File size: 62,966 Bytes
6fa4bc9 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 |
{
"paper_id": "2020",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T01:06:35.822186Z"
},
"title": "Personalized Early Stage Alzheimer's Disease Detection: A Case Study of President Reagan's Speeches",
"authors": [
{
"first": "Ning",
"middle": [],
"last": "Wang",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Stevens Institute of Technology Hoboken",
"location": {
"postCode": "07030",
"region": "NJ"
}
},
"email": "nwang7@stevens.edu"
},
{
"first": "Fan",
"middle": [],
"last": "Luo",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Stevens Institute of Technology Hoboken",
"location": {
"postCode": "07030",
"region": "NJ"
}
},
"email": ""
},
{
"first": "Vishal",
"middle": [],
"last": "Peddagangireddy",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Stevens Institute of Technology Hoboken",
"location": {
"postCode": "07030",
"region": "NJ"
}
},
"email": ""
},
{
"first": "K",
"middle": [
"P"
],
"last": "Subbalakshmi",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Stevens Institute of Technology Hoboken",
"location": {
"postCode": "07030",
"region": "NJ"
}
},
"email": ""
},
{
"first": "R",
"middle": [],
"last": "Chandramouli",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Stevens Institute of Technology Hoboken",
"location": {
"postCode": "07030",
"region": "NJ"
}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Alzheimer's disease (AD)-related global healthcare cost is estimated to be $1 trillion by 2050. Currently, there is no cure for this disease; however, clinical studies show that early diagnosis and intervention helps to extend the quality of life and inform technologies for personalized mental healthcare. Clinical research indicates that the onset and progression of Alzheimer's disease lead to dementia and other mental health issues. As a result, the language capabilities of patient start to decline. In this paper, we show that machine learningbased unsupervised clustering of and anomaly detection with linguistic biomarkers are promising approaches for intuitive visualization and personalized early stage detection of Alzheimer's disease. We demonstrate this approach on 10 year's (1980 to 1989) of President Ronald Reagan's speech data set. Key linguistic biomarkers that indicate early-stage AD are identified. Experimental results show that Reagan had early onset of Alzheimer's sometime between 1983 and 1987. This finding is corroborated by prior work that analyzed his interviews using a statistical technique. The proposed technique also identifies the exact speeches that reflect linguistic biomarkers for early stage AD.",
"pdf_parse": {
"paper_id": "2020",
"_pdf_hash": "",
"abstract": [
{
"text": "Alzheimer's disease (AD)-related global healthcare cost is estimated to be $1 trillion by 2050. Currently, there is no cure for this disease; however, clinical studies show that early diagnosis and intervention helps to extend the quality of life and inform technologies for personalized mental healthcare. Clinical research indicates that the onset and progression of Alzheimer's disease lead to dementia and other mental health issues. As a result, the language capabilities of patient start to decline. In this paper, we show that machine learningbased unsupervised clustering of and anomaly detection with linguistic biomarkers are promising approaches for intuitive visualization and personalized early stage detection of Alzheimer's disease. We demonstrate this approach on 10 year's (1980 to 1989) of President Ronald Reagan's speech data set. Key linguistic biomarkers that indicate early-stage AD are identified. Experimental results show that Reagan had early onset of Alzheimer's sometime between 1983 and 1987. This finding is corroborated by prior work that analyzed his interviews using a statistical technique. The proposed technique also identifies the exact speeches that reflect linguistic biomarkers for early stage AD.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Alzheimer's disease is a serious mental health issue faced by the global population. About 44 million people worldwide are diagnosed with AD. The U.S. alone has 5.5 million AD patients. According to the Alzheimer's association the total cost of care for AD is estimated to be $1 trillion by 2050. There is no cure for AD yet; however, studies have shown that early diagnosis and intervention can delay the onset.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Regular mental health assessment is a key challenge faced by the medical community. This is due to a variety of reasons including social, economic, and cultural factors. Therefore, Internet based technologies that unobtrusively and continually collect, store, and analyze mental health data are critical. For example, a home smart speaker device can record a subject's speech periodically, automatically extract AD related speech or linguistic features, and present easy to understand machine learning based analysis and visualization. Such a technology will be highly valuable for personalized medicine and early intervention. This may also encourage people to sign-up for such a low cost and home-based AD diagnostic technology. Data and results of such a technology will also instantly provide invaluable information to mental health professionals.