My company has been using Jira for production issue tracking for last 6~8 years and as a result, there is a huge amount of production issue details logged in our Jira.

Usually each Jira ticket for any production support issues consist of some useful information such as:

• Error Message
• System Involved
• Root Cause
• Resolution
• Time Taken
• etc

My company has its own team chat service that supports the Chatbot API in Java / Python / etc. I would like to build the smart chatbot (if not AI) that is smart enough to exchange conversation like this in the chatroom:

DevOps) Hey Jirabot, what do you know about this error message? [xyzvxc exception occured at line 82.... ]

Jirabot) Hi there, in which systems did this occur? Can you choose from one of the followings?

1. System A
2. System B

DevOps) 1

Jirabot) Right, it looks like following Jira tickets have experienced the similar issues.. please check the following tickets.

1. Jira-12zx
2. Jira-52123zz
3. Jira-vvvbbb

I would like to ask people with experiences in implementing something similar to this or have any relevant experience in ML / Neural Network / Natural Language Processing the following questions.

1. Is this even possible for non-NPL Expert? Do you think it would be possible to build something like this for the software engineer with 8 years of experience with advanced beginner skills in Machine Learning? (Advanced Beginner in Dreyfus skill model)

I work for the company which is known for difficult programming interview questions. And I have about 8 years of extensive programming skills and have completed Andrew Ng's machine learning course honestly and with good marks. I am in the middle of (half way) completing the Deep Learning course in Udacity and from the course I am learning to work with Tensorflow.

2. What frameworks / technologies can I use? If you think it is possible, what frameworks / technologies do you recommend I look for? Or is there any example of something similar if not exactly the same?

3. If not possible, what area should I focus? If you think it is practically NOT possible for a pure developer to build on, then how and what areas can I focus on improving to be able to build one?

Many Thanks guys,

J

It is possible. NLP has various search engines that are capable of supporting it while indexing the Jira DB. Engines such as Solr or LucidWorks are possible candidates (there are others offcourse).

You probably will have to create some extracting scripts of the data so the engine will be able to use Jira as content source.