`Fukumochi et al.
`
`I 111111111111111111111 lllll lllll 1111111111111111111111111111111111111111111
`US005289375A
`[11] Patent Number:
`[45] Date of Patent:
`
`5,289,375
`Feb. 22, 1994
`
`[75]
`
`[54] TRANSLATION MACHINE
`Inventors: Yoji Fukumochi, lkoma; Hitoshi
`Suzuki, Nara; Shuzo Kugimiya, Nara;
`Ichiko Sata, Nara; Tokuyuki Hirai,
`Yamato-Koriyama, all of Japan
`[73] Assignee: Sharp Kabushiki Kaisha, Osaka,
`Japan
`[21] Appl. No.: 641,738
`Jan. 15, 1991
`[22] Filed:
`Foreign Application Priority Data
`[30]
`Jan. 22, 1990 [JP]
`Japan ...................................... 1-2146
`Jan. 22, 1990 [JP]
`Japan ...................................... 1-2147
`Jan. 26, 1990 [JP]
`Japan ...................................... 1-7116
`Int. Cl.5 •••••••••••••••••••••••• G06F 15/38; G06G 7/60
`[51]
`[52] U.S. Cl ................................................. 364/419.02
`[58] Field of Search .............................. 364/419, 226.4
`References Cited
`U.S. PATENT DOCUMENTS
`4,661,924 4/1987 Okamoto et al. ................... 364/900
`4,706,212 11/1987 Toma .................................. 364/900
`4,791,587 12/1988 Doi ...................................... 364/900
`4,814,988 3/1989 Shiotani et al. ..................... 364/419
`4,821,230 4/1989 Kumano et al. .................... 364/900
`4,887,212 12/1989 Zamora et al. ..................... 364/419
`4,931,936 6/1990 Kugimiya et al. .................. 364/419
`4,953,088 8/1990 Suzuki et al. ....................... 364/419
`4,962,452 10/1990 Nogami et al. ..................... 364/419
`
`(56]
`
`FOREIGN PATENT DOCUMENTS
`0266001 of 0000 European Pat. Off ..
`0274281 of 0000 European Pat. Off ..
`
`MAIN
`CPU
`
`MAIN
`MEMORY
`
`1
`
`2
`
`CRT
`
`3
`
`0403057 12/1990 European Pat. Off ..
`57-69362 4/1982 Japan .
`61-18074 1/1986 Japan .
`61-74069 4/1986 Japan .
`62-271057 11/1987 Japan .
`63-221475 9/1988 Japan .
`1-166258 6/1989 Japan.
`1-233560 6/1989 Japan .
`2211639 of 0000 United Kingdom .
`2211640 ofOOOO United Kingdom.
`
`Primary Examiner-Donald E. McElheny, Jr.
`Assistant Examiner-Laura Brotman
`Attorney, A.gent, or Firm-Nixon & Vanderhye
`
`ABSTRACT
`[57]
`A translation machine having a storing unit for storing
`information of words and grammer rules, the translation
`machine being capable of translating sentences de(cid:173)
`scribed in a source language into a target language
`includes a unit for dividing input sentence described in
`the source language into morphologic elements and for
`generating a mark-inserted sentence by adding a mark
`which represents an information of the detected mor(cid:173)
`phologic elements, a unit connected to the dividing unit
`for analyzing a syntax of the morphologic elements in
`accordance with the information stored in the storing
`unit, and a unit connected to the analyzing unit for
`converting a structure of the syntax obtained in the
`dividing unit into a structure of syntax in the target
`language so that translated sentence is generated on a
`basis of the syntax structure in the target language ob(cid:173)
`tained by the converting unit.
`
`9 Claims, 12 Drawing Sheets
`
`RANSLATION
`MODULE
`
`4
`
`5
`
`CXCT~RES, ~TICAL
`RULES Al-0 TREE-smu:TUFE
`CONVERTING RULES
`
`6
`
`Page 1 of 25
`
`GOOGLE EXHIBIT 1037
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 1 of 12
`
`5,289,375
`
`·-
`
`-
`
`.
`0\
`........
`LL.
`
`i.,-1.0
`
`(D
`
`ii~ ~-
`
`ur~Zf
`~t-~
`gifi U)W
`
`-w~
`t- ...J
`U::,o
`cSa::u
`
`z
`
`0 -t-
`
`<( UJ
`...J
`...J
`(/') ::>
`Zo
`~~
`,
`
`0 a: <! . -~
`' CD
`
`0
`~
`:X::
`
`I
`
`,......M
`
`I-
`a:
`u
`
`--N
`
`..... _
`
`f:c
`zO
`:::E
`-
`<( w
`:::E
`:::E
`
`z :::,
`:i Cl.
