Home and information on tool use. Readability Metrics Cloze Testing User Evaluation Natural Language Processing and Machine Learning based approaches. Help for developers using this tool. More information about research on which this website is based Important legal disclaimer. Read before using this tool.


Enhancing the Readability of Legislation

Outline

These web pages provide demonstration/manual facilities for assessing the readability of legislation (legal rules). Although provided for this purpose, they can also be used for readability assessments of other documents. A key objective of this site is to provide readability research tools to enable researchers to undertake readability research on legislation. These readability tools are available as manual services (i.e. by pasting text into the relevant web page and using the tool), or as a readability tool server which can be accessed via http calls to the service. This latter service is designed for researcher/developers and more information is provided on the help page.

NB: ALL FEATURES ARE IN BETA PHASE TESTING, PLEASE REPORT ANY BUGS OR ISSUES.

Readability Metrics

Readability metrics have been widely used since the 1930's for the purpose of assessing the readability of text. Such metrics have been used to study legislation, but their use is controversial. The most familiar such metric is the Flesh Readability Ease metric, which can be called in Microsoft Word. This is only one of the available metrics. There are hundreds that have been used. Nine different metrics are provided through this site. Visit the metrics page for the manual service, or the help page for information on how to send http calls which return the required metric.

Cloze Testing

Cloze testing is based on an objective test involving a human subject undertaking a cloze test, which tests the ability of the subject to correctly guess a missing word removed at intervals from a text. This test has been shown to objectively measure the readability difficulty of a text.

Subjective User Evaluation

Subjective user evaluations of the difficulty of a text have been shown to be an empirically valid method of assessing the readability of a text. Indeed, it is probably the best such measure, particularly if sufficient empirical data is collected.

Machine Learning and Natural Language Processing

Traditional readabbility metrics were designed to be easily calculated with technologies available at the time they were created. Metrics can typically be calculated by visually counting characters, words, syllables and sentences. These simple features were selected as substitutes for more complex language features that were difficult to calculate. Natural language processing and machine learning have in more recent years been used to develop new measures of text readability. Features include word frequency and morphology, parts of speech frequency, syntactic structure, parse trees.