Trends
Emerging terms in paper titles across the indexed venues
Rising (2022)
vision-language
15 → 66 titles
+340%
vision transformers
21 → 84 titles
+300%
tuning
19 → 70 titles
+268%
token
12 → 41 titles
+242%
vision transformer
17 → 56 titles
+229%
logical
15 → 49 titles
+227%
patch
10 → 32 titles
+220%
test-time
10 → 32 titles
+220%
spotting
11 → 33 titles
+200%
dictionary
13 → 38 titles
+192%
medical image segmentation
11 → 32 titles
+191%
hate speech detection
10 → 27 titles
+170%
deformable
16 → 43 titles
+169%
makes
12 → 32 titles
+167%
learners
17 → 45 titles
+165%
sarcasm detection
11 → 29 titles
+164%
medical image
19 → 50 titles
+163%
2d
16 → 42 titles
+163%
knowledge graph completion
10 → 26 titles
+160%
mixing
16 → 41 titles
+156%
No 2021 baseline (2022)
Terms with no 2021 titles, so there is no growth rate to compute. Ranked by 2022 volume. Most are returning after a gap year rather than appearing for the first time.
condescending language
26 titles
patronizing
25 titles
in-context
23 titles
patronizing and condescending
23 titles
condescending language detection
17 titles
language-image
16 titles
news article
16 titles
complex named
14 titles
in-context learning
14 titles
masked autoencoders
14 titles
complex named entity
13 titles
prompt learning
13 titles
lexical simplification
12 titles
transformer-based models
12 titles
manipulations
11 titles
misogynous
11 titles
vision-based
11 titles
abusive comment
9 titles
cloze
9 titles
detecting signs
9 titles
Declining (2022)
spans
31 → 5 titles
-84%
rating
23 → 6 titles
-74%
language identification
33 → 9 titles
-73%
humor
36 → 11 titles
-69%
toxic
46 → 15 titles
-67%
alzheimer
15 → 5 titles
-67%
graph convolution
18 → 6 titles
-67%
similar
18 → 6 titles
-67%
transitions
15 → 5 titles
-67%
word embedding
17 → 6 titles
-65%