The Arabic DeConversion...
The DeConverter generates target language sentences in natural language from UNL expressions by applying the DeConversion rules. The DeConverter transforms the sentences previously enconverted into UNL expressions (semantic networks made up of a set of binary relations) into a directed hyper-graph called the Node-net. The root node in a Node-net is called the Entry Node and represents the main predicate of the sentence. The DeConverter then applies generation rules to all the nodes in the net in order to generate a node list in the target language. Afterwards, the syntactic structure is determined by applying syntactic rules, and morphemes are generated by applying morphological rules. Hence, the DeConversion process is divided into two stages; the syntactic stage and the morphological stage.
-
The syntactic stage
The syntactic stage deals with the order of words in the node list and can be further divided into two sub-stages:
- The first stage has to do with determining the main predicate of the sentence together with its modifiers representing the main skeleton of the sentence.
- The second stage is concerned with determining the modifiers related to the elements composing the main skeleton.
-
The Morphological stage
The Morphological stage is the final stage in the DeConversion process, it specifies the correct word forms and controls gender, number, person and definiteness agreement. The Morphological stage is concerned with three axes:
- Inserting affixes into the node list to generate the final forms of the words according to both, the linguistic features attached to each in the dictionary, and the attributes attached to the nodes.
- Inserting the prepositions, attributes, and pronouns needed for the Arabic syntactic structure under generation.
- Inserting punctuations and spaces wherever needed.
Achievements of the Arabic DeConversion:
- DeConverting 25 articles from the Encyclopedia of Life Support Systems (EOLSS).
- Preparing an evaluation database (2000 Sentences) from the articles of the Encyclopedia of Life Support Systems (EOLSS).
- Generating the Arabic counterparts of the UNL expressions of 200 English and French titles.
- Verifying the UNL expressions of 800 Arabic titles.
|
The Arabic UNL Dictionary:
The Arabic UNL Dictionary The size of the main general dictionary reached 99,908
entries representing 52,572 universal concepts.