Ori Shapira

27 posts

Ori Shapira

Ori Shapira

@obspp18

NLP research

Bergabung Haziran 2016
155 Mengikuti63 Pengikut
Ori Shapira
Ori Shapira@obspp18·
Web search is changing. At #EMNLP 2021, Ido Dagan gave a keynote about work at @biunlp on multi-text consumption, including "interactive summarization". His slides resemble those in today's presentation by @yusuf_i_mehdi on the new search experience on @bing. Ido said it first😎
Ori Shapira tweet mediaOri Shapira tweet mediaOri Shapira tweet media
English
1
1
18
1.6K
Vered Shwartz
Vered Shwartz@VeredShwartz·
When computing ROUGE score for a single example with multiple predictions (p) and multiple references (r), which of these is more standard to report?
English
2
0
5
3.6K
Ori Shapira
Ori Shapira@obspp18·
@AlvinGrissomII True. That's why we need to first formalize such tasks with evaluation frameworks in a way that the NLP community will understand them (in "NLP words"). Once a task becomes approachable, it will more easily take off. That's exactly what Ido Dagan was advocating in his keynote.
English
0
0
1
0
Yossi Rosenzweig
Yossi Rosenzweig@RosenzweigYossi·
אחרי עשר שנות אקדמיה אני די בטוח שרוב הגוף שלי מורכב מסנדוויץ טונה
עברית
1
0
5
0
Ori Shapira
Ori Shapira@obspp18·
@yanaiela I remember the sense of power when I finally grasped recall/precision, and then feeling I had superpowers when I could use TP/FP/TN/FN freely in speech. I think the latter was reinforced, less through my research, but rather by Covid testing talk a year back :)
English
0
0
1
0
Yanai Elazar
Yanai Elazar@yanaiela·
Yanai, 4th year PhD student in NLP. Today I finally internalized what each of the true/false positive/negative combination mean.
English
4
1
72
0
Ori Shapira
Ori Shapira@obspp18·
@yanaiela זה כאילו נראה מצחיק, אבל בקונטקסט יש כאן הגיון כי זה בטח מקום, כמו משרד, שאנשים דופקים או מצלצלים ורוצים למנוע את הפעולה. זה מראה את חשיבות הסביבה וידע העולם. זה אומר בצורה אימפליציטית "...ואל תצלצל/י בפעמון". זה נשמע כמו מה שאתה עושה במחקר שלך :)
עברית
1
0
4
0
Yanai Elazar
Yanai Elazar@yanaiela·
ואז אנחנו מתפלאים שזה מה שgpt-3 כותב
Yanai Elazar tweet media
עברית
2
0
14
0
Ori Shapira
Ori Shapira@obspp18·
@ani_nenkova The context of this tweet confused me, because "ganon" in Hebrew means nursery school :)
English
1
0
0
0
Ani Nenkova
Ani Nenkova@ani_nenkova·
Langauage and literacy day report from preschool: writing out the word Ganon.
English
1
0
2
0
Ori Shapira
Ori Shapira@obspp18·
@AlexTamkin As you point out, there are many aspects that have advantages/disadvantages with respect to where one is in the PhD (e.g. at the beginning needing more/less of something, and later on the opposite). Fortunately, I have an advisor that dynamically evolved for my needs.
English
0
0
1
0
Alex Tamkin
Alex Tamkin@AlexTamkin·
Hope this thread was helpful, and please feel free to share your own thoughts, whether as an admit, student, or advisor! 🌃12/12
English
12
0
132
0
Alex Tamkin
Alex Tamkin@AlexTamkin·
A quick thread for PhD admits thinking about potential advisors: I see a lot of discussion about "hands-on" vs "hands-off" advisors But I think there are at least 3 underlying dimensions here, each of which is worth considering in its own right: 👇 [THREAD] 1/
