MultiTrans PRISM Analysis Agent Tutorial,


Level I


Other MultiTrans tutorials: Other statistical analysis tool tutorials:
MultiTrans PRISM TextBase Builder Tutorial, Level I  
MultiTrans PRISM Term Extractor Tutorial, Level I  
MultiTrans PRISM TermBase Manager Tutorial, Level I  
MultiTrans PRISM Translation Agent Tutorial, Level I  

MultiTrans PRISM TextBase Agent Tutorial, Level I

MultiTrans PRISM TermBase Agent Tutorial, Level I  



I. Introduction


MultiTrans is a translation environment. Among other features, MultiTrans includes an Analysis Agent, which can compare a source text with the TextBase and TermBase in order to generate statistics based on matching segments found in these two databases and identify candidate terms within this source text. The results generated by the Analysis Agent can indicate how much of the source text will need to be translated from scratch and how much can be reused from the TextBase and TermBase. This kind of analysis can help a user to better estimate the time and effort required to translate a text, and to evaluate how useful a particular TextBase and TermBase are likely to be for the task.


To see how the Analysis Agent fits into the translation process in MultiTrans, consult the MultiTrans Work Flow diagram.


You can find out more about MultiTrans by consulting the MultiCorpora website at http://www.multicorpora.com. You can also read information about MultiTrans's different functions by selecting MultiTrans Help from the Help menu, once you have opened MultiTrans.

II. Getting ready


  1. Save the files you will need for these exercises:
    1. Create a sub-directory called MultiTrans Analysis Agent or another name that you wish. (For instructions, see Creating a sub-directory in Windows.)
    2. Download the file WHO obesity - text to translate. This is a French-language Word document about obesity that is different from the document used to create the TextBase in the MultiTrans TextBase Builder Tutorial, Level I. (See Note 1.)
  2. Locate the files you will need for the tutorial.
    1. If you have completed MultiTrans TextBase Builder Tutorial, Level I and MultiTrans TermBase Manager Tutorial, Level I, your TextBase and TermBase will be saved in their respective folders on the Home (H:) drive (H:\MultiTrans).
    2. If you have not created a TextBase and TermBase, download the folder MultiTrans Bases - Agents, extract the contents and save the TextBase and TermBase in their respective sub-directories in your H:\MultiTrans folder (i.e. H:\MultiTrans\TextBases and H:\MultiTrans\TermBases). (For instructions, see Opening premade bases in MultiTrans.)
  3. Open MultiTrans.
    1. MultiTrans opens and asks you to enter a username. Enter the name of your choice and click the Start button.
    2. The main page you see is referred to as the Start Screen. It displays icons for various MultiTrans functions (e.g. Start SessionAnalysis) and allows you to navigate between them. (See Note 2.)
  4. Open your TextBase and TermBase, choosing French as the source language and English as the target language. (For instructions, see Opening a TextBase and TermBase in MultiTrans.) The bases must be open in order for you to compare the text to be translated with them. 

III. Using the Analysis Agent to analyze a text


The Analysis Agent allows you to compare a text to be translated with the contents of a MultiTrans TextBase and TermBase. The comparison will show you the resemblance between your bases and the text to be translated and can help you in giving accurate delivery dates and quotes to clients that take into account the savings in time and effort that the TextBase and TermBase may offer.


