0
I Use This!
Very Low Activity

Commits : Listings

Analyzed 2 days ago. based on code collected 2 days ago.
Jun 09, 2023 — Jun 09, 2024
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
Handle missing values when exporting to ARFF and we represent nominal values as strings. More... over 6 years ago
Make runtime parameters optional as far as possible to avoid the GUI complaining. More... over 6 years ago
Better name for the ENUM and the class name of the JSON exorter. More... over 6 years ago
Add option to export python data with strings intact. More... over 6 years ago
Implement an exporter for exporting sequences in a JSON format that is readabale by python. This uses a list of one or two elements: the first element is the sequence of feature vectors, the second element is the sequence of targets. The sequence of targets could be missing. The first element is in turn a list, where each element is a list representing the feature vector. The sequence of targets is just a list of target alphabet indices. More... over 6 years ago
Hopefully better representation of START/STOP symbols. Also fixed a small bug where START was used instead of STOP for an ATTRIBUTE. More... almost 7 years ago
Do not create an attribute if the feature4value feature does not exist. More... almost 7 years ago
A little bit towards making MEMM work, perhaps. More... almost 7 years ago
Read Mallet Sequence model correctly. More... almost 7 years ago
Remove unnecessary field; add comment about the trailing slash for URLs indicating directories. More... almost 7 years ago
Better implementation for handling URIs and URLs. More... almost 7 years ago
Fixing #52. Replaced all reading from files by reading from URLs where possible. More... almost 7 years ago
Forgot to commit this important file. More... almost 7 years ago
Restructured code so that the engine is now responsible for choosign and creating the CorpusRepresentation instance. This is now compiling and passes tests. Also renamed some classes to more intuitive names and cleaned up the code a bit. This should now be usable for implementing out-of-memory corpus representations and non-mallet corpus representations. More... almost 7 years ago
Some renaming of classes and methods to use more adequate names in the API. Start implementing the methods and code for making the Engine instances responsible for creating their own CorpusRepresentation and for managing the way how they get initialised after loading. More... almost 7 years ago
Merge branch 'jp170701-seqmulticlass' More... almost 7 years ago
WIP More... almost 7 years ago
Add support for using the internal target field when doing classification. More... almost 7 years ago
Make classification training using a sequence algorithm work. Fix a bug in classification application when using a sequence algorithm. More... almost 7 years ago
In some odd cases, the class needs to get escaped as well. For example, when the class is literally a comma. More... almost 7 years ago
Apparently we need to quote an attribute name if it contains a comma. More... almost 7 years ago
Attempt to handle reverse lookup of the START/STOP features we create. Since these features get created in addition, there is no feature specification so the reverse lookup returns null. For now we just check if we get null because this is a START/STOP feature and in that case do not throw an exception and rather just use the default numeric attribute definition we already have without modifications. More... almost 7 years ago
Remove debug print statement. More... almost 7 years ago
Correct comment. More... almost 7 years ago
Add support for START/STOP features to normal attributes. If the attribute has a within parameter and the start/end offsets are equal then the START/STOP feature is added. More... almost 7 years ago
Add START/STOP features for attribute lists if we have a within type. More... almost 7 years ago
Wordshape: distinguish between lower case and uppercase letters. More... almost 7 years ago
Start on code for word shape features. More... almost 7 years ago
Affix generation seems to work now. More... almost 7 years ago
Add a PR for generating suffix/prefix features. More... almost 7 years ago