Human-Assigned vs. ML-Predicted Technology Taxonomy
20,150 projects|public techport.nasa.gov API|100% portfolio coverage|TREX ML classifier
March 2026
This analysis was generated using Claude (Anthropic) with live access to NASA's public TechPort database via an MCP server. All data comes from the public TechPort REST API
(techport.nasa.gov) and public web sources — no internal or restricted NASA data was used.
NASA's Technology Taxonomy (TX) is one of the key ways to slice the ~20,000-project TechPort portfolio by technology area. But TX assignment is a judgment call — set by the submitting NASA technologist. The depth and consistency of TX submissions varies across programs, from no TX at all to full 3-level codes. Before using it for portfolio analysis, we need to know: is this field consistent, trustworthy, and useful?
What We Found in the API
TechPort provides a human-assigned TX code for each project — the standard classification used in portfolio reviews and gap analyses.
While exploring the API, we discovered TechPort also exposes TREX — NASA's own ML taxonomy classifier. TREX reads a project's title and description and predicts the most appropriate TX code based purely on technology content, with no knowledge of program context.
Its existence means every project has a potential second opinion on its TX classification — one generated by an ML model rather than human judgment.
What We Did
Ran all 20,150 TechPort projects through TREX to generate a complete set of ML-predicted TX codes — enabling a portfolio-wide comparison of human vs. ML classification.
Compared agreement rates at the top-level TX area (e.g., TX08 vs TX09).
Analyzed patterns by technology area and by program to understand where TX is reliable and where it breaks down.
Performed a focused analysis on the Flight Opportunities (FO) program to examine human vs. ML classification for projects our team knows firsthand.
2
The NASA Technology Taxonomy (TX)
NASA's Space Technology taxonomy organizes all technology development into 17 top-level areas. Revised in 2024, the TX taxonomy replaced the earlier TA (Technology Area) system. TX codes are the standard classification used in TechPort for portfolio tracking, gap analysis, and investment planning.
🚀 Both attempt to classify the same thing — the project's technology. The difference is method: human judgment vs. ML model. Where they disagree, the project potentially sits at a taxonomy boundary.
4
The Core Finding
76.3%
Top-level area agreement (human TX vs ML TX)
23.7%
Top-level area mismatch
816
No human TX at all (4.0%)
Of 19,334 projects with a human-assigned TX (20,150 − 816), the ML agrees on the top-level technology area (e.g., both say TX08) for 76.3%. For 23.7%, the ML sees a different area entirely.
What "mismatch" means in practice: a project classified under one technology area by the human may not show up in portfolio reviews for the area the ML thinks it belongs to. For example, a LiDAR sensor project filed under TX04 (Robotics) won't appear in a TX08 (Sensors) portfolio review — even though the core technology is a sensor.
The disagreements are not random — they follow consistent patterns by technology area and by program, which makes them analyzable.
🚀 76% agreement means TX carries real signal. The 24% mismatch is systematic, not random — which makes it interpretable and correctable.
5
TX08 (Sensors) Is the Taxonomy's Gravity Well
Under ML classification, TX08 gains 987 net projects — more than any other area. The ML sees sensor technology embedded in health monitors, robotic perception, comms hardware, and thermal sensors.
From
To TX08
Count
What's happening
TX06 (Health)
TX08
301
Biomedical sensors classified by tech identity
TX12 (Materials)
TX08
136
Materials characterization = sensor work
TX05 (Comms)
TX08
116
Comm receivers are sensor technology
TX11 (Software)
TX08
113
Sensor data processing systems
TX14 (Thermal)
TX08
99
Thermal sensors reclassified
TX02 (Computing)
TX08
93
Sensor signal processing
TX04 (Robotics)
TX08
69
Perception sensors for robots
TX09 (EDL)
TX08
56
Navigation/landing sensors
TX07 (Exploration)
TX08
54
In-situ measurement instruments
🚀 A TX08 portfolio review based on human assignments misses ~1,000 relevant projects.
