TranSpec

Details of Results.zipIncluded are the correct SMILES in the test set of QM9S dataset and the SMILES candidate produced from different models and model-fusion. The parameters of the models are provided in .pt files. <br>Details of Training-Data.zipIncluded are the files of the SMILES, IR and Raman spectra of QM9S and QM14S dataset. For the QM9S’s spectra files, the first column is SMILES and the 2nd to 3001st are spectra intensities corresponding to 0-4000cm-1. For the QM14S’s spectra files, the SMILES are not present but listed in qm14s_SMILES.txt.<br>Details of example.zipConsidering the running of TranSpec is impossible on Figshare. We randomly selected 5000 molecules from QM9S to illustrate how to run the code. <br>Details of Group-Recognition.zipFiles to calculate the accuracy of functional group recognition. All molecules containing the same groups are collected into one same folder, with the Top1 candidates produced from TranSpec.<br>Details of Isomer-Distinguish.zipFiles to calculate the accuracy of isomers distinguishment. The first column is the correct SMILES, while the second to eleventh column are the predicted Top1 to Top10 SMILES candidates. All isomers corresponding to the same elementary compositions are separated with an empty row.