MLTS
Machine Learning Toolkit for SEO
Initial demo notebook here
What are the problems/needs?
What are the particular problems in the community that could be solved via machine learning.
- Generating better titles.
- Generating descriptions for pages. Summarization.
- Generating alt text from images.
- Need to get from the community.
- Create a Twitterbot
What is the overall flow?
- Data Getting
- Data Cleaning and Feature Extraction
- Iteration and Updating
- Optimization
- Models (train / predict)
Roles
- Developing Use Cases
- Evangelism / Community
- Analytics (per Britney: Analytics)
- Coding
- Tutorials
- Documentation / Readability
- Unit Tests / Linting
- Design
Data needs
- Link data
- Analytics
- Scraping
- Ranking data
- Anonymous performance data
Proposed Structure
Most folders include a Todo.txt with some suggested items to start with.
- APIs: Holds glue for various SEO APIs
- Data: Holds datagetter classes for APIs and hosted datasets.
- Docs: Holds the documentation for the repo.
- Models: Holds various models that can be used to train on.
- NPL: Glue for NLP libraries
- Testing: Unit testing and CI
- Tutorials: Holds iPython tutorials in Pytorch and Tensorflow
- Config.py: Holds API keys and configuration data.
- Main.py: The main application file.
- requirements.txt: Python libraries needed to install via Pip.
Original concept gist: (source)