Create a multi-label text classification model to predict the subjects of fictions(literary works) based on their summaries
Budget: $10 – $30 USD
Work Done So Far:
I completed the whole codes in Colab notebook and two files necessary to run the code (data.csv, test.csv).
1. data.csv:
- This file contains the dataset used for training and testing the model.
- It consists of approximately 12,000 bibliographic records with titles, publication years, summaries, and tagged themes.
- I have limited the themes to the following 11 categories: domestic fiction, psychological fiction, religious fiction, etc.
- Each work in the dataset can have one or more of these 11 labels or, in some cases, none at all.
2. test.csv:
- This file contains summaries of novels that I want to classify using the trained model.
- Ultimately, I aim to label each work with the appropriate themes.
Issues (Help Needed): Labeling Accuracy and Coverage:
- The major issue is that more than half of the results in 'the test.csv' file are not labeled at all. I want at least 95% of the fictions to be accurately predicted within the 11 subejct labels.
- Additionally, could you add an "Other" category for works that do not fit any of the 11 labels?
- How can I improve the accuracy of classification model while increasing the proportion of labeled data?
I completed the whole codes in Colab notebook and two files necessary to run the code (data.csv, test.csv).
1. data.csv:
- This file contains the dataset used for training and testing the model.
- It consists of approximately 12,000 bibliographic records with titles, publication years, summaries, and tagged themes.
- I have limited the themes to the following 11 categories: domestic fiction, psychological fiction, religious fiction, etc.
- Each work in the dataset can have one or more of these 11 labels or, in some cases, none at all.
2. test.csv:
- This file contains summaries of novels that I want to classify using the trained model.
- Ultimately, I aim to label each work with the appropriate themes.
Issues (Help Needed): Labeling Accuracy and Coverage:
- The major issue is that more than half of the results in 'the test.csv' file are not labeled at all. I want at least 95% of the fictions to be accurately predicted within the 11 subejct labels.
- Additionally, could you add an "Other" category for works that do not fit any of the 11 labels?
- How can I improve the accuracy of classification model while increasing the proportion of labeled data?