JPH04188276A - Conversion main leading-tye machine translation system - Google Patents
Conversion main leading-tye machine translation systemInfo
- Publication number
- JPH04188276A JPH04188276A JP2318672A JP31867290A JPH04188276A JP H04188276 A JPH04188276 A JP H04188276A JP 2318672 A JP2318672 A JP 2318672A JP 31867290 A JP31867290 A JP 31867290A JP H04188276 A JPH04188276 A JP H04188276A
- Authority
- JP
- Japan
- Prior art keywords
- conversion
- translation
- knowledge
- machine translation
- bilingual
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Machine Translation (AREA)
Abstract
Description
【発明の詳細な説明】
[産業上の利用分野]
この発明は機械翻訳方式に関し、特に、解析処理などの
各処理単位が要求駆動で変換処理から起動されるような
変換主導型機械翻訳方式に関する。[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a machine translation method, and particularly to a conversion-driven machine translation method in which each processing unit such as an analysis process is started from a conversion process based on a request. .
[従来の技術および発明が解決しようとする課題]電子
計算機による機械翻訳のシステムは、ますますその需要
か高まり、研究開発が盛んに行なわれている。特に、最
近では、技術文書たけでなく、医者と患者との会話や目
的指向の電話会話の翻訳など、多様な分野への適用も行
なわれるようになった◇
第4図は従来の解析主導型機械翻訳方式と称される電子
計算機による機械翻訳の手順を説明するためのフロー図
である。第4図において、従来の機械翻訳システムは、
入力部から入力された原言語文が解析部で解析処理され
、変換部で変換処理され、生成部で生成処理されて目的
言語文に翻訳される。このうち、解析処理がシステム全
体の能力(速度、訳の質)を決定していた。[Problems to be solved by conventional techniques and inventions] The demand for machine translation systems using electronic computers is increasing, and research and development are being actively conducted. In particular, in recent years, it has been applied not only to technical documents but also to a variety of fields, such as the translation of conversations between doctors and patients and goal-oriented telephone conversations. It is a flow diagram for explaining the procedure of machine translation by an electronic computer, which is called a machine translation method. In Figure 4, the conventional machine translation system is
A source language sentence input from the input unit is analyzed by the analysis unit, converted by the conversion unit, and generated by the generation unit to be translated into a target language sentence. Among these, the analysis processing determined the overall system capability (speed, quality of translation).
しかしながら、解析主導型機械翻訳方式では、解析処理
が翻訳と独立に設計され、翻訳に必要十分な情報を得ら
れないことが多い。換言すれば、翻訳のために不十分な
情報しか得られず、翻訳の質が悪くなったり、翻訳に不
必要な処理が常時行なわれ、解析処理の速度は遅くなる
ことによって翻訳全体の速度が極端に遅くなったりして
いた。However, in analysis-driven machine translation methods, the analysis process is designed independently from the translation, and it is often not possible to obtain sufficient information for the translation. In other words, insufficient information is obtained for translation, resulting in poor translation quality, and processing unnecessary for translation is constantly performed, slowing down the analysis process and slowing down the overall translation speed. It was becoming extremely slow.
しかも、解析主導型機械翻訳方式では、解析か単語から
構成的に行なわれ、換ざすれば中詰の品詞1品詞および
文法範鴎から新たに文法範鴫を構成する規則の適用の履
歴としての本構造を変換の基礎にするために、大局的な
対応を実現するのが困難であり、特に日本語と英語のよ
うに言葉の成立、したかって、語堂1文法が著しく違う
場合に良質の翻訳は得られなかった。Furthermore, in the analysis-driven machine translation method, analysis is performed constructively from words, in other words, it is performed as a history of the application of rules to newly construct a grammar category from a middle part of speech and a grammar category. Since this structure is used as the basis for conversion, it is difficult to achieve global correspondence, and it is especially difficult to achieve high-quality correspondence when the formation of words and, therefore, the grammar of Word Hall 1 are significantly different, such as in Japanese and English. No translation was available.
それゆえに、この発明の主たる目的は、翻訳速度および
翻訳の質の向上を実現できるような変換主導型機械翻訳
方式を提供することである。Therefore, the main objective of the present invention is to provide a conversion-driven machine translation method that can improve translation speed and translation quality.
[課題を解決するための手段]
入力された原言語文を目的言語文に翻訳する機械翻訳方
式であって、解析処理などの各処理単位が要求駆動で変
換処理から起動されるように構成したものである。[Means for solving the problem] A machine translation method that translates an input source language sentence into a target language sentence, and is configured so that each processing unit such as analysis processing is started from the conversion processing on a request-driven basis. It is something.
[作用]
この発明に係る変換主導型機械翻訳方式は、解析を中心
とすることなく、要求駆動で変換処理から起動すること
により、翻訳速度、翻訳の質の向上を実現する。[Operation] The conversion-driven machine translation method according to the present invention realizes improvements in translation speed and quality by starting from the conversion process on a request-driven basis without focusing on analysis.
[発明の実施例] 第1図はこの発明の一実施例の概略ブロック図である。[Embodiments of the invention] FIG. 1 is a schematic block diagram of an embodiment of the present invention.
第1図を参照して、入力部1はキーボードや文学誌1装
置や音声認1装置などからなり、原言語文が入力されて
同時に単語ごとに分割され、品詞などの情報が辞箸に従
って付与される。入力された原言語文は変換機構2に与
えられる。変換機構2は原言語の単語列または要求駆動
で作られた原言語の解析結果の部分(全体も含む)に関
連した対訳変換知識を対訳変換知識データベース7から
検索部3によって検索し、その関連した対訳変換知識と
入力(原言語の単語列または解析結果)とを類似度計算
部4によって照合し、最も近い対訳変換知識に従って、
部分的な翻訳をつくり出す。Referring to Figure 1, the input unit 1 consists of a keyboard, a literary journal device, a voice recognition device, etc., into which a source language sentence is input and simultaneously divided into words, and information such as parts of speech is added according to the dictionary. be done. The input source language sentence is given to the conversion mechanism 2. The conversion mechanism 2 uses a search unit 3 to search the bilingual conversion knowledge database 7 for bilingual conversion knowledge related to a word string in the source language or a part (including the whole) of the analysis result of the source language created based on a request, and searches for the related bilingual conversion knowledge from the bilingual conversion knowledge database 7. The similarity calculation unit 4 compares the bilingual conversion knowledge obtained with the input (source language word string or analysis result), and calculates according to the closest bilingual conversion knowledge.
Create a partial translation.
このことを再帰的に繰り返す。Repeat this recursively.
一方、解析部5は上述の変換過程で負荷の低い対訳変換
知識のみでは翻訳できない場合にのみ起動される。この
点が従来の解析主導機械翻訳との明確な相違である。変
換機構2の出力は出力部6に与えられ、活用変化1語尾
変化などが行なわれ、表示装置、印−1装置、音声合成
装置などによって、目的言語文が出力される。On the other hand, the analysis unit 5 is activated only in the above-mentioned conversion process when translation cannot be performed using only bilingual conversion knowledge with a low load. This point is a clear difference from conventional analysis-driven machine translation. The output of the conversion mechanism 2 is applied to an output section 6, where conjugation 1, endings, etc. are performed, and a target language sentence is outputted by a display device, a sign-1 device, a speech synthesis device, etc.
第2図は従来の変換方式とこの発明の一実施例による複
数の観点から整理された知識を使った変換方式との相違
を説明するための図である。FIG. 2 is a diagram for explaining the difference between a conventional conversion method and a conversion method using knowledge organized from a plurality of viewpoints according to an embodiment of the present invention.
まず、簡単な例で基本的な翻訳過程を説明する。First, let's explain the basic translation process with a simple example.
(IJ) 日本の首部は東京です。(IJ) The head of Japan is Tokyo.
(I EI Capital of’ Japan i
s Tokyo。(I EI Capital of' Japan i
s Tokyo.
(IJ)から(IE)の翻訳を得るために、検索部3は
対訳変換知識データベース7から次の対訳変換知識を検
索する。In order to obtain the translation from (IJ) to (IE), the search unit 3 searches the bilingual conversion knowledge database 7 for the next bilingual conversion knowledge.