`~ u
`
`Page 2 of 25
`
`
`
`OI
`.....,l
`~
`....
`\0
`00
`~
`....
`OI
`
`~
`~
`
`~ -~
`
`rJl =(cid:173)~
`
`0 ....
`
`~
`~
`~
`
`i
`
`~
`
`~
`
`tt> = f"'t,-
`~
`lot,
`•
`rJJ.
`0 •
`
`LANGUAGE
`TARGET
`
`SECTION
`GENERATING
`SENTENCE
`TRANSLATED
`
`SECTION
`CONVERTING
`
`SECTION
`ANALYING
`SYNTAX
`
`A NALVZING SECTION
`MCJ?PHOLOGIC ELEMENT
`01 CTIO\IARV-CCN5l.l.TING
`
`LANGUAGE
`SO.JRCE
`
`14
`
`13
`
`12
`
`11
`
`Fig.2
`
`Page 3 of 25
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 3 of 12
`
`5,289,375
`
`Fig.3
`
`ST A RT
`
`S1
`
`SET SEN1ENCE - NJMBE
`FLAG TO 1
`
`S5
`INCREMENT
`SENTENCE-NLMBER
`FLAG
`
`52
`
`SIDRE SUBJECT
`SENTENCE IN
`BUFFER 8
`5 3
`
`ADD
`AM BIGUITY-INDICA
`MARK
`
`YES
`
`NO
`
`END
`
`Page 4 of 25
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 4 of 12
`
`5,289,375
`
`START
`
`.---_ ____._ __ _
`
`SET
`WORD- POSITION
`FLAG TO 0
`
`CONSULT
`DICTIONARIES
`
`ANALYZE
`MORPHOLOGIC
`ELEMENTS
`
`Fig.4
`
`511
`
`512
`
`513
`
`YES
`
`515
`
`517
`
`I~REMENT
`OORD-FOSffiCJ.J FLA
`
`NO
`
`STORE AMBIGUITY
`INFORMATION IN
`BUFFER D
`
`516
`
`YES
`
`518
`NSERT AMBIGUITY-INDICATING
`MARKTOINPUTSENlENCE
`BA:EDONCO'-JIDJTCF BUFFER
`
`END
`
`Page 5 of 25
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 5 of 12
`
`5,289,375
`
`Fig.5
`
`START
`
`SET SENTEt-CE- N...t-1BER 5 2 1
`FLAG TO 1
`
`.----------------
`
`STORE SABJECT
`SENTENCE IN
`BUFFER B
`
`522
`
`528
`,----------------
`INCREMENT
`SENTENCE-NUMBER
`FLAG
`
`523
`
`S24
`
`525
`
`SELECT PART
`CF SPEECH
`------L---------26
`REPLACE AMSIGUilY-INDICA Tlt-.G
`MARK WITH SYMBOL FOR
`SPECIFYJNG PART CF SPEB:H
`
`END
`
`Page 6 of 25
`
`
`
`01
`-....l
`CH
`...
`\0
`QC
`~
`...
`01
`
`ft) -°' 0 """ 1--l
`
`ft)
`1:1"'
`rJl
`
`N
`
`~
`
`~ .J:,,
`
`1--l
`~N
`N
`
`'"'d a tt) = f""t,-
`
`•
`rJJ
`L! •
`
`NOUN
`VERB{ADJECTIVE USE)/ ADJECTIVE
`VERB{GERUND)
`VERB{PRESENT PARTICIPLE)
`VERB
`(manufacturing)
`CANDIDATE
`PART-OF-SPEECH
`
`(JAPANESE)
`
`*17Q manufacturing strategies
`the irrplemementation of
`
`The importance of strategies.
`
`(ENGLISH)
`
`Fig. 6
`
`5
`
`4
`
`3
`
`2
`
`1
`
`Page 7 of 25
`
`
`
`U.S. Patent
`sU
`tmt
`
`Feb. 22, 1994
`
`u.m7w«m
`Sheet 7 of 12
`
`5,289,375
`5,289,375
`
`9.n,.mflmmfibm*0mocflbag9.:mfimmzéivAfmzwzwv
`
`M,BcozflcmEmfigm5
`
`
`VI
`0..,
`Ol
`0..,
`...
