notebook.community

Edit and run 





In [48]:



    


import numpy as np
import pandas as pd



    











In [49]:



    


inference_directory = r'C:\mallet\clinton\data\inference'



    











In [50]:



    


elda_df = pd.read_csv(r'{}\estimates.csv'.format(inference_directory))
elda_df.columns = [ 'doc#', 'chibs', 'hm', 'is', 'lengths', 'lr' ]



    











In [51]:



    


elda_df.loc[:, [ 'doc#', 'chibs', 'hm', 'is', 'lr', 'lengths' ]]



    


















    
Out[51]:











  
    

      
      doc#
      chibs
      hm
      is
      lr
      lengths
    

  
  
    

      0
      -1
      -3308.066350
      -4057.316653
      -4833.843998
      -6979.759686
      526.0
    

    

      1
      112
      -2194.937182
      -2856.354385
      -3203.586426
      -3753.796858
      351.0
    

    

      2
      116
      -7483.372249
      -10951.847889
      -12795.477996
      -79225.555694
      1403.0
    

    

      3
      120
      -48.490558
      -55.460270
      -16.491044
      -24.962511
      5.0
    

    

      4
      143
      -13447.325183
      -18445.449022
      -21766.299243
      -44525.262588
      2321.0
    

    

      5
      145
      -97.303550
      -188.407712
      -145.152475
      -116.830846
      20.0
    

    

      6
      163
      -172.382007
      -270.985854
      -204.684733
      -227.548619
      28.0
    

    

      7
      165
      -30.017762
      -85.502313
      -41.103530
      -59.482084
      8.0
    

    

      8
      172
      -56.021985
      -76.144785
      -34.592093
      -52.393771
      8.0
    

    

      9
      191
      -4106.326833
      -6150.510195
      -6433.514249
      -8551.285793
      706.0
    

    

      10
      192
      -2803.084441
      -4094.311340
      -4445.139750
      -5262.427147
      497.0
    

    

      11
      200
      -14.914917
      -86.151111
      -35.155587
      -39.206374
      8.0
    

    

      12
      213
      -892.787334
      -1420.265081
      -1329.312324
      -1599.878574
      158.0
    

    

      13
      225
      -244.775343
      -347.489823
      -324.899533
      -3370.872687
      53.0
    

    

      14
      226
      -125.339999
      -177.194220
      -118.787917
      -142.395256
      18.0
    

    

      15
      24
      -122.562321
      -202.027510
      -121.503106
      -95.786155
      18.0
    

    

      16
      25
      -59.670483
      -141.808724
      -98.170743
      -75.758010
      15.0
    

    

      17
      30
      -1512.635409
      -2211.316185
      -2311.815227
      -3358.740777
      250.0
    

    

      18
      38
      -6768.717885
      -9160.411167
      -11088.509753
      -20538.781602
      1244.0
    

    

      19
      39
      -57.926385
      -75.653664
      -34.567976
      -47.054955
      7.0
    

    

      20
      66
      -19.478866
      -20.253009
      0.274041
      -15.711790
      2.0
    

    

      21
      72
      -265.247270
      -412.648026
      -326.038144
      -353.309789
      45.0
    

    

      22
      81
      -20275.023190
      -28901.223720
      -33727.429473
      -93511.181363
      3643.0
    

    

      23
      87
      -25307.518361
      -34634.673918
      -41478.053965
      -167038.102826
      4485.0
    

    

      24
      94
      -730.092663
      -1163.826986
      -1079.985824
      -1265.651605
      125.0
    

  




















In [52]:



    


## perplexity, per document, probably not usable ...



    











In [53]:



    


def calc_perplexity(estimate, tokens):
    assert len(estimate) == len(tokens)
    
    output = []
    for i in range(0, len(estimate)):
        output.append(
            np.exp(-1 * (estimate[i] / tokens[i])))
    
    return output



    











In [54]:



    


tokens = elda_df.lengths.values.tolist()
elda_df['chibs_perplexity'] = calc_perplexity(elda_df.chibs.values.tolist(), tokens)
elda_df['hm_perplexity'] = calc_perplexity(elda_df.hm.values.tolist(), tokens) 
elda_df['is_perplexity'] = calc_perplexity(elda_df['is'].values.tolist(), tokens) 
elda_df['lr_perplexity'] = calc_perplexity(elda_df.lr.values.tolist(), tokens)

elda_df.loc[:, ['doc#', 'lengths', 'chibs', 'chibs_perplexity', 'hm', 'hm_perplexity', 'is', 'is_perplexity', 'lr', 'lr_perplexity']]



    


















    
Out[54]:











  
    

      
      doc#
      lengths
      chibs
      chibs_perplexity
      hm
      hm_perplexity
      is
      is_perplexity
      lr
      lr_perplexity
    

