[Paper Review] Possessive Pronouns as Determiners in Japanese-to-English Machine Translation

Francis Bond, Kentaro Ogura+1 more|ArXiv.org|Jan 23, 1996

Natural Language Processing Techniques8 references9 citations

TL;DR

This paper proposes a heuristic method for generating possessive pronouns as determiners in Japanese-to-English machine translation, where Japanese lacks overt possessive markers. By combining lexical properties of nouns, noun phrase referentiality, and syntactic context (subject and main verb), the method increases correct possessive pronoun usage by 346 instances (from 263 to 609) in a 6,200-sentence test set, with a modest 83 errors introduced.

ABSTRACT

Possessive pronouns are used as determiners in English when no equivalent would be used in a Japanese sentence with the same meaning. This paper proposes a heuristic method of generating such possessive pronouns even when there is no equivalent in the Japanese. The method uses information about the use of possessive pronouns in English treated as a lexical property of nouns, in addition to contextual information about noun phrase referentiality and the subject and main verb of the sentence that the noun phrase appears in. The proposed method has been implemented in NTT Communication Science Laboratories' Japanese-to-English machine translation system ALT-J/E. In a test set of 6,200 sentences, the proposed method increased the number of noun phrases with appropriate possessive pronouns generated, by 263 to 609, at the cost of generating 83 noun phrases with inappropriate possessive pronouns.

Motivation & Objective

• To address the challenge of generating English possessive pronouns as determiners when no equivalent exists in Japanese.
• To improve the accuracy of possessive pronoun generation in NTT's ALT-J/E Japanese-to-English machine translation system.
• To model the use of possessive pronouns as a lexical property of nouns, informed by syntactic and referential context.
• To reduce the gap between Japanese and English syntactic structures in possessive constructions.

Proposed method

• The method uses a lexical property database associating nouns with their typical possessive pronoun usage in English.
• It analyzes noun phrase referentiality (definite/indefinite) to guide possessive pronoun selection.
• It incorporates syntactic context, including the subject and main verb of the clause, to infer appropriate possessive marking.
• A heuristic rule set applies these features to generate possessive pronouns even when absent in the source Japanese sentence.
• The system is implemented in NTT's ALT-J/E machine translation pipeline for end-to-end evaluation.

Experimental results

Research questions

• RQ1 How can possessive pronouns be generated in English translation when Japanese does not express them?
• RQ2 What linguistic features predict the appropriate use of possessive pronouns in English noun phrases?
• RQ3 Can lexical properties of nouns combined with syntactic context improve possessive pronoun generation in MT?
• RQ4 What is the trade-off between increased correct usage and incorrect generation of possessive pronouns?

Key findings

• The method increased the number of correctly generated possessive pronouns from 263 to 609 in a 6,200-sentence test set.
• The improvement represents a net gain of 346 correctly generated possessive pronouns.
• The system introduced 83 instances of incorrect possessive pronoun usage, indicating a controlled error rate.
• The combination of lexical noun properties and syntactic context significantly enhanced generation accuracy.