A は B です→A is B Cの D−−D of C 日本呻Japan 首部呻CapHal 東京−Toky。A is B → A is B D--D of C Japan Moan Japan Neck groan CapHal Tokyo-Tokyo.
この検索は再帰的に、入力と対訳変換知識データベース
7の対訳変換知識との最大照合(最も大きな範囲をカバ
ーする照合)によって検索部3か行なう。変換は検索と
同時に、トップダウンに行なわれる。この例文の場合
A is B、D of Cis B。This search is recursively performed by the search unit 3 by maximally matching the input with the bilingual translation knowledge in the bilingual translation knowledge database 7 (matching that covers the largest range). Conversion occurs top-down, simultaneously with the search. In this example sentence, A is B, D of Cis B.
Capital of Japan is Tokyo
となる。Capital of Japan is Tokyo
becomes.
次に、高度な例で基本的な翻訳過程について説明する。Next, we will explain the basic translation process with an advanced example.
前述の(IJ)と同じ文型の次の例文(2J)は従来方
式である単語から要素構成原理によって翻訳する解析主
導型機械翻訳方式では翻訳が困難であるが、この発明の
一実施例では簡単に良質の翻訳+2E)が得られること
を示す。The following example sentence (2J), which has the same sentence pattern as the above-mentioned (IJ), is difficult to translate using the conventional analysis-driven machine translation method, which translates words using the principle of element composition, but can be easily translated using an embodiment of the present invention. This shows that a high quality translation +2E) can be obtained.
(2J)料金は現金です。(2J) Fees are in cash.
(2EI Payment 5hould be ma
de by cash。(2EI Payment 5hold be ma
de by cash.
A は B です→A is B
料金→charge
現金→cash
以上の3つの対訳変換知識から得られる英文でchar
ge Is cashは誤った翻訳である。この例文は
、単語から要素構成原理によって翻訳する解析主導型機
械翻訳でも誤ってしまう。しかし、この発明の一実施例
では、適切な対訳変換知識と類似度界によって良質の翻
訳か得られる。日本語文(2J)の場合、後述の対訳変
換知識例で示される対訳変換知識を用いて、英語文(2
E)か得られる。A is B→A is B Charge→charge Cash→cash Charge in English obtained from the above three bilingual conversion knowledge
ge Is cash is an incorrect translation. This example sentence would also be incorrect even in analysis-driven machine translation, which translates from words using the principle of element composition. However, in one embodiment of the present invention, high-quality translation can be obtained using appropriate bilingual conversion knowledge and similarity circles. In the case of a Japanese sentence (2J), use the bilingual conversion knowledge shown in the bilingual conversion knowledge example below to convert the English sentence (2J) into an English sentence (2J).
E) can be obtained.
次に、対訳変換知慮について説明する。対訳変換加工の
一般形は原言語表現SEに対応する目的言語表現TEと
その対応か成立するときの条件Cからなり、この発明の
一実施例では以下の表記法を用いる。Next, bilingual conversion consideration will be explained. The general form of bilingual translation processing consists of a target language expression TE corresponding to a source language expression SE and a condition C for when the correspondence holds, and in one embodiment of the present invention, the following notation is used.
5E−TE (C)
入力文の条件C1nputの下で原言語表現SEに対応
する目的言語表現を求めるとき、次の2つの場合がある
。5E-TE (C) When finding the target language expression corresponding to the source language expression SE under the condition C1nput of the input sentence, there are the following two cases.
■ 原言語表現SEに対応する目的言語表現が1通りの
場合、すぐTEが出力される。■ If there is only one target language expression corresponding to the source language expression SE, TE is immediately output.
■ 同じ原言語表現SEに対応する目的言語表現が異な
る2つ、置、TE2とあり、それぞれの条件がC1,C
2であったとする。このとき、C1nputとC1,C
2の類似度計算によって訳の選択が行なわれる。条件C
1nput、C1゜C2は文脈(より広い範囲の表現)
に関することもある。■ There are two different target language expressions corresponding to the same source language expression SE, position and TE2, and the respective conditions are C1 and C.
Suppose it was 2. At this time, C1nput and C1,C
Translations are selected by the similarity calculation in step 2. Condition C
1nput, C1゜C2 is context (wider range of expression)
It may also be related to.
