You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Apr 16, 2025. It is now read-only.
The code starts with cross validation of the data we have. The model is built and MRR is checked. The actual model is to be built of all the data in data/soccer2 folder.
7
-
Changes for integration with karma need to be done in HybridSTModelHandler.java
7
+
Integration with karma is done in HybridSTModelHandler.java
8
8
The DSL_main.predictSemanticType() function needs to be called from above file. Model needs to be loaded and predictions will be ranked. Once the ranking is done, check for highest probability value. If that probability is above 0.3, recommend that semantic type. If the probability is below 0.3, do not give any recommendations - treat the incoming data as newly seen data and save it. While saving the data, also check whether the data you already have (data/soccer2) surpasses the amount of data you want to hold on the server. If it does, remove certain set % of data rows from each table and then store the new file.
9
-
Minor changes will be required in terms of importing the module into Karma. Test for compatibility with all the running modules. The model will need to be stored in such a way that it can be imported on local on any desktop.
9
+
Once the model is built it is stored in the resources folder. During run time of karma the model is used directly from the resources folder. No re-training is required.
logger.log(Level.INFO, "Feature Extraction Done ! \n Starting model train !");
151
150
DSL_maindsl_obj = newDSL_main(app.modelFilename, featureExtractorObject, true, true, false); // To re-train the model pass the value of load the model as false.
0 commit comments