Skip to content
Snippets Groups Projects
UCSD-Newspapers.ipynb 17.4 KiB
Newer Older
John Graham's avatar
John Graham committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784
{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First let's change our current working directory to the folder where the data is located using `os.chdir`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import os\n",
    "os.chdir(\"./ucsdHistory_raw\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's confirm that the changes are in effect by checking the current working directory:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's open one of the newspaper's files and have a look at the contents. We'll start with the *Student Newspapers*, just because it's listed first alphabetically. We'll list all the files in the directory, open the first one and inspect the content by displaying the first 2000 characters in the file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "journal_name = 'Student Newspapers'\n",
    "path = os.path.join(os.getcwd(), journal_name)\n",
    "files = os.listdir(path)\n",
    "files[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`os.listdir()` returns the names of the files but not their path, which we need to be able to open them. For now we'll just use `os.path.join()` again."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "file = os.path.join(path, files[0])\n",
    "with open(file, 'r') as fp:\n",
    "    content = fp.read()\n",
    "    \n",
    "content[:2000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Woops. Encoding error. Let's fix this by specifying the encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "file = os.path.join(path, files[0])\n",
    "with open(file, 'r', encoding = 'utf8') as fp:\n",
    "    content = fp.read()\n",
    "    \n",
    "content[:2000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, let's get rid of the hyphenation using regular expressions (for which we'll need the **re** package). We'll create a specific function for it:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import re\n",
    "def preprocess(text):\n",
    "    return re.sub('\\-\\n+', '', text)\n",
    "\n",
    "content = preprocess(content)\n",
    "content[:2000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we want to be able to load all the files in the directory and apply the preprocessing to them. To do that, we'll first define a function that will return the full path of all the files in the directory and will makes sure that we are only dealing with text files (so that we don't try to open a folder or another filetype that would cause an error):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def load_files(directory):\n",
    "    path = os.path.join(os.getcwd(), directory)\n",
    "    files = os.listdir(path)\n",
    "    files = [os.path.join(path, _) for _ in files if '.txt' in _]\n",
    "    return files\n",
    "\n",
    "load_files(path)[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Once we have all the files, we'll create a new function `load_data` to each file one by one, preprocess it and store the result in memory in an array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def load_data(files):\n",
    "    X = []\n",
    "    for file in files:\n",
    "        with open(file, 'r', encoding = 'utf8') as fp:\n",
    "            content = fp.read()\n",
    "            content = preprocess(content)\n",
    "        X.append(content)\n",
    "    return X"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's actually run the function and get the data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "files = load_files(journal_name)\n",
    "documents = load_data(files)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "len(documents)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we'll vectorize the documents. We'll only keep terms that appear in less than 80% of the documents and at least twice in the corpus. We'll limit the number of features to 5000."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vectorizer = TfidfVectorizer(max_df=0.8, min_df=2, max_features=5000, stop_words='english')\n",
    "X = vectorizer.fit_transform(documents)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, we'll apply non-negative matrix factorization to do topic modeling. We'll try with 25 topics first:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.decomposition import NMF\n",
    "nmf = NMF(n_components=25, init = 'nndsvd').fit(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's have a look at nmf.components_ to understand the output and try to interprete the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "np.shape(nmf.components_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "nmf.components_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "nmf.components_[0].argsort()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "vocabulary = vectorizer.get_feature_names()\n",
    "print(vocabulary[2499])\n",
    "print(vocabulary[4960])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "nmf.components_[0][2499]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "nmf.components_[0][4960]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "nmf.components_[0].argsort()[::-1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's now print the top 10 words for the first topic:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "[vocabulary[i] for i in nmf.components_[0].argsort()[::-1][:10]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's create a function that will allow us to do that for all the topics:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def display_topics(model, vocabulary, no_words):\n",
    "    for index, topic in enumerate(model.components_):\n",
    "        print(\"Topic\", index, \":\")\n",
    "        print(\" \".join([vocabulary[i] for i in topic.argsort()[::-1][:no_words]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "display_topics(nmf, vocabulary, 15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note that we can also use nmf to get the distribution of each of our topic within a specific document (what percentage of document a concerns topic 1, 2 or 3? ) by normalizing the results of the NMF over the corpus:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "doc_topics = nmf.transform(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "np.shape(doc_topics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "doc_topics[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sum(doc_topics[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "doc_topics = doc_topics / np.sum(doc_topics, axis=1, keepdims=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "doc_topics[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's have a look at what metadata we can work with by opening the csv file corresponding to the corpus we're looking at. We will use the **pandas** package to load the csv file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "# we've stored the name of the journal in a variable earlier, we can reuse this here so that we can more easily switch to another corpus in the future\n",
    "df = pd.read_csv(journal_name + '.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's look at the columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And the first entries..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We notice that the filenames listed do not exactly correspond to the names of the raw text files we have used before, nor are they listed in the same order:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "files[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For that reason we will need to process our filenames and match them to the appropriate row in the dataframe as well as to the appropriate topic distribution in the NMF matrix we created earlier from the corpus.\n",
    "\n",
    "We notice that the Object Unique ID of each row can be found in the filenames of the text files we have, so let's use that to match the raw text to the rows of the csv. Let's do a test with the first file:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "file_id = files[0].split('/')[-1].split('-')[0]\n",
    "file_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df['Object Unique ID'].str.contains(file_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df[df['Object Unique ID'].str.contains(file_id)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's try to extract the date for this particular row. There are several date columns (BeginDate, EndDate, ...) but they all seem to be equivalent, so let's go with BeginDate..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_date = df['Begin date'][df['Object Unique ID'].str.contains(file_id)]\n",
    "test_date"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are getting the right value, but the format of the output tells us that the data type returned is not exactly standard. Let's confirm this:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "type(test_date)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is a variable type specific to pandas, so we might run into trouble when trying to perform standard operations over it or sending it to some package's function. Let's make sure we only get the value and not the **pandas** Series cell:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_date = df['Begin date'][df['Object Unique ID'].str.contains(file_id)].values\n",
    "print(test_date)\n",
    "print(type(test_date))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Almost there..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_date = df['Begin date'][df['Object Unique ID'].str.contains(file_id)].values[0]\n",
    "print(test_date)\n",
    "print(type(test_date))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ok. So now we have a standard string. Which is good, but not ideal in case we want to perform some calculations on the date (substraction or extracting the month, year...). We can use the **dateparser** package to help us:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import dateparser\n",
    "parsed = dateparser.parse(test_date)\n",
    "print(type(parsed))\n",
    "print(parsed)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A little parsing mistake here but we can handle that later. Let's iterate over all our files and store the results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "yearbook = defaultdict(list)\n",
    "\n",
    "for i, file in enumerate(files):\n",
    "    doc_year = dateparser.parse(df['Begin date'][df['Object Unique ID'].str.contains(file.split('/')[-1].split('-')[0])].values[0]).year\n",
    "    \n",
    "    # handles the 2k bug\n",
    "    if doc_year > 2018:\n",
    "        doc_year -= 100\n",
    "    \n",
    "    # fetches the corresponding distribution for the document. our files and topic distribution matrix have the same index\n",
    "    topic_distribution = doc_topics[i]\n",
    "    \n",
    "    # stores the result in an array\n",
    "    yearbook[doc_year].append(topic_distribution)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's compute the mean distribution of topics for each year (maybe absolute values would be better) and store the result in a dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "finals = {v: np.mean(yearbook[v], axis = 0) for v in yearbook}\n",
    "res = pd.DataFrame.from_dict(finals)\n",
    "res.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's transpose and plot the data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "res.transpose().plot(figsize=(12,12))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}