Ticket Tagger

Machine learning driven issue classification bot. Add to your repository now!

License

Ticket Tagger automatically predicts and labels issue types.
Copyright (C) 2018,2019,2020  Rafael Kallis

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <https://www.gnu.org/licenses/>. 

Development

notice:

• nodejs ^8.3.x is required to compile/install dependencies
• wget is required for fetching datasets

get started:

git clone https://github.com/rafaelkallis/ticket-tagger ticket-tagger
cd ticket-tacker

# install appropriate nodejs version
nvm install 8
nvm use 8

# compile/install dependencies
npm install

# fetch dataset
npm run dataset

# run benchmark
npm run benchmark

# run linter
npm run lint

# run tests
npm test

# run server
npm start

experiments:

For each experiment, we need a dataset that allows to test the stated hypothesis, as well as a baseline dataset which contains the same amount of labelled issues.

Does a repository specific dataset affect the model's performance?

# run baseline-issues benchmark
npm run dataset:vscode:baseline
npm run benchmark

# run vscode-issues benchmark
npm run dataset:vscode
npm run benchmark

Does a (spoken) language specific dataset affect the models perfomrnace?

# run baseline-issues benchmark
npm run dataset:english:baseline
npm run benchmark

# run english-issues benchmark
npm run dataset:english
npm run benchmark

Do code snippets affect the models perfomrnace?

# run baseline-issues benchmark
npm run dataset:nosnip:baseline
npm run benchmark

# run nosnip-issues benchmark
npm run dataset:nosnip
npm run benchmark

generate dataset:

A dataset (with 10k bugs, 10k enhancements and 10k questions) can be downloaded using npm run dataset. The dataset was generated using github archive's which can be accessed through google BigQuery.

Add the query below to your BigQuery console and adjust if needed (e.g., add __label__ prefix to labels, etc.).

SELECT
  label, CONCAT(title, ' ', REGEXP_REPLACE(body, '(\r|\n|\r\n)',' '))
FROM (
  SELECT
    LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.labels[0].name')) AS label,
    JSON_EXTRACT_SCALAR(payload, '$.issue.title') AS title,
    JSON_EXTRACT_SCALAR(payload, '$.issue.body') AS body
  FROM
    [githubarchive:day.20180201],
    [githubarchive:day.20180202],
    [githubarchive:day.20180203],
    [githubarchive:day.20180204],
    [githubarchive:day.20180205]
  WHERE
    type = 'IssuesEvent'
    AND JSON_EXTRACT_SCALAR(payload, '$.action') = 'closed' )
WHERE 
  (label = 'bug' OR label = 'enhancement' OR label = 'question')
  AND body != 'null';

run serverless app:

You need a .env file in order to run the github app. The file should look like this:

GITHUB_CERT=/path/to/cert.private-key.pem
GITHUB_SECRET=123456
GITHUB_APP_ID=123
PORT=3000

Note: When running app in production, environment variables should be provided by host.

references:

• Building GitHub Apps
• Fasttext