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Several studies show that subtle linguistic changes are observed even at the early stages of AD. In (Forbes-McKay and Venneri, 2005) , more than 70% of AD patients scored low in a picture description task. Therefore, a critical research question is: can spontaneous temporal language impairments caused by AD be detected at an early stage of the disease? Relation between AD, language functions and language domain are summarized in (Szatloczki et al., 2015) . In (Venneri et al., 2008) , a significant correlation between the lexical attributes characterising residual linguistic production and the integrity of regions of the medial temporal lobes in early AD patients is observed. Specific lexical and semantic deficiencies of AD patients at early to moderate stages are also detected in verbal communication task in (Boy\u00e9 et al., 2014) . Therefore, in this paper, we explore a machine-learning based clustering and data visualization technique to identify linguistic changes in a subject over a time period. The proposed methodology is also highly personalized since it observes and analyzes the linguistic patterns of each individual separately using only his/her own linguistic biomarkers over a period of time.",
"cite_spans": [
{
"start": 100,
"end": 132,
"text": "(Forbes-McKay and Venneri, 2005)",
"ref_id": "BIBREF4"
},
{
"start": 433,
"end": 458,
"text": "(Szatloczki et al., 2015)",
"ref_id": "BIBREF13"
},
{
"start": 464,
"end": 486,
"text": "(Venneri et al., 2008)",
"ref_id": "BIBREF14"
},
{
"start": 820,
"end": 839,
"text": "(Boy\u00e9 et al., 2014)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "First, we explore a machine learning algorithm called t-distributed stochastic neighbor embedding (t-SNE) (Maaten and Hinton, 2008) . t-SNE is useful in dimensionality reduction suitable for visualization of high-dimensional datasets. It calculates the probability that two points in a highdimensional space are similar, computes the corresponding probability in a low-dimensional space, and minimizes the difference between these two probabilities for mapping or visualization. During this process, the sum of Kullback-Leibler divergences (Liu and Shum, 2003) over all the data points is minimized. Our hypothesis is that highdimensional AD-related linguistic features when visualized in a low-dimensional space may quickly and intuitively reveal useful information for early diagnosis. Such a visualization will also help individuals and general medical practitioners (who are first points of contact) to assess the situation for further tests. Second, we investigate two unsupervised machine learning techniques, one class support vector machine (SVM) and isolation forest, for temporal detection of linguistic abnormalities indicative of early-stage AD. These proposed approaches are tested on President Reagan's speech dataset and corroborated with results from other research in the literature.",
"cite_spans": [
{
"start": 106,
"end": 131,
"text": "(Maaten and Hinton, 2008)",
"ref_id": "BIBREF9"
},
{
"start": 540,
"end": 560,
"text": "(Liu and Shum, 2003)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper is organized as follows. Background research is discussed in Section 2, Section 3 presents the Reagan speech data set used in this paper, data pre-processing techniques that were applied and AD-related linguistic feature selection rationale and methodology, Section 4 contains the clustering and visualization of the handcrafted linguistic features and anomaly detection to infer the onset of AD and to detect the time period of changes in the linguistic statistical characteristics, and experimental results to demonstrate the proposed method. In Section 5, we describe the machine learning algorithms for detecting anomalies from personalized linguistic biomarkers collected over a period of time to identify early-stage AD. Concluding remarks are given in Section 6.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The \"picture description\" task has been widely studied to differentiate between AD and non-AD or control subjects. In this task, a picture (the \"cookie theft picture\") is shown and the subject is asked to describe it. It has been observed that subjects with AD usually convey sparse information about the picture ignoring expected facts and inferences (Giles et al., 1996) . AD patients have difficulty in naming things and replace target words with simpler semantically neighboring words. Understanding metaphors and sarcasm also deteriorate in people diagnosed with AD (Rapp and Wild, 2011).",
