Skip to content

Commit

Permalink
initial version
Browse files Browse the repository at this point in the history
  • Loading branch information
@tblock
tblock committed Feb 26, 2019
0 parents commit 8d6e1e0
Show file tree
Hide file tree
Showing 26 changed files with 63,952 additions and 0 deletions.
21 changes: 21 additions & 0 deletions LICENSE
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
MIT License

Copyright (c) 2019 Timo Block

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
13 changes: 13 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
# thesis-data

This repository contains supplementary scripts and data for my thesis.
I classified the [Ten Thousand German News Articles Dataset](https://tblock.github.io/10kGNAD/) with four text classifiers. Namely a Support Vector Machine, facebook's [fastText](https://fasttext.cc) libary, a TensorFlow neuronal net and the [ULMFiT](https://arxiv.org/abs/1801.06146) method.


The scripts can be run in a [Google Colab IPython Notebook](https://colab.research.google.com).

- SVM [[view]](https://github.com/tblock/thesis-data/blob/master/reproduce_SVM.ipynb) [[run]](https://colab.research.google.com/github/tblock/thesis-data/blob/master/reproduce_SVM.ipynb)
- fastText [[view]](https://github.com/tblock/thesis-data/blob/master/reproduce_fastText.ipynb) [[run]](https://colab.research.google.com/github/tblock/thesis-data/blob/master/reproduce_fastText.ipynb)
- TensorFlow [[view]](https://github.com/tblock/thesis-data/blob/master/reproduce_TensorFlow.ipynb) [[run]](https://colab.research.google.com/github/tblock/thesis-data/blob/master/reproduce_TensorFlow.ipynb)
- ULMFiT [[view]](https://github.com/tblock/thesis-data/blob/master/reproduce_ULMFiT.ipynb) [[run]](https://colab.research.google.com/github/tblock/thesis-data/blob/master/reproduce_ULMFiT.ipynb)

1,028 changes: 1,028 additions & 0 deletions fastText-data/fastText_test.csv

Large diffs are not rendered by default.

256 changes: 256 additions & 0 deletions reproduce_SVM.ipynb
Original file line number Diff line number Diff line change
@@ -0,0 +1,256 @@
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "REP_SVM.ipynb",
"version": "0.3.2",
"provenance": [],
"collapsed_sections": [],
"toc_visible": true
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
}
},
"cells": [
{
"metadata": {
"id": "5_EKIwLlcII7",
"colab_type": "text"
},
"cell_type": "markdown",
"source": [
"## Support Vector Machine to classify the [Ten Thousand German News Articles Dataset](https://github.com/tblock/10kGNAD)\n",
"This Notebook contains the code to reproduce the results in my thesis.\n",
"The code reproduces the exact results.\n",
"\n",
"Run all cells consecutively."
]
},
{
"metadata": {
"id": "84eqgmr2RZhY",
"colab_type": "text"
},
"cell_type": "markdown",
"source": [
"### Enviroment Setup "
]
},
{
"metadata": {
"id": "oyzQUOgwcy2W",
"colab_type": "code",
"outputId": "3a646b13-70f0-4492-9dde-a8141fd261a0",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 53
}
},
"cell_type": "code",
"source": [
"# load the dataset and generate subsets\n",
"!rm -rf 10kGNAD lowshot\n",
"!git config --global advice.detachedHead false\n",
"!git clone -q --branch v1.1 https://github.com/tblock/10kGNAD.git && echo \"downloaded dataset\"\n",
"!mkdir lowshot\n",
"!cp 10kGNAD/train.csv .\n",
"!python 10kGNAD/code/generate_lowshot_sets.py > /dev/null && echo \"generated train subsets\""
],
"execution_count": 0,
"outputs": [
{
"output_type": "stream",
"text": [
"downloaded dataset\n",
"generated train subsets\n"
],
"name": "stdout"
}
]
},
{
"metadata": {
"id": "3wuK2dHCcHf9",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"import glob\n",
"import numpy as np\n",
"import pandas as pd\n",
"from sklearn.svm import LinearSVC\n",
"from sklearn.pipeline import Pipeline\n",
"from sklearn.feature_extraction.text import CountVectorizer\n",
"from sklearn.feature_extraction.text import TfidfTransformer\n",
"from sklearn.preprocessing import StandardScaler"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "63NWI1D4RiJb",
"colab_type": "text"
},
"cell_type": "markdown",
"source": [
"### Train Models"
]
},
{
"metadata": {
"id": "9b1C4PVtt4LL",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# load test set\n",
"df_test = pd.read_csv('10kGNAD/test.csv', header=None, sep=';', quotechar=\"'\", names=['label', 'text'])"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "fIRl3HQvutRJ",
"colab_type": "code",
"outputId": "67b95baf-1ecb-40b4-991b-90b798bd60d2",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1637
}
},
"cell_type": "code",
"source": [
"filenames = sorted(glob.glob(\"lowshot/*.csv\"))\n",
"\n",
"for filename in filenames: # for each subset\n",
" \n",
" df_train = pd.read_csv(filename, header=None, sep=';', quotechar=\"'\", names=['label', 'text'])\n",
"\n",
" # build the classifier pipeline\n",
" lsvc_classifier = Pipeline([\n",
" ('vect', CountVectorizer()),\n",
" ('tfidf', TfidfTransformer(\n",
" sublinear_tf=True # Apply sublinear tf scaling, i.e. replace tf with 1 + log(tf).\n",
" )),\n",
" ('clf', LinearSVC(\n",
" dual=False,\n",
" C=1.6,\n",
" class_weight=\"balanced\"\n",
" ))\n",
" ])\n",
"\n",
" lsvc_classifier.fit(df_train['text'], df_train['label']) # train the classifier\n",
" predicted = lsvc_classifier.predict(df_test['text']) # predict the test set \n",
" acc = np.mean(predicted == df_test['label']) # calculate the accuracy\n",
" \n",
" print(filename[16:-4],\"%.2f\" % float((100 - acc*100)), sep=\" -> \") # print the error rate "
],
"execution_count": 0,
"outputs": [
{
"output_type": "stream",
"text": [
"0_0.01_0 -> 40.56\n",
"0_0.01_1 -> 39.40\n",
"0_0.01_2 -> 36.87\n",
"0_0.01_3 -> 38.13\n",
"0_0.01_4 -> 42.02\n",
"0_0.01_5 -> 40.56\n",
"0_0.01_6 -> 39.59\n",
"0_0.01_7 -> 38.33\n",
"0_0.01_8 -> 38.04\n",
"0_0.01_9 -> 39.59\n",
"1_0.02_0 -> 32.10\n",
"1_0.02_1 -> 27.72\n",
"1_0.02_2 -> 30.64\n",
"1_0.02_3 -> 32.78\n",
"1_0.02_4 -> 28.21\n",
"1_0.02_5 -> 29.38\n",
"1_0.02_6 -> 30.06\n",
"1_0.02_7 -> 28.99\n",
"1_0.02_8 -> 31.81\n",
"1_0.02_9 -> 32.10\n",
"2_0.05_0 -> 22.76\n",
"2_0.05_1 -> 24.22\n",
"2_0.05_2 -> 20.53\n",
"2_0.05_3 -> 21.89\n",
"2_0.05_4 -> 22.86\n",
"2_0.05_5 -> 22.57\n",
"2_0.05_6 -> 22.76\n",
"2_0.05_7 -> 24.51\n",
"2_0.05_8 -> 21.69\n",
"2_0.05_9 -> 22.28\n",
"3_0.075_0 -> 18.97\n",
"3_0.075_1 -> 19.26\n",
"3_0.075_2 -> 20.72\n",
"3_0.075_3 -> 20.43\n",
"3_0.075_4 -> 20.53\n",
"3_0.075_5 -> 19.94\n",
"3_0.075_6 -> 19.94\n",
"3_0.075_7 -> 19.84\n",
"3_0.075_8 -> 19.65\n",
"3_0.075_9 -> 20.14\n",
"4_0.1_0 -> 19.36\n",
"4_0.1_1 -> 18.68\n",
"4_0.1_2 -> 17.02\n",
"4_0.1_3 -> 18.77\n",
"4_0.1_4 -> 17.32\n",
"4_0.1_5 -> 17.61\n",
"4_0.1_6 -> 18.48\n",
"4_0.1_7 -> 18.68\n",
"4_0.1_8 -> 17.61\n",
"4_0.1_9 -> 17.02\n",
"5_0.2_0 -> 14.98\n",
"5_0.2_1 -> 15.47\n",
"5_0.2_2 -> 14.88\n",
"5_0.2_3 -> 15.66\n",
"5_0.2_4 -> 15.27\n",
"5_0.2_5 -> 15.37\n",
"5_0.2_6 -> 16.73\n",
"5_0.2_7 -> 15.56\n",
"5_0.2_8 -> 16.15\n",
"5_0.2_9 -> 15.47\n",
"6_0.5_0 -> 13.13\n",
"6_0.5_1 -> 13.04\n",
"6_0.5_2 -> 13.23\n",
"6_0.5_3 -> 12.45\n",
"6_0.5_4 -> 13.42\n",
"6_0.5_5 -> 12.65\n",
"6_0.5_6 -> 13.13\n",
"6_0.5_7 -> 13.91\n",
"6_0.5_8 -> 13.04\n",
"6_0.5_9 -> 12.84\n",
"7_0.75_0 -> 11.28\n",
"7_0.75_1 -> 11.87\n",
"7_0.75_2 -> 12.06\n",
"7_0.75_3 -> 12.65\n",
"7_0.75_4 -> 12.35\n",
"7_0.75_5 -> 11.58\n",
"7_0.75_6 -> 12.35\n",
"7_0.75_7 -> 12.06\n",
"7_0.75_8 -> 12.06\n",
"7_0.75_9 -> 12.26\n",
"8_1.0_0 -> 11.48\n",
"8_1.0_1 -> 11.48\n",
"8_1.0_2 -> 11.48\n",
"8_1.0_3 -> 11.48\n",
"8_1.0_4 -> 11.48\n",
"8_1.0_5 -> 11.48\n",
"8_1.0_6 -> 11.48\n",
"8_1.0_7 -> 11.48\n",
"8_1.0_8 -> 11.48\n",
"8_1.0_9 -> 11.48\n"
],
"name": "stdout"
}
]
}
]
}
Loading

0 comments on commit 8d6e1e0

Please sign in to comment.