Open SwiftWinds opened 2 years ago
precision recall f1-score support
geo 0.82 0.86 0.84 1531
gpe 0.83 0.89 0.86 652
org 0.55 0.58 0.56 803
per 0.71 0.75 0.73 728
tim 0.81 0.78 0.79 849
micro avg 0.75 0.78 0.77 4563
macro avg 0.74 0.77 0.76 4563
weighted avg 0.75 0.78 0.77 4563
Time Elapsed: 5784.2341272830 seconds
Time Elapsed: 96.403902121384 minutes
Time Elapsed: 1.6067317020230 hours
Following a tutorial, CPU-only -- no GPU. CPU @ Max 1.80 GHz Results are using 10% of someone else's dataset, not as dirty as our dataset though (theirs is very clean). Download the dataset I tested with: https://www.kaggle.com/namanj27/ner-dataset
Find and read research papers on named entity recognition of products (or named entity recognition of simply recommendations in general <- it would be amazing if you could find papers on this, but there might not be any such papers; it seems pretty hard to do) from online forums and discuss with team the methods used and which might be best to use (e.g., pros and cons of each one)