How do I train my sentiment analysis model with my own data?

[nawneel]

How To train model with training_data . what is command and how do i start using sentiment for analysing my twitter data.please give some insight on this.

[Inteliment_Splu]

hadoop@master:/opt/splunk/etc/apps/sentiment/bin$ $SPLUNK_HOME/bin/python train.py train Demo/ Demo_Demo
Training Directory: ../training_data/Demo/train

loaded 0 reviews from ../training_data/Demo/train
Error: float division by zero

Can anyone help me out in solving this error

[_jgpm_]

this is an answer a long time coming, but you have to point to the right "train".
python has to be declared in the location SPLUNKHOME/bin/python
train has to be declared in the location SPLUNKHOME/etc/apps/sentiment/bin/train.py
demo has to be declared in the location SPLUNKHOME/etc/apps/sentiment/training_data/demo
model name has to be declared as lowercase "demo"

Then it works. It should look something like this:
Training Directory: SPLUNKHOME/etc/apps/sentiment/training_data/demo/train
names = ['SPLUNKHOME/etc/apps/sentiment/training_data/demo/train/-1.txt', SPLUNKHOME/etc/apps/sentiment/training_data/demo/train/1.txt']

+++++4k
++++len data = 9501
corpdir = 'SPLUNKHOME/etc/apps/sentiment/training_data/demo/train'

loaded 9501 reviews from SPLUNKHOME/etc/apps/sentiment/training_data/demo/train
Original Token Count: 5876
After removing rare: 5493
After removing weakly correlated: 156
...

[_jgpm_]

correction, your pwd has to be 'SPLUNKHOME/etc/apps/sentiment/bin' for it to work without errors.

[marcoscala]

Hi! just trained my own set!!!

1. Move into the "$SPLUNK_HOME/etc/apps/sentiment/bin" directory
2. type "$SPLUNK_HOME/bin/python train.py train "root_set_dir" "Model_name"

where
"root_set_dir" is the name ofthe first segment of the dir you created under "sentiment/training_data" containing the train subdir with the sample data.

if you have:
/opt/splunk/etc/apps/sentiment/training_data/my_model/train
you have to run the following command:

"$SPLUNK_HOME/bin/python train.py train my_model/ my_model

Marco Scala

[carasso · Accepted solution]

$ python sentiment/bin/train.py

Usage: train

For example, "python sentiment/bin/train.py mydata mymodel"

(I have a small question in my head if you have to be in the bin directory when you run it)

See the sentiment/training_data directory for sample data and how it's laid out. See the README.

[marcoscala]

David,
you forgot the "train" command after train.py:

"$SPLUNK_HOME/bin/python train.py train "root_set_dir" "Model_name"

it really needs "train" again!

Marco

[carasso]

$ python sentiment/bin/train.py
Usage: train
For example, "python sentiment/bin/train.py mydata mymodel"
(I have a small question in my head if you have to be in the bin directory when you run it)

See the sentiment/training_data directory for sample data and how it's laid out. See the README.