How the SSI reviews work.
This page explains the fields used on each paper page. It stays inside repository data, and it does not invent peer-review status, scores, or outside validation.
Core fields
review_state
Shows whether the page is still a draft, a full-text draft, or something further along.
review_priority
Marks which papers deserve attention first inside the SSI review queue.
review_confidence
States how certain the judgment is when the source record is incomplete or strong.
review_basis
Explains whether the judgment came from full text, existing expert seeds, or another recorded basis.
source_coverage
Shows how much source material was available when the review record was written.
axes_moved
Lists the field dimensions the paper genuinely advances, such as wearability or open vocabulary.
axes_unresolved
Lists open problems that remain visible in the review instead of being hidden.
axes_regressed
Records tradeoffs or regressions when a paper improves one axis but weakens another.
claim_evidence_alignment
Captures how closely the paper's claims match the available evidence.
overclaim_risk
Flags where the language should stay cautious because the evidence does not justify stronger wording.
deployment_readiness_gap
Highlights what still blocks practical use outside the paper setting.
generalization_gap
Highlights where the result still needs broader testing before it should be treated as general.
Entry papers in use
The three current entry pages are all full-text reviewed with high source coverage.
SilentSpeller: Towards mobile, hands-free, silent speech text entry using electropalatography
basis: full_text+existing_expert_seed; coverage: high; moved: large_vocabulary, unseen_word_generalization, walking_robustness, live_text_entry, hands_free_texting
SottoVoce: An Ultrasound Imaging-Based Silent Speech Interaction Using Deep Neural Networks
basis: full_text+existing_expert_seed; coverage: high; moved: ecosystem_integration, speech_regeneration, smart_speaker_control, user_adaptation_loop
NasoVoce: A Nose-Mounted Low-Audibility Speech Interface for Always-Available Speech Interaction
basis: full_text+existing_expert_seed; coverage: high; moved: wearability, discreet_speech, noise_robust_capture, ai_voice_agent_alignment, multimodal_fusion
Reading rule
If a field is missing, it stays missing.
The paper pages prefer omission over guessed metadata. That keeps the archive reproducible.