Kian is the neural network designed to serve Wikidata.
Project description
Kian
================
Kian is the neural network designed to serve Wikidata.
This library in current shape adds statements to items based on categories in Wikis (Wikipedia, Wikisource, etc.)
In order to add statements or parse articles you need to have pywikibot-core installed.
How to run this code for a run on classifying humans based on Persian Wikipedia:
::
python scripts/initiate_model.py -n faHuman -w fawiki -p P31 -v Q5
python scripts/train_model.py -n faHuman
python scripts/evaluate.py -n faHuman #To see AUC and fitness parameters
python scripts/parser.py -lang:fa -newpages:100 -n faHuman
Bottlenecks of speed in Kian are:
1) Loading category links from Wikipedia. Since it caches them training different models from one wiki tends to work better
2) Training the model. Since it's an ANN and ANNs are resource consuming to train, this may take a while but depends on Wiki you are working with.
Dependency injection
--------------------
Kian is usable for any kind of training, you can simply inject the training set and get the result:
.. code-block:: python
>>> from kian import Kian
>>> bot = Kian(training_set=the_training_set)
>>> bot.train()
>>> bot.theta
>>> bot.finalize(path='path_to_save_results')
Authors
=======
Amir Sarabadani:
* Ladsgroup<AT>gmail.com
================
Kian is the neural network designed to serve Wikidata.
This library in current shape adds statements to items based on categories in Wikis (Wikipedia, Wikisource, etc.)
In order to add statements or parse articles you need to have pywikibot-core installed.
How to run this code for a run on classifying humans based on Persian Wikipedia:
::
python scripts/initiate_model.py -n faHuman -w fawiki -p P31 -v Q5
python scripts/train_model.py -n faHuman
python scripts/evaluate.py -n faHuman #To see AUC and fitness parameters
python scripts/parser.py -lang:fa -newpages:100 -n faHuman
Bottlenecks of speed in Kian are:
1) Loading category links from Wikipedia. Since it caches them training different models from one wiki tends to work better
2) Training the model. Since it's an ANN and ANNs are resource consuming to train, this may take a while but depends on Wiki you are working with.
Dependency injection
--------------------
Kian is usable for any kind of training, you can simply inject the training set and get the result:
.. code-block:: python
>>> from kian import Kian
>>> bot = Kian(training_set=the_training_set)
>>> bot.train()
>>> bot.theta
>>> bot.finalize(path='path_to_save_results')
Authors
=======
Amir Sarabadani:
* Ladsgroup<AT>gmail.com
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
kian-0.2.0.tar.gz
(12.7 kB
view details)
Built Distributions
kian-0.2.0-py3.4.egg
(57.9 kB
view details)
kian-0.2.0-py2.7.egg
(56.6 kB
view details)
File details
Details for the file kian-0.2.0.tar.gz
.
File metadata
- Download URL: kian-0.2.0.tar.gz
- Upload date:
- Size: 12.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 2c46450cd9546430ff4282ca79cd786403b58f9ce306c18a592f492c8bfbe879 |
|
MD5 | 7805758143e9a8e0c673ec5a91f9075b |
|
BLAKE2b-256 | 326c9b5c900a738c72336961b6dfe97290f6a777e384aa23f805cb068b337621 |
File details
Details for the file kian-0.2.0-py3.4.egg
.
File metadata
- Download URL: kian-0.2.0-py3.4.egg
- Upload date:
- Size: 57.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7b624b3042f9c2aa31143fb784127b8ea0aa9b44a3ff864938f8a20d73c8eb76 |
|
MD5 | 985e536e200147679061f82be02eec56 |
|
BLAKE2b-256 | a73010aa2c76b0b55cd08d90b00abfd89ad4f5630a56bd5b813eba3e665ade35 |
File details
Details for the file kian-0.2.0-py2.7.egg
.
File metadata
- Download URL: kian-0.2.0-py2.7.egg
- Upload date:
- Size: 56.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8f7f2b09d01d2583701833b28117c5f4021aac6c300ee5651dcef3667cd8b339 |
|
MD5 | 64c5cf4e0c2ddc8b3310b58e0e82812d |
|
BLAKE2b-256 | 6367375ab4dd749bfefeb6af80cbce7975f9f14c4a3101005e754160cab7e35f |