Score Tracker

This page describes *ScoreTracker classes.

class artm.score_tracker.SparsityPhiScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of Phi sparsity.
  • zero_tokens - number of zero rows in Phi.
  • total_tokens - number of all rows in Phi.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.SparsityThetaScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of Theta sparsity.
  • zero_topics - number of zero rows in Theta.
  • total_topics - number of all rows in Theta.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.PerplexityScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of perplexity.
  • raw - raw values in formula for perplexity (in case of one class id).
  • normalizer - normalizer values in formula for perplexity (in case of one class id).
  • zero_tokens - number of zero p(w|d) = sum_t p(w|t) p(t|d) (in case of one class id).
  • transaction_typename_info - array of structures, each structure contains raw, normalizer zero_tokens and transaction_typename name (in case of several transaction types)Note, that in the case of non-transaction model transaction type is equal @default_transaction.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.TopTokensScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • num_tokens - number of requested top tokens.
  • coherence - each element of list is a dict, key - topic name, value - topic coherence counted using top-tokens
  • average_coherence - average coherencies of all scored topics.
  • tokens - each element of list is a dict, key - topic name, value - list of top-tokens
  • weights - each element of list is a dict, key - topic name, value - list of weights of corresponding top-tokens (weight of token == p(w|t))
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.TopicKernelScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • tokens - each element of list is a dict, key - topic name, value - list of kernel tokens
  • size - each element of list is a dict, key - topic name, value - kernel size
  • contrast - each element of list is a dict, key - topic name, value - kernel contrast
  • purity - each element of list is a dict, key - topic name, value - kernel purity
  • coherence - each element of list is a dict, key - topic name, value - topic coherence counted using kernel tokens
  • average_size - average kernel size of all scored topics.
  • average_contrast - average kernel contrast of all scored topics.
  • average_purity - average kernel purity of all scored topics.
  • average_coherence - average coherencies of all scored topics.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.ItemsProcessedScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - numbers of processed documents.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.ThetaSnippetScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • document_ids - each element of list is a list of ids of returned documents.
  • snippet - each element of list is a dict, key - doc id, value - list with corresponding p(t|d) values.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.TopicMassPhiScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of ratio of sum_t n_t of scored topics.and all topics
  • topic_mass - each value is a dict, key - topic name, value - topic mass n_t
  • topic_ratio - each value is a dict, key - topic name, value - topic ratio
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.ClassPrecisionScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of ratio of correct predictions.
  • error - numbers of error predictiona.
  • total - numbers of all predictions.
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.
class artm.score_tracker.BackgroundTokensRatioScoreTracker(score)
__init__(score)
Properties:
  • Note: every field is a list of info about score on all synchronizations.
  • value - values of part of background tokens.
  • tokens - each element of list is a lists of background tokens (can be acceced if ‘save_tokens’ was True)
  • Note: every field has a version with prefix ‘last_’, means retrieving only info about the last synchronization.