Dialog state tracking is one of the key sub-tasks of dialog management, which defines the representation of dialog states and updates them at each moment on a given on-going conversation. To provide a common testbed for this task, the first Dialog State Tracking Challenge (DSTC) was organized. More recently, Dialog State Tracking Challenges 2 & 3 and Dialog State Tracking Challenge 4 have been successfully completed.

In the fifth challenge, we will continue evaluating the dialog state tracking task on human-human dialogs. Different from DSTC4, in this challenge we will focus on cross-language DST. In addition to this main task, we also propose a series of pilot tracks for the core components in developing end-to-end dialog systems based on the same dataset.

More details about the challenge can be found from the official website (http://workshop.colips.org/dstc5/index.html).

This repository provides the resources including evaluation scripts, ontology, and handbooks for the challenge participants.

Clone this repository into your working directory.

$ git clone https://github.com/seokhwankim/dstc5.git
$ cd dstc5/

Download the archived datasets from the link provided to each registered participant and extract the files into 'data/' directory.

$ tar xvfz dstc5_train_dev.tar.gz

Install python-levenshtein and fuzzywuzzy which is a prerequisite for the baseline tracker.

$ pip install python-levenshtein
$ pip install fuzzywuzzy

Run the baseline tracker (with method 1).

$ python scripts/baseline.py --dataset dstc5_dev --dataroot data --trackfile baseline_dev.json --ontology scripts/config/ontology_dstc5.json --method 1

Check the structure and contents of the tracker output.

$ python scripts/check_main.py --dataset dstc5_dev --dataroot data --ontology scripts/config/ontology_dstc5.json --trackfile baseline_dev.json
Found no errors, trackfile is valid

Evaluate the output.

$ python scripts/score_main.py --dataset dstc5_dev --dataroot data --trackfile baseline_dev.json --scorefile baseline_dev.score.csv --ontology scripts/config/ontology_dstc5.json

Print out the summarized results.

$ python scripts/report_main.py --scorefile baseline_dev.score.csv

                       featured metrics
--------------------------------------------------------------
                    |   all.schedule1    |   all.schedule2   |
--------------------------------------------------------------
segment.accuracy    |     0.0411147      |     0.0542986     |
slot_value.precision|     0.1769352      |     0.2536232     |
slot_value.recall   |     0.1104536      |     0.1431493     |
slot_value.fscore   |     0.1360049      |     0.1830065     |



                                    basic stats
-----------------------------------------------------------------------------------
                  dataset : dstc5_dev
                 sessions : 2
          total_wall_time : 24.7001729012
               utterances : 3130
  wall_time_per_utterance : 0.0078914290419

You can get the latest updates and participate in discussions on DSTC mailing list

To join the mailing list, send an email to: (listserv@lists.research.microsoft.com) putting "subscribe DSTC" in the body of the message (without the quotes). To post a message, send your message to: (dstc@lists.research.microsoft.com).