`0
`C ~
`.Q 1n
`
`-
`
`acpkaE
`
`t:J\
`. ..._
`
`Page 8 0f 25
`
`_J
`
`~ z
`w
`~
`
`Ol
`
`0.., ro ~ ... VI -0
`
`0..,
`
`F
`-
`
`
`
`mflmmwmbmmctguflncmtac
`
`... tt, Ol
`.., C
`C ·(cid:173)
`a.,
`'-
`E ::,
`0.., u
`- .,, o.-
`-~ i
`
`.!:
`I
`.., C
`
`N
`
`...:r
`
`U"')
`
`Page 8 of 25
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 8 of 12
`
`5,289,375
`
`Fig.8
`
`START
`
`r------'----.;551
`STOOE lNFUT SENIDJCE
`IN BUFFER A
`
`SET 'MJRD-POSITION
`FLAG 10 ZERO
`
`S52
`
`NO
`
`S54
`
`YES
`SAVE WORD POSIDON
`ANJ SFECIFIED SENTEN:E
`PATTERN 1N BUFFER 8
`
`555
`
`DELETE SENTENCE-RmE
`SVMOOL FFOv1
`OJRRENTLY- FOlNIEO WORD
`~SAVE IN BUFFER C
`
`YES
`
`END
`
`S58
`
`INCREMENT
`\\ORD POSITDN
`FLAG
`
`S56
`
`SAVE
`CURRENLY-PClNT'EO
`\\ORO IN BUFFER C
`
`Page 9 of 25
`
`
`
`(.II
`-l
`CH
`....
`\0
`00
`N
`....
`(.IJ
`
`~
`1--l
`
`\C) s.
`f ft) ....
`
`~
`1--l
`J~
`?"
`~
`
`~
`
`~ = f'+-
`~
`"'C
`•
`rl.J
`•
`C:!
`
`SERU SONO SAPOTO KINOUSURU
`;,t d) 1t ,t~-+ ~-ftg, ,Q
`-tt. -?>
`function
`
`makes the suppcrt
`
`SONO SISUTEMU
`:,f_ (f)
`
`:,;-:J..jL,..
`
`system
`
`the
`
`VERB CEFJNITIVE NClJN VERB 1
`
`VERB
`I
`I
`PHRASE
`VERB
`
`~
`PHRA5E
`NOUN
`
`VERB 5
`
`I
`
`DEFINITIVE NOUN
`
`NOUN A-IRASE
`
`A
`
`VERV PHRASE
`
`SENTENCE
`
`Fig.9
`
`Page 10 of 25
`
`
`
`... w
`\0
`00
`~
`...
`UI
`
`UI
`-l
`
`"'"' N
`0 ....
`"'"' 0
`~
`00 =(cid:173)~
`
`"'"' ~
`
`~
`
`~
`
`...,N
`N
`
`('D = ~
`~ = ~
`
`•
`Cf'J.
`0 •
`
`j ;t~· -~ •it
`
`SONO SAPOTO
`~ 0)
`
`KINOU
`
`SURJ
`i~
`
`SISUTEMJ
`V"J..1./4
`
`SONO
`¥. d)
`
`function
`
`support
`
`the
`
`makes
`
`system
`
`the
`
`NOUN A PHRASE
`I
`
`t\OUN
`
`DEFINITIVE NOJN
`
`NOUN
`
`VEVB 5
`
`I
`
`VERB
`
`VERB PHRASE
`
`SENTENCE
`
`Fig. /0
`
`"
`
`DEFINITIVE NOUN
`
`NOLN PHRASE
`
`A
`
`Page 11 of 25
`
`
`
`U1
`.......
`CH
`...
`\0
`00
`~
`...
`U1
`
`.... N
`0 ....
`....
`~ -....
`00 =-~
`
`.,. ....
`~
`
`N
`N
`
`.,:..
`IC
`IC
`
`~ • r.ri
`
`•
`
`""d a ff) = "'"
`
`KINOU
`~-~E.
