Methods of ranking search results for searches based on multiple search concepts carried out in multiple databases

• Ranking methods for reordering search hits reviewed.
• Particular problems when searches involve multiple concepts explored.
• Pivot ranking, using both full-text and abstract databases, expounded.
• Pivot ranking enables high recall and high precision whilst avoiding complex processing.
• Resulting high efficiency and flexibility in viewing and assessing hits explained.

If a patent prior art search produces several hundred results, it can be annoying when the best document turns out to be the last one. Ideally the most relevant search result should be brought to the top of the list. This article discusses the particular problems which searches for multiple concepts entail, explains what ranking is and compares some ways to rank, or reorder, search results. In particular, the article deals with a special technique which examiners at the European Patent Office (EPO) can use and which has been found to work well, especially when there are several search concepts and several technical fields to be searched. This ranking technique, called Horváth–Materne ranking or pivot ranking, is based on the assumption that the best documents will not only mention the search concepts in full-text databases, but will also mention at least some of the concepts in corresponding abstract databases. The more concepts present in the abstracts, the greater the probability that the document is relevant. In principle the technique could be used to rank results in any database where both full-text and abstracts are available. The technique can be broadened by including an automatic concept extraction.