search

iangow / se_features

Linguistic features derived from StreetEvents
1 stars 3 forks source link

Compare `liwc_orig` and `liwc_alt` for random texts #36

Closed Yvonne-Han closed 4 years ago

Yvonne-Han commented 4 years ago

So far, liwc_orig and liwc_alt produce the same results (except for total word count, as mentioned in #15) for the text displayed here.

The task here is to randomly select several utterances (e.g., file_name, section, context, speaker_number, etc.) (say, 10?) and verify whether liwc_orig and liwc_alt always behave in the same way.

Yvonne-Han commented 4 years ago

10 randomly selected utterances for testing:

image

Yvonne-Han commented 4 years ago

Confirmed that liwc_orig and liwc_alt produces very similar results for random texts. (For those very rare exceptions - I think it is actually caused by LIWC's inconsistencies with their documentation so I'm not able to fix them at this stage).