6
Mismatch Rate by Technology Area
TX08 - Sensors
9.4%
TX01 - Propulsion
13.9%
TX03 - Power
15.4%
TX15 - Aeronautics
16.1%
TX09 - EDL
22.7%
TX12 - Materials
24.7%
TX04 - Robotics
27.9%
TX11 - Software
31.6%
TX06 - Human Health
32.0%
TX10 - Autonomous
34.9%
TX14 - Thermal
41.5%
TX13 - Ground
43.3%
TX02 - Computing
56.9%
TX16 - Air Traffic
77.3%
Mismatch %
7
Classification Quality by Program
🚀 Science programs classify best. Center innovation funds classify worst.
FO at 74.8% match — mid-range, consistent with its broad technology scope
AFRC CIF: 59.8% mismatch + 40.2% granularity gap — most projects mis- or under-classified
8
Classification Quality by Program
94%
SAT
93%
IIP
13%
87%
APRA
15%
85%
HRP
18%
82%
STRG
22%
78%
SBIR/ STTR
25%
75%
FO
26%
74%
GCD
38%
62%
NIAC
43%
57%
EPSCoR
60%
40%
AFRC CIF
Portfolio avg. 76.3%
Match %Mismatch %
9
TX Distribution: Major STMD Programs (Human vs. ML)
How each program's technology profile shifts under ML classification. X-axis: TX01–TX17.
SBIR/STTR (12,272)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
GCD (468)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
STRG (1,102)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
FO (430)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
HumanML
10
TX Distribution: High-Mismatch Programs
Programs with the highest mismatch rates show the most dramatic shifts. X-axis: TX01–TX17.
NIAC (327, 37% mismatch)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
EPSCoR (346, 43% mismatch)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
AFRC CIF (139, 44% mismatch)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
HRP (184, 15% mismatch)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
EPSCoR: 70 projects shift TX08→TX11 (human says Sensors, ML says Software/Data Science). AFRC CIF: TX14 nearly disappears, absorbed by TX15 and TX08.
11
Dual View: How the Portfolio Shifts Under ML Classification
Some areas grow significantly, others shrink — revealing where technology is hidden.
Area
Human-Assigned
ML-Predicted
Delta
Shift
TX08 - Sensors
4,127
5,114
+987
Absorbs from all areas
TX01 - Propulsion
1,849
2,170
+321
Thermal + ATM projects
TX15 - Aeronautics
728
932
+204
ATM & flight systems
TX03 - Power
1,135
1,248
+113
Fuel cells from other areas
TX07 - Exploration
849
954
+105
ISRU & surface ops
TX06 - Health
2,338
2,072
-266
Sensors & materials out
TX14 - Thermal
981
718
-263
To propulsion & sensors
TX02 - Computing
534
286
-248
To sensors & software
TX16 - Air Traffic
181
66
-115
To propulsion & aero
TX13 - Ground/Launch
386
292
-94
To materials & propulsion
12
Dual View: Human vs. ML Portfolio Distribution
TX01 Propulsion
TX02 Computing
TX03 Power
TX04 Robotics
TX05 Comms
TX06 Health
TX07 Exploration
TX08 Sensors
TX09 EDL
TX10 Autonomous
TX11 Software
TX12 Materials
TX13 Ground
TX14 Thermal
TX15 Aero
TX16 ATM
TX17 GN&C
Human-AssignedML-Predicted
13
The Granularity Gap: 613 Under-Specified Projects
3.2% of projects are assigned a top-level TX code only (e.g., 'TX12') with no sub-area. The ML provides the missing sub-classification for every one.
What the ML fills in:
Human: TX12 (Materials)
ML: TX12.4.1 (Manufacturing Processes)
Human: TX08 (Sensors)
ML: TX08.1.5 (Lasers)
Human: TX06 (Human Health)
ML: TX06.3.1 (Medical Diagnosis)
Worst granularity gap programs:
Program
Count
Gap Rate
MEP
17
58.6%
AOSP
6
54.5%
AFRC CIF
41
40.2%
SCaN
11
35.5%
MSFC CIF
49
30.1%
MSFC IRAD
44
29.3%
MCO
31
25.4%
14
Flight Opportunities
TX Deep Dive
15
Flight Opportunities: TX Deep Dive
FO's 24.7% mismatch rate matches the portfolio average — but the patterns are FO-specific.
