Data handling containers
Pipeline
class Pipeline(exp=None, name=None, out=None)
Shell for experiment pipeline storing and handling.
| Parameters | Type | Doc |
|---|---|---|
| exp | class, optional, default None | Instance of Experimen with fitted pipes. If not supplied, name and source should be set. |
| out | tuple, optional, default None | Tuple with storage options, can be "json" (json serialization),or "db" (for database storage, requires blitzdb). |
Methods
| Function | Doc |
|---|---|
| _make_tab | Tabular level experiment representation. |
| save | Save experiment and classifier in format specified. |
| load | Load experiment and classifier from source specified. |
| classify | Given a data point, return a (label, probability) tuple. |
_make_tab
_make_tab()
Tabular level experiment representation.
Generates a table-level representation of an experiment. This stores JSON native information ONLY, and is used for the experiment table in the front-end, as deserializing a lot of experiments will be expensive in terms of loading times.
save
save()
Save experiment and classifier in format specified.
load
load()
Load experiment and classifier from source specified.
classify
classify(data)
Given a data point, return a (label, probability) tuple.
CSV
class CSV
Quick and dirty csv loader.
| Parameters | Type | Doc |
|---|---|---|
| text | integer | Index integer of the .csv where the text is located. |
| parse | integer, optional, default None | Index integer of the .csv where the annotations are provided. Currently it assumes that these are per instance a list of, for every word, (token, lemma, POS). Frog and spaCy are implemented to provide these for you. |
| header | boolean, optional, default False | If the file has a header, you can skip it by setting this to true. |
Methods
| Function | Doc |
|---|---|
| iter | Standard iter method. |
| next | Iterate through csv file. |
iter
__iter__()
Standard iter method.
next
__next__()
Iterate through csv file.