kuhumcst/affixtrain

Using supervised learning, create a set of affix rules for use by the CSTlemma lemmatiser.

affixtrain's Issues

Missing 0/flexrules.short_XC and "Cannot open file training_data_V/t for writing" error

Dear @BartJongejan, thank you for your work on Affixtrain and for making it available on Github. I would like to try it with Sumerian, but I have stumbled upon the following issues. The Affixtrain I am using is a Windows executable I built yesterday with your code.

1. The program runs on the example data, but flexrules.short_XC is not being created in /0/ (contains the files flexrules.short_A_XC, flexrules.short_ADV_XC, flexrules.short_N_XC, flexrules.short_NUM_XC, and flexrules.short_V_XC) or elsewhere.

2. When I am trying to test Affixtrain on Sumerian data (packed in a file with the same structure as short in the example folder), it first seemingly runs as expected, creating the numbered folders and a plenty of files. However, unlike when it runs with the example data, a lot of messages like
   !f1 (output tmp\out--_XC-N:DN_affixsingleshotincSize_5_9856_1.txt) PRESS ENTER
   appear. The program stops with the following message:
   Open files: 1721
Single tag input file: Cannot open file training_data_V/t for writing (wb). Exiting.
   Just as in 1, The expected flexrules.training_data_XC is also nowhere to be found.

I'd be very grateful if you could tell me what might be the cause of those, what I could do wrong and how to fix it. Many thanks in advance!

Segmentation fault

Dear @BartJongejan, thank you for your work. I wanted to create flex rules for cstlemma, but I run into problem when "./affixtrain -i short -n FLOO" ends with segmentation fault.

Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x000000000041354d in node::cleanup (this=0x0, parent=0x0, recursionDepth=0) at node.cpp:770
770 if (this->IfPatternSucceeds)

Thank you very much.