All aboard the Semantic Bandwagon
The document discusses different types of semantics including natural language semantics, lexical semantics, and compositional semantics. It also discusses how search engines currently use statistical methods rather than semantic methods. While statistics have been effective, the document argues for semantic markup, especially for non-textual data like tables, microarray data, experimental protocols, and chemical structures, which do not come in sentences. Semantic markup of chemical structures could allow for better indexing and searching of literature for chemists. The document poses questions about how letting words stand for themselves and semantic markup could be achieved and what non-textual data and external resources an organization may have.
All aboard the Semantic Bandwagon
- 1. All aboard the semantic bandwagon… … hold on a moment Colin Batchelor 20101202
- 2. What do people mean by semantics? Semantic web Specifying how web service shall speak unto web service Natural-language semantics Lexical semantics Compositional semantics
- 3. What do search engines do? Words stand only for themselves (Sp ä rck-Jones) You shall know a word by the company it keeps (Firth) Search engines don’t use semantic methods.
- 4. Statistics versus semantics A counterblast: Alon Halevy, Peter Norvig and Fernando Pereira, “ The Unreasonable Effectiveness of Data ”, Intelligent Systems , 2009, 24 , 8. But STM publishers don’t just publish text.
- 5. Beyond text Data which doesn’t come in sentences: Tables in articles Microarray data Experimental protocols Crystallographic data Chemical structures
- 6. The case for chemical markup There are very well established and understood methods for indexing and searching chemical structures. Chemical structures are the chief route into the literature for many chemists. We can extract chemical structures at least semi-automatically from: Molecule names Author-supplied graphics
- 7. 5 questions for you What could you achieve through “letting words stand for themselves”? What do you hope to achieve through semantic markup? What kinds of non-textual data do you have access to? What can you do with your non-textual data? What external resources could be opened to your readers?
- 8. Any questions for me?