paddy74 / lowletorfeats Goto Github PK

Implement the methods for Jelinek-Mercer smoothing.

• LMIR.jm()
• FeatureCollector option

Implement the methods for Bayesian smoothing using Dirichlet priors.

• LMIR.dir()
• FeatureCollector option

Create the following test cases

• FeatureCollector
• FeatureKey
• Document
• TFIDF
• Okapi
• LMIR

test_main results in an exception std::out_of_range during map call of .at.

Should be covered by exception handling with try/catch statements.

Fix all GCC -Wconversion and -Wtypeconversion warnings.

Look into what feature set is generally best for the preset feature set

A default text analyzer is required in the FeatureCollector class for the private member variable anlyzToTfMapLenPair.

One of the following options might be pursued:

• A very simple in-project analyzer.
• Custom out-of-project analyzer (see paddy74/textalyzer.
  • Requires abstraction of the token type instead of using std::string

Having a lot of strange issues with the utils::strSplit function where it will not work if implemented under the following conditions:

• Defined in .cpp files
• Defined without template
• Defined directly in function using utils::strSplit
• Defined in same file as function using utils::strSplit

It only seems to work only when defined in utils.hpp using template<typename T>.

Receive a Floating point exception (core dumped) error upon attempting to create an instance of the FeatureKey class. This occurs in the constructor when attempting to assign any member variables from the type maps.

This is some issue with calling any operation on one of the unordered_map type maps.

Implement the methods for absolute discounting smoothing.

• LMIR.abs()
• FeatureCollector option