-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
734 lines (636 loc) · 50.6 KB
/
index.html
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
<!DOCTYPE html>
<html lang="en" >
<!-- Head -->
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" > <!-- Metadata, OpenGraph and Schema.org -->
<!-- Standard metadata -->
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<title>Maria Teleki</title>
<meta name="author" content="Maria Teleki">
<meta name="description" content="Personal website for Maria Teleki.">
<meta name="keywords" content="natural-language-processing, portfoilio-website, maria-teleki, researcher, computer-science">
<!-- Fonts & Icons -->
<!-- <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/@fortawesome/[email protected]/css/all.min.css" integrity="sha256-mUZM63G8m73Mcidfrv5E+Y61y7a12O5mW4ezU3bxqW4=" crossorigin="anonymous"> -->
<link rel="stylesheet" href="./assets/fontawesome-free-5.15.4-web/css/all.min.css"/>
<!-- <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/css/academicons.min.css" integrity="sha256-i1+4qU2G2860dGGIOJscdC30s9beBXjFfzjWLjBRsBg=" crossorigin="anonymous"> -->
<link rel="stylesheet" href="./assets/academicons-1.9.4/css/academicons.min.css"/>
<!-- <link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700|Roboto+Slab:100,300,400,500,700|Material+Icons"> -->
<link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700|Roboto+Slab:100,300,400,500,700|Material+Icons">
<!-- Bootstrap & MDB -->
<!-- <link href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css" rel="stylesheet" integrity="sha256-DF7Zhf293AJxJNTmh5zhoYYIMs2oXitRfBjY+9L//AY=" crossorigin="anonymous"> -->
<link href="./assets/bootstrap-4.6.1-dist/css/bootstrap.min.css" rel="stylesheet">
<!-- <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/css/mdb.min.css" integrity="sha256-jpjYvU3G3N6nrrBwXJoVEYI/0zw8htfFnhT9ljN3JJw=" crossorigin="anonymous"> -->
<link href="./assets/MDB-Free_4.20.0/css/mdb.min.css" rel="stylesheet">
<!-- Local Stylesheet -->
<link rel="stylesheet" href="./main.css">
</head>
<!-- Body -->
<body class="fixed-top-nav">
<!-- Header -->
<header>
<!-- Nav Bar -->
<nav id="navbar" class="navbar navbar-light navbar-expand-sm fixed-top z-depth-0"> <!-- change to bg-white to make it white again -->
<div class="container">
<!-- <div class="navbar-brand social">
Add social icons or brand content here
</div> -->
<!-- Toggler for Mobile -->
<button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation">
<span class="sr-only">Toggle navigation</span>
<span class="icon-bar top-bar"></span>
<span class="icon-bar middle-bar"></span>
<span class="icon-bar bottom-bar"></span>
</button>
<!-- Navbar Links -->
<div class="collapse navbar-collapse" id="navbarNav">
<ul class="navbar-nav ms-auto" style="font-size: medium;">
<li class="nav-item"><a class="nav-link" href="#top">About</a></li>
<li class="nav-item"><a class="nav-link" href="#news_id">News</a></li>
<li class="nav-item"><a class="nav-link" href="#publications_id">Publications</a></li>
<li class="nav-item"><a class="nav-link" href="#education_id">Education</a></li>
<li class="nav-item"><a class="nav-link" href="#teaching_id">Teaching & Mentoring</a></li>
<li class="nav-item"><a class="nav-link" href="#service_id">Service</a></li>
<li class="nav-item"><a class="nav-link" href="#awards_id">Awards</a></li>
<li class="nav-item"><a class="nav-link" href="#invitedtalks_id">Invited Talks</a></li>
<li class="nav-item"><a class="nav-link" href="#work_id">Experience</a></li>
<li class="nav-item"><a class="nav-link" href="#more_id">More</a></li>
</ul>
</div>
</div>
</nav>
</header>
<!-- Content -->
<div class="container mt-5">
<!-- about.html -->
<div class="post">
<header class="post-header">
<h1 style="display: inline;" class="post-title"><span class="font-weight-bold">Maria</span> Teleki <a href="https://translate.google.com/?hl=en&tab=TT&sl=en&tl=hu&text=Maria%20Teleki&op=translate"><img alt=":arxiv:" src="./img/noun-speaker-7514306-6C5B7B.png" height="35" width="35"></a></h1>
<!-- <p style="font-size: medium;">The best way to contact me is via email: <code class="language-plaintext highlighter-rouge">[email protected]</code>.</p> -->
</header>
<article style="margin-top: 20px;">
<div class="profile float-left"><img src="./img/me.jpg" class="img-fluid rounded" width="auto" height="auto" alt="prof_pic.jpg" onerror="this.onerror=null; $('.responsive-img-srcset').remove();">
</div>
<div class="clearfix">
<p style="font-size: medium;">Howdy! I’m a third-year PhD Student in <a href="https://engineering.tamu.edu/cse/index.html" rel="external nofollow noopener noopener noreferrer" target="_blank">Computer Science</a>
at <a href="https://www.tamu.edu/" rel="external nofollow noopener noopener noreferrer" target="_blank">Texas A&M University</a>
(gig em!),
advised by <a href="https://people.engr.tamu.edu/caverlee" rel="external nofollow noopener noopener noreferrer" target="_blank">Prof. James Caverlee</a>.