"cite_spans": [
{
"start": 352,
"end": 372,
"text": "(Giles et al., 1996)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "Machine learning based classifier design to differentiate between AD and non-AD subjects is an active area of research. A significant correlation between dementia severity and linguistic measures such as confrontation naming, articulation, wordfinding ability, and semantic fluency exists. Some studies have reported a 84.8 percent accuracy in distinguishing AD patients from healthy controls using temporal and acoustic features ( (Rentoumi et al., 2014) ; (Fraser et al., 2016) ).",
"cite_spans": [
{
"start": 432,
"end": 455,
"text": "(Rentoumi et al., 2014)",
"ref_id": "BIBREF12"
},
{
"start": 458,
"end": 479,
"text": "(Fraser et al., 2016)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "Our study differs from the prior work in several ways. Prior work attempt to identify linguistic features and machine learning classifiers for differentiating between AD and non-AD subjects from a corpus (e.g., Dementia Bank (MacWhinney, 2007)) containing text transcripts of interviews with patients and healthy control. In this paper, we first analyze the speech transcripts (over several years) of a single person (President Ronald Reagan) and visualize time-dependent linguistic information using t-SNE. The goal is to identify visual clues for linguistic changes that may indicate the onset of AD. Note that such an easy to understand visual representation depicting differences in linguistic patterns will be useful to both a common person and a general practitioner (who is the first point of contact for majority of patients). Since most general practitioners are not trained mental health professionals the impact of such a visualization tool will be high, especially at the early stages of AD. Sigificant AD-related linguistic biomarkers derived from t-SNE analysis are then used in two unsupervised clustering algorithms for detecting AD-related temporal linguistic anomalies. This provides an estimate for the time when early-stage AD symptoms are beginning to be observed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "3 Reagan Speeches: Data Collection, Pre-processing, and Feature Selection",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "We describe the data collection, pre-processing, and feature engineering methodologies in this section. Ronald Reagan was the 40th president (served from 1981 to 1989) of the United States of America. He was an extraordinary orator, a radio announcer, actor, and the host for a show called \"Gen-eral Electric Theatre.\" Clearly, for being successful in these professional domains one needs to have good memory, consciousness, intuition, command over the language, and ease of communicating with a large audience. Reagan officially announced that he had been diagnosed with AD on November 5, 1994. But it was speculated that his cognitive abilities where on the decline even while in office (Gottschalk et al., 1988) . Therefore, analyzing his speech transcripts for early signs of AD may reveal interesting patterns, if any. The Reagan Library is the repository of presidential records for President Reagan's administration. We download his 98 speeches from 1980 to 1989 as shown in Table 1 . We removed special characters, tags, and numbers and kept only the words from each speech transcript. The resulting data was then lemmatized and tokenized. Part-of-speech (POS) features: People diagnosed with AD use more pronouns than nouns. Therefore, POS features such as the number of pronouns and the pronouns-to-nouns ratio are important. We identified adverbs, nouns, verbs, and pronouns for each speech transcript with natural language processing (NLP) tools. POS tags having at least 10 occurrences were selected to compute their percentage ratios. Similarly, words that have at least a frequency of 10 were selected and their occurrence percentages were computed.",
"cite_spans": [
{
"start": 689,
"end": 714,
"text": "(Gottschalk et al., 1988)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [
{
"start": 982,
"end": 989,
