Detext Versions Save

Correct 'feature_names' as a required field https://github.com/linkedin/detext/pull/42 Update smart-arg to 0.1.1 https://github.com/linkedin/detext/pull/44 Added references https://github.com/linkedin/detext/pull/45 https://github.com/linkedin/detext/pull/46

Changes:

• Use external dependency smart-arg instead of the temporary arg_suite solution (#38 )
• Fix bugs related to text processing (#40 )

test automatic release

Currently DeText's design for sparse feature has simple modeling power for sparse features.

1. only linear model is applied on sparse features
2. there's no interaction between sparse features and dense features (model_score = dense_score + sparse_score)

DeText v1.2.0 resolves the above limitation on sparse feature by

1. computing dense representation of sparse features
2. allowing interactions between sparse features and wide features

More specifically, the model architecture changes from

dense_score = dense_ftrs -> MLP
sparse_score = sparse_ftrs -> Linear
final_score = dense_score + sparse_score

to

sparse_emb_ftrs = sparse_ftrs -> Dense(sp_emb_size)
all_ftrs = (dense_ftrs, sparse_emb_ftrs) -> Concatenate
final_score= all_ftrs -> MLP

See https://github.com/linkedin/detext/pull/12 for changes

expose tfrecord dataset transformation function for LinkedIn usage (#10 )