Topic Modeling using BERT
Budget: $250 – $750 USD
I have two datasets containing approximately 28,000 comments in total on Excel sheets. The average length of the comments is 282 characters. I need to classify the comments into 3, 4, and 5 topics using Bidirectional Encoder Representations from Transformers (BERT). That is, you need to classify each comment into 3 topics, then into 4 topics, and then 5 topics using BERT.
First, I'll provide you the smaller dataset with 2,879 comments. You can use the dataset to create the topics and share the results with me, including the code. Once I verify the code and output, I'll provide you the second dataset containing 25,391 comments. The comments in both datasets are extracted from the same source, so they should be of similar topics.
Deliverables:
(i) An Excel file containing the topic label of each comment (e.g., Topic 1, Topic 2) and their probabilities. I have provided a template for the output file; (ii) 10 keywords characterizing each topic; (iii) The topic modeling code.
First, I'll provide you the smaller dataset with 2,879 comments. You can use the dataset to create the topics and share the results with me, including the code. Once I verify the code and output, I'll provide you the second dataset containing 25,391 comments. The comments in both datasets are extracted from the same source, so they should be of similar topics.
Deliverables:
(i) An Excel file containing the topic label of each comment (e.g., Topic 1, Topic 2) and their probabilities. I have provided a template for the output file; (ii) 10 keywords characterizing each topic; (iii) The topic modeling code.