The Indonesian language (Bahasa Indonesia) has a number of uncommon characteristics, such as a great amount of derivational affixes. There are so many combinations of affixes and stems in Bahasa Indonesia that ambiguities often arise. To record all words into a word dictionary is almost impossible because it will make the size of the word dictionary huge and processing time very long. We propose a method to analyze the morphology of Indonesian words by using part-of-speech (POS) tagged data, an affix rule table and minimum connectivity costs to solve the problems mentioned above. Experiments showed that our system achieved a good analysis result (more than 97% accuracy).

View full abstract

In this paper, we will analyse The Elements of Euclid by Head-driven phrase structure grammar (HPSG). In ancient Greek senteces, the word orders are rather free and instead phrase structures strongly depend upon the agreement of features; this is the reason why we adopt a unification-based grammar. In addition to ordinary grammar rules, we add several rules which concerns ellipsis and crossed-dependency. We show that the grammar could cover about 79% for 1154 sentences of The Elements.

View full abstract

In this paper we propose a transfer and generation approach implemented in a prototype Japanese-Chinese machine translation system, jaw/Chinese. We developed the transfer-based system in two phases. The first phase covers Chinese translation solutions for propositional contents and the second covers those for Japanese function word after predicate (FWAP). In phase one, a given Japanese sentence is parsed, propositional contents in it are matched with transfer rules into Chinese semantic elements, and generated into a Chinese sentence. We designed three types of the transfer rules to keep records of Japanese patterns and their semantically corresponding Chinese expressions: base type, content-word addition type and function-word addition type. In phase two, The FWAP is partitioned into four groups, “causative etc”, “tense etc”, “modalities etc” and “conjunctions” by their functions. We designed a table format for translation rules of the former three groups, based on their correspondent Chinese expression components. The translation for “conjunctions” is dealt as function-word addition type. The Chinese semantic elements in phase one and expression components are arranged word order into a linear string. The evaluation shows the approach is effective and acceptable within machine translation.

View full abstract

Paraphrasing is a very important technique which is required by a lot of NLP applications. This paper represents a dictionary-based method of recognizing and paraphrasing periphrastic phrases and overlapping phrases. A periphrastic phrase is a phrase in which a verb does not represent an action, or a noun does not represent an agent/object of an action. An overlapping phrase is a phrase in which there is overlapping meaning between a noun and a verb. Most of these phrases have equivalent phrases which are expressed in a simpler form. The result of recognition and paraphrasing was assessed by two human judges. The result of recognition was evaluated through precision and recall, and the average precision and recall were 78% and 52% respectively. The average accuracy of paraphrasing was 91%.

View full abstract

We propose a method of constructing an example-based machine translation (EBMT) system that exploits a content-aligned bilingual corpus. First, the sentences and phrases in the corpus are aligned across the two languages, and the pairs with high translation confidence are selected and stored in the translation example database. Then, for a given input sentences, the system searches for fitting examples based on both the monolingual similarity and the translation confidence of the pair, and the obtained results are then combined to generate the translation. Our experiments on translation selection showed the accuracy of 82% demonstrating the basic feasibility of our approach.

View full abstract