"text": "Table 1",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "The full set of POS features we used were: (1) number of pronouns, (2) pronoun-noun ratio, (3) number of adverbs, (4) number of nouns, (5) number of verbs, (6) pro-noun frequency rate, (7) noun frequency rate, (8) verb frequency rate, (9) adverb frequency rate, (10) word frequency rate, and (11) word frequency rate without excluding stop words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "Vocabulary Richness: AD patients show a decline in their vocabulary range. Therefore, vocabulary richness metrics: Honore's Statistic (HS), Sichel Measure (SICH), and Brunet's Measure (BM) were calculated for each speech. Higher values of Honore's and Sichel measures indicate greater vocabulary richness. But a higher value corresponds to low vocabulary richness for the Brunet's measure.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "Readability Measures: We computed two readability measures, namely, Automated Readability Index (ARI) and Flesch-Kincaid readability (FKR) score. A higher ARI indicates complex speech with rich vocabulary whereas lower Flesch-Kincaid score indicates rich vocabulary. Figure 1 shows the correlation matrix of the chosen linguistic features computed for the Reagan speech dataset. Note that some of the features are highly correlated, therefore, we pruned the feature set to the following 9 features: ",
"cite_spans": [],
"ref_spans": [
{
"start": 267,
"end": 275,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Background",
"sec_num": "2"
},
{
"text": "We selected the t-SNE machine learning technique for clustering and visulaization of linguistic features extracted from Reagan's speeches. t-SNE is better at creating a single map for revealing structures at many different scales important for highdimensional data that lie on several different, but related, low-dimensional manifolds. This implies that it can capture much of the local structure of the high-dimensional data very well, while also revealing global structure such as the presence of clusters at several scales (van der Maaten and Hinton 2008).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "If there are AD-related linguistic patterns then t-SNE may reveal them as clusters.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "In t-SNE, the high-dimensional Euclidean distances between datapoints are converted into conditional probabilities representing similarities between them. For example, the similarity of data point x j with x i is the conditional probability p j|i that x i would pick x j as its neighbor. Neighbors are picked in proportion to their probability density under Student t-distribution centered at x i in the low-dimensional space. Then the Kullback-Leibler divergence between a joint probability distribution, P, in the high-dimensional space and a joint probability distribution, Q, in the low-dimensional space is then minimized:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "minKL(P ||Q) = i j p ij log p ij q ij (1)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "t-SNE has two tunable hyperparameters: perplexity and learning rate. Perplexity allows us to balance the weighting between local and global relationships of the data. It gives a sense of the number of close neighbors for each point. A perplexity value between 5 and 50 is recommended. The learning rate for t-SNE is usually in the interval [10, 1000] . For a high learning rate, each point will approximately be equidistant from its nearest neighbours. For a low rate, there will be few outliers and therefore the points may look compressed into a dense cloud. Since t-SNE's cost function is not convex different initializations can produce different results. After tuning t-SNE's hyperparameters for the dataset, we chose perplexity value equal to 4 and learning rate equal to 100.",
"cite_spans": [
{
"start": 340,
"end": 344,
"text": "[10,",
"ref_id": null
},
{
"start": 345,
"end": 350,
"text": "1000]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "Pronoun-to-noun ratio: Figure 2 shows t-SNE based clustering of the speech transcripts, speeches sharing similar patterns are clustered together. Besides, the radius of each circle is proportional to the pronoun-to-noun ratio. Notice that the cluster on the left side of the graph contains speeches (from 1983 to 1987) have higher values of the ratio. Recall that higher pronoun-to-noun ratio is an indicator of early stage AD. Pronoun frequency: Fig. 3 indicates clustering and low-dimensional visualization results when the size of each circle is proportional to the pronoun frequency. We again see that the cluster on the left side of the map contains speeches with higher pronoun frequency, another indicator of early stage AD. Readability score: From Fig. 4 we observe that the speeches from 1983-1987 have lower readability scores.",
"cite_spans": [
{
"start": 299,
"end": 318,
"text": "(from 1983 to 1987)",
"ref_id": null
}
],
"ref_spans": [
{
"start": 23,
"end": 31,
"text": "Figure 2",
"ref_id": "FIGREF1"
},
{
"start": 447,
"end": 453,
"text": "Fig. 3",
"ref_id": "FIGREF2"
},
{
"start": 756,
"end": 762,
"text": "Fig. 4",
"ref_id": "FIGREF3"
}
],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "Word repetition frequency: The word repetition frequency map in Fig. 5 shows that three speeches standout. These three speeches have a higher repetition of high frequency words. Interestingly, two of these speeches have word lengths smaller than the mean length of all the speeches. From these clustering results the speeches identified as showing early signs of AD are:",
