2026-10-13T21:06:04Z https://www.scidb.cn/oai

10.57760/sciencedb.013o5 2026-10-10T16:42:03Z

2026-10-10 An infrared terminology dataset linking physical definitions with bilingual labels and machine readable quantity profiles doi:10.57760/sciencedb.013o5 en This infrared terminology dataset is a reusable reference data product for infrared data authors and AI applications. It connects physical definitions, Chinese-English labels, machine-readable quantity profiles and source provenance. Developed from the unpublished Chinese terminology manuscript led by Dingbo Kuang, it preserves 1,843 numbered records, including 1,618 fourth-level terms with complete available English explanations, 255 native equation objects and 26 content tables. All 406 identified quantity-bearing records (265 quantities and 141 quantity families) have 563 field profiles with recommended names, units or explicit unit-selection rules, and measurement context. These are terminology and metadata reference data; the supplied experimental observations are explicitly synthetic teaching examples.This Version 3 revision deposits main package 1.4.0-reuse and application companion 1.2.0-reuse. The corresponding files are IR_Term_ScienceDB_Main_1.4.0_Reuse_20260927.zip and IR_Term_Applications_1.2.0_Reuse_20260927.zip, identified by 00_CURRENT_RELEASE_1.4.0.txt. They extend the scientific snapshot in the main 1.3.0-ai-ready and application 1.1.0-ai-ready packages published in Version 1 on 24 September 2026. Version 2 updated the descriptive metadata but retained the earlier five attachments; Version 3 supplies the corresponding new main and application packages. Earlier packages and release pointers remain historical snapshots. The 27 September 2026 pre-deposit statements inside the ZIPs document their build-time status; the repository page and the release pointer dated 29 September 2026 govern subsequent deposit status. Dataset DOI: https://doi.org/10.57760/sciencedb.013o5; CSTR: https://cstr.cn/31253.11.sciencedb.013o5. Availability of this revision follows the actual ScienceDB review and publication outcome.The public browser at https://infrared-terminology.sklu.chatgpt.site provides Chinese/English label and identifier search, English definitions, native equations, quantity profiles, source citations and JSON export. A separate view searches the image-free text of all 623 Chinese-source pages by physical PDF page number, with source fingerprints. Page-level Chinese material is not asserted to be aligned with individual English definitions. The packages include browser source and a static build for local reuse. Seven unpublished source documents are catalogued; complete originals remain in the separate Restricted ScienceDB record with assigned DOI 10.57760/sciencedb.013op, which was still shown as unregistered on 27 September 2026. The public package retains six originals and the Chinese image-free text surrogate.Two executable author workflows demonstrate radiance/spectral-radiance field selection, unit and context checks, and traceable bounded NeXus export/readback. Their 31 checks passed. Existing input schemas and blank JSON templates cover all 563 profiles; the earlier 3,180 input tests remain documented. The Excel experiment template preserves seven worksheets, 563 profiles, 401 formulas and blank observation fields. Updated manuscripts and a reuse guide explain when and how to cite the dataset, stable term identifiers and specific source evidence. Passing software checks does not establish the scientific correctness of user-supplied context, universal unit conversion or full official NXDL conformance.The executed RAG and training Notebook supplies 3,468 exact source text spans from 1,618 terms, grouped training inputs, 1,618 non-gold instruction candidates and an actual word-bigram teaching model. The 1,604 inherited groups produce 1,308/152/158 train/validation/test concepts, with no declared group crossing partitions; this is not a guarantee against every form of semantic leakage. The retrieval tutorial demonstrates three cited definition answers and two refusals, preserving record, chunk and character-span provenance. It makes no external model calls and is not unrestricted generative question answering. All eight code cells executed successfully and 31 reuse tests passed. These examples support preparation for continued pretraining, supervised fine-tuning and an external infrared RAG knowledge base; they do not establish an independently evaluated infrared language model, expert-approved translations or a gold training corpus. All 563 profiles still require independent expert review.Authors: Dingbo Kuang, Jiaxiong Fang and Wei Lu, all at the Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai, China. Dingbo Kuang was responsible for the original Chinese terminology and scientific content; Jiaxiong Fang first proposed and organized the cataloguing; Wei Lu completed the data-article work based on the Chinese materials. Corresponding author: Wei Lu, luwei@mail.sitp.ac.cn. This work received no project funding. Author-contributed main and application content is offered under CC BY 4.0 with Open access; the separate original-document record retains Restricted access. infrared terminology; physical definitions; bilingual labels; quantity profiles; physical units; NeXus Dingbo Kuang Jiaxiong Fang Wei Lu PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank