This repository contains a few Dialog Datasets that I retrieved from various sources. Please cite/acknowledge the original authors or creators:
FRIENDS TV show dataset cleaned and partitioned. I obtained this data from https://github.com/fangj/friends/ and cleaned it to a nicer TSV format. No information on whom to cite (guessing this is in public domain now!)
I obtained this from this link and formatted to nicer TSV format
@InProceedings{Danescu-Niculescu-Mizil+Lee:11a,
author={Cristian Danescu-Niculescu-Mizil and Lillian Lee},
title={Chameleons in imagined conversations:
A new approach to understanding coordination of linguistic style in dialogs.},
booktitle={Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics, ACL 2011},
year={2011}
}
TSVs are nicer to work with, so lets make this as a standard :)
- A corpus file can have many dialogs
- An empty line (new line
\n) separates dialogs. - Each dialog can have many utterances ( an utterance is a turn)
- Each turn can have three fields, separated by tab character
\t- first column is unique identifier for turn.
- second column is character or agent name.
- third column is a text utterance (a sequence of words).
Here is an example of two dialogs:
0101_002 monica there 's nothing to tell he 's just some guy i work with
0101_003 joey c 'mon , you 're going out with the guy there 's gotta be something wrong with him
0101_004 chandler all right joey , be nice . so does he have a hump ? a hump and a hairpiece ?
0101_005 phoebe wait , does he eat chalk ?
0101_007 phoebe just , cause , i don 't want her to go through what i went through with carl- oh
0101_008 monica okay , everybody relax . this is not even a date . it 's just two people going out to dinner and- not having sex .
0101_009 chandler sounds like a date to me .
P.S. If you have a new dataset, please send me a link by creating an issue or send it via pull request. Goal is to collect lots of data, so we can train interesting neural models