The Korean parser supports two kinds of sentences:
Korean verbs are inflected for politeness level, tense, aspect, voice and sense.
The verb dictionary provides the plain (colloquial) and the polite forms of the verbs.
There are only two tenses in Korean, past and non-past. The non-past covers both the present and the future tense.
The verb aspect denotes the conjugations for perfect, progressive, potential and must forms. The perfect aspect is the stative form of the verb. The must form denotes necessity.
The verb sense indicates whether the verb inflects for a positive or a negative statement. For each of the above, the verbs are inflected by suffixing some ending based on which verb group they belong to.
The verbs do not inflect for gender or person.
The verb voice refers to whether the verb is an active or passive mode. In Korean, the passive form in turn inflects for politeness level and aspect, and is therefore treated as a verb in its own right. The passive verb forms are entered by using their dictionary form in the verb dictionary.
The postpositions 이다 (ida) and 아니다 (anida) have also been included as part of the verb dictionary.
For compound verbs where only the second verb is inflected, for example 유명하다 which is the composite of 유명 and 하다, 하다 is taken to be the active verb. For such (noun + verb) verbs, there is no need to enter the compound verbs separately as long as the active verb belongs in the verb dictionary.
The following are the verb forms present in the verb dictionary:
The automatic verb conjugations works for majority of the verbs. The conjugations for irregular verbs will have to be entered manually. See Configure list of recognized verbs for more information.
Adjectives in Korean behave very much like verbs. For example, all adjectives are conjugated for the politeness level, sentence sense and aspects mentioned above. Thus, the adjectives are also entered using the verb dictionary. In spite of all the similarities with verbs, the adjective conjugation has a few peculiarities. For example, adjectives do not conjugate for progressive aspect. Therefore, the text boxes corresponding to progressive aspect will always be empty for adjectives. When using the verb editor, if you select the dictionary form as being an adjective, these dissimilarities are handled by the verb conjugator.
If you follow the sentence structure guidelines here for creating Korean sentences, then the verbs or the adjectives in the sentence will always end in -da (다). Verb forms that end in characters other than da are not handled. These include imperative form, inquisitive form, connective and, connective if and certain propositive form.
If adjectives ending in 하다 are used in sentences, then these adjectives must first be entered in the verb editor. This is because the verbs and the adjectives that end in hada behave differently. The system has no way of differentiating between a verb and an adjective unless they are already a part of the verb dictionary. The verbs ending in hada behave correctly because they use the inflections of hada as endings. On the other hand, with adjectives in plain present positive form, the ending gets changed slightly (the ending hada is used as opposed to한다).
In an SOV sentence, the verb is taken to be the active verb. If adjectives are present within the sentence, they are not inflected.
In a verbless sentence, the adjectives are inflected based on the noun + verb combination or the postposition used at the end of the sentence.
When parsing sentences in Oracle Policy Modeling, the following parts of the sentence are underlined:
NOTES:
The following verb inflections are currently not handled.
The first three forms are unlikely to occur in the Oracle Policy Automation rulebase framework. For the fourth and fifth verb forms, Oracle Policy Automation has an existing framework for expressing conditionals and clausal relationships when developing a rulebase. As such, these verb inflections are redundant. For expressing 'must', the sentence should be rephrased, for example using the verb 'obligated'.
The parser only supports limited number of counter words such as those for age and number of people.
For example, look at the following sentences.
The person is eligible if the person pays tax.
이 사람은 세금을 납부한다면 자격이 있다
In Oracle Policy Modeling this should be written as two separate sentences where the first one is formatted as the conclusion and the second one as the level 1 condition.
The person is eligible.
이 사람은 자격이 있다
The person pays tax.
이 사람은 세금을 납부한다
The person is retired and the person’s age is greater than 65.
이 사람은 은퇴하였으며 나이는 65세 이상이다
The above sentence should be broken down into two separate discrete sentences.
The person is retired and
이 사람은 은퇴하였다 그리고
The person’s age is greater than 65
이 사람의 나이는 65세 이상이다
Here the sentences represent two conditions that need to occur simultaneously. This will be reflected by the 'and' (그리고) rather than inflecting the verb. Thus, if there are sentences where verb forms that are not covered by the verb editor are used, one should try to rewrite them as separate attributes especially when the sentences are clausal in nature.
This refers to creating question, uncertain and unknown forms for sentences .
If you want to use "what" or "who" in the question form for variables (eg 'the person’s age'), it is really hard to infer the correct form of the words to use since they depend on the semantics of the sentence in question.
As a rough rule if the last word (in this case '나이') ends without a tail consonant, just add "는" after the word. So in this case, the sentence changes to "그 사람의 나이는?".
However, if the last word ends with a tail consonant (for example '생일' or '수입'), you need to add "은" after the word. For instance, 'the person's birthday' is '그 사람의 생일은?' and 'the person's income' is '그 사람의 수입은?'.