</p>
<p>
<b>My research focuses on algorithms for natural language processing in spoken contexts.</b>
I work with large language models, automatic speech recognition systems, and psycholinguistic theories.
</p>
<p>
My work is supported by an Avilés-Johnson Fellowship.
</p>
<a style="font-size: xx-large;" href="mailto:[email protected]" title="Email"><i class="fa fa-envelope"></i></a>
<a style="font-size: xx-large;" href="./pdf/Maria_Teleki_CV.pdf" title="CV" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="ai ai-s ai-cv-square"></i></a>
<a style="font-size: xx-large;" href="https://www.linkedin.com/in/mariateleki" title="LinkedIn" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="fab fa-s fa-linkedin"></i></a>
<a style="font-size: xx-large;" href="https://scholar.google.com/citations?hl=en&user=aImLrNcAAAAJ" title="Google Scholar" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="ai ai-s ai-google-scholar"></i></a>
<a style="font-size: xx-large;" href="https://dblp.org/pid/359/3074.html" title="DBLP" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="ai ai-s ai-dblp"></i></a>
<a style="font-size: xx-large;" href="https://github.com/mariateleki" title="GitHub" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="fab fa-s fa-github"></i></a>
<a style="font-size: xx-large;" href="https://www.youtube.com/@MariaTelekiCS" title="YouTube" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="fab fa-s fa-youtube"></i></a>
<a style="font-size: xx-large;" href="https://medium.com/@mariatelekiresearch" title="Medium" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="fab fa-s fa-medium-m"></i></a>
<!-- <a href="https://orcid.org/0009-0006-7015-4015" title="ORCiD" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="ai ai-s ai-orcid"></i></a> -->
<!-- ACL Anthology https://aclanthology.org/people/m/maria-teleki/-->
<!-- Open Review https://openreview.net/profile?id=~Maria_Teleki1 -->
<!-- <a href="." title="arXiv" rel="external nofollow noopener noopener noreferrer" target="_blank"><i class="ai ai-xs ai-arxiv"></i></a> -->
<!-- <p class="desc">PhD Student in <a href="https://engineering.tamu.edu/cse/index.html" rel="external nofollow noopener noopener noreferrer" target="_blank">Computer Science</a> at <a href="https://www.tamu.edu/" rel="external nofollow noopener noopener noreferrer" target="_blank">Texas A&M University</a>.</p> -->
</div>
<div class="news">
<h2 id="news_id">News</h2>
<div class="table-responsive" style="height:100px; overflow-y:scroll;">
<table class="table table-sm table-borderless">
<tr>
<th scope="row">Feb 4, 2025</th>
<td>Excited to guest lecture this week for CSCE 670 on <i>IR <a href="./pdf/week05-class06.pdf">Evaluation</a> and Learning to Rank! 📊</i></td>
</tr>
<tr>
<tr>
<th scope="row">Dec 19, 2024</th>
<td>Our survey on <a href="https://arxiv.org/pdf/2412.14352"><i>LLM Inference-Time Self-Improvement</i></a> is up on arXiv! <img alt=":arxiv:" src="./img/arxiv.png" height="20" width="20"></td>
</tr>
<tr>
<th scope="row">Nov 18, 2024</th>
<td>Our paper -- <a href="./pdf/Masculine_Defaults_via_Gendered_Discourse_in_Podcasts_and_Large_Language_Models.pdf"><i>Masculine Defaults via Gendered Discourse in Podcasts and Large Language Models</i></a> -- was accepted to ICWSM 2025! <img class="emoji" title=":tada:" alt=":tada:" src="https://github.githubassets.com/images/icons/emoji/unicode/1f389.png" height="20" width="20"></td>
</tr>
<tr>
<th scope="row">Oct 7, 2024</th>
<td>Gave a talk at the <a href="https://vbma.biz">Texas Tech University - School of Veterinary Medicine VBMA Club</a>! Check it out here: <i><a href="./pdf/The Other AI.pdf">The Other AI: An Intuitive Understanding of Artificial Intelligence</a></i>. <img alt=":txtech:" src="./img/txtech.png" height="20" width="20"> </td>
</tr>
<tr>
<th scope="row" style="width: 120px;">June 4, 2024</th>
<td>Our work was accepted to INTERSPEECH!<img alt=":interspeech2024:" src="./img/interspeech24.png" height="20" width="20"></td>
</tr>
<tr>
<th scope="row" style="width: 120px;">April 19, 2024</th>
<td>Had a great time meeting and learning from so many awesome people at the <a href="https://cra.org/cra-wp/grad-cohort-for-women/">CRA-WP Grad Cohort for Women</a> in Minneapolis, MN! <img alt=":CRA-WP:" src="./img/CRA-WP.png" height="20" width="20"></td>
</tr>
<tr>
<th scope="row">Mar 14, 2024</th>
<td>We had 2 papers accepted to LREC-COLING on disfluency and language modeling! <img class="emoji" title=":tada:" alt=":tada:" src="https://github.githubassets.com/images/icons/emoji/unicode/1f389.png" height="20" width="20"></td>
</tr>
</table>
</div>
</div>
<h2 id="publications_id">Publications
</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm">
<!-- ITSI Survey Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="85%" src="./img/survey2.png">
</td>
<td style="font-size: large;">A Survey on LLM Inference-Time Self-Improvement<br>
<b style="font-size: medium;">Xiangjue Dong,* <b><u>Maria Teleki</u></b>,* and James Caverlee</div></b><br>
