The two search engines we picked are Google and DuckDuck Go. After going through some trials we found that Google’s searches seemed to be faster and handled our searches better. Google does tend to frequently cluster search results to be more open by offering multiple interpretations; there is less clustering on DuckDuck Go. Google has better disambiguation and clustering compared to DuckDuck Go. Google is also better with complex, content-sensitive questions. 

We picked a sample semantically unambiguous query, Hubble Space Telescope launch date, on both search engines. Considering the first 20 search results, Google’s top results typically include reliable sources such as NASA’s official website and science organizations and the information is highlighted in snippets from the websites and articles. DuckDuck Go also returned relative results like links to NASA and other reputable sources. Both of these search engines provided accurate information regarding the questions about the Hubble launch date and their consistency among both search engines confirms the information’s reliability. Google seemed to have done a better job ay prioritizing highly authoritative sources, making sure that users get information from credible organizations. 

Recall is the number of relevant documents retrieved divided by the total number of relevant documents available, it measures how many of the available relevant results a search engine actually returns. Theoretically recall can be measured if you know all the documents in the dataset, know which ones are relevant for each query, and if you can count how many of those relevant ones the search engine returns, but this is not practical or easy at all in practice especially because there is no access to the full index and there is no list that is agreed upon on what is a relevant source and what is not.