"cite_spans": [],
"ref_spans": [
{
"start": 64,
"end": 70,
"text": "Fig. 5",
"ref_id": "FIGREF4"
}
],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "\u2022 03-23-1983: Address to the Nation on National Security (\"Star Wars\" SDI Speech)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "\u2022 \u2022 he started showing signs of early AD well before the official announcement in 1994",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "\u2022 it is highly likely that he developed AD sometime between 1983 and 1987",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "\u2022 over time, President Reagan's showed a significant reduction the number of unique words but a significant increase in conversational fillers and non-specific nouns",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "\u2022 the proposed method identifies specific speeches that exhibit linguistic markers for AD Some of these findings are corroborated by prior research that analyzed his interviews and compared them with President Bursh's public speeches, (Berisha et al., 2015) .",
"cite_spans": [
{
"start": 235,
"end": 257,
"text": "(Berisha et al., 2015)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Clustering and Visualization of Linguistic Features",
"sec_num": "4"
},
{
"text": "t-SNE-clustering-based approach provides visualization of speeches that are statistically different (\"anomalies\"). But we need an automated method to identify these anomalies for early signs of AD. Therefore, we investigate a one-class support vector machine (SVM) anomaly detector (Erfani et al., 2016) . This method is useful in practice when majority of a subject's speeches over several years would be (statistically) typical of a healthy control (\"normal\") until he/she begins to exhibit early signs of AD. Our hypothesis is that early stage AD will begin to reveal itself as statistical anomalies in linguistic feature space. In this section we investigate this hypothesis. We designed a one class SVM with the following hyperparameter values (the choices of these values are not discussed for the sake of clarity and focus), \u03bd = 0.5 (an upper bound on the fraction of training errors and a lower bound of the fraction of support vectors), kernel=rbf (radial basis function) and \u03b3 = 1 (number of features\u00d7variance of features) (kernel coefficient). Figure 6 shows that the one class SVM detector identified several speeches from 1984 to 1986 as anomalous. That is, President Reagan's speeches in these years are different from the previous years in the linguistic feature space. Therefore, it is likely that:",
"cite_spans": [
{
"start": 282,
"end": 303,
"text": "(Erfani et al., 2016)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [
{
"start": 1055,
"end": 1063,
"text": "Figure 6",
"ref_id": "FIGREF5"
}
],
"eq_spans": [],
"section": "Linguistic Anomaly Detector for AD",
"sec_num": "5"
},
{
"text": "\u2022 he started showing signs of early AD well before the official announcement in 1994",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Linguistic Anomaly Detector for AD",
"sec_num": "5"
},
{
"text": "\u2022 the onset of AD start from 1984 and became more pronounced in 1985 and 1986, which is corroborated by prior research (e.g., (Berisha et al., 2015) ) that analyzed his interviews.",
"cite_spans": [
{
"start": 126,
"end": 148,
"text": "(Berisha et al., 2015)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Linguistic Anomaly Detector for AD",
"sec_num": "5"
},
{
"text": "One class SVM learns the profile of non-AD speeches as \"normal\" over a period of time and detects anomalies to signal the onset of AD. But in many practical instances long historical data may not be available for a subject. In this case, we must identify anomalies explicitly instead of learning what is normal. Isolation forest (Ding and Fei, 2013) is an unsupervised machine learning algorithm that is applicable for this purpose.",
"cite_spans": [
{
"start": 329,
"end": 349,
"text": "(Ding and Fei, 2013)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Linguistic Anomaly Detector for AD",
"sec_num": "5"
},
{
"text": "We applied the isolation forest algorithm on the speech dataset and tuned its hyperparameters. Figure 8 show the results. Figure 7 shows the ten speeches that were isolated as anomalies. The corresponding dates of these speeches are seen in Figure 8 . We observe that a few speeches between 1984 and 1988 as linguistic anomalies. This time period overlaps significantly with the results of once class SVM and t-SNE clustering. ",
"cite_spans": [],
"ref_spans": [
{
"start": 95,
"end": 103,
"text": "Figure 8",
"ref_id": "FIGREF8"
},
{
"start": 122,
"end": 130,
"text": "Figure 7",
"ref_id": "FIGREF7"
},
{
"start": 241,
"end": 249,
"text": "Figure 8",
"ref_id": "FIGREF8"
}
],
"eq_spans": [],
"section": "Linguistic Anomaly Detector for AD",
"sec_num": "5"
},
{