<a href="https://arxiv.org">arXiv 2024<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://arxiv.org/pdf/2412.14352" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-7" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-7" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/dongxiangjue/Awesome-LLM-Self-Improvement"" role="button">GitHub</a>
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="" role="button">Poster</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video Transcript</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Slides</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">AAAI Archive Link</a> -->
<div class="collapse" id="abstract-text-7">
<div class="card card-body z-depth-0" style="font-size: small;">
Techniques that enhance inference through increased computation at test-time have recently gained attention. In this survey, we investigate the current state of LLM Inference-Time Self-Improvement from three different perspectives: Independent Self-improvement, focusing on enhancements via decoding or sampling methods; Context-Aware Self-Improvement, leveraging additional context or datastore; and Model-Aided Self-Improvement, achieving improvement through model collaboration. We provide a comprehensive review of recent relevant studies, contribute an in-depth taxonomy, and discuss challenges and limitations, offering insights for future research.
</div>
</div>
<div class="collapse" id="bib-text-7">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{dong24_survey,<br>
title = {A Survey on LLM Inference-Time Self-Improvement},<br>
author = {Xiangjue Dong and Maria Teleki and James Caverlee},<br>
year = {2024},<br>
booktitle = {arXiv}<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- End ITSI Survey Paper-->
<!-- Gendered Discourse Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="85%" src="./img/icwsm25_diagram.png">
</td>
<td style="font-size: large;">Masculine Defaults via Gendered Discourse in Podcasts and Large Language Models<br>
<b style="font-size: medium;"><b><u>Maria Teleki</u></b>, Xiangjue Dong, Haoran Liu, and James Caverlee</div></b><br>
<a href="https://www.icwsm.org/2025/index.html">ICWSM 2025<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/Masculine_Defaults_via_Gendered_Discourse_in_Podcasts_and_Large_Language_Models.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-6" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-6" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/mariateleki/masculine-defaults"" role="button">Code</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.gendered-discourse.net" role="button">Project Website</a>
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="" role="button">Poster</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video Transcript</a> -->
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/MasculineDefaultsSlides.pdf" role="button">Slides</a>
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">AAAI Archive Link</a> -->
<div class="collapse" id="abstract-text-6">
<div class="card card-body z-depth-0" style="font-size: small;">
Masculine defaults are widely recognized as a significant type of gender bias, but they are often unseen as they are under-researched (Cheryan and Markus 2020). Masculine defaults involve three key parts: (i) the cultural context, (ii) the masculine characteristics or behaviors, and (iii) the reward for, or simply acceptance of, those masculine characteristics or behaviors. In this work, we study discourse-based masculine defaults, and propose a twofold framework for (i) the large-scale discovery and analysis of gendered discourse words in spoken content via our Gendered Discourse Correlation Framework (GDCF); and (ii) the measurement of the gender bias associated with these gendered discourse words in LLMs via our Discourse Word-Embedding Association Test (D-WEAT). We focus our study on podcasts, a popular and growing form of social media, analyzing 15,117 podcast episodes. We analyze correlations between gender and discourse words -- discovered via LDA and BERTopic -- to automatically form gendered discourse word lists. We then study the prevalence of these gendered discourse words in domain-specific contexts, and find that gendered discourse-based masculine defaults exist in the domains of business, technology/politics, and video games. Next, we study the representation of these gendered discourse words from a state-of-the-art LLM embedding model from OpenAI, and find that the masculine discourse words have a more stable and robust representation than the feminine discourse words, which may result in better system performance on downstream tasks for men. Hence, men are rewarded for their discourse patterns with better system performance by one of the state-of-the-art language models -- and this embedding disparity constitutes a representational harm and a masculine default.
