PRIV F41 Content-Based Feeds

Explanation

Video and pics

Content-based Filtering

• Text-filter for feedWP
• “Throw away” posts that do not contain keywords
• GUI to enter that are accepted
• Target Words: Definable List: Example (“SNB”, “Swiss”)
• when word “SNB” “Swiss National Bank” is found then load post
• Occurrences of target words inside the post, Example: 3 target words –> get the post –> Category SNB&CHF
• Example Author: (“Pieter Pulitz”): 1 Target Word in the list –> get the post –> Category “Pieter Pulitz”
• One rule per category

Google search part of GUI

- Click to enlarge

Solution Architecture

• Performance: Reading posts in the feed is no Performance issue
• Reason: FeedWP has problems with writing to DB.
• But reading post by posts only requires RAM, we have enough RAM on the server
• Cron every 30 minutes, only Zerohedge
• Case insensitive “Swiss National Bank”, “swiss National Bank”, “snb”, “SNb”

Limitations

Reloading posts

We cannot reload old posts. The filter must be configured correctly when we load.

Example: Zerohedge Google Search

 Add post only when Keyword matches

We introduce 3 new parameters:

• Searcgh in Comments
• Search in Channel ID
• Search in Channel Name
• Search in Content
• Minimum Views

They are used for Videos filtering only.

Current Functionality not entirely described

Current text: Add post in site and assign to category when keywords match

new Text: Add post to Site and Assign Category only when keywords match. Otherwise do not load