Natural language ticket similarity matching

Hey! I’m working on a project to group similar Helpdesk tickets together. The idea is that when a new ticket is created, the system would recommend the top 5 best matching historical tickets. I have a dataset with the historical ticket title and description that are in natural language.

One option I’m currently studying is using doc2vec to transform them all into embeddings and calculate their distance, but I’m not sure if this will work. How would you recommend approaching this problem? Any advice or learning materials much appreciated!

Data sample:

Title Description
Noise from the engine During startup, there is a high-pitch noise for a few …
Indicator light not working The indicator light for sensor 114 is not showing the …