</div>
</div>
<div class="collapse" id="bib-text-6">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{teleki25_icwsm,<br>
title = {Masculine Defaults via Gendered Discourse in Podcasts and Large Language Models},<br>
author = {Maria Teleki and Xiangjue Dong and Haoran Liu and James Caverlee},<br>
year = {2025},<br>
booktitle = {ICWSM 2025}<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- End Gendered Discourse Paper-->
<!-- Comparing ASR Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="95%" src="./img/interspeech_diagram.png">
</td>
<td style="font-size: large;">Comparing ASR Systems in the Context of Speech Disfluencies<br>
<b style="font-size: medium;"><b><u>Maria Teleki</u></b>, Xiangjue Dong, Soohwan Kim, and James Caverlee</div></b><br>
<a href="https://interspeech2024.org/">INTERSPEECH 2024<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.isca-archive.org/interspeech_2024/teleki24_interspeech.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-5" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-5" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/mariateleki/Comparing-ASR-Systems" role="button">Code</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.comparing-asr-systems.com" role="button">Project Website</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/Resized-Interspeech-2024-Poster.pdf" role="button">Poster</a>
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Video Transcript</a> -->
<!-- <a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="." role="button">Slides</a> -->
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.isca-archive.org/interspeech_2024/teleki24_interspeech.html" role="button">ISCA Archive Link</a>
<div class="collapse" id="abstract-text-5">
<div class="card card-body z-depth-0" style="font-size: small;">
In this work, we evaluate the disfluency capabilities of two automatic speech recognition systems -- Google ASR and WhisperX -- through a study of 10 human-annotated podcast episodes and a larger set of 82,601 podcast episodes. We employ a state-of-the-art disfluency annotation model to perform a fine-grained analysis of the disfluencies in both the scripted and non-scripted podcasts. We find, on the set of 10 podcasts, that while WhisperX overall tends to perform better, Google ASR outperforms in WIL and BLEU scores for non-scripted podcasts. We also find that Google ASR's transcripts tend to contain closer to the ground truth number of edited-type disfluent nodes, while WhisperX's transcripts are closer for interjection-type disfluent nodes. This same pattern is present in the larger set. Our findings have implications for the choice of an ASR model when building a larger system, as the choice should be made depending on the distribution of disfluent nodes present in the data.
</div>
</div>
<div class="collapse" id="bib-text-5">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{teleki24_interspeech,<br>
title = {Comparing ASR Systems in the Context of Speech Disfluencies},<br>
author = {Maria Teleki and Xiangjue Dong and Soohwan Kim and James Caverlee},<br>
year = {2024},<br>
booktitle = {Interspeech 2024},<br>
pages = {4548--4552},<br>
doi = {10.21437/Interspeech.2024-1270},<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- End Comparing ASR Paper-->
<!-- Quantifying Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="95%" src="./img/Quantifying.png">
</td>
<td style="font-size: large;">Quantifying the Impact of Disfluency on Spoken Content Summarization<br>
<b style="font-size: medium;"><b><u>Maria Teleki</u></b>, Xiangjue Dong, and James Caverlee</div></b><br>
<a href="https://lrec-coling-2024.org/">LREC-COLING 2024<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://aclanthology.org/2024.lrec-main.1175.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-4" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-4" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/mariateleki/Quantifying-Impact-Disfluency" role="button">Code</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/Quantifying LREC-COLING Poster.pdf" role="button">Poster</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.youtube.com/watch?v=iTlJiEHN5Rk" role="button">Video</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/Quantifying LREC-COLING Slides.pdf" role="button">Slides</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://aclanthology.org/2024.lrec-main.1175/" role="button">ACL Anthology Link</a>
<div class="collapse" id="abstract-text-4">
<div class="card card-body z-depth-0" style="font-size: small;">
Spoken content is abundant -- including podcasts, meeting transcripts, and TikTok-like short videos. And yet, many important tasks like summarization are often designed for written content rather than the looser, noiser, and more disfluent style of spoken content. Hence, we aim in this paper to quantify the impact of disfluency on spoken content summarization. Do disfluencies negatively impact the quality of summaries generated by existing approaches? And if so, to what degree? Coupled with these goals, we also investigate two methods towards improving summarization in the presence of such disfluencies. We find that summarization quality does degrade with an increase in these disfluencies and that a combination of multiple disfluency types leads to even greater degradation. Further, our experimental results show that naively removing disfluencies and augmenting with special tags can worsen the summarization when used for testing, but that removing disfluencies for fine-tuning yields the best results. We make the code available at https://github.com/mariateleki/Quantifying-Impact-Disfluency.