次に、要求駆動について説明する。対訳変換知識は複数
の観点から構成されている。この実施例では、簡単のた
め、単語列1文型、依存構造の3つの観点を取上げる。Next, request driving will be explained. Bilingual translation knowledge is composed of multiple perspectives. In this embodiment, for the sake of simplicity, we will take up three viewpoints: a single word string sentence type and a dependency structure.
左から右に順に、深くなる(負荷が多(なる)。変換主
導型機械翻訳は、第2図(b)に示すように浅い知識が
優先される。From left to right, the depth increases (the load increases). In conversion-driven machine translation, shallow knowledge is given priority as shown in Figure 2 (b).
第2図(a)・に示すような従来の解析主導型機械翻訳
とは異なり、浅い観点で適合する知識がない場合のみ解
析部5が起動される。Unlike conventional analysis-driven machine translation as shown in FIG. 2(a), the analysis unit 5 is activated only when there is no relevant knowledge from a shallow perspective.
次に、対訳変換知識例について説明する。Next, an example of bilingual conversion knowledge will be explained.
■ 目的言語表現TEが1通りの場合、すぐTEが出力
される。■ If there is only one target language expression TE, the TE is output immediately.
ありがとう ございます→thank Youわからな
い点→a queuion
■ 目的言語表現TEが複数の場合
(1)TEが原言語の類似度(CinputがC1、C
2のいずれに類似しているか)によって選択される。
′
AはBです一+Paymenl 5hould be
made by B。Thank you→thank You What I don't understand→a question ■ When there are multiple target language expressions TE (1) TE is the similarity of the source language (C1, C
2).
' A is B + Paymenl 5hold be
Made by B.
((A、B)〜(料金、現金))
AはBです→Th1s 1SB
((A、B)〜(こちら、事務局))
AはBです−A is B
((A、B)〜(名前、〈固有名詞〉)、・・・)(2
)TEか文脈によって選択される。((A, B) ~ (Fee, cash)) A is B → Th1s 1SB ((A, B) ~ (Here, the secretariat)) A is B - A is B ((A, B) ~ ( name, <proper noun>),...)(2
) selected by TE or context.
はい→YeS (前文がYN疑問文) はい→hell。Yes → Yes (The preamble is a YN question) Yes → hell.
(入力が最初の発話)
はい−”yes、 plea、se
(前文が「お聞きしたいことがあるのですが」と類似)
検索部3は索引によって高速化される。以下に例示する
ような原言語表現SEそのものまたはその要素の単語の
集合など種々のものが索引の見aしとして有効である。(The input is the first utterance) Yes-"yes, please, se (The preamble is similar to "I have something to ask you.") The search unit 3 is speeded up by the index. Various expressions such as the source language expression SE itself or a set of words of its elements as exemplified below are effective as the index a.
■ 単語列 もしもし、ありがとう ございます、わからない 点 この索引は特に定型表現に適している。■ Word string Hello, thank you, I don't understand. This index is particularly suited to fixed expressions.
■ 文型(機能語列)
〜は〜です、〜たい の です が
基本的な入力の多くがこの観点で処理されるので、この
索引は特に重要である。■ Sentence pattern (function word sequence) ~ is ~, ~ is ~, but since most of the basic input is processed from this perspective, this index is particularly important.
■ 構造(部分依存構造)
(名詞 (連体形の動詞・・・))
この索引は連体形埋込み文の翻訳などのように、−船釣
な対訳変換知識規則のために有効である。■ Structure (partial dependence structure) (noun (verb in adnominal form...)) This index is effective for bilingual translation knowledge rules that are difficult to understand, such as the translation of embedded sentences in adnominal form.
第3図は比較的複雑な例文の翻訳過程を模式的に示した
図である。FIG. 3 is a diagram schematically showing the translation process of a relatively complicated example sentence.