"text": "A set of nine linguistic biomarkers for AD was identified. Two complementary unsupervised machine learning methods were applied on the linguistic features extracted from President Reagan's 98 speeches given between 1980 to 1989. The first method, t-SNE, identified and visualized speeches indicating early onset of AD. A higher pronoun usage frequency, lower readability scores, higher repetition of high frequency words were revealed to be the key characteristics of potential AD-related speeches. A subset speeches from 1983 to 1987 were detected to possess these characteristics.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "6"
},
{
"text": "The second machine learning method, one class SVM, learned what is \"normal\" (i.e., non-AD speech) to detect anomalies in speeches over a period of time. This approach detected several speeches between 1984 and 1986 as potential ADrelated. Since normal speech may not be available historically we applied the isolation forest algorithm that explicitly detects anomalies without learning what is normal. This detected 10 speeches from 1983 to 1987 as AD-related.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "6"
},
{
"text": "From the experimental analysis our conclusion is that President Reagan had signs of AD sometime between 1983 and 1988. This conclusion corroborates results from other studies in the literature. Note that that President Reagan had AD was publicly disclosed only in November 1994.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "6"
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Tracking discourse complexity preceding alzheimer's disease diagnosis: a case study comparing the press conferences of presidents ronald reagan and george herbert walker bush",
"authors": [
{
"first": "Visar",
"middle": [],
"last": "Berisha",
"suffix": ""
},
{
"first": "Shuai",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Amy",
"middle": [],
"last": "Lacross",
"suffix": ""
},
{
"first": "Julie",
"middle": [],
"last": "Liss",
"suffix": ""
}
],
"year": 2015,
"venue": "Journal of Alzheimer's Disease",
"volume": "45",
"issue": "3",
"pages": "959--963",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Visar Berisha, Shuai Wang, Amy LaCross, and Julie Liss. 2015. Tracking discourse complexity preced- ing alzheimer's disease diagnosis: a case study com- paring the press conferences of presidents ronald rea- gan and george herbert walker bush. Journal of Alzheimer's Disease, 45(3):959-963.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Contrastive conversational analysis of language production by alzheimer's and control people",
"authors": [
{
"first": "Ma\u00eft\u00e9",
"middle": [],
"last": "Boy\u00e9",
"suffix": ""
},
{
"first": "Natalia",
"middle": [],
"last": "Grabar",
"suffix": ""
},
{
"first": "Thi",
"middle": [
"Mai"
],
"last": "Tran",
"suffix": ""
}
],
"year": 2014,
"venue": "MIE",
"volume": "",
"issue": "",
"pages": "682--686",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ma\u00eft\u00e9 Boy\u00e9, Natalia Grabar, and Thi Mai Tran. 2014. Contrastive conversational analysis of language pro- duction by alzheimer's and control people. In MIE, pages 682-686.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "An anomaly detection approach based on isolation forest algorithm for streaming data using sliding window. IFAC Proceedings Volumes",
"authors": [
{
"first": "Zhiguo",
"middle": [],
"last": "Ding",
"suffix": ""
},
{
"first": "Minrui",
"middle": [],
"last": "Fei",
"suffix": ""
}
],
"year": 2013,
"venue": "",
"volume": "46",
"issue": "",
"pages": "12--17",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhiguo Ding and Minrui Fei. 2013. An anomaly de- tection approach based on isolation forest algorithm for streaming data using sliding window. IFAC Pro- ceedings Volumes, 46(20):12-17.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Highdimensional and large-scale anomaly detection using a linear one-class svm with deep learning. Pattern Recognition",
"authors": [
{
"first": "M",
"middle": [],
"last": "Sarah",
"suffix": ""
},
{
"first": "Sutharshan",
"middle": [],
"last": "Erfani",
"suffix": ""
},
{
"first": "Shanika",
"middle": [],
"last": "Rajasegarar",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Karunasekera",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Leckie",
"suffix": ""
}
],
"year": 2016,
"venue": "",
"volume": "58",
"issue": "",
"pages": "121--134",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sarah M Erfani, Sutharshan Rajasegarar, Shanika Karunasekera, and Christopher Leckie. 2016. High- dimensional and large-scale anomaly detection us- ing a linear one-class svm with deep learning. Pat- tern Recognition, 58:121-134.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Detecting subtle spontaneous language decline in early alzheimer's disease with a picture description task",
"authors": [
{
"first": "Katrina",
"middle": [
"E"
],
"last": "Forbes-Mckay",
"suffix": ""
},
{