</div>
</div>
<div class="collapse" id="bib-text-4">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{teleki-etal-2024-quantifying-impact,<br>
title = "Quantifying the Impact of Disfluency on Spoken Content Summarization",<br>
author = "Teleki, Maria and<br>
Dong, Xiangjue and<br>
Caverlee, James",<br>
editor = "Calzolari, Nicoletta and<br>
Kan, Min-Yen and<br>
Hoste, Veronique and<br>
Lenci, Alessandro and<br>
Sakti, Sakriani and<br>
Xue, Nianwen",<br>
booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",<br>
month = may,<br>
year = "2024",<br>
address = "Torino, Italia",<br>
publisher = "ELRA and ICCL",<br>
url = "https://aclanthology.org/2024.lrec-main.1175",<br>
pages = "13419--13428",<br>
abstract = "Spoken content is abundant {--} including podcasts, meeting transcripts, and TikTok-like short videos. And yet, many important tasks like summarization are often designed for written content rather than the looser, noiser, and more disfluent style of spoken content. Hence, we aim in this paper to quantify the impact of disfluency on spoken content summarization. Do disfluencies negatively impact the quality of summaries generated by existing approaches? And if so, to what degree? Coupled with these goals, we also investigate two methods towards improving summarization in the presence of such disfluencies. We find that summarization quality does degrade with an increase in these disfluencies and that a combination of multiple disfluency types leads to even greater degradation. Further, our experimental results show that naively removing disfluencies and augmenting with special tags can worsen the summarization when used for testing, but that removing disfluencies for fine-tuning yields the best results. We make the code available at https://github.com/mariateleki/Quantifying-Impact-Disfluency.",<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- End Quantifying Paper-->
<!-- DACL Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="95%" src="./img/DACL.png"></div>
</td>
<td style="font-size: large;">DACL: Disfluency Augmented Curriculum Learning for Fluent Text Generation<br>
<b style="font-size: medium;">Rohan Chaudhury, <b><u>Maria Teleki</u></b>, Xiangjue Dong, and James Caverlee</div></b><br>
<a href="https://lrec-coling-2024.org/">LREC-COLING 2024<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://aclanthology.org/2024.lrec-main.385.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-3" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-3" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/Rohan-Chaudhury/Generating-Fluent-Text-through-Curriculum-Learning-And-Disfluency-Augmentation" role="button">Code</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/DACL LREC-COLING Poster.pdf" role="button">Poster</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://youtu.be/8VIDlocdaco?si=35wXmBLOMNbrYz1X" role="button">Video</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/DACL Presentation.pdf" role="button">Slides</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://aclanthology.org/2024.lrec-main.385/" role="button">ACL Anthology Link</a>
<div class="collapse" id="abstract-text-3">
<div class="card card-body z-depth-0" style="font-size: small;">
Voice-driven software systems are in abundance. However, language models that power these systems are traditionally trained on fluent, written text corpora. Hence there can be a misalignment between the inherent disfluency of transcribed spoken content and the fluency of the written training data. Furthermore, gold-standard disfluency annotations of various complexities for incremental training can be expensive to collect. So, we propose in this paper a Disfluency Augmented Curriculum Learning (DACL) approach to tackle the complex structure of disfluent sentences and generate fluent texts from them, by using Curriculum Learning (CL) coupled with our synthetically augmented disfluent texts of various levels. DACL harnesses the tiered structure of our generated synthetic disfluent data using CL, by training the model on basic samples (i.e. more fluent) first before training it on more complex samples (i.e. more disfluent). In contrast to the random data exposure paradigm, DACL focuses on a simple-to-complex learning process. We comprehensively evaluate DACL on Switchboard Penn Treebank-3 and compare it to the state-of-the-art disfluency removal models. Our model surpasses existing techniques in word-based precision (by up to 1%) and has shown favorable recall and F1 scores.
</div>
</div>
<div class="collapse" id="bib-text-3">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{chaudhury-etal-2024-dacl-disfluency,<br>
title = "{DACL}: Disfluency Augmented Curriculum Learning for Fluent Text Generation",<br>
author = "Chaudhury, Rohan and<br>
Teleki, Maria and<br>
Dong, Xiangjue and<br>
Caverlee, James",<br>
editor = "Calzolari, Nicoletta and<br>
Kan, Min-Yen and<br>
Hoste, Veronique and<br>
Lenci, Alessandro and<br>
Sakti, Sakriani and<br>
Xue, Nianwen",<br>
booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",<br>
month = may,<br>
year = "2024",<br>
address = "Torino, Italia",<br>
publisher = "ELRA and ICCL",<br>
url = "https://aclanthology.org/2024.lrec-main.385",<br>
pages = "4311--4321",<br>
abstract = "Voice-driven software systems are in abundance. However, language models that power these systems are traditionally trained on fluent, written text corpora. Hence there can be a misalignment between the inherent disfluency of transcribed spoken content and the fluency of the written training data. Furthermore, gold-standard disfluency annotations of various complexities for incremental training can be expensive to collect. So, we propose in this paper a Disfluency Augmented Curriculum Learning (DACL) approach to tackle the complex structure of disfluent sentences and generate fluent texts from them, by using Curriculum Learning (CL) coupled with our synthetically augmented disfluent texts of various levels. DACL harnesses the tiered structure of our generated synthetic disfluent data using CL, by training the model on basic samples (i.e. more fluent) first before training it on more complex samples (i.e. more disfluent). In contrast to the random data exposure paradigm, DACL focuses on a simple-to-complex learning process. We comprehensively evaluate DACL on Switchboard Penn Treebank-3 and compare it to the state-of-the-art disfluency removal models. Our model surpasses existing techniques in word-based precision (by up to 1{\%}) and has shown favorable recall and F1 scores.",<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- End DACL Paper-->
<!-- Co2PT Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="95%" src="./img/diagram3.png"></div>
</td>
<td style="font-size: large;">Co2PT: Mitigating Bias in Pre-trained Language Models through Counterfactual Contrastive Prompt Tuning<br>
<b style="font-size: medium;">Xiangjue Dong, Ziwei Zhu, Zhuoer Wang, <b><u>Maria Teleki</u></b>, and James Caverlee</div></b><br>
<a href="https://2023.emnlp.org/program/">Findings of EMNLP 2023<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/co2pt-paper.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-2" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-2" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://github.com/dongxiangjue/Co2PT" role="button">Code</a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://aclanthology.org/2023.findings-emnlp.390/" role="button">ACL Anthology Link</a>
<div class="collapse" id="abstract-text-2">
<div class="card card-body z-depth-0" style="font-size: small;">
Pre-trained Language Models are widely used in many important real-world applications. However, recent studies show that these models can encode social biases from large pre-training corpora and even amplify biases in downstream applications. To address this challenge, we propose Co2PT, an efficient and effective debias-while-prompt tuning method for mitigating biases via counterfactual contrastive prompt tuning on downstream tasks. Our experiments conducted on three extrinsic bias benchmarks demonstrate the effectiveness of Co2PT on bias mitigation during the prompt tuning process and its adaptability to existing upstream debiased language models. These findings indicate the strength of Co2PT and provide promising avenues for further enhancement in bias mitigation on downstream tasks.