`
`SONO SAPOTO
`'i (fJ
`11,·-1'
`
`TSUKLRU
`fp ~
`
`SISUTEM
`97'=tA
`
`SONO
`)id")
`
`function
`
`support
`
`the
`
`makes
`
`system
`
`the
`
`~
`
`NOUN
`
`DEFINITIVE NOUN
`
`NOUN PHRASE
`
`~
`
`VERB PHRASE
`
`VERB 3
`
`VERB
`I
`
`~
`
`CEFINITINE NOUN
`
`NOlJ'.I PHRASE
`
`SENTENCE
`
`----
`
`Fig. I I
`
`Page 12 of 25
`
`
`
`U.S. Patent
`
`Feb. 22, 1994
`
`Sheet 12 of 12
`
`5,289,375
`
`Fig. 12
`
`START
`
`S101
`
`C~K HEAD CHARCTER
`OF WORD
`
`NO
`
`S104
`CALO..JLAlE AS A::R:ENl'At:E 'r-'0/v ~
`a= S8XN) OR LA1ER ~ HA VE
`LPF£RCASES AT RESPB:T1'IE HEADS
`
`YES
`
`S106
`
`S107
`
`SET TO
`LFPER:ASES MODE
`
`SET TO
`STANDARD MOOE
`
`END
`
`Page 13 of 25
`
`
`
`1
`
`TRANSLATION MACHINE
`
`5,289,375
`
`2
`part of each English word and its Japanese equivalent in
`a dictionary-consulting buffer. If the English word has
`two or more parts of speech, two or more combinations
`of each part of the English word and each Japanese
`5 equivalent are stored in
`the dictionary-consulting
`buffer. Then, the next process is to read the combina(cid:173)
`tions for each word out of the dictionary-consulting
`buffer in accordance with the predetermined rules and
`to generate the Japanese equivalents on a basis of the
`combinations. One example of those Japanese equiva(cid:173)
`lents is such that;
`
`BACKGROUND OF THE INVENTION
`I. Field of the Invention
`The present invention relates to a translation machine
`which is capable of generating a natural translated sen(cid:173)
`tence. More particularly, the invention relates to the
`translation machine which is capable of translating an
`input sentence in accordance with a sentence pattern of 10
`the input sentence by an interaction with an operator.
`2. Description of the Related Art
`The inventors of the present invention know that
`there have been traditionally proposed some translation
`machines. A known translation machine mainly in- 15
`eludes a CPU, a translation module, and storage units
`for storing a main dictionary, a user dictionary, gram(cid:173)
`matical rules and tree-structure converting rules. In
`operation, the above-mentioned translation machine is
`adapted to input a sentence described in a source Ian- 20
`guage from a keyboard into a translation module under
`the control of the CPU and to translate the sentence
`described in the source language into a sentence de(cid:173)
`scribed in a target language in the translation module
`using the main and user dictionaries and the grammati- 25
`cal and tree-structure converting rules.
`In executing the translation from English to Japanese,
`for example, the above-mentioned translation module
`serves to divide the input sentence described in English
`word by word and to obtain grammatical information 30
`such as a Japanese equivalent to each word or a part of
`speech of the Japanese equivalent by consulting the
`dictionaries stored in the storage unit. If the subject
`word has two or more parts of speech, the Japanese
`equivalent matching to each part of speech can be out- 35
`put. As such, in translating the English sentence into a
`Japanese equivalent, the above-mentioned translation
`module generates two or more Japanese equivalent
`candidates on a basis of the part of speech obtained by
`consulting the dictionaries. It results in that the English 40
`sentence is ambiguously interpreted.
`For example, when the English sentence "Time flies
`like an arrow" is input, some Japanese equivalents are
`provided depending on the interpretation of the part of
`speech of each word. That is, "Time" may be inter- 45
`preted as a noun or a verb, "flies" may be interpreted as
`a noun or a verb, "like" may be interpreted as a noun or
`a preposition. The resulting Japanese equivalents are as
`follows:
`"(ya-no-youna-jikan-bae) (meaning that "ar- 50
`I. "
`row-analogous flies (insects) called as time")
`2. "
`"(ya-no-youni-hae-wo-hakare) (meaning that
`"Measure flies as if an arrow moves through air")
`3. " "(ya-no-youna-hae-wo-hakare) (meaning that
`"Measure arrow-analogous flies")
`4. " "(jikan-bae-wa-ya-wo-konomu) (meaning that
`"Flies called as time are fond of an arrow")
`5. " "(jikan-wa-ya-noyouni-tobu) (meaning
`"Time passes as if an arrow moves through air")
`A selection of a proper Japanese equivalent from 60
`these Japanese equivalent candidates depends on an
`operator's judgment.
