Heriot-WattUniversity/dialog_babi

Dialog bAbI tasks data

In this directory is the set of 6 tasks for testing end-to-end dialog systems in the restaurant domain as described in the paper "Learning End-to-End Goal-Oriented Dialog" by Bordes & Weston (http://arxiv.org/abs/1605.07683). The aim is that each task tests a unique aspect of dialog. Tasks are designed to complement the set of 20 bAbI tasks for story understanding already released with the paper "Towards AI Complete Question Answering: A Set of Prerequisite Toy Tasks" by Weston et al. (http://arxiv.org/abs/1502.05698).

Data

For each task, there are 1000 dialogs for training, 1000 for development and 1000 for testing. For tasks 1-5, we also include a second test set (with suffix -OOV.txt) that contains dialogs including entities not present in training and development sets.

The file format for each task is as follows: ID user_utterance [tab] bot_utterances

The IDs for a given dialog start at 1 and increase. When the IDs in a file reset back to 1 you can consider the following sentences as a new dialog. When the bot speaks two times in a row, we used the special token "" to fill in for the missing user utterance.

For example (for task 1):

1 hi	hello what can i help you with today
2 can you make a restaurant reservation with italian cuisine for six people in a cheap price range	i'm on it
3 <SILENCE>	where should it be
4 rome please	ok let me look into some options for you
5 <SILENCE>	api_call italian rome six cheap

The goal of the tasks is to predict the bot utterances, that can be sentences or API calls (sentences starting with the special token "api_call").

Along with the train, dev and test sets, we also include a knowledge base file (dialog-babi-kb-all.txt) that contain all entities appearing in dialogs for tasks 1-5. We also include a file containing the candidates to select the answer from (dialog-babi-candidates.txt) for tasks 1-5, that is simply made of all the bot utterances in train, dev, test for these tasks.

Task 6 is a bit different since its data comes from the Dialog State Tracking Challenge 2 (http://camdial.org/~mh521/dstc/), which we modified to convert it into the same format as the other tasks. There is no OOV test set associated with this task and the knowledge base (dialog-babi-task6-dstc2-kb.txt) is imperfect. This task has its own candidates file (dialog-babi-task6-dstc2-candidates.txt).

License

This dataset is released under Creative Commons Attribution 3.0 Unported license. A copy of this license is included with the data.

Contact

The author of this porting is Alessandro Suglia and he has only made available the dataset via Huggingface datasets. For more details on the dataset and baselines, see the paper "Learning End-to-End Goal-Oriented Dialog" by Antoine Bordes and Jason Weston (http://arxiv.org/abs/1605.07683). For any information, contact Antoine Bordes : abordes (at) fb (dot) com .