+
location = {Seattle, Washington, USA},
|
102 |
+
year={2020}
|
103 |
+
}
|
104 |
+
```
|
105 |
+
|
106 |
|
107 |
### License
|
108 |
|