Stanford Politeness Corpus (Stack Exchange)¶
A collection of requests from Stack Exchange, annotated with politeness (6,603 utteranecs).
Distributed together with: A Computational Approach to Politeness with Application to Social Factors. Cristian Danescu-Niculescu-Mizil, Moritz Sudhof, Dan Jurafsky, Jure Leskovec, Christopher Potts. ACL, 2013.
Each utterance corresponds to a Stack Exchange request. For each utterance, we provide:
id: row index of the request given in the original data release.
speaker: the author of the utterance.
conversation_id: id of the first utterance in the conversation this utterance belongs to, which in this case is the id of the utterance itself
reply_to: None. In this dataset, each request is seen as a full conversation, and thus all utterances are at the ‘root’ of the conversations
text: textual content of the utterance.
Metadata for each utterance is inherited from the general CMV corpus:
Normalized Score: Normalized politeness score computed based on annotations.
Binary: A binarized politeness label where 1=”polite”, 0=”neutral”, -1 = “impolite”.
Annotations: the original annotations from Amazon Mechanical Turkers for the given utterance. Ratings are on a 1-25 scale.
parsed: dependency-parsed version of the utterance text
To download directly with ConvoKit:
>>> from convokit import Corpus, download >>> corpus = Corpus(filename=download("stack-exchange-politeness-corpus"))
For some quick stats:
>>> len(corpus.get_utterance_ids()) 6603
ConvoKit’s Stanford Politeness Corpus is governed by the CC BY license v4.0. Copyright (C) 2017-2020 The ConvoKit Developers.