</div>
</div>
<div class="collapse" id="bib-text-2">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{dong-etal-2023-co2pt,<br>
title = "{C}o$^2${PT}: Mitigating Bias in Pre-trained Language Models through Counterfactual Contrastive Prompt Tuning",<br>
author = "Dong, Xiangjue and<br>
Zhu, Ziwei and<br>
Wang, Zhuoer and<br>
Teleki, Maria and<br>
Caverlee, James",<br>
editor = "Bouamor, Houda and<br>
Pino, Juan and<br>
Bali, Kalika",<br>
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",<br>
month = dec,<br>
year = "2023",<br>
address = "Singapore",<br>
publisher = "Association for Computational Linguistics",<br>
url = "https://aclanthology.org/2023.findings-emnlp.390",<br>
doi = "10.18653/v1/2023.findings-emnlp.390",<br>
pages = "5859--5871",<br>
abstract = "Pre-trained Language Models are widely used in many important real-world applications. However, recent studies show that these models can encode social biases from large pre-training corpora and even amplify biases in downstream applications. To address this challenge, we propose Co$^2$PT, an efficient and effective *debias-while-prompt tuning* method for mitigating biases via counterfactual contrastive prompt tuning on downstream tasks. Our experiments conducted on three extrinsic bias benchmarks demonstrate the effectiveness of Co$^2$PT on bias mitigation during the prompt tuning process and its adaptability to existing upstream debiased language models. These findings indicate the strength of Co$^2$PT and provide promising avenues for further enhancement in bias mitigation on downstream tasks.",
}
</code>
</div>
</div>
</td>
</tr>
<!-- End Co2PT Paper-->
<!-- Alexa Prize Paper-->
<tr>
<td style="width: 35%; vertical-align: top;">
<img class="preview rounded z-depth-0" width="95%" src="./img/taskbot.png"></div>
</td>
<td style="font-size: large;">Howdy Y’all: An Alexa TaskBot<br>
<b style="font-size: medium;">Majid Alfifi, Xiangjue Dong, Timo Feldman, Allen Lin, Karthic Madanagopal, Aditya Pethe, <b><u>Maria Teleki</u></b>, Zhuoer Wang, Ziwei Zhu, James Caverlee</div></b><br>
<a href="https://www.amazon.science/alexa-prize/proceedings/howdy-yall-an-alexa-taskbot">Alexa Prize TaskBot Challenge Proceedings 2022<br></a>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="./pdf/howdy-yall-paper.pdf" role="button">Paper</a>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#abstract-text-1" aria-expanded="false" aria-controls="collapseExample">Abstract</button>
<button class="btn btn-light btn-sm z-depth-0 text-dark" type="button" data-toggle="collapse" data-target="#bib-text-1" aria-expanded="false" aria-controls="collapseExample">Bib</button>
<a class="btn btn-light btn-sm z-depth-0 text-dark" style="text-decoration: none;" href="https://www.amazon.science/academic-engagements/ten-university-teams-selected-to-participate-in-alexa-prize-taskbot-challenge" role="button">Amazon Science Link</a>
<div class="collapse" id="abstract-text-1">
<div class="card card-body z-depth-0" style="font-size: small;">
In this paper, we present Howdy Y’all, a multi-modal task-oriented dialogue agent developed for the 2021-2022 Alexa Prize TaskBot competition. Our design principles guiding Howdy Y’all aim for high user satisfaction through friendly and trustworthy encounters, minimization of negative conversation edge cases, and wide coverage over many tasks. Hence, Howdy Y’all is built upon a rapid prototyping platform to enable fast experimentation and powered by four key innovations to enable this vision: (i) First, it combines a rules, phonetic matching, and a transformer-based approach for robust intent understanding. (ii) Second, to accurately elicit user preferences and guide users to the right task, Howdy Y’all is powered by a contrastive learning search framework over sentence embeddings and a conversational recommender for eliciting preferences. (iii) Third, to support a variety of user question types, it introduces a new data augmentation method for question generation and a self-supervised answer selection approach for improving question answering. (iv) Finally, to help motivate our users and keep them engaged, we design an emotional conversation tracker that provides empathetic responses to keep users engaged and a monitor of conversation quality.