  1. From the Start Screen, click on the Analysis icon. (If you are currently viewing the TextBase or TermBase, return to the Start Screen by selecting this option from the View menu.) The Analysis Agent dialogue box opens.
  2. Specify the features of this analysis.
    1. Complete the options in the upper left-hand portion of the Analysis Agent dialogue box.
      1. Ensure French is selected as the Source language.
      2. Ensure English is selected as the Target language.
      3. Leave the default setting next to Most frequently used documents.
      4. Leave the default setting next to the Pre-translation editor dropdown menu. 
    2. Complete the options in the upper right-hand portion of the dialogue box.
      1. Ensure the boxes next to TextBase and TermBase, under Compare to, are checked, so that MultiTrans generates results that reflect both of these bases.
      2. Ensure the box next to TextBases matches, under Extract, is checked, so that MultiTrans identifies correspondences between the source text and the TextBase.
      3. Ensure the box next to Terminology, under Extract, is checked, so that MultiTrans identifies candidate terms found in the source text. 
    3. Modify the matching options.
      1. Click on the Matching Options button at the right of the dialogue box. The Matching Options dialogue box opens.
      2. Click on the Penalties and Bonuses tab at the top of the dialogue box. 
      3. Each option allows you to adjust the match percentage by increasing it (i.e. a bonus) or decreasing it (i.e. a penalty). For now, ensure all sliders are at 0% (i.e. no penalty and no bonus).
      4. Click the OK button.
    4. In the lower half of the box, identify the source text you would like to analyze.
      1. Click on the Add documents icon (a green plus sign) (see Screenshot 1) beside the Documents to Analyze field.
      2. Locate the sub-directory where you saved the WHO obesity - text to translate file.
      3. Click on the file name and then on Open. The file name will appear in the box below the browse field.
      4. Click the Start button.
  3. The results of the analysis will appear. Notice that the Process Bar has changed (as compared to when the TextBase Search or TermBase Editor are displayed). (See Note 3.)
  4. Explore the four tabs along the bottom of the window.
    1. The Statistics tab shows statistics on the resemblances between the text you are analyzing and the contents of the TextBase and TermBase.
      1. Examine the MultiTrans Performance - TextBase section. The statistics indicate how many segments (usually sentences), sub-segments, and paragraphs found in the text are identical or similar to those in the TextBase. (The resemblance is expressed in a percentage of word forms in the text to be translated that are also found — in the same order — in the TextBase.) (See Note 4.)
      2. Examine the MultiTrans Performance - TermBase section. The statistics show how many items found in the text have corresponding term records in the TermBase.
      3. Based on these statistics, do you think this TextBase and/or TermBase will be helpful for translating the text? Why or why not?
    2. The Terminology Extractor tab lists the expressions of two or more consecutive words that repeat within the file you have analyzed.
      1. Examine the features of the expressions in this list.
        • A paperclip symbol to the left of an expression indicates that the term is already in the TermBase. (See Screenshot 2.)
        • A T to the left of an expression indicates that the term was found in the TextBase. (See Screenshot 2.)
      2. Add units to the TermBase.
        • Identify an expression that you want to add to the TermBase.
        • Click in the Input Translation (English) column, next to the expression.
        • In the text field that opens, type in your translation.
        • Repeat these steps for as many terms as you like.
        • From the Analysis Agent menu, select Insert in TermBase and then All terms with translations. Or, to send only a selection of one or more terms, choose the Selected terms with translations option.
        • The terms are automatically added to the open TermBase. 
    3. The TermBase Extraction tab lists all expressions that were found in the open TermBase and the analyzed file, along with the corresponding translation.
      1. How does this extraction compare to the term extraction that is done when you create a TextBase? (Use the results obtained in the MultiTrans Term Extractor Tutorial, Level I as a basis for comparison, if you like.)
    4. The TextBase Extraction tab displays all full and fuzzy matches (75% resemblance and higher) found in the documents in the open TextBase.
      1. It shows you the source text segment (from the document analyzed), the target text segment (from your TextBase), and the match percentage so that you can see which segments you may not need to spend time translating since they are already in the TextBase.
      2. How much of the text that you have just analyzed has some kind of match in the TextBase? How close are the matches?

