Topic Ranking

Lingospot’s topic ranking algorithm uses a complex formula with several input parameters to determine a weighted score for every topic on a given page. At the core of our ranking intelligence is a proprietary knowledge database of over 30 million topics, and the topics that each is most closely related to. This database has been compiled through crawling and linguistically analyzing over two billion pages and is used to compute the pair-wise “semantic distance” between all the concepts on a page, so closely related topics are easily identified.

Additionally, meta-data for each topic is retrieved, such as whether the topic appears in Wikipedia, in a particular ontology, whether it’s a city, a publicly traded company, etc. Next, the semantic distance of the topic to your site’s content signature is computed, as a measure of how relevant each topic is to the site. The closer this measure is, the more likely your readers would find related content of interest. Next, the historical interaction data of each topic is calculated, including hovers, clicks, time spent, etc, so topics that generate more activity are more likely to be chosen. Finally, the available content for each topic is calculated by searching your entire site for recent content related to each topic.

Taking all these inputs into consideration, a score is computed for each topic, and the topic list is sorted to choose the top N topics (where N is controlled by limits set by the publisher through the Lingospot console).

Topic ranking is a process similar to that used by major search engines that present the top 10 results for a given query, instead of showing 100 results in unranked order.

Next Steps