search

SpectData / MONAH

Multi-Modal Narratives for Humans
MIT License
3 stars 0 forks source link

Bug - M/F in Talkturn No. Column #22

Closed joshkyh closed 3 years ago

joshkyh commented 3 years ago

Hi Marriane,

Suppose I sidestepped the issue in the #21 ; in tfp.combine_prosodyfeatures, I ran into another problem at dfr = pd.merge(dfr, df, how='outer', on=['audio_id', 'speaker', 'talkturn no'])

See below screenshot image

Why do I see M/F in talkturn number when it should be integers..?

joshkyh commented 3 years ago

df_vokaturi in line #76 of talktun_family_prosody has broken talkturn no column.

This is due to bad talkturn_vokaturi.csv which is created in line #123 of execute_vokaturi.py

We suspect that it is due to the hardcoding of underscore positions, whihc is risky given that we don't know what underscore naming convention will be in user supplied files.

mfmakahiya commented 3 years ago

Hi Joshua! I have fixed the issue with the talkturn number without hardcoding underscore positions.

joshkyh commented 3 years ago

That is great @mfmakahiya , could you please follow through to try run lines 183 and 184 of https://github.com/SpectData/MONAH/blob/dev-pitch/Python/Data_Exploration/rescore_all_iemocap_pitchvol.py to see that there is no further error?

mfmakahiya commented 3 years ago

Hi Joshua! I'm just having some difficulty with the azure storage blob. I can't seem to import the BlobServiceClient package.

joshkyh commented 3 years ago

Let's migrate to Linux, please recreate the MONAH environment using Linux.yml. That would save us trouble in the long run.

joshkyh commented 3 years ago

Done with 25a40e4