`For automating this part of translation, the above(cid:173)
`mentioned translation machine works as follows. When
`an English sentence is input, the above-mentioned trans- 65
`lation module takes the steps of dividing the English
`sentence into words by consulting the dictionaries for
`each word, and storing the combination of each speech
`
`that
`
`Input English
`Sentence:
`Part of Speech:
`Japanese equivalent
`candidate:
`
`Time
`
`flies
`
`an arrow.
`
`Noun
`
`Noun
`
`Preposition Noun
`
`(ya-no-youna-jikan-bae)
`(meaning that "Arrow-analogous flies
`(insect) called as time")
`
`If the candidate is not proper, the operator instructs
`the CPU to accept or correct the part of speech for each
`word, thereby the accepted or the corrected part of
`speech is fixed. The above-mentioned translation mod(cid:173)
`ule serves to read the part of speech for another word
`and to reorganize the combination of parts of speech of
`each word composing the English sentence by reading
`another part of speech for another word from the dic(cid:173)
`tionary-consulting buffer. The translating process of the
`reorganized combination results in offering the next
`combination of the parts of speech of the English sen-
`tence and its Japanese equivalent candidate. The above
`mentioned translation module produces the operation
`by the interaction with the operator. The above-men(cid:173)
`tioned repeated operations result in offering a high(cid:173)
`quality translated sentence.
`However, the above-mentioned translation machine
`cannot grasp which word is ambiguous considering a
`result of an interpretation, thereby it reads one combina(cid:173)
`tion of a speech part of each word composing the input
`sentence described in English and its Japanese equiva(cid:173)
`lent depending on the predetermined rules stored in the
`dictionary-consulting buffer, and translates based on the
`combination and generates the translated equivalent
`candidate. Then, the operator accepts or corrects the
`part of speech of each English word by watching the
`translated equivalent candidate on the CRT. It results in
`removing ambiguity of the input sentence in an interac-
`tive manner with the operator.
`The above-mentioned translation machine is designed
`to remove interpretative ambiguity of an input sentence
`by the operator's trial and error, thereby it requires the
`operator to keep an interaction with the machine, and as
`55 a result, the overall process of the translation is ineffi(cid:173)
`cient.
`Further, the above-mentioned translation machine is
`designed to define the syntax structure of the input
`sentence on a basis of the grammatical rules stored in
`the memory. It results in that the translation machine
`outputs two or more translated equivalents when the
`verb contained in the input sentence has two or more
`sentence patterns. Hence, a user has to do the annoying
`operation of selecting a proper translated equivalent
`from the output translated equivalents.
`In the following part, a process of finding a title of
`each English (source language) word for consulting the
`dictionaries used for the above-mentioned translation
`
`Page 14 of 25
`
`
`
`3
`machine will be described in detail. Normally, in the
`dictionaries, the title of the English word is registered in
`lowercase. However, the word placed at the head of the
`input sentence has an uppercase at the head. Hence, the
`word placed at the head of the input sentence may be S
`determined not to be registered in the dictionaries. For
`avoiding it, before consulting the dictionaries for the
`sentence-head word, the uppercase placed at the head
`of the word is converted into the lowercase. This dic(cid:173)
`tionary-consulting mode is referred to as a standard 10
`mode.
`On the other hand, for a specific sentence consisting
`of uppercases only such as a telex description, the stan(cid:173)
`dard mode is not effective. Hence, for such a sentence
`including the uppercases in the sentence-head word as 15
`well as the other words, all the uppercases included in
`the sentence are converted into the lowercases before
`consulting the dictionaries. This dictionary-consulting
`mode is referred to as an uppercase mode.
`The above-mentioned translation machine requires an 20
`operator to select the standard mode or the uppercase
`mode in accordance with the input sentence.
`the
`For setting the dictionary-consulting mode,
`above-mentioned translation machine automatically sets
`it as the standard mode in a case that no mode is set by 25
`the operator. When the standard mode is selected by the
`operator, the translation machine sets it as the standard
`mode. When the uppercase mode is selected by the
`operator, the translation machine sets it as the uppercase
`mode.
`However, an ordinary sentence has both lowercases
`and uppercases. For example, a proper noun often con(cid:173)
`tains an uppercase. At the standard mode, therefore, the
`translation machine does not convert the uppercases
`into the lowercases except that placed at the head of the 35
`sentence. The word having an uppercase is determined
`not to be registered in the dictionaries only if the upper(cid:173)
`case does not appear at the head of the sentence. As a
`result, disadvantageously, the translation machine out(cid:173)
`puts the original English word as its translated equiva- 40
`lent. It means that it is necessary to set the dictionary(cid:173)
`consulting mode as the uppercase mode for making
`effective use of the contents stored in the dictionaries.
