Abstract--Although gold nanorods have been the subject of much research, the pathways for controlling their shape and thereby their optical properties remain largely heuristically understood. Although it is apparent that the simultaneous presence of and interaction between various reagents during synthesis control these properties, computational and experimental approaches for exploring the synthesis space can be either intractable or too time-consuming in practice. This motivates an alternative approach leveraging the wealth of synthesis information already embedded in the body of scientific literature by developing tools to extract relevant structured data in an automated, high-throughput manner. To that end, we present an approach using the powerful GPT-3 language model to extract structured multi-step seed-mediated growth procedures and outcomes for gold nanorods from unstructured scientific text. GPT-3 prompt completions are finetuned to predict synthesis templates in the form of JSON documents from unstructured text input with an overall accuracy of 86%. The performance is notable, considering the model is performing simultaneous entity recognition and relation extraction. We present a dataset of 11,644 entities extracted from 1,137 papers, resulting in 268 papers with at least one complete seed-mediated gold nanorod growth procedure and outcome for a total of 332 complete procedures. In the last three semiconductor technology,[11, 12] biomedicine,[13, 14] and decades, chemists have developed the ability to synthesize cosmetics.[15] The suitability of a nanoparticle for a particular anisotropic metal nanoparticles in a controllable and re-application depends on its morphology and size, which correspond to different plasmonic properties.[16,

This completion can be formatted as either English sentences or a more structured schema such as a list of JSON documents. Large language models (LLMs) such as GPT-3 [12], PaLM To use this method, one only has to define the desired [25], Megatron [26], OPT [27], Gopher [28], and FLAN [29] output structure--for example, a list of JSON objects with a have been shown to have remarkable ability to leverage semantic predefined set of keys--and annotate 100 500 text passages information between tokens in natural language sequences using this format. GPT-3 is then fine-tuned on these of varying length. They are particularly adept at examples, and the resulting model is able to accurately extract sequence-to-sequence (seq2seq) tasks, where a text input is desired information from text and output information in used to seed a text response from the model. In this paper the same structured representation as shown in Figure 1.