Get morphy from sentences using LSTM

Example of morphy.

Table of Contents

• Table of Contents
• Data
• Requirements
• Launch the code
  • Mode
  • Example
• Model
  • POS prediction
    • GET_POS
  • MORPHY prediction
    • SUPERTAG
    • SEPARATE
    • FUSION
• Results

Data

• find the data here: http://hdl.handle.net/11234/1-5287

• Download ud-treebanks-v2.13.tgz

• use tar -xvzf ud-treebanks-v2.13.tgz to unstack the tgz file

• and move data in order have a path like:

  .
  ├── ...
  ├── data
  │ ├── UD_Abaza-ATD
  │ │ ├── abd_atb-ud-test.conllu
  │ │ ├── abd_atb-ud-test.txt
  │ │ └── ...
  │ ├── UD_Afrikaans-AfriBooms
  │ │ ├── af_afribooms-ud-dev.conllu
  │ │ ├── af_afribooms-ud-dev.txt
  │ │ ├── af_afribooms-ud-test.conllu
  │ │ ├── af_afribooms-ud-test.txt
  │ │ ├── af_afribooms-ud-train.conllu
  │ │ ├── af_afribooms-ud-train.txt
  │ │ └── ...
  │ ├── ...
  │ └── ...

Requirements

To run the code you need python (We use python 3.9.13) and packages that is indicate in requirements.txt. You can run the following code to install all packages in the correct versions:

pip install -r requirements.txt

Launch the code

The main.py script is the main entry point for this project. It accepts several command-line arguments to control its behavior:

• --mode or -m: This option allows you to choose a mode between 'train', 'baseline', 'test', and 'infer'.
• --config_path or -c: This option allows you to specify the path to the configuration file for training. The default is config/config.yaml.
• --path or -p: This option allows you to specify the experiment path for testing, prediction, or generation.
• --task or -t: This option allows you to specify the task for the model. This will overwrite the task specified in the configuration file for training.

Mode

Here's what each mode does:

• train: Trains a model using the configuration specified in the --config_path and the task specified in --task.
• baseline: Runs the baseline benchmark test.
• test: Tests the model specified in the --path. You must specify a path.
• infer: Runs inference using the model specified in the --path It will run inference on some exemple sentences located in "infer/infer.txt" and put the results in "infer/configname_infer.txt". If the path is 'baseline', it will run the baseline inference. You must specify a path.

Example

Here's an example of how to use the script to train a model:

python main.py --mode train --config_path config/config.yaml --task get_pos

This command will train a model using the configuration specified in config/config.yaml with a task=get_pos.

Here's an example of how to run a test on the experiment separete:

python main.py --mode test --path logs/separete

Here is an exemple of how to run inference using the baseline model:

python main.py --mode infer --path baseline

Model

POS prediction

input shape: $B \times K$ output shape: $B \times K \times 19$ where $B$ is batch size, $K$ sequence size and $19$ the number of POS classes

GET_POS

MORPHY prediction

input shape: $B \times K$ output shape: $B \times K \times 28 \times 13$ where $B$ is batch size, $K$ sequence size and $28$ the number of MORPHY classes and $13$ the maximun of number possibilities of one morphy.

SUPERTAG

SEPARATE

FUSION

Results

model	crossentropy	accuracy micro	accuracy macro
GET_POS	0.204	0.944	0.816

Table 1: Test results for pos prediction

model	crossentropy	accuracy micro	all good
BASELINE	-	0.980	0.791
SUPERTAG	1.700	0.436	0.002
SEPARATE	1.70	0.893	0.046
FUSION	1.698	0.884	0.154

Table 2: Test results for morphy prediction