430
FO projects
24.7%
Top-level mismatch (human vs ML TX)
75.3%
Top-level agreement
10
No human TX provided
Top Reclassification Flows in FO
Human TX
ML TX
Count
Pattern
TX04 (Robotics)
TX07 (Exploration)
9
Lunar tools filed under manipulation → ISRU
TX06 (Health)
TX08 (Sensors)
5
Biomedical instruments → sensor tech
TX09 (EDL)
TX08 (Sensors)
5
Landing sensors → sensor/laser tech
TX05 (Comms)
TX08 (Sensors)
5
Receivers and optical systems → sensors
TX04 (Robotics)
TX08 (Sensors)
4
Perception hardware → sensor tech
TX14 (Thermal)
TX01 (Propulsion)
3
Cryo thermal mgmt → propulsion systems
🚀 Dominant FO pattern: lunar surface tools filed under Robotics (TX04) are reclassified to ISRU (TX07). And sensor technology is undercounted, just like portfolio-wide.
Top FO areas (human): TX06 (81), TX01 (61), TX08 (57), TX09 (42) | Top FO areas (ML): TX06 (76), TX08 (69), TX01 (68), TX07 (37)
Capability gap mapping — matches technology identity to shortfalls
Use Human TX for...
"Which program funded this and why?"
Organizational accountability — who claimed the work
Solicitation tracing — maps back to SBIR subtopics, BAAs
Program-level portfolio reviews within a known scope
Use Both for...
Finding hidden connections — the 4,587 mismatches are taxonomy boundary projects
Investment analysis — where program intent diverges from technology reality
Dual-tagging as standard practice for future submissions
The Bottom Line
For technology portfolio analysis — gap mapping, capability assessments, investment prioritization — the ML classification is the more reliable primary lens. It has no coverage gaps, no granularity gaps, and classifies by what the technology IS rather than who funded it. The human TX remains essential context for understanding organizational intent and funding flows.
🚀 Yes, TX is useful for portfolio analysis — especially with dual classification.
2021
TX01 Propulsion Systems
TX02 Flight Computing & Avionics
TX03 Aerospace Power & Energy Storage
TX04 Robotic Systems
TX05 Communications, Navigation & OD
TX06 Human Health, Life Support & Habitation
TX07 Exploration Destination Systems
TX08 Sensors & Instruments
TX09 Entry, Descent & Landing
TX10 Autonomous Systems
TX11 Software, Modeling & Simulation
TX12 Materials, Structures & Manufacturing
TX13 Ground, Test & Surface Systems
TX14 Thermal Management Systems
TX15 Flight Vehicle Systems
TX16 Air Traffic Management & Range Tracking
TX17 Guidance, Navigation & Control
22
TX01 Propulsion Systems
TX02 Flight Computing & Avionics
TX03 Aerospace Power & Energy Storage
TX04 Robotic Systems
TX05 Communications, Navigation & OD
TX06 Human Health, Life Support & Habitation
TX07 Exploration Destination Systems
TX08 Sensors & Instruments
TX09 Entry, Descent & Landing
TX10 Autonomous Systems
TX11 Software, Modeling & Simulation
TX12 Materials, Structures & Manufacturing
TX13 Ground, Test & Surface Systems
TX14 Thermal Management Systems