`If the dictionary-consulting mode is always set to the
`uppercase mode, the uppercase placed at the head of a 45
`proper noun, such as the name of a person, is converted
`into the lowercase even though it is not desired. In this
`case, if the dictionaries have a translated equivalent
`word to the proper noun, it results in disadvantageously
`outputting an erroneous Japanese equivalent as de- so
`scribed below.
`Input word: Mr. Brown
`Translated equivalent
`Standard mode: Brown
`equivalent of "Mr.")
`(Chyairo.shi: "Chyairo"
`Uppercase mode:
`is a Japanese equivalent of a brown color)
`Both of the above-mentioned disadvantages in the
`dictionary-consulting mode result in that the standard
`mode or uppercase mode is indiscriminately set without 60
`considering each word of the input sentence.
`
`(Shi: meaning a Japanese
`
`SUMMARY OF THE INVENTION
`It is therefore an object of the present invention to
`provide a translation machine which is capable of inclu- 65
`sively removing interpretative ambiguity of an input
`sentence before doing the translation process for en(cid:173)
`hancing the efficiency of the translation process.
`
`5,289,375
`
`4
`The object of the present invention can be achieved
`by a translation machine having a storing unit for stor(cid:173)
`ing information of words and grammar rules, the trans(cid:173)
`lation machine being capable of translating sentences
`described in a source language into a target language
`includes a unit for dividing input sentence described in
`the source language into a morphologic elements and
`for generating a mark-inserted sentence by adding a
`mark which represents an information of the detected
`morphologic elements, a unit connected to the dividing
`unit for analyzing a syntax of the morphologic elements
`in accordance with the information stored in the storing
`unit, and a unit connected to the analyzing unit for
`converting a structure of the syntax obtained in the
`dividing unit into a structure of syntax obtained in the
`dividing unit into a structure of syntax in the target
`language so that translated sentence is generated on a
`basis of the syntax structure in the target language ob(cid:173)
`taill,ed by the converting unit.
`Preferably, the dividing unit includes a unit for de(cid:173)
`tecting the morphologic elements which are functioned
`as two or more parts of speech among the morphologic
`elements, and a unit for generating a mark-inserted
`sentence by adding a mark which represents an informa(cid:173)
`tion of parts of the speech of the detected morphologic
`elements.
`More preferably, the dividing unit further includes a
`unit for detecting the mark-inserted morphologic ele(cid:173)
`ments from the mark-inserted sentence, a unit for ob-
`30 taining two or more parts of the speech of the mark(cid:173)
`inserted morphologic elements and for displaying the
`part of speech candidates consisting of the obtained two
`or more parts of the speech.
`Further preferably, the dividing unit includes a unit
`for replacing the mark included in the mark-inserted
`sentence with the part of speech symbol for specifying
`a proper part of the speech selected from the part of
`speech candidates and for generating a sentence with
`the part of speech symbol inserted.
`The analyzing unit is preferably adapted to analyze
`the syntax based on the sentence with the part of speech
`symbol generated by the generating unit.
`The information preferably includes a part of speech,
`a tense, a person, a numeral, a translated equivalent, and
`a part of speech of the translated equivalent.
`The source language is English and the target lan(cid:173)
`guage is Japanese, preferably.
`The translation machine according to the first aspect
`of the invention works as follows.
`When a sentence is input into the dividing unit, the
`input sentence is divided into morphologic element
`series so as to obtain the information such as a part of
`speech for each element and its translated equivalent.
`Then, the detecting unit detects a morphologic element
`55 having two or more parts of speech from the divided
`morphologic elements. If such a morphologic element is
`detected, the mark-inserted sentence generating unit
`serves to add a mark representing the information about
`two or more parts of speech included in the detected
`morphologic element for generating a mark-inserted
`sentence.
`Next, the mark-inserted morphologic element detect-
`ing unit detects the mark-inserted morphologic element
`from the mark-inserted se~tence generated by the mark(cid:173)
`inserted sentence generating unit. And, the part-of(cid:173)
`speech-candidate-series displaying unit serves to seek
`two or more parts of speech included in the detected
`mark-inserted morphologic element based on the marks
`
`Page 15 of 25
`
`
`
`25
`
`s
`and display them on the display unit. the specific part of
`speech is selected from the speech part candidate series
`displayed on the display unit. The unit for generating a
`sentence with
`the part-of-speech symbol
`inserted
`therein serves to replace the mark included in the mark- 5
`inserted sentence with the part-of-speech symbol for
`specifying the selected part of speech, thereby generat(cid:173)
`ing the sentence with the part-of-speech symbol in(cid:173)
`serted thereto.
