A Productivity+PLUSapplication   
from X.Systems, Inc.   

Controlled language, represented by such formal specifications as AECMA Simplified English as well as vocabularies developed and used by individual companies and industries, is a growing method of dealing with the demands of language translation and use by differing audiences.

While controlled language is highly effective in its use, it presents the writer with a difficult authoring task. Writing in a tightly controlled vocabulary, often with as few as 800 basic words and tight controls on terminology/grammar, can be time-consuming and error prone.

Language Manager addresses this difficulty by coupling The Arbor Text editor with a powerful terminology control engine to provide near real time analysis and correction support. LM was carefully designed to provide authors and editors with the tools necessary to achieve maximum productivity in the demanding Controlled Language environment.

Language Manager Design Goals:

Language Manager is designed to make the writer's task easier, more productive and less stressful. Studies have shown that near-immediate feedback to the author facilitates more effective correction by signaling errors while the current content is fresh in mind. In addition, the reinforcement of coupling successful correction with the original thought process generates a reduction in first-time errors as writing continues.

With LM, control of the language analysis process always resides with the writer. Analysis results returned from the terminology engine are made available to the user in either "immediate correction" or "review later" mode, enabling effective control and scheduling of the review and correction process commensurate with the type of data, the current schedule requirements and the volume and nature of errors identified.

Modes of operation:
The LM user may select from several modes of operation:

Check current element - causing Arbor Text to send only the tagged element in which the insertion point resides. This allows the user to select any logical segment of the content, from the entire file down to the smallest element containing a full semantic unit (sentence or equivalent.) The element is returned from the terminology engine as individual sentences, each of which is then displayed in the LM Correction Panel, setting the stage for immediate correction if the user desires. At any point during this process, the user may elect to terminate the correction process, replacing all content in the file for later resumption.

Check selection - allowing the user to select a single sentence within an element for analysis. The system returns the sentence for analysis in the panel after which it is returned to the position from which it was sent. The diagram below illustrates the flow of data and procedss in a Check Current Element or Check Selection transaction:

Continue Correction - If a sentence has been returned to the editor with one or more uncorrected errors, this function allows the correction process to begin with only those remaining errors displayed. In this mode, the content is not resent to the terminology engine, minimizing its operating load. Instead, the partially corrected sentence is taken from the stored correction inside the document file and reloaded into the correction panel as illustrated below.

Reference Display - When the user has made all appropriate corrections in the LM Correction Panel, there may be additional grammatical restructuring to be completed in Arbor Text. This often happens when errors such as passive voice or noun clusters are encountered. While this work is in process in Arbor Text, the user may call for a view of the correction panel showing all of the original errors, analysis and suggestions. In this mode, the panel may be navigated but no changes made through it. The diagram below illustrates this mode of operation.

Incorporate correction - When the user is satisfied that the sentence has been properly modified, this function may be used to remove all change tagging and make the new version a normal part of the file. Optionally, this ability may be granted only to certain users, ensuring that senior editors review all pending corrections before acceptance.

Incorporate and Recheck - This function incorporates the new version of the sentence, then automatically selects and re-sends it to the terminology engine to ensure its validity. If errors are detected, the correction process is restarted as described above.

Cancel correction - If a pending correction is deemed unacceptable for any reason, this function returns the sentence to its original, pre-corrected state.

Communication between the Epic Editor and the Terminology Engine:

The user invokes LM process in the editor, selecting from the options described above.

Arbor Text sends designated data to the terminology engine as defined by the selected option. All tagging within and surrounding the designated data is sent intact. If multiple version tagging exists within the selected element, the user is asked for which version the analysis should be done (see Version Manager.) Only a single version may be sent at one time.

The terminology engine receives data from the interface module, analyzes it and generates the following data for return to Language Manager:

Original tagged elements containing errors, with all initial tagging intact and the addition of a defined tag surrounding any word or group of words failing the grammar test. This tag contains the following elements:

• sentence length
• error type code (up to three or each error)
• replacement suggestions; up to six, separated by commas.

If the element sent to the terminology engine contains multiple sentences or other semantic units to be analyzed, the returned data contains processing instructions delimiting each sentence or linguistic unit to provide positive separation.

Processing of data returned from the Terminology Engine:

Upon receiving the returned data element, Language Manager takes the following actions:

If no error elements exist in the returned element, the correction panel is not activated and no change is made to the Arbor Text open data file.

If any errors are returned, Language Manager activates the Correction Panel, illustrated above, displaying the first sentence returned. At this point, all sentences returned by the terminology engine are separated for individual correction and replacement in the document file.

For each sentence returned, the user may start correction immediately or defer it until later by clicking the "sentence finished" button on the panel and responding "yes" to the resulting "uncorrected errors OK?" query. Corrections made be made by double clicking on a suggestion for the active error, by entering desired replacement text in the space provided (entering ".." here deletes the error word/s) or by clicking on the "ignore" button to indicate that the error has been noted and no replacement changes are being made.

As each sentence is reviewed and corrected, the panel displays the next sentence until all have been corrected at which time the element is returned to Arbor Text as a series of Red Line changes, one for each sentence.

Any uncorrected errors are stored as part of the change to facilitate continuation of the correction process at the user's discretion.

If the author needs assistance with correction, the Examples database, illustrated in figure LM-3, may be activated to display all instances of correction of the word and error type currently highlighted in the LM panel

When the Red Line change element containing the pending correction is eitheriincorporated or canceled, all tagging is removed and the text set to the new or original versions of the text respectively.

Examples Database:

Language Manager provides a unique support feature, the Examples Database and display. As each correction is made to a sentence, the original and corrected versions are automatically presented to a designated administrator for evaluation as a potential correction example. If the correction illustrates a useful correction approach to a particular error, it is entered into the Examples Database, keyed to the selected error type and particular error word or text. Sentences containing multiple errors may be entered into the database multiple times, keyed to different errors and error text.

As this database grows, it represents a valuable resource for writers facing similar problems, especially in view of its focus on how the particular organization deals with those problems. Figure LM-3 illustrates an instance in which the user has confronted an error for which the system could not make a suggestion, and found the answer in the examples database.

If the correct text is multiple words, the user may copy it from the example display and paste it in the "replacement text" window for entry into the sentence. If, instead, the illustration is concerned with sentence structure, it may be used during sentence correction via the "Reference Display" mode of the correction panel.

return to software page