(3J)わからない点がございましたらいつでもお聞き
ください
(3E) If 7u havea questio
n please ask usat an71ime
■ 入力と最大照合する対訳変換知識を選択する。入力
と「〜たら〜ください」という原言語表現SEを持つ対
訳変換知識がステップ(1)で最大照合し、その対訳変
換知識の目的言語表現TErlf−please〜Jに
従って、ステップ(1′)で部分的に翻訳される。(3J) If you have any questions, please feel free to ask. (3E) If 7u have a question
n please ask usat an71ime ■ Select the bilingual conversion knowledge that matches the input at most. In step (1), the input and the bilingual conversion knowledge with the source language expression SE of "~tara~ please" are maximally matched, and in step (1'), according to the target language expression Terlf-please~J of the bilingual conversion knowledge, the bilingual conversion knowledge is partially compared. translated into words.
単語列1文型など浅い観点の知識で適合するものかない
場合は、解析部5が起動され、その結果と構想の観点の
知識とを照合する。たとえば「分から ない 点」は浅
い対訳変換知識に照合するので、解析部5は起動されな
いが、「分から ない ない 本」は浅い対訳変換知識
に照合しないので解析部5が起動され、連体形埋込み文
のための構造の観点での対訳変換知識によって翻訳され
る。If there is no matching knowledge from a shallow perspective, such as a single word string sentence pattern, the analysis unit 5 is activated and the result is compared with the knowledge from the concept perspective. For example, ``Points I don't understand'' is checked against shallow bilingual conversion knowledge, so the analysis unit 5 is not activated, but ``I don't understand books'' is not checked against shallow bilingual conversion knowledge, so the analysis unit 5 is activated, and translated by bilingual translation knowledge in terms of structure.
「〜」に適合した入力の部分に対して再帰的に同じ手続
が適用され、必要な場合は文脈が参照される。The same procedure is applied recursively to the portion of the input that matches "~", with context referenced if necessary.
[発明の効果コ
以上のように、この発明によれば、解析処理を中心とす
ることなく、要求駆動で変換処理から解析処理を起動す
ることにより、翻訳速度および翻訳の質の向上を実現で
きる。より具体的には、単語から要素構成原理によって
翻訳するのではなく、単語列9文型、構造など大局的な
翻訳知識によって良質の翻訳を実現でき、要求駆動によ
る高速な翻訳を実現でき、システムの知識の作成、改良
が容易となる。[Effects of the Invention] As described above, according to the present invention, translation speed and quality can be improved by starting analysis processing from conversion processing driven by demand, without focusing on analysis processing. . More specifically, instead of translating words based on element composition principles, we can achieve high-quality translation by using global translation knowledge such as word strings9 sentence types and structures, and we can achieve high-speed translation driven by demand, and improve system performance. It becomes easier to create and improve knowledge.
第1図はこの発明の一実施例の概略ブロック図である。
第2図は従来の変換方式とこの発明の一実施例による複
数の観点から整理された知識を使った変換方式との相違
を説明するための図である。
第3図は比較的複雑な文を変換主導型機械翻訳方式で翻
訳する過程を模式的に示した図である。第4図は従来の
解析主導型機械翻訳方式の手順を示すフロー図である。
図において、1は入力部、2は変換機構、3は検索部、
4は類似度計算部、5は解析部、6は出力部、7は対訳
変換知識データベースを示す。
特許出願人 株式会社エイ・ティ・アール自動翻訳電話
研究所 て\
代理人弁理士深見久部−ビ ・。
(ほか2名) ”””
第2図
入力 出力 入力
出力第3図
入力文
分かう ない 点 が ございまし
たら (1つでも h聞き くたさ(−八
たら B
(ださ(−(11、、、、、、、、。11.1.16
1116.011.11.1111.。6.111..
11.1.1.、、、、、、、、、、、、□、、、、、
、、、、、、、1,1.、、、、、、.1.、□、2.
