Search Engine Evaluation

Sammanfattning: With the prevalence of modern search engines, users today are growing more and moreaccustomed to fast and relevant search results. This has resulted in web applications alsowanting to improve their search experience when querying their data. Luckily, there are searchengines on the market already that are ready to be plugged in to an existing application in orderto query data, some open source while others are proprietary products. In this thesis, two suchsearch engines, Elasticsearch and Azure Cognitive Search, will be evaluated in terms ofdifferent soft and hard metrics with the intention of providing insight into which of them best suitsthe needs of a Uppsala based software company, Caspeco AB. A simple prototype is developed that queries the different implementations of the searchengines. Both engines have indexed the same JSON file and are hosted on the Azure cloud.The query results are produced with Powershell scripts that measure the time it takes for acertain API call to execute. In addition to a performance study, a survey for the Caspeco employees is done. The surveyinvestigates opinions about the current solution at Caspeco and is followed up with a shortdemo consisting of the survey respondents testing the prototype. The results show that Elasticsearch comes out on top in regards to query speed,customizability, ease of use and maintenance. In terms of speed both products are retrievingsearch hits in the order of milliseconds but Elasticsearch is faster in most cases. However, thereis a drastic difference in certain queries that leaves Azure Cognitive Search behind Elasticsearch.