Hyperdeep is a proof of concept of the deconvolution algorithm applied on text. It use a standard CNN for text classification, and include one layer of deconvolution for the automatic detection of linguistic marks. This software is integrated on the Hyperbase web platform (http://hyperbase.unice.fr) - a web toolbox for textual data analysis.
- python3
- keras + tensorflow
Datas are stored in the data/ folder. The training set should be splited into phrases of fixed length (50 words by default). And each phrase should have a label name at the begining of the line. A label is written : LABELNAME Hyperdeep is distributed with an example of corpus named Campagne2017 (in data/Campagne2017). This is a french corpus of the 2017 presidential election in france. There are 5 main candidates encoded with the labels (Melenchon, Hamon, Macron, Fillon, LePen).
You can train a skipgram model (word2vec) by using the command : $ python hyperdeep.py skipgram -input data/Campagne2017 -output bin/Campagne2017.vec This command will create a file named Campagne2017.vec in the folder bin (create the folder if needed). This txt is the vectors file of the training data.
the skipgram model can be tested by using the command (example with the word France) : $ python hyperdeep.py nn bin/Campagne2017.vec France
To train the classifier: $ python3 hyperdeep.py train -input data/Campagne2017 -output bin/Campagne2017 The command will create bin/Campagne2017.index for the vocabulary and bin/Campagne2017 for the model
Then you can make predictions on new text. There is an example in bin/Campagne2017.test. It's a discourse of E. Macron as french president on 31th december 2017. Hyperdeep will split the discourse in fixed length phrases and should predict most of the phrase a E. Macron $ python hyperdeep.py predict bin/Campagne2017 bin/Campagne2017.vec data/Campagne2017.test
The predict command line will create a result file in the folder result/ (create the folder if needed). This file is a json format file where you can find the activation score given by the deconvolution for each word. An example of results is given in result/Campagne2017.test.res
python3 -m venv .venv
. .venv/bin/activate
python3 -m pip install -U --force-reinstall keras==1.1.4
python3 -m pip install -U --force-reinstall keras==2.0.0
python3 -c "import keras; print(keras.__version__);"
mkdir bin/
python hyperdeep.py skipgram -input data/Campagne2017 -output bin/Campagne2017.vec
python hyperdeep.py nn bin/Campagne2017.vec France
[('rayonne', 0.625952959060669), ('fière', 0.6240432262420654), ('puissante', 0.6158779263496399), ('coexistence', 0.615835428237915), ('gagnante', 0.6157932281494141), ('fraternelle', 0.6106721758842468), ('façonné', 0.6029340028762817), ('traversé', 0.6028076410293579), ('forte', 0.6014634370803833), ('mosaïque', 0.5977412462234497)]
TRAINING CLASSIFIER
python3 hyperdeep.py train -input data/Campagne2017 -output bin/Campagne2017
TEST CLASSIFIER
python hyperdeep.py predict bin/Campagne2017 bin/Campagne2017.vec data/Campagne2017.test
----------------------------
DECONVOLUTION
----------------------------
(3, 128, 1, 512)
(3, 128, 1, 512)
DECONVOLUTION SHAPE : (101, 50, 128, 1)
101
cat results/Campagne2017.test.res | json_pp
[
{
"sentence" : "::11.593507766723633 le:16.891237258911133 choix:47.93613815307617 du:35.58413314819336 peuple:65.93293762207031 français:64.55390930175781 ,:55.15997314453125 votre:42.49549865722656 choix:56.94794845581055 par:34.339454650878906 lequel:29.816349029541016 vous:30.37843132019043 m':26.36418914794922 avez:34.59223175048828 <PAD>:69.55484008789062 votre:45.25927734375 confiance:60.548038482666016 et:36.16371154785156 avec:28.190982818603516 elle:36.694664001464844 votre:40.45203399658203 impatience:94.9781723022461 ,:56.47802734375 vos:33.44449234008789 exigences:32.770057678222656 ,:31.24873924255371 vos:32.61974334716797 attentes:28.31665802001953 ;:101.72959899902344 j':23.938312530517578 en:21.642822265625 suis:23.090648651123047 pleinement:25.462093353271484 conscient:31.470001220703125 .:29.107559204101562 Aussi:54.604217529296875 ,:56.7678337097168 depuis:38.937042236328125 mon:22.68429946899414 élection:45.78916549682617 en:22.658348083496094 mai:34.11579513549805 dernier:46.26791000366211 ,:46.37293243408203 je:28.92486572265625 me:28.759069442749023 suis:36.3964729309082 attaché:44.55897903442383 à:19.251060485839844 simplement:12.999951362609863 ",
"prediction" : [
5.47782474313863e-05,
0.00010245582961943,
0.997055888175964,
0.00185899913776666,
0.000927872606553137
]
},
…
]
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent