icaruseffect / tt-rss-naive-bayes Goto Github PK

run as daemon

1. for scored and read articles:
   article ----> extract features -----> train classifier
2. for unread articles:
   article ----> extract features -----> classify article ------> give score ------> save score

-word of bags from all articles
-adding articles to list
--processing content to feature

manages articles - refactored

1.) creating specific rows in ttrss table for
a) compressed content
b) trained to filter (True/False)
c) features

Write scores to db

creates a surrogate for an article from the database
memory efficient:

• only takes features passed to the surrogate for reduced memory consumption
• holds the processed score

cannot feed the created tags/words to the bayes filter from nltk project..

-managing bayes filter:
--categorization
--labeling
--training

create recommendations based on k-nearest-neighbor algorithm

make the whole program threaded to speed up the classification of new articles:
change article-processing to a worker based layout
#11

-create bag of words per language
--create trainingset based on articles per language

Create a module to classify with Reuters articles from nltk.corpus

create database manager with functions:
-read article from database
-write article to database

transparent handling of bag of words with following features:
create a table for mysql_manager to handle the bag of words with word count
a) check if table exists
no -> create table
b) add table
c) get bag of words -> write table
d) read table -> write bag of words
e) update table

Add tagging as recommended based on k-nearest-neighbor algorithm (#9)

Happens for read and unread list.
Possible reasons :
Empty field(s)