MVA - Algorithms for Speech and NLP Lab 2 (NLP)

Source Code for the second lab of the NLP MVA master's course. (E. Dupoux, B. Sagot)

Requirements :

To firstly run the code, you need to install the requirements:

pip install -r requirements.txt

Use the System :

(NB : Multiprocessing doesn't work on Windows based system, please use n_jobs = 1)

Results on the test split :

To get results on the test split (last 10%) of sequoia-corpus. run :

bash run.sh

This will run the script with only one processor. For multiprocessing, you can use the arg n_jobs :

to run with 12 processors for example, run :

bash run.sh --n_jobs 12

Results on new sentences :

To do so, you need to specify that the script runs in test mode, specify the input file and the name of the output file:

To use 12 processors, on a sentence file named test_sentences.txt, and train on all the sequoia-corpus for example, you can run:

bash run.sh --n_jobs 12 --train-frac 1.0 --test-mode --input test_sentences.txt

Description of all arguments :

• data-file : path to the parse data file.

• train-frac : the train percentage, (default: 0.9).

• emb-file : path to the pickled word embedding for the oov module.

• lev-cands : number of levenstein candidates to search for (default: 2).

• emb-cands : number of embedding candidates to search for (default: 20).

• alpha : coefficient for the bigram linear interpolation (default: 0.8).

• test-mode : call this argument if you want to test on new sentences, if not, the default behavior is to train on a fraction of the data and test on the rest.

• input : path to the test sentences if test-mode is True.

• output : the path to the result file. if test_mode is true : it will store the parse of the test sentences, else, it will store the results on the test split.

• n_jobs : Number of processors to use, -1 means use all processors, in Windows, multiprocessing doesn't work, go for n_jobs = 1.