Skip to main content
. 2020 Apr 6;21(11):3133–3160. doi: 10.1093/pm/pnaa061

Figure 8.

Figure 8

Examples of artificial, synthesized PubMed abstracts generated using a gated recurrent unit-based deep learning network. These results highlight the common patterns of content and organization found in human-published abstracts pertaining to pain. A scaling factor known as temperature adjusted the balance between the conservative and obvious vs the experimental and error-prone.