IV. Saving the results of an analysis

  1. Save a copy of the analysis session as a package that you can re-open in its original format in MultiTrans. (See Note 5.)
    1. From the Analysis Agent menu, choose the Save Analysis session… option.
      1. Click the arrow of the Save in drowndown menu, and locate and select the folder where your TextBase is saved (i.e. H:\MultiTrans\TextBases).
      2. Type WHO obesity analysis session, or another name that you wish, in the File name field.
      3. Click the Save button.
  2. Close the Analysis Agent.
    1. From the Analysis Agent menu, select Close Analysis Agent. You will return to your open TextBase and TermBase.
  3. View the files you have just saved.
    1. From the Process menu, select Open existing Analysis session.
    2. In the dialogue box that opens, find the sub-directory where you have just saved your analysis session and click on the WHO obesity analysis session.mta file.
    3. Click Open. The Analysis Agent window opens and the results are displayed.

V. Wrapping up

  1. Close the TextBase, TermBase and Analysis Agent.
    1. From the File menu, select Close Session.
  2. Close MultiTrans.
    1. From the File menu, select Exit.
  3. Make a copy of your files as a backup, or transfer them to another computer.
    1. In the Home (H:) drive, create a sub-directory called WHO_Bases_yyyymmdd, (replacing the series of letters at the end with today's date), or use another name that you prefer. (For instructions, see Creating a sub-directory in Windows.)
    2. Copy your TermBase and paste it in the sub-directory you have just created.
      1. Locate the TermBase in the H:\MultiTrans\TermBases sub-directory.
      2. Copy and paste it by using Ctrl+C and Ctrl+V, or by choosing Copy and Paste from the contextual menu that appears when you right-click on a file or folder.
    3. Copy your TextBase folder and paste it in the sub-directory you have just created. This folder will also contain the .mta analysis session saved in the previous section. (See Warning 1.)
      1. Locate this folder in the H:\MultiTrans\TextBases sub-directory.
      2. Copy and paste it by using Ctrl+C and Ctrl+V, or by choosing Copy and Paste from the contextual menu that appears when you right-click on a file or folder.
    1. Make a compressed folder that contains the sub-directory created in step 3a. (For instructions, see Creating a compressed folder in Windows.)
    2. Copy this compressed folder to a USB key. Or, if the folder is less than 2 MB, send a copy as an attachment to your email.






NOTE 1: The Analysis Agent will analyze the file, but will not change it, so there is no need to make a copy of the WHO obesity - text to translate document at this point. Other MultiTrans functions — such as the Translation Agent — will write over your file, so in general it is a good idea to make sure you have a backup copy when you use those functions.














NOTE 2: You will not be able to access the Translation Agent, TextBase Agent, or TermBase Agent from the Start Screen. For information about how to access and use these tools, consult the corresponding tutorials: MultiTrans Translation Agent Tutorial, Level I; MultiTrans TextBase Agent Tutorial, Level I; MultiTrans TermBase Agent Tutorial, Level I.  





























Screenshot 1 coming soon.






NOTE 3: If you do not see the Process Bar, click on Process Bar from the View menu.








NOTE 4: The TermBase matches will likely appear as 0; this does not mean that there were no term matches. This value is 0 as there are no segments in your TermBase that match segments in the text analyzed.








Screenshot 2 coming soon.


























NOTE 5: You may also save the results as an HTML package. These files can be opened without using MultiTrans. For instructions, see Saving the results of an analysis as an HTML package.

























WARNING 1: It is very important that you copy everything in your TextBase folder. This includes the sub-directories Content and Indexes, as well as the TextBase file ending in .tcs. All of these files are needed to open a TextBase in MultiTrans.







VI. Questions for reflection


  • While doing this tutorial, what were your first impressions of terminology management in MultiTrans? 
  • How is the Analysis Agent useful? Do you think this could help you in your professional translation career? How does it compare to the Analysis option in other tools you have tried?
  • Did you find that the Analysis Agent results gave you a good indication of how the TextBase and TermBase could help you with the translation? Do you think it could help you give a client a more accurate estimate of the time and work involved in doing a translation?


Tutorial created and updated by the CERTT team. (2011-04-13)

Tutorial updated for Prism by Trish Van Bolderen. (2012-08-09)