`Next, the syntax analyzing unit serves to analyze the 10
`syntax of the morphologic clement series based on the
`resulting symbol-inserted sentence. The syntax struc(cid:173)
`ture given by the syntax analyzing unit is converted into
`the target-language syntax structure in the converting
`unit. Based on the resulting target-language syntax 15
`structure, the translated sentence generating unit serves
`to generate the translated sentence.
`As described above, before analyzing the syntax of
`the morphologic clement series, the present translation
`machine can inclusively check the morphologic ele- 20
`mcnt having two or more parts of speech and the con(cid:173)
`tent of each part of speech. It results in being able to
`inclusively select the proper one of the two or more
`parts of speech included in the morphologic element for
`a short time.
`It is a second object of the present invention to pro(cid:173)
`vide a translation machine which can provide a proper
`translated equivalent by specifying a sentence pattern of
`an input sentence.
`The second object of the present invention can be 30
`achieved by a translation machine having a storing unit
`for storing information of words and grammar rules, the
`translation machine being capable of translating sen(cid:173)
`tences described in a source language into a target lan(cid:173)
`guage includes a unit for dividing input sentence de- 35
`scribed in the source language into morphologic ele(cid:173)
`ments, a unit connected to the dividing unit for analyz(cid:173)
`ing a syntax of the morphologic elements in accordance
`with the information stored in the storing unit, a unit
`connected to the analyzing unit for converting a struc- 40
`ture of the syntax obtained in the analyzing unit into a
`structure of syntax obtained in the analyzing unit into a
`structure of syntax in the target language, and a unit
`connected to the converting unit for generating a trans(cid:173)
`lated sentence based on the target language syntax 45
`structure obtained by the converting unit.
`Preferably, the dividing unit includes a unit for add(cid:173)
`ing a symbol so as to specify a sentence pattern to a
`word included in the input sentence described in the
`source language, a unit for detecting whether or not the 50
`sentence pattern symbol is added in the input sentence
`during a translating operation, a unit for storing a word
`to which the symbol is added and a sentence pattern
`specified by the symbol at a time when the sentence
`pattern symbol is detected, and a unit for defining a 55
`structure of the input sentence by referring to the con(cid:173)
`tent of the storage unit.
`More preferably the sentence pattern detected by the
`symbol adding unit is selected from any one of a first
`pattern sentence consisting of a subject and a verb, a 60
`second pattern sentence consisting of a subject, a verb
`and a complement, a third pattern sentence consisting of
`a subject, a verb and an object, a fourth pattern sentence
`consisting of a subject, a verb, a direct object and indi(cid:173)
`rect object, a fifth pattern sentence consisting of a sub- 65
`jcct, a verb, an object and a complement.
`The storing unit is capable of storing grammatical
`rules regarding a plurality of sentence patterns which
`
`5,289,375
`
`6
`are preferably selected in accordance with a result of
`the dividing unit.
`The plurality of sentence patterns include a sentence
`consisting of a noun phrase only, a sentence consisting
`of a verb phrase only, and sentence consisting of a noun
`and a verb phrases, preferably.
`The translation machine designed according to the
`second aspect of the invention works as follows.
`When a sentence is input to the machine, the dividing
`serves to analyze the input sentence into morphologic
`clements and to detect whether or not a sentence-pat(cid:173)
`tern-specifying symbol (referred to as a sentence-pat(cid:173)
`tern symbol) is included in the morphologic elements. If
`the symbol is detected, the word to which the symbol is
`added and the sentence pattern specified by the symbol
`arc stored in a storage unit. By referencing the content
`of the storage unit, it is possible to define the syntax
`structure of the input sentence. If, therefore, the input
`sentence has the sentence-pattern symbol, the transla(cid:173)
`tion machine can output the translated equivalent de(cid:173)
`fined on the sentence specified by the symbol.
`As mentioned above, the present translation machine
`is designed to insert the sentence-pattern symbol to an
`input sentence in advance, detect the symbol in doing
`the actual translation, and analyze the syntax of the
`input sentence based on the specified sentence pattern,
`thereby being able to more easily output an accurate
`translated equivalent.
`Further, the specification of the sentence pattern
`results in reducing the syntax candidates to be analyzed.