English
21
399
1.9K
0
Ori Shapira
Ori Shapira@obspp18·
@ani_nenkova They're lucky to have such a mind and a mensch. Wishing you much success and true fulfillment. You're a real role model!
English
1
0
1
0
Ani Nenkova
Ani Nenkova@ani_nenkova·
I feared starting a new position during the pandemic but find it energizing and a great antidote against covid fatigue. This month I joined Adobe research as the head of lab/Maryland. I look forward to working with the fantastic team, Adobe colleagues and our academic partners.
English
57
7
452
0
Ori Shapira
Ori Shapira@obspp18·
@RTomMcCoy If we knew how to measure actual NLU, let alone automatically, it seems to me like we would also know how to make much better NLP models.
English
0
0
1
0
Tom McCoy
Tom McCoy@RTomMcCoy·
Summarization is ROUGE, MT is BLEU. Can automatic metrics Ever measure NLU?
English
8
18
215
0
Ori Shapira
Ori Shapira@obspp18·
@yoavgo אולי ד"ר הדר רונן תהיה מעוניינת, או שהיא תוכל לקשר למישהו אחר שמתאים.
עברית
0
0
2
0
(((ل()(ل() 'yoav))))👾
אני מחפש מישהי שאוהבת אתגרי ux וטובה בזה, שיצא לה לעבוד/לחשוב בעבר על רכיבי קלט מורכבים, לעבודה לא שגרתית. בקצרה יש לנו שפת שאילתא מורכבת, ואנחנו צריכים לבנות ממשק שיחשוף אותה למשתמשים באופן אינטואיטיבי. צריך לאפיין query builder, אבל אני לא מכיר רכיבים סטנדרטיים להשען עליהם.
עברית
10
9
23
0
Ori Shapira
Ori Shapira@obspp18·
@NatalieShapira יש לילדים שלנו נטייה לשאול שאלה בפורמט של השלמה, לדוגמא: "צריך לנסוע כי", ולצפות לתשובה. אני אומר להם, "אל תעשה לי language model, תשאל אותי שאלה!". אני מרגיש שזה מגיע מתוך עצלות מנטאלית וזה מפריע לי. יכולת ניסוח שאלה זה כלי חשוב.
עברית
1
0
2
0
Natalie Shapira
Natalie Shapira@NatalieShapira·
"אמא למה את עושה לי language model?" (כי שאלתי איפה משהו וקיבלתי תשובה: "במקום" והמשכתי: "שהוא...?") הילד בן 11 והמקסימום שחשפתי בפניו זה מושגים מתורת הקבוצות, אינסוף, עוצמות ובעיות NPC @obspp18 יש לך קשר לזה? כששני ההורים דוקטורנטים למדעי המחשב במעבדה לעיבוד שפות...
עברית
4
1
17
0
Ori Shapira
Ori Shapira@obspp18·
@VeredShwartz Nice! Any non-accepts? I wonder if you've managed to submit even more papers :) Shana Tova!
English
1
0
1
0
Vered Shwartz
Vered Shwartz@VeredShwartz·
My co-authors and I had 4 papers accepted at #emnlp2020, and 2 papers accepted at Findings :) Here are some of them (others will be added later):
English
10
9
210
0
J
J@assoulix·
@yoavgo In our fields like whisky or wine best paper have to endure the maturation period. Best paper award should be done on last year or 2 years then who would have some metric of impact with co citation for example.Moreover those that were'nt reproducible should be banned a posterior
English
1
0
2
0
(((ل()(ل() 'yoav))))👾
even less popular opinion: "best paper" awards should be abolished as well. (yes i know it helps their recipients win jobs against candidates from other fields who have such awards. but it's like, 2-3 people per year? so random)
Djamé..@zehavoc

unpopular opinion: I hate the "outstanding" reviewer awards (even though I got it once for no particular reasons afaict). It's like a giant "meh" thrown to the face of the less outstanding reviewers who threw idk how many hours reviewing papers.

English
7
2
71
0