Pull Request Review Comments Dataset

<p><strong>Pull&nbsp;Request&nbsp;Review&nbsp;Comments&nbsp;(PRRC)&nbsp;Datasets</strong></p> <p>Two&nbsp;datasets&nbsp;have&nbsp;been&nbsp;created&nbsp;from&nbsp;the&nbsp;<a href="https://www.gharchive.org/">gharchive</a>&nbsp;website.&nbsp;The&nbsp;<a href="https://docs.github.com/en/developers/webhooks-and-events/github-event-types#pullrequestreviewcommentevent">Pull&nbsp;Request&nbsp;Review&nbsp;Comment&nbsp;Event</a>&nbsp;was&nbsp;selected&nbsp;from&nbsp;the&nbsp;set&nbsp;of&nbsp;available&nbsp;GitHub&nbsp;events.&nbsp;This&nbsp;dataset&nbsp;has&nbsp;been&nbsp;created&nbsp;for&nbsp;CARA:&nbsp;Chatbot&nbsp;for&nbsp;Automating&nbsp;Repairnator&nbsp;Actions&nbsp;as&nbsp;part&nbsp;of&nbsp;a&nbsp;master&#39;s&nbsp;thesis&nbsp;at&nbsp;KTH,&nbsp;Stockholm.</p> <p>First,&nbsp;a&nbsp;source&nbsp;dataset&nbsp;was&nbsp;downloaded&nbsp;from&nbsp;gharchive.&nbsp;That&nbsp;dataset&nbsp;ranges&nbsp;from&nbsp;January&nbsp;2015&nbsp;to&nbsp;December&nbsp;2020.&nbsp;It&nbsp;consisted&nbsp;of&nbsp;54,021,838&nbsp;PRRCs&nbsp;and&nbsp;is&nbsp;over&nbsp;18&nbsp;Gigabytes&nbsp;in&nbsp;size.&nbsp;It&nbsp;took&nbsp;over&nbsp;120&nbsp;hours&nbsp;to&nbsp;download&nbsp;all&nbsp;the&nbsp;data&nbsp;files&nbsp;and&nbsp;extract&nbsp;PRRC&nbsp;from&nbsp;it.&nbsp;From&nbsp;this&nbsp;source&nbsp;dataset,&nbsp;two&nbsp;subsets&nbsp;were&nbsp;derived:</p> <ol> <li>&nbsp;Pull&nbsp;Request&nbsp;Review&nbsp;Comments&nbsp;Dataset:&nbsp;This&nbsp;is&nbsp;the&nbsp;dataset&nbsp;of&nbsp;the&nbsp;comments&nbsp;from&nbsp;the&nbsp;latest100,000&nbsp;threads&nbsp;in&nbsp;the&nbsp;source&nbsp;dataset&nbsp;from&nbsp;gharchive.</li> <li>&nbsp;Pull&nbsp;Request&nbsp;Review&nbsp;Threads&nbsp;Dataset:&nbsp;This&nbsp;is&nbsp;the&nbsp;dataset&nbsp;of&nbsp;comments&nbsp;that&nbsp;were&nbsp;concatenated&nbsp;together&nbsp;if&nbsp;they&nbsp;were&nbsp;from&nbsp;the&nbsp;same&nbsp;thread (in&nbsp;chronological&nbsp;order).</li> </ol> <p>&nbsp;</p> <p><strong>Description</strong></p> <p>The&nbsp;dataset&nbsp;is&nbsp;stored&nbsp;in&nbsp;the&nbsp;<a href="https://jsonlines.org/">JSONLines&nbsp;format</a>,&nbsp;as&nbsp;was&nbsp;the&nbsp;source&nbsp;dataset&nbsp;from&nbsp;gharchive.</p> <p>For&nbsp;PRRC&nbsp;events,&nbsp;the&nbsp;source&nbsp;dataset&nbsp;contains&nbsp;the&nbsp;fields&nbsp;`comment_id`,&nbsp;`commit_id`,&nbsp;`url`,&nbsp;`author`,&nbsp;`created_at`,&nbsp;and&nbsp;`body`.&nbsp;</p> <ul> <li>`comment_id`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;specifies&nbsp;the&nbsp;ID&nbsp;GitHub&nbsp;uses&nbsp;for&nbsp;that&nbsp;comment.</li> <li>`commit_id`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;specifies&nbsp;the&nbsp;ID&nbsp;of&nbsp;the&nbsp;commit&nbsp;proposed&nbsp;in&nbsp;the&nbsp;pull&nbsp;request.</li> <li>`url`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;specifies&nbsp;the&nbsp;url&nbsp;to&nbsp;the&nbsp;comment&nbsp;in&nbsp;a&nbsp;pull&nbsp;request&nbsp;thread.</li> <li>`author`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;lists&nbsp;the&nbsp;username&nbsp;of&nbsp;the&nbsp;author&nbsp;of&nbsp;the&nbsp;pull&nbsp;request.</li> <li>`created_at`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;specifies&nbsp;the&nbsp;time&nbsp;at&nbsp;which&nbsp;the&nbsp;pull&nbsp;request&nbsp;comment&nbsp;was&nbsp;created.</li> <li>`body`&nbsp;is&nbsp;the&nbsp;field&nbsp;which&nbsp;describes&nbsp;the&nbsp;contents&nbsp;of&nbsp;the&nbsp;PRRC.</li> </ul> <p>The&nbsp;threads&nbsp;dataset&nbsp;contains&nbsp;the&nbsp;fields&nbsp;`url`&nbsp;and&nbsp;`body`&nbsp;which&nbsp;contain&nbsp;similar&nbsp;information&nbsp;as&nbsp;described&nbsp;above.&nbsp;However,&nbsp;the&nbsp;body&nbsp;field&nbsp;differs:&nbsp;it&nbsp;is&nbsp;a&nbsp;concatenation&nbsp;of&nbsp;all&nbsp;the&nbsp;PRRCs&nbsp;in&nbsp;a&nbsp;pull&nbsp;request&nbsp;thread.&nbsp;The&nbsp;comments&nbsp;dataset&nbsp;contains&nbsp;the&nbsp;fields&nbsp;&nbsp;`comment_id`,&nbsp;`commit_id`,&nbsp;`url`,&nbsp;`author`,&nbsp;`created_at`,&nbsp;and&nbsp;`body`.&nbsp;They&nbsp;are&nbsp;the&nbsp;same&nbsp;fields&nbsp;from&nbsp;the&nbsp;initial&nbsp;dataset.</p> <p>&nbsp;</p> <p><strong>Construction</strong></p> <p>We&nbsp;used&nbsp;the&nbsp;<a href="https://fasttext.cc/docs/en/language-identification.html">fasttext</a>&nbsp;model&nbsp;published&nbsp;by&nbsp;Facebook&nbsp;to&nbsp;detect&nbsp;the&nbsp;language&nbsp;of&nbsp;the&nbsp;PRRC.&nbsp;Only&nbsp;those&nbsp;PRRCs&nbsp;in&nbsp;English&nbsp;were&nbsp;preserved.&nbsp;We&nbsp;also&nbsp;removed&nbsp;any&nbsp;PRRC&nbsp;or&nbsp;thread&nbsp;whose&nbsp;size&nbsp;exceeded&nbsp;128&nbsp;Kilobytes.</p>