TX15 Flight Vehicle Systems
TX16 Air Traffic Management & Range Tracking
TX17 Guidance, Navigation & Control
Backup: Full TX Taxonomy (TX01–TX02)
TX01: Propulsion Systems (2,178)
TX01.1 Chemical Space Propulsion (760)
TX01.1.1 Liquid Rocket Engines (162)
TX01.1.2 Solid Rocket Motors (120)
TX01.1.3 Hybrid Propulsion (341)
TX01.1.4 Cold/Warm Gas Systems (32)
TX01.1.5 Monopropellant Systems (77)
TX01.1.6 Propellant Storage (11)
TX01.1.7 Pressure & Feed Systems (14)
TX01.1.8 Combustion Stability (3)
TX01.2 Electric Space Propulsion (565)
TX01.2.1 Ion/Gridded Thrusters (137)
TX01.2.2 Hall Thrusters (400)
TX01.2.3 Pulsed Plasma Thrusters (25)
TX01.2.4 Electrospray (3)
TX01.3 Aero Propulsion (568)
TX01.3.1 Air-Breathing Propulsion (344)
TX01.3.2 Turbomachinery (24)
TX01.3.4 Combustors & Nozzles (45)
TX01.3.5 Propulsion Control (4)
TX01.3.6 Inlets (4)
TX01.3.8 Propellant Mgmt (24)
TX01.3.9 Electric Propulsion Components (83)
TX01.3.10 Nuclear Propulsion (5)
TX01.3.11 Green Propellants (31)
TX01.3.12 Propulsion Testing (4)
TX01.4 Advanced Propulsion (285)
TX01.4.1 Nuclear Thermal (66)
TX01.4.2 Solar Sail (11)
TX01.4.3 Solar Electric (134)
TX01.4.4 Laser/Beam Propulsion (66)
TX01.4.7 Tethers (8)
TX02: Flight Computing & Avionics (285)
TX02.1 Avionics Component Technologies (200)
TX02.1.1 Processors (20)
TX02.1.2 Memory (6)
TX02.1.3 FPGAs (36)
TX02.1.4 Data Buses (12)
TX02.1.5 Single-Board Computers (47)
TX02.1.6 Rad-Hard Electronics (48)
TX02.1.7 Power Electronics (3)
TX02.1.8 Analog/Mixed-Signal (28)
TX02.2 Avionics Systems & Subsystems (64)
TX02.2.1 Integrated Avionics (6)
TX02.2.2 Flight Software (26)
TX02.2.4 Comm Hardware (5)
TX02.2.5 CDH (8)
TX02.2.6 Instrumentation (8)
TX02.2.7 Displays & Controls (4)
TX02.2.8 Timing Systems (7)
TX02.3 Avionics Tools, Models & Analyses (21)
TX02.3.1 Design Tools (7)
TX02.3.2 V&V (10)
TX02.3.4 Modeling (4)
ML-predicted TX classifications across 20,150 projects. Counts show projects per sub-area. NASA TechPort taxonomy (17 L1, 94 L2, 384 L3 nodes).
23
Backup: Full TX Taxonomy (TX03–TX04)
TX03: Aerospace Power & Energy Storage (1,248)
TX03.1 Power Generation & Conversion (545)
TX03.1.1 Photovoltaic Electrical Power (337)
TX03.1.2 Heat Sources (32)
TX03.1.3 Static Energy Conversion (11)
TX03.1.4 Dynamic Energy Conversion (155)
TX03.1.6 Other Advanced Concepts (10)
TX03.2 Energy Storage (402)
TX03.2.1 Electrochemical: Batteries (257)
TX03.2.2 Electrochemical: Fuel Cells (108)
TX03.2.3 Advanced Energy Storage (37)
TX03.3 Power Mgmt & Distribution (301)
TX03.3.1 Management and Control (103)
TX03.3.2 Distribution and Transmission (26)
TX03.3.3 Electrical Power Conversion (139)
TX03.3.4 Advanced Electronic Parts (33)
TX04: Robotic Systems (801)
TX04.1 Sensing & Perception (60)
TX04.1.1 Sensing for Robotic Systems (11)
TX04.1.2 State Estimation (23)
TX04.1.3 Onboard Mapping (19)
TX04.1.4 Object/Event Recognition (7)
TX04.2 Mobility (347)