</div>
</div>
<div class="collapse" id="bib-text-1">
<div class="card card-body z-depth-0">
<code class="language-plaintext highlighter-rouge" style="font-size: small;">
@inproceedings{University2022,<br>
author={Alfifi, Majid and Dong, Xiangjue and Feldman, Timo and Lin, Allen and Madanagopal, Karthic and Pethe, Aditya and Teleki, Maria and Wang, Zhuoer and Zhu, Ziwei and Caverlee, James},<br>
title = {Howdy Y’all: An Alexa TaskBot},<br>
year = {2022},<br>
url = {https://www.amazon.science/alexa-prize/proceedings/howdy-yall-an-alexa-taskbot},<br>
booktitle = {Alexa Prize TaskBot Challenge Proceedings},<br>
}
</code>
</div>
</div>
</td>
</tr>
<!-- Alexa Prize Paper-->
</table>
</div>
<div class="education">
<h2 id="education_id">Education</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm" id="edu_table">
<style>
#edu_table td {
padding: 1px;
vertical-align: top;
}
</style>
<tr>
<td style="width: 20%;"><i>(2022 - Present)</i></td>
<td><b>PhD Computer Science at Texas A&M University</b></td>
</tr>
<tr>
<td style="width: 20%;"><i>(2017 - 2022)</i></td>
<td><b>B.S. Computer Science at Texas A&M University</b> -- <i>Summa Cum Laude</i></td>
</tr>
</table>
</div>
</div>
<div class="teaching">
<!-- SYMBOLS:
Clover = ♣
Big Star = ★
Triangle = ▲
Diamond = ◆
-->
<h2 id="teaching_id">Teaching & Mentoring</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm">
<p style="font-family: 'Roboto', cursive; font-size: 14px;">
★ indicates that the student was an author on a published paper during the mentorship.
♣ indicates that the student had no publications prior to mentorship.
▲ indicates that the student completed their thesis during the mentorship.
◆ indicates that the student received course credit as part of the mentorship (i.e. CSCE 485, CSCE 691).<br>
</p>
<b>MS Students</b>
<br><a href="https://www.linkedin.com/in/chaudhury-rohan/">Rohan Chaudhury</a> [★♣▲] – <i>First Employment: Amazon</i>
<br><a href="https://www.linkedin.com/in/saitejas-janjur/">Sai Janjur</a> [♣]
<br><br>
<b>Undergraduate Students</b>
<br><a href="https://www.linkedin.com/in/soohwan-kim-8724801bb/">Soohwan Kim</a> [♣◆] – <i>First Employment: UPS</i>
<br><a href="https://olivergrabner.github.io/">Oliver Grabner</a> [♣]
<br><a href="https://www.linkedin.com/in/thomas-docog-789236280/">Thomas Docog</a> [♣]
</table>
</div>
</div>
<div class="service">
<h2 id="service_id">Service</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm">
Program Committee (Reviewer) for <a href="https://aclrollingreview.org/"><b>ACL ARR</b></a>: Aug '24, Oct '24, Dec '24<br>
Program Committee (Reviewer) for <a href="https://www.icwsm.org/2024/index.html/"><b>ICWSM</b></a>: Jan '24, May '24, Sep '24<br>
External Program Committee (External Reviewer) for <a href="https://recsys.acm.org/recsys24/"><b>RecSys</b></a>: '24
</table>
</div>
</div>
<div class="education">
<h2 id="awards_id">Awards</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm" style="border-spacing: 0; border-collapse: collapse;" id="award_table">
<style>
#award_table td {
padding: 1px;
vertical-align: top;
}
</style>
<tr>
<td style="width: 20%;"><i>(2022-2026)</i></td>
<td>Dr. Dionel Avilés ’53 and Dr. James Johnson ’67 Fellowship in Computer Science and Engineering</b></td>
</tr>
<tr>
<td style="width: 20%;"><i>(Spring 2024)</i></td>
<td><a href="https://cra.org/cra-wp/grad-cohort-for-women/">CRA-WP Grad Cohort for Women</a></b></td>
</tr>
<tr>
<td style="width: 20%;"><i>(Spring 2024)</i></td>
<td>Department of Computer Science & Engineering Travel Grant</b></td>
</tr>
<tr>
<td style="width: 20%;"><i>(2017-2021)</i></td>
<td>President's Endowed Scholarship</b></td>
</tr>
<tr>
<td style="width: 20%;""><i>(2018)</i></td>
<td>Bertha & Samuel Martin Scholarship</b></td>
</tr>
</table>
</div>
</div>
<div class="Invited Talks">
<h2 id="invitedtalks_id">Invited Talks</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm" id="talk_table">
<style>
#talk_table td {
padding: 1px;
vertical-align: top;
}
</style>
<tr>
<td style="width: 20%;"><i>(Spring 2025)</i></td>
<td style="white-space: pre-wrap;"><i>Guest Lectures on IR <a href="./pdf/week05-class06.pdf">Evaluation</a> and Learning to Rank</i> @ Texas A&M University - CSCE 670</td>
</tr>
<tr>
<td style="width: 20%;"><i>(Fall 2024)</i></td>
<td style="white-space: pre-wrap;"><i><a href="./pdf/The Other AI.pdf">The Other AI: An Intuitive Understanding of Artificial Intelligence</a></i> @ <a href="https://vbma.biz">Texas Tech University - School of Veterinary Medicine VBMA Club</a></td>
</tr>
</table>
</div>
</div>
<div class="work">
<h2 id="work_id">Experience</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm">
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">RetailMeNot</b><br>
<b style="font-size: large;">Software Engineering Intern</b><br>