"first": "Annalena",
"middle": [],
"last": "Venneri",
"suffix": ""
}
],
"year": 2005,
"venue": "Neurological sciences",
"volume": "26",
"issue": "4",
"pages": "243--254",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Katrina E Forbes-McKay and Annalena Venneri. 2005. Detecting subtle spontaneous language decline in early alzheimer's disease with a picture description task. Neurological sciences, 26(4):243-254.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Linguistic features identify alzheimer's disease in narrative speech",
"authors": [
{
"first": "C",
"middle": [],
"last": "Kathleen",
"suffix": ""
},
{
"first": "Jed",
"middle": [
"A"
],
"last": "Fraser",
"suffix": ""
},
{
"first": "Frank",
"middle": [],
"last": "Meltzer",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Rudzicz",
"suffix": ""
}
],
"year": 2016,
"venue": "Journal of Alzheimer's Disease",
"volume": "49",
"issue": "2",
"pages": "407--422",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kathleen C Fraser, Jed A Meltzer, and Frank Rudzicz. 2016. Linguistic features identify alzheimer's dis- ease in narrative speech. Journal of Alzheimer's Dis- ease, 49(2):407-422.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Performance on the boston cookie theft picture description task in patients with early dementia of the alzheimer's type: missing information",
"authors": [
{
"first": "Elaine",
"middle": [],
"last": "Giles",
"suffix": ""
},
{
"first": "Karalyn",
"middle": [],
"last": "Patterson",
"suffix": ""
},
{
"first": "John R",
"middle": [],
"last": "Hodges",
"suffix": ""
}
],
"year": 1996,
"venue": "Aphasiology",
"volume": "10",
"issue": "4",
"pages": "395--408",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Elaine Giles, Karalyn Patterson, and John R Hodges. 1996. Performance on the boston cookie theft pic- ture description task in patients with early dementia of the alzheimer's type: missing information. Apha- siology, 10(4):395-408.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Presidential candidates and cognitive impairment measured from behavior in campaign debates",
"authors": [
{
"first": "A",
"middle": [],
"last": "Louis",
"suffix": ""
},
{
"first": "Regina",
"middle": [],
"last": "Gottschalk",
"suffix": ""
},
{
"first": "Ronda",
"middle": [],
"last": "Uliana",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Gilbert",
"suffix": ""
}
],
"year": 1988,
"venue": "Public Administration Review",
"volume": "",
"issue": "",
"pages": "613--619",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Louis A Gottschalk, Regina Uliana, and Ronda Gilbert. 1988. Presidential candidates and cognitive impair- ment measured from behavior in campaign debates. Public Administration Review, pages 613-619.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Kullbackleibler boosting",
"authors": [
{
"first": "Ce",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Hueng-Yeung",
"middle": [],
"last": "Shum",
"suffix": ""
}
],
"year": 2003,
"venue": "IEEE Computer Society Conference on Computer Vision and Pattern Recognition",
"volume": "1",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ce Liu and Hueng-Yeung Shum. 2003. Kullback- leibler boosting. In 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recog- nition, 2003. Proceedings., volume 1, pages I-I. IEEE.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Visualizing data using t-sne",
"authors": [
{
"first": "Laurens",
"middle": [],
"last": "Van Der Maaten",
"suffix": ""
},
{
"first": "Geoffrey",
"middle": [],
"last": "Hinton",
"suffix": ""
}
],
"year": 2008,
"venue": "Journal of machine learning research",
"volume": "9",
"issue": "",
"pages": "2579--2605",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-sne. Journal of machine learning research, 9(Nov):2579-2605.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "The talkbank project",
"authors": [
{
"first": "Brian",
"middle": [],
"last": "Macwhinney",
"suffix": ""
}
],
"year": 2007,
"venue": "Creating and digitizing language corpora",
"volume": "",
"issue": "",
"pages": "163--180",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Brian MacWhinney. 2007. The talkbank project. In Creating and digitizing language corpora, pages 163-180. Springer.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Nonliteral language in alzheimer dementia: a review",
"authors": [
{
"first": "M",
"middle": [],
"last": "Alexander",
"suffix": ""
},
{
"first": "Barbara",
"middle": [],
"last": "Rapp",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Wild",
"suffix": ""
}
],
"year": 2011,
"venue": "Journal of the International Neuropsychological Society",
"volume": "17",
"issue": "2",
"pages": "207--218",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alexander M Rapp and Barbara Wild. 2011. Nonlit- eral language in alzheimer dementia: a review. Jour- nal of the International Neuropsychological Society, 17(2):207-218.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Features and machine learning classification of connected speech samples from patients with autopsy proven alzheimer's disease with and without additional vascular pathology",