TX04.2.1 Below-Surface Mobility (22)
TX04.2.2 Above-Surface Mobility (19)
TX04.2.3 Small-Body/Microgravity (13)
TX04.2.4 Surface Mobility (226)
TX04.2.5 Robot Navigation (10)
TX04.2.6 Collaborative Mobility (57)
TX04.3 Manipulation (179)
TX04.3.1 Dexterous Manipulation (52)
TX04.3.2 Grappling (45)
TX04.3.3 Contact Dynamics (8)
TX04.3.4 Sample Acquisition (74)
TX04.4 Human-Robot Interaction (83)
TX04.4.1 Multimodal Interaction (67)
TX04.4.2 Distributed Collaboration (8)
TX04.4.3 Remote Interaction (8)
TX04.5 AR&D (88)
TX04.5.1 Relative Nav Sensors (22)
TX04.5.2 R&D Algorithms (18)
TX04.5.3 RPOC Systems (5)
TX04.5.5 Capture Mechanisms (27)
TX04.5.6 Robot Control for Capture (8)
TX04.5.7 RPOC Modeling & Testing (8)
TX04.6 System Integration (44)
TX04.6.1 Modularity & Interfaces (18)
TX04.6.2 Modeling & Simulation (16)
TX04.6.3 Robot Software (10)
ML-predicted TX classifications across 20,150 projects. Counts show projects per sub-area. NASA TechPort taxonomy (17 L1, 94 L2, 384 L3 nodes).
24
Backup: Full TX Taxonomy (TX05–TX06)
TX05: Comms, Nav & Orbital Debris (913)
TX05.1 Optical Comms (280)
TX05.1.1 Free-Space Optical (51)
TX05.1.3 Atmospheric Optical (33)
TX05.1.4 Optical Terminals (36)
TX05.1.6 Optical Ground Stations (153)
TX05.1.7 Quantum Comms (7)
TX05.2 RF Comms (295)
TX05.2.1 Proximity RF (42)
TX05.2.2 Near-Earth RF (69)
TX05.2.4 Deep-Space RF (38)
TX05.2.6 RF Ground Systems (134)
TX05.2.7 Cognitive/Adaptive RF (12)
TX05.3 Internetworking (28)
TX05.3.1 Delay-Tolerant Networking (15)
TX05.3.3 Space Internet (13)
TX05.4 PNT (95)
TX05.4.1 Timekeeping (32)
TX05.4.2 Navigation Systems (63)
TX05.5 Revolutionary Concepts (192)
TX05.5.1 X-Ray Nav (72)
TX05.5.2 Quantum Sensing (108)
TX05.5.3 Debris Tracking (12)
TX05.6 Ground-Based OD (23)
TX05.6.1 Ground OD Systems (23)
TX06: Human Health, Life Support & Habitation (2,045)
TX06.1 ECLSS & Habitation (557)
TX06.1.1 Air Revitalization (176)
TX06.1.2 Water Recovery (214)
TX06.1.3 Waste Management (86)
TX06.1.4 Habitation Systems (81)
TX06.2 EVA Systems (271)
TX06.2.1 EVA Suits (105)
TX06.2.2 EVA Mobility (100)
TX06.2.3 EVA Tools & Interfaces (66)
TX06.3 Human Health & Performance (674)
TX06.3.1 Medical Diagnostics (354)
TX06.3.2 Exercise & Countermeasures (69)
TX06.3.3 Behavioral Health (106)
TX06.3.4 Pharmacology (12)
TX06.3.5 Food & Nutrition (92)
TX06.3.6 Sleep & Circadian (25)
TX06.3.7 Microbiome (16)
TX06.4 Monitoring & Safety (257)
TX06.4.1 Fire Safety (179)
TX06.4.2 Contamination & Toxicology (57)
TX06.4.3 Acoustic Monitoring (21)
TX06.5 Radiation (209)
TX06.5.1 Space Radiation Monitoring (19)
TX06.5.2 SPE Forecasting (15)
TX06.5.3 Shielding (112)
TX06.5.4 Radiation Risk Models (34)
TX06.5.5 Biological Countermeasures (29)
TX06.6 Human Systems Integration (77)
TX06.6.1 HSI (77)
ML-predicted TX classifications across 20,150 projects. Counts show projects per sub-area. NASA TechPort taxonomy (17 L1, 94 L2, 384 L3 nodes).
