UGTag - a moprhological tagger for Ukrainian language

Description

Morphological Dictionaries

Getting the Program

Description

UGTag is a program for annotation texts with morphological and syntactical information.  It was written within Polish-Ukrainian Parallel Corpus project, coordinated by dr Natalia Kotsyba. Main features of the program include:

• cross-platformness - being written in Java, UGTag is cross-platform. It will work on all system where Java and SWT (Standard Widget Toolkit, a library used by UGTag to create user interface) are ported, and these are more than a dozen of operating systems, like Windows, Linux, BSD-like clones and MacOS. 
• intuitive and user-friendly graphical interface - program is very easy to use. Just open the text you want to tag, and the program will start to annotate it automatically. 
• syntax coloring - the user can tune how the tagged text will look like.  For example, words that were not recognized by the tagger (more on that later) can be shown underscored with red lines, to draw the attention. Or, words with multiple interpretations can be shown in different font/with different color from those that have only one interpretation
• manual disambiguation - for words with multiple interpretations the user can manually select the correct one (i.e. disambiguate) by simply selecting the one from context menu 
• embedded sentence editor - along with tagger, UGTag includes sentence analyzer, which automatically determines where are the borders between sentences. Moreover, it allows the user to manually re-layout sentences.  To split a sentence in two, you just select the "Split Sentence" command, and the click on the place  where you want to put the sentence break. To join sentences, you select the text, click the "Join Sentence" command, and -  voila, all sentence mark within selection were removed
• list of unrecognized words - during tagging phase UGTag makes the list of words it was not able to recognize. The user can later edit the list, to manually add interpretations of those words. The information can either be stored in that document only or be recorded in the user dictionary (see below).
• console, GUI and Web clients - UGTag allows the user to interact with itself in three different ways. The GUI client (mainly discussed here) is used for interactive text annotating; console client is used for processing large amounts of text without user intervention. Finally, Web client allows the users to tag their own texts using their Web browsers. 

Figure 1. UGTag GUI client with small text file opened. Sentence marks and syntax highlighting is switched off.

Figure 2. Dialog for editing visual appearance of tagged text. Words with zero (i.e. unrecognized) one or more interpretations can are distinguished, as well as punctuation marks or numbers. The user can tune font face, size and style (bold, italic or underline), as well as background and/or text  colors.

Morphological Dictionaries

UGTag has  a dictionary-based tagger. Dictionaries are stored in binary format (for faster loading), but they can be  as well imported from an XML format. UGTag default dictionary is based on Ukrainian Grammatical Dictionary, which was created under guidance of dr. Igor Shevchenko fom ULIF NANU. It was, however, significantly (and I mean really that) reorganized, to be compatible with MultiLex-East specifications. Work on the Ukrainian part of the specifications took considerable amount of time. 

Since the UGS data is not publicply available, UGTag cannot be readily released to wide public. There are plans, however, to make it an open-code. Having program in hand, users are encouranged to make their own dictionaries. 

Figure 3. Dictionary list editor. Users can create new dictionaries, load existing ones from external files, and edit dictionaries' contents.

Figure 4. Dictionary editor. The user can change the name, optional descriptions and the list of authors.

Figure 5. Dictionary content editor. Basically a dictionary is a list of lemmas and a list of entries. Each entry is in essence a word form with bounded grammatical interpretation and reference to  word lemma. 

Extending Dictionaries

Sometimes data present in the dictionary is not enough. In this case UGTag makes a list of words that were not found in dictionaries and presents them in the following dialog.

Figure 6. Dialog for editing unrecognized word list. Fortunately, all words were found.

Figure 7. Dialog for adding interpretation of unknown words directly to user dictionary. 

Getting the program

The project is hosted on SourceForge.net, https://sourceforge.net/projects/ugtag

Last updated, 17 October, 2009,  by Andriy Mykulyak. See also PolUkr and project time-line.