Grammatical Consideration
- The translator performs by mapping of the following Japanese grammatical form using the Church [1] procedure.
Noun phrase(subject) wa Noun Phrase(indirect /direct object) (o/ ga) Verb Phrase.
Into its English form
Noun phrase(subject) Verb Phrase Noun Phrase(indirect/direct objects)
The above grammatical forms are modified to take into account the inclusion of
compound phrases and conjunctions. The Japanese grammatical form are clearly
shown in the third output of the log file.
- Typical snippet is shown here. The English translation of the Japanese words are placed in sequence. The first word val_2 denotes a sequence of words placed inside brackets (…) which are translated but not mapped, val_2 is treated as a Noun word.
27 NP bracks [ bracksP {wa bracksS (wa o VB
28 val_2 NP bracks
29 such ADJ bracks
30 concept N bracks
31 basis N bracks
32 wa P bracks ] bracksP }
33 , PU bracks
34 natural_language N bracks [ bracksP {o
35 input_signal N bracks
36 o P bracks ] bracksP }
37 decide V bracks bracksV < bracksP {VB
38 make V bracks
39 NP bracks [ bracksP {m bracksS (m VB bracksV > bracksP } bracksS )
40 on behalf of CNJ bracks ]
41 well ADV bracks [
42 definition N bracks
43 are being V bracks ] bracksV < bracksP {VB bracksV > bracksP } bracksS )
44 . PU bracks [ ] bracksP {None bracksS (None bracksP } bracksS )
- The parsed sentence is formed by the Church [1] procedure. The key is explained in the sequence in which it is formed. Each opening bracket is closed by its complement.
- NP bracks [… bracks ], This is a series of Noun Phrases separated by post-position particles. The sequence of a set of ‘no’ (‘of’) ending phrases are inverted as part of the subsequent mapping process.
- VB bracksV <…bracksV. This is a Verb Phrase made up of a series of verbs. Possibly preceded by an adverb or a conjunction.
- bracksP {… bracksP } are used to enclose either a Noun Phrase or a Verb Phrase. It has the type of phrase described immediately after the opening bracket. Noun Phrases are shown as ‘wa’,’o’,’ga’ type.
- Compound sentences made up by a starting Noun Phrase with missing ‘wa’ or ‘o’ or ‘ga’ articles but containing verbs are labeled as a first hybrid ‘n’. The second hybrid phrase is labeled ‘m’ .
- bracksS (… bracksS ). This completes a formed sentence. The types of phrases given by the enclosed bracksP are shown immediately after the opening bracksS.
Note that the bracksS are also used to enclose the period ‘.’ PU punctuation on its own.
- Broadly speaking, a bracksS (wa, o, VB bracksS) sentence is mapped into its English format
bracksS( wa, VB, o)
The Verb Phrase is first mapped into its position. At the same time all verbs in the hybrid Noun Phrases are mapped to the starting position on the ‘n’ or ‘m’ phrase. The position of a pair of verbs are reversed.
Then each Noun Phrase wa or o or ga is subsequently mapped for its ‘no’ particle endings ‘P’.
- In this example we have two sentences ie a compound sentence..
BracksS (wa, o, VB), bracksS(m, VB)
The second sentence starts with an m phrase that in turn, starts with a conjunction. The Verb is in its transitive or Gerund form, ie. ‘being’ ‘running’ etc.
- The complete mapping procedure in JEMAP is an evolving algorithm that maps the Japanese grammatical form into a good English form that preserves the original Japanese meaning. It is the Final form in which most of the Semantic based translators are currently unable to produce on a consistent basis.
General Considerations and Good Practice
- Japanese sentences as defined by a closing period, tend to be long. That is because related sentences are strung together by commas rather than periods. In order to get shorter sentences that are mapped into their proper English forms, it is sometimes useful to exchange a period for a comma in the Japanese text.
- The first sequence in the log_test.txt file will show where a well placed period can be made. Similarly a long starting Noun Phrase can often use a ‘wa’ demarcation which then divides the Noun Phrase in two for a better subsequent mapping of the Verb Phrase.
- A combined wa and period insertion is particularly useful in the hybrid phrases ‘n’ or ‘m’ because there are no requirements in English to imbed the verbs in their transitive or Gerund form in a compound sentence. The insertion would simply create two simple English sentences with the same meaning.
[References]
1. Bond, Francis,”Translating the Untranslatable, A solution to the problem of generating
English Determiners.”, CLSI publications, Stanford, 2005.
2. Church, K. W., “A Stochastic Parts Program and Noun Phrase Parser for Unrestricted
Text,” Proc. Second Conf. on Applied Natural Language Processing, Austin, TX, 1988,
pp 136-143.
3. Francis Bond, Hitoshi Isahara, Sanae Fujita, Kiyotaka Uchimoto, Takayuki Kuribayashi
and Kyoko Kanzaki,”Enhancing the Japanese Wordnet” 7th Workshop on Asian
Language Resources in conjunction with the ACL-IJCNLP 2009, Singapore.
4. The Honyaku Japanese Translator, XLsoft Corp., www.xlsoft.com.
5. WhiteSmoke, Writing Software| Grammar Software| Grammar Checker ,
www.whitesmoke.com.