Paper Function Taxonomy

24 categories describing what each KG paper actually does in the field. Discovered by GPT-5.5-pro, classified by Claude Sonnet 4.6 across 17,499 papers in 35 minutes.

Nearly 1 in 3 KG papers introduces a new model or method. Model innovation dominates at 30%, followed by application demonstrations (17%) and graph construction (12%). The field produces 22× more new methods than it validates existing ones — only 1.3% of papers do empirical benchmarking.

All 24 Categories

model innovation

5,080 (29.0%)

application demonstration

3,028 (17.3%)

graph construction

2,151 (12.3%)

reasoning enablement

1,052 (6.0%)

field synthesis

1,013 (5.8%)

construction automation

831 (4.7%)

resource release

651 (3.7%)

system implementation

587 (3.4%)

formal foundations

482 (2.8%)

methodological guidance

307 (1.8%)

access enablement

268 (1.5%)

agenda framing

256 (1.5%)

empirical benchmarking

230 (1.3%)

benchmark creation

223 (1.3%)

resource_release

160 (0.9%)

quality assessment

153 (0.9%)

knowledge integration

149 (0.9%)

community stewardship

145 (0.8%)

schema design

139 (0.8%)

diagnostic analysis

115 (0.7%)

explainability support

110 (0.6%)

model_innovation

91 (0.5%)

training_optimization

77 (0.4%)

training optimization

54 (0.3%)

safety auditing

48 (0.3%)

efficiency engineering

47 (0.3%)

agenda_framing

24 (0.1%)

human evaluation

17 (0.1%)

construction_automation

7 (0.0%)

application_demonstration

3 (0.0%)

system_implementation

1 (0.0%)

Year Trends (Top 5)

Year	model innovation	application demonstration	graph construction	reasoning enablement	field synthesis
2015	25	17	11	13	6
2016	39	27	20	10	9
2017	69	48	38	21	10
2018	122	72	58	28	18
2019	228	118	117	53	42
2020	355	189	154	50	84
2021	505	237	195	66	95
2022	572	249	232	99	123
2023	746	381	300	101	155
2024	889	572	348	166	199
2025	1107	760	459	285	186
2026	373	264	186	137	58