25
Backup: Full TX Taxonomy (TX07–TX08)
TX07: Exploration Destination Systems (941)
TX07.1 ISRU (670)
TX07.1.1 O2/Propellant from Regolith (253)
TX07.1.2 Water Extraction (55)
TX07.1.3 Construction & Manufacturing (339)
TX07.1.4 Resource Characterization (23)
TX07.2 Surface Infrastructure (223)
TX07.2.1 Habitation (39)
TX07.2.2 Surface Power (19)
TX07.2.3 Surface Comms (29)
TX07.2.4 Dust Mitigation (23)
TX07.2.5 Surface Mobility Systems (113)
TX07.3 Mission Operations (48)
TX07.3.2 Mission Planning (24)
TX07.3.5 Surface EVA Operations (24)
TX08: Sensors & Instruments (5,114)
TX08.1 Remote Sensing Instruments (3,835)
TX08.1.1 Passive Optical/UV/Vis (1,467)
TX08.1.2 Active Optical/Lidar (67)
TX08.1.3 Passive Microwave/RF (457)
TX08.1.4 Active Microwave/Radar (680)
TX08.1.5 Passive IR (1,107)
TX08.1.6 X-Ray/Gamma (57)
TX08.2 Observatories (519)
TX08.2.1 Space Telescopes (452)
TX08.2.2 Ground Observatories (39)
TX08.2.3 Suborbital Platforms (28)
TX08.3 In Situ Instruments (760)
TX08.3.1 Chemical Analyzers (237)
TX08.3.2 Biological Detectors (82)
TX08.3.3 Particle Analyzers (14)
TX08.3.4 Geophysical Instruments (384)
TX08.3.5 Weather Instruments (10)
TX08.3.6 Radiation Detectors (33)
ML-predicted TX classifications across 20,150 projects. Counts show projects per sub-area. NASA TechPort taxonomy (17 L1, 94 L2, 384 L3 nodes).
26
Backup: Full TX Taxonomy (TX09–TX10)
TX09: Entry, Descent & Landing (801)
TX09.1 Aeroassist & Atmospheric Entry (197)
TX09.1.1 Rigid Aeroshells (116)
TX09.1.2 Deployable Aeroshells (58)
TX09.1.3 Inflatable Decelerators (23)
TX09.2 Descent (62)
TX09.2.1 Parachute Systems (55)
TX09.2.2 Other Descent (7)
TX09.3 Landing (45)
TX09.3.1 Landing Systems (45)
TX09.4 Vehicle Systems (497)
TX09.4.1 Flight Dynamics (4)
TX09.4.4 TPS Materials (28)
TX09.4.5 Thermal Protection Systems (349)
TX09.4.6 Structural Systems (58)
TX09.4.7 GN&C for EDL (58)
TX10: Autonomous Systems (448)
TX10.1 Situational & Self-Awareness (63)
TX10.1.1 Health Monitoring (20)
TX10.1.2 Fault Detection (24)
TX10.1.4 Environment Perception (19)
TX10.2 Reasoning & Acting (341)
TX10.2.1 Planning & Scheduling (7)
TX10.2.2 Activity Planning (46)
TX10.2.3 Contingency Management (23)
TX10.2.4 Autonomous GN&C (111)
TX10.2.5 Multi-Agent Coordination (57)
TX10.2.6 Onboard Autonomy (41)
TX10.2.7 V&V of Autonomous Systems (56)
TX10.3 Collaboration & Interaction (18)
TX10.3.3 Human-Machine Interaction (8)
TX10.3.4 Collaborative Decision (10)
TX10.4 Engineering & Integrity (26)
TX10.4.1 Software Assurance (26)
ML-predicted TX classifications across 20,150 projects. Counts show projects per sub-area. NASA TechPort taxonomy (17 L1, 94 L2, 384 L3 nodes).
27
Backup: Full TX Taxonomy (TX11–TX12)
TX11: Software, Modeling, Simulation & Info Processing (954)