About PRISM

PRISM is an acronym for apPRoximate Intelligent String Matcher.

It comprises a webservice developed for matching a written sentence to a set of alternative sentences, and uses some 'intelligence' (i.e. natural language processing techniques) to accomplish this goal. It aligns each of the sentence pairs (sentence-alternative) on the level of their tokens, computes the similarity between the aligned tokens (based on Levenshtein's edit distance), and annotates the tokens linguistically (lemmatization, part-of-speech tagging, and part-of-speech categorization). What you get back for each of the sentence pairs is this aligned annotation, as well as a similarity measure (expressed in percentage).

In other words, PRISM computes the distance between one sentence and a set of alternative sentences.

PRISM can be used to analyze learner language in written (semi)-open tasks, to compare a learner's written response with a set of predicted (correct and incorrect) responses, and to generate corrective feedback on spelling, semantics, and morphology. It is not very suitable for giving feedback on errors related to syntax.

This webservice is not to be confused with the PRISM surveillance programme that was erected by the U.S. Government's National Security Agency. We are not using 'your data' (i.e. whatever you enter into the demo page or submit to the webservice) to collect information about you, nor are we exploiting your data in any way that may harm your privacy.

As to the name, we were first. The NSA ought to know. :-)

Disclaimer and rights

The PRISM webservices and demo page are part of a research project. They are provided "as is", and may be modified or taken off-line at any time. All rights reserved by Ruben Lagatie (Dept. of Computer Sciences KU Leuven) and Frederik Cornillie (Dept. of Linguistics KU Leuven).

PRISM is currently maintained by Frederik Cornillie (frederik dot cornillie at kuleuven hyphen kulak dot be). We provide no support.

PRISM uses the Natural Language Toolkit.