LID

Language IDentification system

• Insipred from Carbajal's Experiment 6 in the modeling chapter

Replicating Results from de Seyssel & Dupoux, 2020

1. Fill out the kaldi_setup/cmd.sh and kaldi_setup/path.sh accordingly with your setup.

2. Create a wavs directory and add link to all wav files from EMIME in data/emime/wavs (On Oberon, you can use the data/emime/wavs_oberon.txt file containing the path of all used wavs to create the directory.

3. Link all files from kaldi_setup and abx into appropriate local folders (this is where the output of your experiments will go). Consider creating this local folders into a place with enough storage.

   mkdir -p local/kaldi_setup
   mkdir -p local/abx
   cd local/kaldi_setup && ln -s ../../kaldi_setup/* . && cd ../..
   cd local/abx && ln -s ../../abx/* . && cd ../..
   

4. Activate Conda Environment. TODO : Give info on env

5. Run local/run_cogsci.sh : cd local/kaldi_setup && local/run_cogsci.sh

6. Run the ABX tests : cd ../abx/ && ./run_by_spk.sh EMIME-controlled

7. Retrieve the scores : `./retrieve_scores.sh EMIME-controlled > EMIME-controlled.byspk.scores.txt

Note : Results are prone to minor changes due to the randomness in computing MFCC features

---

de Seyssel, M. & Dupoux, E. (2020). Does bilingual input hurt? A simulation of language discrimination and clustering using i-vectors. In Proceedings for the Annual Meeting of the Cognitive Science Society 2020