Building a Better Search Engine for Semantic Scholar
A "tell-all" account of why improving our academic search engine was not as simple as we had hoped, including the following problems and their solutions: dirty data, feature engineering, proper evaluation, posthoc correction.
Source: medium.com