Week 8

Meeting 10

(August 4th, 2022)

GSOC 2022 weekly update

Attendees

• Gaurav Mishra
• Shaheem Azmal M MD
• Vivek Kumar
• Kaushlendra Pratap
• Ayush Bhardwaj
• Vasudev Maduri
• Sushant Kumar
• Rohit Pandey
• Feng Wenhan
• Soham Banerjee
• Samuel Dushimimana
• Krishna Mahato

Updates

• Doc2vec Model requires tagged document for training the model, such that it can make a vector for each document that can be further use to calculate cosine similarity for the license statement. Due to lack of memory in my computer I was not able to tag whole dataset. So currently for checking the working of model, this week I have trained a smaller part of dataset on Doc2vec Model.
• Created agent for the same trained model on atarashi and got the accuracy score of 8 percent.

 Total files scanned = 100
 Successfully matched = 8

      ++++++++++++++++++ Result ++++++++++++++++++
      ++++++++++++++++++++++++++++++++++++++++++++
      ---> Total time elapsed: 13.02 Seconds  <---
      ---> Accuracy: 8.0%                     <---
      ++++++++++++++++++++++++++++++++++++++++++++
      ++++++++++++++++++++++++++++++++++++++++++++

• Working of agent is shown below:

{
  "file": "/home/shushant/check.py",
  "results": [
    {
      "description": "",
      "shortname": "BSD-1-Clause",
      "sim_score": 0.5472090244293213,
      "sim_type": "semanticTextSim"
    }
  ]
}

Conclusion and Further Plans

• Trainig has been done on a smaller part of dataset. Will train Doc2vec Model on whole dataset and see if any further improvement is needed or not.
• Will start working on the documentation for final evalutation.