Semantic search engine
A semantic search engine is a search engine that takes natural language as input and tries to capture the semantics of a question. It then searches for suitable answers in its database, the semantics of which the search engine has analyzed (usually in advance). Ideally, the answer consists of individual statements instead of entire documents, as is common with keyword-oriented search engines.
Meaningful search engines
A search for the term 'hits', for example, in a search engine based on key terms can only display websites that contain the searched word or that are referred to with this term. A search query for 'hits' to one of the already existing meaningful search engines, on the other hand, also uses terms that are often mentioned on the web in connection with' hits', even if they are not mentioned in the query itself (e.g. 'mp3', ' Songtexte ',' Musikvideo '), and accordingly outputs websites that do not contain the originally searched word' Hits'.
Search engines with processing of natural language search queries
In natural language, changing just one word can completely change the meaning of the question. Example:
- When did Martin Luther King die?
- How did Martin Luther King die?
A search engine based purely on keywords would provide almost the same references for both queries. Understood as a natural question, the questions achieve significantly different answers. A successful semantic search engine will detect this difference and be able to provide a document containing the date of death in response to the first question. In the second case, the answer should focus on the assassination attempt.
In natural language, the same question can be put into different words. Example:
- When did Martin Luther King die?
- What Day Was Martin Luther King Assassinated?
- State the date of death of Martin Luther King.
The need for information is the same for all formulations. A semantic search engine should therefore present the same answer to all of them. A keyword-oriented search engine, on the other hand, will present different locations and may not find any hits at all.
The developments in the processing of general, natural language inputs have so far been limited to fairly simple queries and mostly to the evaluation of the English language. Examples of complex (multiply conditional, proportional measures using, and characterized by fuzziness) are:
- "Show me a list of all works by German-language writers who lived in Paris des fin de siècle and were born after 1850."
- "Show me all the suppliers of a product similar to XY who have been on the market for more than 5 years, who have since been largely rated as very good by their customers, and who, if possible, are also above average inexpensive."
Another approach to answering complex, natural language questions pursues a much stronger content structure formation in a web that is currently barely structured in terms of content (see semantic web ).
Examples
- Bing (hardly any semantic approaches recognizable)
- Google , especially with the "Knowledge Graph"
- GoPubMed , semantic search engine for the biomedical domain
- Swoogle semantic search engine that can search documents, terms and data on the semantic web
- Wolfram Alpha , the “answer machine” from mathematician Stephen Wolfram , currently only available in English, with an emphasis on the exact sciences
- AskWiki , semantic search engine for the data stock of the German-language Wikipedia with input of the query by language, discontinued in 2013