"authors": [
{
"first": "Vassiliki",
"middle": [],
"last": "Rentoumi",
"suffix": ""
},
{
"first": "Ladan",
"middle": [],
"last": "Raoufian",
"suffix": ""
},
{
"first": "Samrah",
"middle": [],
"last": "Ahmed",
"suffix": ""
},
{
"first": "Celeste A De",
"middle": [],
"last": "Jager",
"suffix": ""
},
{
"first": "Peter",
"middle": [],
"last": "Garrard",
"suffix": ""
}
],
"year": 2014,
"venue": "Journal of Alzheimer's Disease",
"volume": "42",
"issue": "s3",
"pages": "3--17",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Vassiliki Rentoumi, Ladan Raoufian, Samrah Ahmed, Celeste A de Jager, and Peter Garrard. 2014. Fea- tures and machine learning classification of con- nected speech samples from patients with autopsy proven alzheimer's disease with and without addi- tional vascular pathology. Journal of Alzheimer's Disease, 42(s3):S3-S17.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Speaking in alzheimer's disease, is that an early sign? importance of changes in language abilities in alzheimer's disease. Frontiers in aging neuroscience",
"authors": [
{
"first": "Greta",
"middle": [],
"last": "Szatloczki",
"suffix": ""
},
{
"first": "Ildiko",
"middle": [],
"last": "Hoffmann",
"suffix": ""
},
{
"first": "Veronika",
"middle": [],
"last": "Vincze",
"suffix": ""
},
{
"first": "Janos",
"middle": [],
"last": "Kalman",
"suffix": ""
},
{
"first": "Magdolna",
"middle": [],
"last": "Pakaski",
"suffix": ""
}
],
"year": 2015,
"venue": "",
"volume": "7",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Greta Szatloczki, Ildiko Hoffmann, Veronika Vincze, Janos Kalman, and Magdolna Pakaski. 2015. Speak- ing in alzheimer's disease, is that an early sign? importance of changes in language abilities in alzheimer's disease. Frontiers in aging neuro- science, 7:195.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "The anatomical bases of semantic retrieval deficits in early alzheimer's disease",
"authors": [
{
"first": "Annalena",
"middle": [],
"last": "Venneri",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "William",
"suffix": ""
},
{
"first": "Heidi",
"middle": [
"M"
],
"last": "Mcgeown",
"suffix": ""
},
{
"first": "Chiara",
"middle": [],
"last": "Hietanen",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Guerrini",
"suffix": ""
},
{
"first": "W",
"middle": [],
"last": "Andrew",
"suffix": ""
},
{
"first": "Michael F",
"middle": [],
"last": "Ellis",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Shanks",
"suffix": ""
}
],
"year": 2008,
"venue": "Neuropsychologia",
"volume": "46",
"issue": "2",
"pages": "497--510",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Annalena Venneri, William J McGeown, Heidi M Hietanen, Chiara Guerrini, Andrew W Ellis, and Michael F Shanks. 2008. The anatomical bases of semantic retrieval deficits in early alzheimer's dis- ease. Neuropsychologia, 46(2):497-510.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"num": null,
"uris": null,
"text": "Correlation matrix of the linguistic features.",
"type_str": "figure"
},
"FIGREF1": {
"num": null,
"uris": null,
"text": ": 2-dimensional visualization of speech transcripts where size of each circle in the map is proportional to the pronoun-to-noun ratio.",
"type_str": "figure"
},
"FIGREF2": {
"num": null,
"uris": null,
"text": "2-dimensional visualization of speech transcripts where size of each circle in the map is proportional to the pronoun frequency.",
"type_str": "figure"
},
"FIGREF3": {
"num": null,
"uris": null,
"text": "2-dimensional visualization of speech transcripts where size of each circle in the map is proportional to the readability score.",
"type_str": "figure"
},
"FIGREF4": {
"num": null,
"uris": null,
"text": "2-dimensional visualization of speech transcripts where size of each circle in the map is proportional to the word repetition frequency.",
"type_str": "figure"
},
"FIGREF5": {
"num": null,
"uris": null,
"text": "Speeches from 1984 to 1986 are detected as abnormal or anomalies.",
"type_str": "figure"
},
"FIGREF7": {
"num": null,
"uris": null,
"text": "Isolation forest algorithm for AD detection.",
"type_str": "figure"
},
"FIGREF8": {
"num": null,
"uris": null,
"text": "Year-wise speeches identified as anomalous.",
"type_str": "figure"
},
"TABREF0": {
"text": "President Reagan's speech dataset",
"type_str": "table",
"num": null,
"content": "<table><tr><td colspan=\"2\">Year No. of speeches</td></tr><tr><td>1980</td><td>6</td></tr><tr><td>1981</td><td>8</td></tr><tr><td>1982</td><td>11</td></tr><tr><td>1983</td><td>12</td></tr><tr><td>1984</td><td>14</td></tr><tr><td>1985</td><td>13</td></tr><tr><td>1986</td><td>14</td></tr><tr><td>1987</td><td>10</td></tr><tr><td>1988</td><td>9</td></tr><tr><td>1989</td><td>1</td></tr></table>",
"html": null
}
}
}
} |