This toy machine translation program is based on the syntax-directed translation approach built around a shift-reduce parser. The key idea is the same as the ones used in computer language translation in compilers.
Unlike computer languages, human languages are ambiguous. In order to analyze all the available parses, the program implements a manual backtracking/alternative selection mechanism. Manually, the user can explore the entire parse space. The trace of operation is kept in a configuration stack. This stack is provided in addition to the parse stack, which is used for the shift-reduce operations. Thus, the program is not a direct implementation of a Push-Down Automaton.
One obvious question is whether and how we can automatize the process so that the program can evolve into a real machine translation system. Furthermore, the user must feel frustrated using this program. Analyze why would naturally lead to a variety of key concepts in Natural Language Processing.

Loading example grammar files: The default example file is "http://www.tcnj.edu/~komagat/MT0/Example1.txt". There is another example file called "Example2.txt".
Loading other grammar files: Text files on the server "www.tcnj.edu/" can be referenced in the same way. Files cannot be loaded from other servers.
Modifying the grammar: The grammar in the grammarArea can be edited. In order to use the modified grammar, you must press RESET.
Entering input sentence: Enter a sentence so that it would match the terminal symbols (case sensitive). Unless punctuation symbols are actually part of the grammar, do not use them.

The grammar format must follow the pattern "X {Y} -> A B C", where X, A, B, and C are grammatical categories (called "type" in the program) and Y is the semantic interpretation associated with X. There must be exactly one category on the left-hand side (LHS). There must be at least one category on the RHS. There is no preset limit on the number of categories on the RHS. The semantic interpretation section can contain any string. However, the categories that appear on the RHS will be replaced with the semantic representation associated with the category. Below is an example rule from "Example2.txt". The effect of this rule is that the categories NP and VP will be combined to form a single category S with the corresponding semantics "NP wa VP". If NP and VP have semantic representations "Furby" and "neru", the interpreted semantic representation for S will be "Furby wa neru".

A dictionary entry typically appears with a single category on the RHS as shown below.

If the semantic effect of a word is not reflected in the target language, the interpretation section can be left empty. However, a pair of curly brackets are still required.

If there are multiple instances of the same category on the RHS and they are associated with distinct semantic representations, the categories must be distinguished by indices in the form of "_#", where # is a numeral. This is necessary to ensure that the corresponding semantic representations can be used correctly in the semantic translation part between "{" and "}".

In the following example, commas "," are used to match commas in the input (they must be space-delimited). Since these commas are not associated with a particular semantics, they are not distinguished by "_#". Note that the comma in the interpretation section has nothing to do with the commas on the RHS. In general, you can insert any string (except for some special symbols such as "}" and ">") necessary to generate desired semantic representation.

Much like commas, "and" is not defined in "Example2.txt". When "and" appears in a pattern like "NP and NP", it is used as a special symbol to activate this rule as in the example below. Note that the semantics of "and" is embedded in this rule and the corresponding interpretation "to" is generated by the rule, not derived from the lexical entry "and". However, there are other ways to do the same thing.

Convention: Use lowercase for terminals (dictionary entries) and upper case for nonterminals (grammar rules).

<End>