Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

Commit

Permalink
Release BST human annotations (#3405)
Browse files Browse the repository at this point in the history
* Add in human annotations

* Fix SHA

* Note
  • Loading branch information
EricMichaelSmith authored Jan 26, 2021
1 parent c11a46c commit c2c800e
Show file tree
Hide file tree
Showing 2 changed files with 8 additions and 1 deletion.
1 change: 1 addition & 0 deletions parlai/tasks/blended_skill_talk/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -25,5 +25,6 @@ The following files are downloaded when calling the BST dataset (for instance, w
- `topic_to_persona_list.txt`: A list of WoW topics and the persona strings that they correspond to, used by the PersonaTopicifierTeachers to ensure that WoW topics are relevant given persona strings and vice versa
- `ed_persona_topicifier__train__both_sides.json`, `ed_persona_topicifier__train__experiencer_only.json`, `ed_persona_topicifier__valid__experiencer_only.json`, and `ed_persona_topicifier__test__experiencer_only.json`: cached files of EmpatheticDialogues conversations with WoW topics and persona strings added to contexts, used by `EDPersonaTopicifierTeacher`. EmpatheticDialogues conversations consist of a Speaker describing a situation and a Listener responding empathetically: the choice of file used depends on whether we are considering conversation turns in which the previous utterance comes from either the Speaker or the Listener (`both_sides`) or from the Speaker (i.e. experiencer) only (`experiencer_only`)
- `safe_personas.txt`: A list of safe personas, used with BlendedSkillTalk in interactive mode with `--safe-personas-only True`
- `human_annotations.json`: The set of ~700 per-turn human annotations of conversations in the BlendedSkillTalk validation set. Conversation turns were annotated to indicate whether they exhibited the modes of knowledge, empathy, personal situations, and/or personal background.

Tags: #BlendedSkillTalk, #All, #ChitChat
8 changes: 7 additions & 1 deletion parlai/tasks/blended_skill_talk/build.py
Original file line number Diff line number Diff line change
Expand Up @@ -60,12 +60,18 @@
'2ee292aa0006ea002e9b23d4f7326fe9e17514ce5793d31fd8d679035d4366a7',
zipped=False,
),
build_data.DownloadableFile(
'http://parl.ai/downloads/blended_skill_talk/human_annotations.json',
'human_annotations.json',
'fe76a989d4fdcfa4a5beb214054069ed0abf45f3530eaeb8df38e75f7bb14795',
zipped=False,
),
]


def build(opt):
dpath = os.path.join(opt['datapath'], 'blended_skill_talk')
version = 'v1.4'
version = 'v1.5'

if not build_data.built(dpath, version_string=version):
print('[building data: ' + dpath + ']')
Expand Down

0 comments on commit c2c800e

Please sign in to comment.