、+m++++=++++nm+++、tf :fb
Jo 111101116111101011.1
611.。
II A
please B(
1°)□、111101111.11.1011.11
100.11.110111111..0.、、、、、
、、、、、、、、、、、、、□、、、、、、、、、、、
、、、、、、、、、、、、、、、。
you have C
+2’l 、、
、、、、、、、、、、、、、、、、、、、、、、、、、
、= 10.=、−01611,==−1
,=−=−16,1a Question
(3“l
”””””””
0.”1.”1.”’+5’ )
−一一一一一出力文
u youhave aques
lion please askus
a+any+ime第4図FIG. 1 is a schematic block diagram of an embodiment of the present invention. FIG. 2 is a diagram for explaining the difference between a conventional conversion method and a conversion method using knowledge organized from a plurality of viewpoints according to an embodiment of the present invention. FIG. 3 is a diagram schematically showing the process of translating a relatively complex sentence using the conversion-driven machine translation method. FIG. 4 is a flow diagram showing the procedure of a conventional analysis-driven machine translation method. In the figure, 1 is an input section, 2 is a conversion mechanism, 3 is a search section,
Reference numeral 4 indicates a similarity calculation section, 5 an analysis section, 6 an output section, and 7 a bilingual conversion knowledge database. Patent applicant: A.T.R. Automatic Translation Telephone Research Institute Co., Ltd. Patent attorney: Kube Fukami. (2 others) “”” Figure 2 Input Output Input
Output Figure 3: There are some points where the input sentence cannot be understood.
Tara (I'm tired of listening to even one thing (-8)
Tara B
(dasa(-(11,,,,,,,.11.1.16
1116.011.11.1111. . 6.111. ..
11.1.1. ,,,,,,,,,,,,□,,,,,
,,,,,,1,1. ,,,,,,. 1. , □, 2.
, +m++++=++++nm+++, tf :fb
Jo 111101116111101011.1
611. . II A
Please B(
1°)□, 111101111.11.1011.11
100.11.110111111. .. 0. ,,,,,,
,,,,,,,,,,,,,□,,,,,,,,,,,,
,,,,,,,,,,,,,,. you have C +2'l,,
,,,,,,,,,,,,,,,,,,,,,,,,,,,
,=10. =, -01611, ==-1
,=-=-16,1a Question (3"l
”””””””
0. "1."1. ”'+5')
-11111 output sentence u youhave aques
lion please ask
a+any+imeFigure 4
Claims (7)
て、 解析処理などの各処理単位が要求駆動で変換処理から起
動されることを特徴とする、変換主導型機械翻訳方式。(1) A conversion-driven machine translation method that translates an input language sentence, and is characterized in that each processing unit such as analysis processing is started from the conversion processing based on a request.
語の対訳変換知識から決定されることを特徴とする、請
求項第1項記載の変換主導型機械翻訳方式。(2) The conversion-driven machine translation method according to claim 1, wherein the request is determined from bilingual conversion knowledge of the source language and the target language built into the conversion process.
ることを特徴とする、請求項第2項記載の変換主導型機
械翻訳方式。(3) The conversion-driven machine translation method according to claim 2, wherein the bilingual conversion knowledge is constructed from a plurality of viewpoints.
る解析処理が存在することを特徴とする、請求項第3項
記載の変換主導型機械翻訳方式。(4) The conversion-driven machine translation method according to claim 3, characterized in that there is an analysis process for mapping input to a representation of the viewpoint for each of the viewpoints.
観点ごとに対訳変換知識が構造化されていることを特徴
とする、請求項第4項記載の変換主導型機械翻訳方式。(5) The conversion-driven machine translation method according to claim 4, wherein the bilingual conversion knowledge is structured for each viewpoint from the one with the least load to the one with the most load of each analysis process.
荷の少ない方から順に使用され、観点nで入力と照合す
る知識がない場合のみ、観点(n+1)の表現に写像す
る解析処理を起動して、負荷の多い観点(n+1)の対
訳変換知識を使用することを特徴とする、請求項第5項
記載の変換主導型機械翻訳方式。(6) The bilingual conversion knowledge structured for each viewpoint is used in order of load, and only when there is no knowledge to match the input at viewpoint n, an analysis process is performed to map it to the expression of viewpoint (n+1). 6. The conversion-driven machine translation method according to claim 5, wherein the conversion-driven machine translation method is activated and uses the bilingual conversion knowledge of viewpoints (n+1) with a large load.