  
  
    

      0
      -1
      526.0
      -3308.066350
      538.668054
      -4057.316653
      2238.429445
      -4833.843998
      9796.862762
      -6979.759686
      5.792591e+05
    

    

      1
      112
      351.0
      -2194.937182
      519.767858
      -2856.354385
      3421.251781
      -3203.586426
      9200.626470
      -3753.796858
      4.411601e+04
    

    

      2
      116
      1403.0
      -7483.372249
      207.231441
      -10951.847889
      2455.341915
      -12795.477996
      9136.970030
      -79225.555694
      3.342223e+24
    

    

      3
      120
      5.0
      -48.490558
      16286.820873
      -55.460270
      65647.445166
      -16.491044
      27.064119
      -24.962511
      1.473045e+02
    

    

      4
      143
      2321.0
      -13447.325183
      328.246071
      -18445.449022
      2827.643792
      -21766.299243
      11825.142942
      -44525.262588
      2.144645e+08
    

    

      5
      145
      20.0
      -97.303550
      129.693954
      -188.407712
      12337.338630
      -145.152475
      1418.880901
      -116.830846
      3.443100e+02
    

    

      6
      163
      28.0
      -172.382007
      471.774088
      -270.985854
      15963.596684
      -204.684733
      1495.429945
      -227.548619
      3.383738e+03
    

    

      7
      165
      8.0
      -30.017762
      42.615594
      -85.502313
      43817.525298
      -41.103530
      170.364667
      -59.482084
      1.694699e+03
    

    

      8
      172
      8.0
      -56.021985
      1099.651030
      -76.144785
      13603.713839
      -34.592093
      75.490860
      -52.393771
      6.987000e+02
    

    

      9
      191
      706.0
      -4106.326833
      335.736612
      -6150.510195
      6073.988880
      -6433.514249
      9069.082812
      -8551.285793
      1.820985e+05
    

    

      10
      192
      497.0
      -2803.084441
      281.465233
      -4094.311340
      3782.161637
      -4445.139750
      7661.347491
      -5262.427147
      3.967135e+04
    

    

      11
      200
      8.0
      -14.914917
      6.451836
      -86.151111
      47519.189849
      -35.155587
      80.999937
      -39.206374
      1.343968e+02
    

    

      12
      213
      158.0
      -892.787334
      284.448650
      -1420.265081
      8014.594725
      -1329.312324
      4506.919393
      -1599.878574
      2.497957e+04
    

    

      13
      225
      53.0
      -244.775343
      101.332044
      -347.489823
      703.741963
      -324.899533
      459.518807
      -3370.872687
      4.185281e+27
    

    

      14
      226
      18.0
      -125.339999
      1057.151499
      -177.194220
      18847.269391
      -118.787917
      734.601908
      -142.395256
      2.726700e+03
    

    

      15
      24
      18.0
      -122.562321
      905.980562
      -202.027510
      74888.120000
      -121.503106
      854.206128
      -95.786155
      2.046811e+02
    

    

      16
      25
      15.0
      -59.670483
      53.411827
      -141.808724
      12758.014108
      -98.170743
      695.559233
      -75.758010
      1.561058e+02
    

    

      17
      30
      250.0
      -1512.635409
      424.342807
      -2211.316185
      6941.441494
      -2311.815227
      10376.105652
      -3358.740777
      6.834871e+05
    

    

      18
      38
      1244.0
      -6768.717885
      230.693859
      -9160.411167
      1577.623014
      -11088.509753
      7432.317861
      -20538.781602
      1.480203e+07
    

    

      19
      39
      7.0
      -57.926385
      3925.299203
      -75.653664
      49398.052412
      -34.567976
      139.530375
      -47.054955
      8.305901e+02
    

    

      20
      66
      2.0
      -19.478866
      16973.911126
      -20.253009
      24996.836579
      0.274041
      0.871952
      -15.711790
      2.580903e+03
    

    

      21
      72
      45.0
      -265.247270
      362.993083
      -412.648026
      9604.203408
      -326.038144
      1401.491198
      -353.309789
      2.569146e+03
    

    

      22
      81
      3643.0
      -20275.023190
      261.249107
      -28901.223720
      2788.776211
      -33727.429473
      10489.677809
      -93511.181363
      1.405360e+11
    

    

      23
      87
      4485.0
      -25307.518361
      282.224252
      -34634.673918
      2258.227114
      -41478.053965
      10385.569436
      -167038.102826
      1.495354e+16
    

    

      24
      94
      125.0
      -730.092663
      344.034281
      -1163.826986
      11054.754525
      -1079.985824
      5652.688726
      -1265.651605
      2.496457e+04
    

  




















In [ ]:

Content source: pepper-johnson/Erudition

Similar notebooks:

notebook.community | gallery | about