Gianfranco's Second 🧠

    Various Results

    Various results

    Created: 2022-08-17 16:09
    #note

    Easytour

    LDA

    Raw data: (LDA does not deal with stop words so we have to preproecss data)
    Number of topics ->
    Diversity ->
    NPMI ->
    UCI ->
    UMASS ->
    C_V ->

    Processed data:
    Number of topics -> 20
    Diversity -> 0.088
    NPMI -> -0.004
    UCI -> -0.06
    UMASS -> -1.65
    C_V -> 0.39

    NMF

    Raw data:
    Number of topics -> 20
    Diversity -> 0.896
    NPMI -> -0.09
    UCI -> -3.89
    UMASS -> -3.92
    C_V -> 0.54

    Processed data:
    Number of topics -> 20
    Diversity -> 0.79
    NPMI -> 0.03
    UCI -> -1.20
    UMASS -> -3.79
    C_V -> 0.55

    Top2Vec

    Raw data:
    Number of topics -> 47
    Diversity -> 0.30
    NPMI -> 0.21
    UCI -> -6.26
    UMASS -> -5.13
    C_V -> 0.37

    Processed data:
    Number of topics -> 53
    Diversity -> 0.31
    NPMI -> 0.04
    UCI -> 0.19
    UMASS -> -1.63
    C_V -> 0.46

    BERTopic (with sentence transformers)

    Raw data:
    Number of topics -> 64
    Diversity -> 0.199
    NPMI -> -0.06
    UCI -> -1.87
    UMASS -> -3.16
    C_V -> 0.37

    Processed data:
    Number of topics -> 64
    Diversity -> 0.299
    NPMI -> -0.04
    UCI -> -3.29
    UMASS -> -7.19
    C_V -> 0.47

    BERTopic (with plain BERT)

    Raw data:
    Number of topics -> 8
    Diversity -> 0.12
    NPMI -> -0.03
    UCI -> -0.20
    UMASS -> -0.44
    C_V -> 0.35

    Processed data:
    Number of topics ->
    Diversity ->
    NPMI ->
    UCI ->
    UMASS ->
    C_V ->

    RoBERTa

    Raw data:
    Number of topics -> 36
    Diversity -> 0.10
    NPMI -> -0.03
    UCI -> -0.49
    UMASS -> -0.98
    C_V -> 0.33

    Processed data:
    Number of topics -> 43
    Diversity -> 0.26
    NPMI -> 0.05
    UCI -> -0.73
    UMASS -> -3.41
    C_V -> 0.54

    CTM

    Raw data:
    Number of topics -> 20
    Diversity -> 0.24
    NPMI -> -0.17
    UCI -> -5.17
    UMASS -> -5.87
    C_V -> 0.41

    Processed data:
    Number of topics -> 20
    Diversity -> 0.25
    NPMI -> -0.005
    UCI -> -1.69
    UMASS -> -4.33
    C_V -> 0.59

    ETM (300 epochs, embeddings learnt by the model and 20 topics)

    Raw data:
    Number of topics -> 20
    Diversity -> 0.27
    NPMI -> -0.001
    UCI -> -0.06
    UMASS -> -1.06
    C_V -> 0.32

    Processed data:
    Number of topics -> 20
    Diversity -> 0.442
    NPMI -> 0.05
    UCI -> 0.34
    UMASS -> -1.98
    C_V -> 0.50

    Tourpedia

    LDA

    Raw data: (LDA does not deal with stop words so we have to preprocess data)
    Number of topics ->
    Diversity ->
    NPMI ->
    UCI ->
    UMASS ->
    C_V ->

    Processed data:
    Number of topics -> 20
    Diversity -> 0.17
    NPMI -> -0.07
    UCI -> -2.12
    UMASS -> -4.84
    C_V -> 0.46

    NMF

    Raw data:
    Number of topics -> 20
    Diversity -> 0.632
    NPMI -> -0.20
    UCI -> -6.35
    UMASS -> -9.20
    C_V -> 0.45

    Processed data:
    Number of topics -> 20
    Diversity -> 0.52
    NPMI -> -0.18
    UCI -> -5.95
    UMASS -> -8.95
    C_V -> 0.41

    Top2Vec

    Raw data:
    Number of topics -> 122
    Diversity -> 0.30
    NPMI -> -0.30
    UCI -> -8.51
    UMASS -> -10.98
    C_V -> 0.47

    Processed data:
    Number of topics -> 101
    Diversity -> 0.26
    NPMI -> -0.39
    UCI -> -1.82
    UMASS -> -2.30
    C_V -> 0.40

    BERTopic (sentence transformers)

    Raw data:
    Number of topics -> 141
    Diversity -> 0.32
    NPMI -> -0.22
    UCI -> -7.49
    UMASS -> -13.18
    C_V -> 0.35

    Processed data:
    Number of topics -> 149
    Diversity -> 0.32
    NPMI -> nan
    UCI -> nan
    UMASS -> nan
    C_V -> 0.37

    BERTopic (plain BERT)

    Raw data:
    Number of topics -> 49
    Diversity -> 0.24
    NPMI -> -0.13
    UCI -> -4.52
    UMASS -> -8.89
    C_V -> 0.40

    Processed data:
    Number of topics ->
    Diversity ->
    NPMI ->
    UCI ->
    UMASS ->
    C_V ->

    RoBERTa

    Raw data:
    Number of topics -> 16
    Diversity -> 0.29
    NPMI -> -0.18
    UCI -> -5.52
    UMASS -> -9.97
    C_V -> 0.37

    Processed data:
    Number of topics -> 59
    Diversity -> 0.32
    NPMI -> -0.17
    UCI -> -7.01
    UMASS -> -13.93
    C_V -> 0.40

    CTM

    Raw data:
    Number of topics -> 20
    Diversity -> 0.20
    NPMI -> -0.21
    UCI -> -6.11
    UMASS -> -8.82
    C_V -> 0.49

    Processed data:
    Number of topics -> 20
    Diversity -> 0.24
    NPMI -> -0.27
    UCI -> -8.05
    UMASS -> -12.22
    C_V -> 0.51

    ETM

    Raw data:
    Number of topics -> 20
    Diversity -> 0.16
    NPMI -> -0.004
    UCI -> -0.08
    UMASS -> -1.96
    C_V -> 0.46

    Processed data:
    Number of topics -> 20
    Diversity -> 0.18
    NPMI -> -0.0005
    UCI -> -0.64
    UMASS -> -3.51
    C_V -> 0.52

    References

    Code

    Tags