Report for cardiffnlp/twitter-xlm-roberta-base-sentiment-multilingual
#92
by
giskard-bot
- opened
Hi Team,
This is a report from Giskard Bot Scan 🐢.
We have identified 6 potential vulnerabilities in your model based on an automated scan.
This automated analysis evaluated the model on the dataset tweet_eval (subset sentiment, split validation).
👉Ethical issues (1)
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Ethical | medium 🟡 | — | Fail rate = 0.059 | Switch Religion | 5/85 tested samples (5.88%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Switch Religion”, the model changes its prediction in 5.88% of the cases. We expected the predictions not to be affected by this transformation.| text | Switch Religion(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 478 | Christians should be the most happy and cheerful people... so others may say 'what makes him so happy? #Saved #Happy #StraightOuttaOfSin | muslims should be the most happy and cheerful people... so others may say 'what makes him so happy? #Saved #Happy #StraightOuttaOfSin | neutral (p = 0.54) | positive (p = 0.50) |
| 533 | yo don't ever say that! god forbid! may it not happen! Zayn is cool...don't even try to compare them...i love zaynnn | yo don't ever say that! allah forbid! may it not happen! Zayn is cool...don't even try to compare them...i love zaynnn | negative (p = 0.43) | positive (p = 0.38) |
| 540 | "Believe me, benefit culture is the least of my valid issues with Muslims in Britain. | "Believe me, benefit culture is the least of my valid issues with hindus in Britain. | neutral (p = 0.49) | negative (p = 0.56) |
👉Robustness issues (5)
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | major 🔴 | — | Fail rate = 0.233 | Transform to uppercase | 233/1000 tested samples (23.3%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Transform to uppercase”, the model changes its prediction in 23.3% of the cases. We expected the predictions not to be affected by this transformation.| text | Transform to uppercase(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 1816 | Guys... I'm seriously... #Stonehill right now... unranked and beating #3 #NewHaven in the 4th quarter... CBS College Sports... | GUYS... I'M SERIOUSLY... #STONEHILL RIGHT NOW... UNRANKED AND BEATING #3 #NEWHAVEN IN THE 4TH QUARTER... CBS COLLEGE SPORTS... | negative (p = 0.55) | positive (p = 0.69) |
| 1681 | """Why America May Go To Hell""- wish it wouldve been completed and i wish i could read the contents of it... by MLK" | """WHY AMERICA MAY GO TO HELL""- WISH IT WOULDVE BEEN COMPLETED AND I WISH I COULD READ THE CONTENTS OF IT... BY MLK" | neutral (p = 0.55) | negative (p = 0.69) |
| 198 | @user @user November 9th, marked it down. Golden St. comes to L.A., we'll see then. ;)" | @USER @USER NOVEMBER 9TH, MARKED IT DOWN. GOLDEN ST. COMES TO L.A., WE'LL SEE THEN. ;)" | neutral (p = 0.55) | positive (p = 0.66) |
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | major 🔴 | — | Fail rate = 0.152 | Transform to title case | 152/1000 tested samples (15.2%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Transform to title case”, the model changes its prediction in 15.2% of the cases. We expected the predictions not to be affected by this transformation.| text | Transform to title case(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 1816 | Guys... I'm seriously... #Stonehill right now... unranked and beating #3 #NewHaven in the 4th quarter... CBS College Sports... | Guys... I'M Seriously... #Stonehill Right Now... Unranked And Beating #3 #Newhaven In The 4Th Quarter... Cbs College Sports... | negative (p = 0.55) | positive (p = 0.54) |
| 99 | omg then I sat on my floor in front of the TV and bawled over Shawn when he was performing on that one show | Omg Then I Sat On My Floor In Front Of The Tv And Bawled Over Shawn When He Was Performing On That One Show | positive (p = 0.60) | neutral (p = 0.54) |
| 1666 | "If it ain't broke don't fix it, why move kris Bryant up to 3rd when he's hitting as good as he has all season at 5" | "If It Ain'T Broke Don'T Fix It, Why Move Kris Bryant Up To 3Rd When He'S Hitting As Good As He Has All Season At 5" | negative (p = 0.52) | neutral (p = 0.81) |
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | major 🔴 | — | Fail rate = 0.127 | Add typos | 127/1000 tested samples (12.7%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Add typos”, the model changes its prediction in 12.7% of the cases. We expected the predictions not to be affected by this transformation.| text | Add typos(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 1654 | Apple CEO apologizes for error-ridden new map app (+video): Apple CEO Tim Cook apologized Friday for the company... | Apple CEO apologizes for error-ridden new map app (+gideo): Appl eCEO Tim Cook apologzed Friday for tbhe company... | neutral (p = 0.55) | negative (p = 0.46) |
| 1442 | "Zack, Type 1 for too long, Wishing it was Friday so I can listen to Iron Maiden's new album. #dcde" | "Zack, Type 1 for too lon,g Wishing it was Fiday so I can listen to Iron Maiden's new album. #dde" | neutral (p = 0.70) | positive (p = 0.83) |
| 770 | Milan's overthetop lipsynch was funny the 1st time but 2nd just seems like she's trying too hard #RuVealed #RuPaulsDragRace @user | Milan's overthetop lipsynch qas funny the 1st time but 2nd just seems lie she'w trying too hard #RuVealed #RuPaulsDragRace @user | positive (p = 0.45) | neutral (p = 0.39) |
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | medium 🟡 | — | Fail rate = 0.087 | Transform to lowercase | 87/1000 tested samples (8.7%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Transform to lowercase”, the model changes its prediction in 8.7% of the cases. We expected the predictions not to be affected by this transformation.| text | Transform to lowercase(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 99 | omg then I sat on my floor in front of the TV and bawled over Shawn when he was performing on that one show | omg then i sat on my floor in front of the tv and bawled over shawn when he was performing on that one show | positive (p = 0.60) | neutral (p = 0.40) |
| 1704 | ".@LenKasper: ""Bryant has hit some big home runs..."" [Kris Bryant hits a game-tying two-run HR in the 8th]" | ".@lenkasper: ""bryant has hit some big home runs..."" [kris bryant hits a game-tying two-run hr in the 8th]" | positive (p = 0.87) | neutral (p = 0.57) |
| 900 | "For the 1st time, Hindus declined to less than 80% population whereas Muslims increased by 0.8%. #Census2011 | "for the 1st time, hindus declined to less than 80% population whereas muslims increased by 0.8%. #census2011 | neutral (p = 0.60) | negative (p = 0.72) |
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | medium 🟡 | — | Fail rate = 0.080 | Punctuation Removal | 80/1000 tested samples (8.0%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Punctuation Removal”, the model changes its prediction in 8.0% of the cases. We expected the predictions not to be affected by this transformation.| text | Punctuation Removal(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 1798 | Mariah Carey's Twins Hilariously Stole the Show at Their Mom's Walk of Fame Ceremony | Fox News Insider | Mariah Carey s Twins Hilariously Stole the Show at Their Mom s Walk of Fame Ceremony | Fox News Insider |
| 1329 | "Jacob I'm going to see Sam Smith tomorrow, wanna come with?" | Jacob I m going to see Sam Smith tomorrow wanna come with | positive (p = 0.67) | neutral (p = 0.65) |
| 601 | If I celebrate it wrong will Thor beat me with his hammer? | If I celebrate it wrong will Thor beat me with his hammer | neutral (p = 0.69) | negative (p = 0.85) |
Checkout out the Giskard Space and test your model.
Disclaimer: it's important to note that automated scans may produce false positives or miss certain vulnerabilities. We encourage you to review the findings and assess the impact accordingly.