Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering

Data supporting the paper entitled "Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering" by Maciej P. Polak and Dane Morganhttps://arxiv.org/abs/2303.05352<br>BulkModulus_test_database_MPPolak_DMorgan.xlsx - dataset of bulk modulus text passages and sentences used for methods assessment.<br>CriticalCoolingRates_MGs_database_MPPolak_DMorgan.xlsx - a database of critical cooling rates of metallic glasses. The data is presented in three versions and (described in detail in the paper), i.e. "raw", "cleaned", and "standardized". The critical cooling rate additionally includes manually extracted data serving as ground truth for tests, in sheets labeled as "manual". In addition, a "standardized_MG" database is included, which limits the results to metallic glasses only, together with "standardized_tables_MG" for values extracted from tables, and "Figure_Classification" which contains Figure numbers, captions, and DOIs of their source documents.<br>YieldStrength_HEAs_database_MPPolak_DMorgan.xlsx - a database of yield strengths in the context of high entropy alloys. The data is presented in three versions and (described in detail in the paper), i.e. "raw", "cleaned", and "standardized". In addition, a "standardized_HEA" database is included, which limits the results to HEAs only, together with "standardized_tables_HEA" for values extracted from tables, and "Figure_Classification" which contains Figure numbers, captions, and DOIs of their source documents.<br><br><br>ChatExtract_Code_MPPolak_DMorgan.zip - These files contain the ChatExtract code with a short example and instructions.