ど大局的な対応を含み、同時に多くの訳語選択のための
制約を満足することを特徴とする、請求項第2項記載の
変換主導型機械翻訳方式。(7) The bilingual conversion knowledge includes global correspondence such as word strings, phrases, sentence patterns, and structures, and simultaneously satisfies constraints for selecting many translation words. Conversion-driven machine translation method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2318672A JP2769919B2 (en) | 1990-11-21 | 1990-11-21 | Conversion-driven machine translation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2318672A JP2769919B2 (en) | 1990-11-21 | 1990-11-21 | Conversion-driven machine translation |
Publications (2)
Publication Number | Publication Date |
---|---|
JPH04188276A true JPH04188276A (en) | 1992-07-06 |
JP2769919B2 JP2769919B2 (en) | 1998-06-25 |
Family
ID=18101744
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
JP2318672A Expired - Fee Related JP2769919B2 (en) | 1990-11-21 | 1990-11-21 | Conversion-driven machine translation |
Country Status (1)
Country | Link |
---|---|
JP (1) | JP2769919B2 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6161083A (en) * | 1996-05-02 | 2000-12-12 | Sony Corporation | Example-based translation method and system which calculates word similarity degrees, a priori probability, and transformation probability to determine the best example for translation |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62163177A (en) * | 1986-01-14 | 1987-07-18 | Toshiba Corp | Mechanical translating device |
JPS6386071A (en) * | 1986-09-30 | 1988-04-16 | Nippon Telegr & Teleph Corp <Ntt> | Translation system for natural language |
-
1990
- 1990-11-21 JP JP2318672A patent/JP2769919B2/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62163177A (en) * | 1986-01-14 | 1987-07-18 | Toshiba Corp | Mechanical translating device |
JPS6386071A (en) * | 1986-09-30 | 1988-04-16 | Nippon Telegr & Teleph Corp <Ntt> | Translation system for natural language |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6161083A (en) * | 1996-05-02 | 2000-12-12 | Sony Corporation | Example-based translation method and system which calculates word similarity degrees, a priori probability, and transformation probability to determine the best example for translation |
Also Published As
Publication number | Publication date |
---|---|
JP2769919B2 (en) | 1998-06-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6278968B1 (en) | Method and apparatus for adaptive speech recognition hypothesis construction and selection in a spoken language translation system | |
US6243669B1 (en) | Method and apparatus for providing syntactic analysis and data structure for translation knowledge in example-based language translation | |
US6282507B1 (en) | Method and apparatus for interactive source language expression recognition and alternative hypothesis presentation and selection | |
US6356865B1 (en) | Method and apparatus for performing spoken language translation | |
US6266642B1 (en) | Method and portable apparatus for performing spoken language translation | |
US6721697B1 (en) | Method and system for reducing lexical ambiguity | |
JP4050755B2 (en) | Communication support device, communication support method, and communication support program | |
US5490061A (en) | Improved translation system utilizing a morphological stripping process to reduce words to their root configuration to produce reduction of database size | |
US7873508B2 (en) | Apparatus, method, and computer program product for supporting communication through translation between languages | |
JP2000353161A (en) | Method and device for controlling style in generation of natural language | |
JPH01501977A (en) | language translation system | |
Bonneau-Maynard et al. | Results of the French Evalda-Media evaluation campaign for literal understanding | |
JP5002271B2 (en) | Apparatus, method, and program for machine translation of input source language sentence into target language | |
Sen et al. | Bangla natural language processing: A comprehensive review of classical machine learning and deep learning based methods | |
Oflazer et al. | Turkish and its challenges for language and speech processing | |
WO2007105615A1 (en) | Request content identification system, request content identification method using natural language, and program | |
JPH04188276A (en) | Conversion main leading-tye machine translation system | |
Ryu et al. | Transformer‐based reranking for improving Korean morphological analysis systems | |
Alam et al. | A finite-state morphological analyzer for Saraiki | |
Dua et al. | An amalgamation of integrated features with DeepSpeech2 architecture and improved spell corrector for improving Gujarati language ASR system | |
Lin et al. | A mandarin to taiwanese min nan machine translation system with speech synthesis of taiwanese min nan | |
Švec et al. | Transfer learning of transformers for spoken language understanding | |
Gebremeskel | Ge’ez POS Tagger Using Hybrid Approach | |
JP2902343B2 (en) | Language analysis system and method | |
Yomie Nzeuhang et al. | Application of the Multilingual Acoustic Representation Model XLSR-53 for the Transcription of Ewondo |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
LAPS | Cancellation because of no payment of annual fees |