Delilah is a computer program for semantic parsing, and generating well-formed Dutch sentences from logic, based on categorial grammar.
Copyright © 1995- Crit Cremers & Maarten Hijzelendoorn, Leiden University Centre for Linguistics (LUCL)
Delilah consists of two parsers, one generator, and a shared lexicon.
Only the chart parser is available.
The parsers analyse Dutch sentences, and return syntactic trees, detailed HPSG-style feature templates, and semantic formulae. To operate the parser, type a Dutch sentence in the input box. No difference is made between uppercase and lowercase letters. A sentence may end on a '.', an '!' or a '?'. Then, click on the button. The sentence will be analyzed. The arrow button in the input box displays a randomly selected list of example sentences. The list is refreshed after each analysis. Select one of them, and click Parse! to have it analysed. For reference, see the button in the Lexicon tab, to get a list of all available word forms that the parser can handle.
The generator produces well-formed Dutch sentences, driven by semantic concepts. To operate the generator, type one or more concept names in the input box, and click on the button. An arbitrary sentence will be generated. The arrow button in the input box display a randomly selected list of example concept names. The list is refreshed after each sentence generation. Select a concept name from the droplist, add one or more concept names, and press Generate! to generate a sentence, based on these concepts. The generator will repeat generating sentences on the basis of some randomly chosen concepts of the previously generated sentence, each time the Generate! button is pressed. This is what the option called Stories does. One can think of the successively generated sentences as a story. The option called Just Sentence tells the generator to generate sentences only on the basis of the concepts that are entered by the user. For reference, see the button in the Lexicon tab, to get a list of all available concepts.
A generated sentence can be fed to the parser by clicking the button. Thus, one can check the grammatical convergence of the parsing and generating engines.
The output of the parser and of the generator can be fine-tuned by using one or more View Options in the Control Panel tab. By default, all options are selected. They give detailed information about the parsed or generated sentence. The option called Syntactic Trees turns on/off syntactic trees in the output. The option Argument Structure turns on/off combined trees, containing syntactic, semantic and phonological labels. The option Templates turns on/off HPSG-style templates in the output. The option Semantics turns on/off semantic formulae in the output.
Viewing without frames
The output of the parsers and the generator can be shown in different ways, depending on the capabilities of your browser. When your browser supports frames, the main window is subdivided into two frames. The user interface is in the upper frame, and the output will come in the lower frame. The detection of the support of frames is done automatically, but the result can be overruled for technical or educational purposes. When your browser does not support frames, the user interface will be shown in one page, and the output will be displayed on another page, which is the same for each request. This "page" might be a tabbed page in the same browser window of modern browsers.
The "Output" tab is not really necessary, but can be used to save the output file more easily. The Output tab is only visible when there is something to display in the output frame.
The Lexicon, which is accessible through the Lexicon tab, has different entrypoints. The button presents a list of all semantic concepts in alphabetic order. The button displays a list of all word forms in alphabetic order. By pressing the a list of all lemmas is displayed in alphabetical order. Finally, all lexical templates associated with the word form can be displayed by entering it in the Lexicon field and by pressing the button. Beware: some word forms have dozens of templates! It is not possible to add entries to the lexicon.
Cookies are unused.
Output File Format
The generated output files comply with HTML 4.01 Strict validation rules, provided by W3C, the World Wide Web Consortium (see http://validator.w3.org).
FAIL: an internal problem prevented the parser from analyzing the input sentence
FAIL: an internal problem prevented the generator to produce a well-formed sentence
FAIL: unknown word form
Program failures or bad user input.
TIME-OUT: the allowed processing time was too short for the parser to analyze the input sentence
TIME-OUT: the allowed processing time was too short for the generator to produce a well-formed sentence
TIME-OUT: the allowed processing time was too short to produce all lexical templates
A session cannot last longer than 60 seconds.
Problem with creating, validating, or transforming an output file.
Delilah is programmed in SICStus Prolog The webinterface is best viewed with Mozilla Firefox, or Google Chrome.
Created: April 17, 2007, Last modified: October 4, 2018