In a parsing table for an expression grammar, the reduce action is used to combine grammar rules to reduce a portion of the input string into a non-terminal symbol. When the parser encounters a reduce action in the parsing table, it replaces a set of symbols on the top of the stack with the non-terminal symbol that corresponds to the grammar rule being applied. This helps in simplifying the input string and moving towards the final goal of parsing the entire input.

Seppo Sippu has written:

'Parsing theory' -- subject(s): Formal languages, Parsing (Computer grammar)

Paola Merlo has written:

'Parsing with principles and classes of information' -- subject(s): Computational linguistics, Parsing (Computer grammar)

To show that a grammar is LALR but not SLR, you can construct a parsing table for the grammar and demonstrate that there are conflicts present in the SLR parsing table that are resolved in the LALR parsing table. Specifically, LALR parsers have larger look-ahead sets than SLR parsers which can resolve these conflicts. In other words, LALR parsers are able to distinguish between more parser states compared to SLR parsers, allowing them to handle more complex grammars.

J. van de Koot has written:

'An essay on grammar-parser relations' -- subject(s): Competence and performance (Linguistics), Generative grammar, Parsing (Computer grammar)