`Hence, the translation machine has an advantage that
`the processing time is made shorter.
`It is a third object of the present invention to provide
`a translation machine which can automatically set the
`dictionary-consulting mode to an optimal mode consid(cid:173)
`ering the characters of the input sentence.
`The third object of the present invention can be
`achieved by a translation machine having a storing unit
`for storing information of words and grammar rules, the
`translation machine being capable of translating sen(cid:173)
`tences described in a source language into a target lan(cid:173)
`guage includes a unit for dividing input sentence de(cid:173)
`scribed in the source language into morphologic ele(cid:173)
`ments, a unit connected to the dividing unit for analyz(cid:173)
`ing a syntax of the morphologic elements in accordance
`with the information stored in the storing unit, a unit
`connected to the analyzing unit for converting a struc(cid:173)
`ture of the syntax obtained in the analyzing unit into a
`structure of syntax in the target language, a unit con(cid:173)
`nected to the converting unit for generating a translated
`sentence based on the target language syntax structure
`obtained by the converting unit, and a unit connected to
`the generating unit for moduling the translated sentence
`so that translated sentence is compared with the input
`sentence.
`Preferably, the moduling unit is capable of inclu(cid:173)
`sively checking the input sentence described in the
`source language, the inclusively checking including a
`step of dividing the input sentence into words, a step of
`checking the word having two or more parts of the
`speech from the words, and a step of inserting a mark to
`the detected word.
`More preferably, the moduling unit is capable or
`removing an ambiguity from the input senter,ce, the
`ambiguity removing including a step of detecting the
`mark-inserted word, a step of obtaining the two or more
`parts of the speech indicated by the mark, a step of
`displaying the two or more parts of the speech as candi-
`
`Page 16 of 25
`
`
`
`5,289,375
`
`25
`
`7
`dates so as to specify a proper one ·or the candidates, and
`a step of replacing the mark with a symbol for indicat(cid:173)
`ing the part of the speech specified by an operator.
`Further preferably, the dividing unit includes a unit
`for determining whether or not an uppercase exists in 5
`character series of the morphologic elements, a unit for
`calculating a proportion of the morphologic elements
`having an uppercase, and a unit for setting a dictionary
`consulting mode from a standard mode to an uppercase
`mode in accordance with a result obtained by the char- 10
`acter determining unit.
`The setting unit is preferably adapted to set the dic(cid:173)
`tionary consulting mode from the standard mode to an
`uppercase mode on a basis of the proportion calculated 15
`by the calculating unit.
`The setting unit is further preferably capable of con(cid:173)
`verting the uppercase at the word head into a lowercase
`at the standard mode before consulting the dictionaries.
`The setting unit is more preferably capable of con- 20
`verting all the uppercases contained in the word into
`the lowercases at the uppercase mode before consulting
`the dictionaries.
`The translation machine designed according to the
`third aspect of the invention works as follows.
`When the sentence is input to the dividing unit, it is
`divided into the morphologic element series. Then, the
`character determining unit determines whether or not
`an uppercase exists in all the morphologic elements
`(words). If yes, the rate calculating unit serves to calcu- 30
`late a percentage of morphologic elements which have
`an uppercase in the input sentence. Depending on the
`result given by the character determining unit and the
`percentage given by the rate calculating unit, the die- 35
`tionary-consulting mode is automatically switched from
`the standard mode to the uppercase mode or vice versa.
`Based on the switchable dictionary-consulting mode,
`the translated equivalent to the morphologic element
`and its part of speech are obtained.
`Next, the syntax analyzing unit serves to analyze the
`syntax of the morphologic element series. And, the
`converting unit converts the source-language syntax
`structure into the target-language one. Then, the trans(cid:173)
`lated equivalent generating unit generates the translated 45
`equivalent based on the target-language syntax struc(cid:173)
`ture.
`The dividing unit performs the dictionary-consulting
`operation at the optimal mode set according to the
`morphology of the characters composing the input 50
`sentence. When the input sentence contains uppercases,
`is automatically
`the dictionary-consulting mode
`switched to the uppercase mode in a case that the per(cid:173)
`centage is equal to or more than a predetermined value.
`The translation machine can analyze the input sen(cid:173)
`tence containing many uppercases except the input
`sentence having an upper case placed at the head of the
`input sentence at a time when the morphologic elements
`of the input sentence are registered in the dictionaries. 60
`When the input sentence contains few uppercases such
`as a proper noun rarely appearing ther