Austin, TX<br>
May 2021 - August 2021
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;">Used Amazon SageMaker and spaCy to get BERT embeddings for concatenated coupon titles and descriptions. <b>Analyzed the relationship between each dimension of the BERT embeddings and uCTR</b> using Spearman's correlation coefficient, and used principal component analysis to find dimensions with stronger correlations. Created a plan to evaluate these dimensions as possible features for the Ranker algorithm--which does store page coupon ranking--using offline analysis and A/B testing. Taught Data Science Guilds about neural networks, word embeddings, and spaCy.</td>
</tr>
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">The Hi, How Are You Project</b><br>
<b style="font-size: large;">Volunteer</b><br>
Austin, TX<br>
May 2020 - Dec 2020
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;"><b>Developed the “Friendly Frog” Alexa Skill</b> with the organization at the beginning of the COVID-19 pandemic to promote mental health by reading uplifting Daniel Johnston lyrics and the organization’s “Happy Habits.”</td>
</tr>
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">RetailMeNot</b><br>
<b style="font-size: large;">Software Engineering Intern</b><br>
Austin, TX<br>
May 2020 - August 2020
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;"><b>Developed the “RetailMeNot DealFinder” Alexa Skill</b> to help users activate cash back offers. Presented on Alexa Skill Development at the Data Science Sandbox with both Valassis and RetailMeNot teams.</td>
</tr>
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">Texas A&M University</b><br>
<b style="font-size: large;">Peer Teacher</b><br>
Dec 2018 - Dec 2019
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;">Helped students with programming homework and answered conceptual questions by hosting office hours and <b>assisting at lab sessions for CSCE 121 and 181</b>. Created notes with exercises and examples to work through as a group during CSCE 121 reviews.</td>
</tr>
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">Silicon Labs</b><br>
<b style="font-size: large;">Applications Engineering Intern</b><br>
Austin, TX<br>
May 2019 - August 2019
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;"><b>Designed and implemented the Snooper library</b> using pandas to (1) systemize IC bus traffic snooping (I2C, UART, SPI, etc.) across different snooping devices (Saleae, Beagle, etc.), and (2) translate the traffic to a human-readable form for debugging purposes. Responded to multiple tickets from customers using the library.</td>
</tr>
<tr>
<td style="width: 35%; vertical-align: top;">
<b style="font-size: large;">The Y (YMCA)</b><br>
<b style="font-size: large;">Afterschool Instructor</b><br>
Sep 2016 - July 2017
</td>
<td style="white-space: pre-wrap; vertical-align: top; font-size: medium;">Taught multiple weekly classes at local elementary schools for the YMCA Afterschool program, and authored Lego Mindstorms Robotics and “Crazy Science” instruction manuals for the program.</td>
</tr>
</table>
</div>
<div class="more">
<h2 id="more_id">More</h2>
<div class="table-responsive-sm">
<table class="table table-borderless table-sm" id="edu_table">
I have a dog named Apollo 🐶
Pics: <a href="img/apollo-smile.png">[smiling]</a>
<a href="img/apollo-pinetrees.png">[pine-trees]</a>
<a href="img/apollo-generated1.png">[generated-1]</a>
<a href="img/apollo-generated2.png">[generated-2]</a>.
</table>
</div>
</article>
<!-- Footer -->
<footer class="fixed-bottom bg-dark text-white">
<div class="container mt-0">© Copyright 2024 Maria Teleki. Based on <a href="https://github.com/alshedivat/al-folio" rel="external nofollow noopener noopener noreferrer" target="_blank">al-folio</a> theme. <a href="https://github.com/mariateleki/mariateleki.github.io" rel="external nofollow noopener noopener noreferrer" target="_blank">Code available at my GitHub.</a></div>
</footer>
<!-- JavaScripts -->
<!-- jQuery -->
<!-- <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/jquery.min.js" integrity="sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4=" crossorigin="anonymous"></script>-->
<script src="./assets/jquery/dist/jquery.min.js"></script>
<!-- Bootsrap & MDB scripts -->
<!-- <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.bundle.min.js" integrity="sha256-fgLAgv7fyCGopR/gBNq2iW3ZKIdqIcyshnUULC4vex8=" crossorigin="anonymous"></script> -->
<script src="./assets/bootstrap-4.6.1-dist/js/bootstrap.bundle.min.js"></script>
<!-- <script src="https://cdn.jsdelivr.net/npm/[email protected]/js/mdb.min.js" integrity="sha256-NdbiivsvWt7VYCt6hYNT3h/th9vSTL4EDWeGs5SN3DA=" crossorigin="anonymous"></script> -->
<script src="./assets/MDB-Free_4.20.0/js/mdb.min.js"></script>
</body>
</html>