JPH0283664A - Reference part qualifying and analyzing system - Google Patents

Abstract

PURPOSE:To separate exaggeration reference from another reference by checking the presence/absence of a reference part in an input text, counting the number of words in the reference part of a text in which the reference part exists, and qualifying the reference part including one word as the exaggeration reference. CONSTITUTION:A pre-processing using spell-checking 5 and pre-editing 6 is applied on the input text 4 obtained by either file input, the input of a keyboard 2, or that of an OCR 3. An output text 11 obtained by a translation part 7 is edited by using translation information by post-editing 8, and the input text 4 and the output text 11 are printed by using a printer 12. The presence/absence of the reference part in the input text is checked, and the number of words in the reference part of the text in which the reference part exists is counted, and the reference part is judged as the exaggeration reference when the number of words is one. In such a way, it is possible to separate the exaggeration reference not having been separated until now from another reference.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for recognizing and analyzing quoted parts, and more particularly to a morphological analysis unit and a syntactic analysis unit in machine translation.

■Technical rights In general, the unit of analysis for syntactic analysis is one sentence. When performing analysis, it is desirable that the analysis range be short. In other words, if a shorter range is to be analyzed, the number of applications and combinations of analysis rules can be reduced, making the analysis easier.
Ambiguity generated during the analysis process can be minimized.

However, actual texts do not only consist of short sentences; on the contrary, many of them consist of long sentences. Therefore, if we consider a partial analysis range within a single sentence called a block and divide the analysis range by specifying parts within a single sentence, the analysis range will be limited and the number of application times and combinations of analysis rules will be significantly reduced. This reduces unnecessary rule applications and combinations, improves analysis efficiency and speed, eliminates ambiguities, and improves analysis accuracy. If this partial analysis range can be specified automatically, the precision and speed of syntax analysis will definitely increase.

The present invention has been made in view of the above-mentioned circumstances.
To be able to distinguish between emphatic quotations and other quotations, which have not been particularly distinguished up until now, and to be able to distinguish between arbitrary and volitional quotations and title quotations, which have not been distinguished so far.
To enable analysis to be performed without specifying an ineffective range by recognition of emphasized citations, and to enable more accurate, fast and efficient analysis to be performed not only in the case of arbitrary citations but also in the case of title citations. It was made for the purpose of

Structure-11 The present invention has been made to achieve the above objects.

(1) The morphological analysis part of a natural language analysis system such as machine translation checks whether there is a quotation part in the input text, counts the number of words in the quotation part for sentences with quotation parts, and calculates the number of words by 1. (2) In the morphological analysis section of a natural language analysis system such as machine translation,
Checks the presence or absence of a quotation part in the input text, checks for the presence or absence of punctuation marks (periods, commas) immediately before and after the quotation mark that closes the quotation part for sentences with quotation parts,
(3) The next processing step of the morphological analysis unit in a natural language analysis system such as machine translation. The syntactic analysis unit, which is equipped with syntactic analysis and grammar rules for analysis, analyzes the highlighted quotation recognized in (1) above, assuming that there are no quotation marks. (4) In a natural language analysis system such as machine translation, the syntactic analysis unit, which is the next processing step after the morphological analysis unit, is equipped with syntactic analysis grammar rules for analysis. , for the two types of quotation parts certified in (2) above, before analyzing one sentence, which is the unit of analysis, the quotation parts are partially analyzed, and arbitrary quotation parts are analyzed. The sentence is analyzed based on the result, assuming that the cited part has a required syntactic role in the sentence, and the title quotation part is used to analyze the sentence even if the analysis result does not become a noun phrase.
This method is characterized in that the sentence is analyzed by assuming that the syntactic role of the quoted part in the sentence is a noun phrase. Hereinafter, the present invention will be explained based on examples.

Thus, the present invention is an improvement to the method of analyzing cited portions in a sentence as blocks (partial analysis ranges).

There are actually different types of quotation parts, and in the present invention, they are divided into three parts: an emphasized quotation part, an optional quotation part, and a title quotation part, and the morphological analysis part distinguishes each part and passes them to the syntactic analysis part. By performing different analysis processes on these three quotation parts in the syntax analysis section, it is possible to improve the efficiency and accuracy of syntax analysis.

FIG. 1 is a block diagram showing an embodiment of a translation device equipped with a dictionary lookup method according to the present invention. In the figure, 1 is a CRT, 2 is a keyboard, 3 is an 0CR54 is an input document, 5 is a spell check unit, 6 is a pre-editing section, 7 is a translation main section, 8 is a post-editing section, 9 is a dictionary, 10 is a grammar rule, 11 is an output document, 12 is a printer, and the information obtained by file input, keyboard input, or OCR input is The input sentences are preprocessed using spell checking and pre-editing, the output sentences obtained by the translation section are edited using the translation information in post-editing, and the input and output sentences are printed using a printer. .

Figure 2 is a diagram showing the flow of the translation main body.
The translation unit) 7 is roughly divided into four processes: morphological analysis, syntactic analysis, conversion, and generation. First, the morphological analysis unit performs dictionary lookup of the input text. The parsing section obtains information about each word and performs parsing according to grammatical rules.
Create a tree structure from the analysis results. The conversion unit transforms the tree structure of the input language into the tree structure of the output language. The generation section translates the obtained tree structure node by node.

The present invention belongs to the above-mentioned morphological analysis section and syntactic analysis section, and here, the input text is assumed to be English text. The morphological analysis section performs the processing shown in FIG. 3 on the input text.

FIG. 3 is a diagram for explaining the flow of quotation recognition in morphological analysis, in which the presence or absence of quotation marks is checked. If there are two quotation marks, the part between them is recognized as the quotation part. Next, count the number of words in the quotation.

If the number of words is 1, that quotation is set as the emphasized quotation. For other quoted parts, check whether there is a punctuation mark (period, comma) immediately after the quotation mark that closes the quoted part,
If there is a period or comma, the quoted part is recognized as an optional quoted part.

When there is no period or comma, the quoted part is recognized as the title quoted part. The quoted part information that has been certified is passed from the morphological analysis section to the syntactic analysis section.

FIG. 4 is a diagram showing the flow of the syntax analysis section, which analyzes each sentence. In this embodiment, the target sentence is analyzed from the bottom up from the end of the sentence using context-free parsing grammar rules. Parsing usually ends when all possibilities are exhausted and rules are applied until the target sentence is finally combined into a single grammatical code representing a sentence or the like. Usually, the final grammatical code is a sentence (SE: 5 sentences)
It is. If we consider a case where a sentence has a quotation that is a partial analysis range (block), there are three types of quotation: emphasized quotation, optional quotation, and title quotation. An emphatic quotation is a word enclosed in quotation marks to emphasize a specific word, and there is only one word inside the quotation mark, and the word inside the quotation and the word outside the quotation are combined. is not semantically delimited by quotation marks, and it is difficult to say that it is an analysis range that has a particular unit that should be delimited. Therefore, in the case of an emphasized quotation, the analysis process is performed without specifically treating it as a partial analysis range. An arbitrary quotation is a quotation with punctuation marks before and after the quotation mark that closes the quotation, and is a typical quotation, with a clear break in meaning between the inside and outside of the quotation.

To analyze a sentence with this quotation, proceed from the end of the sentence. When the analysis progresses to the arbitrary quotation part, which is a partial analysis range (block), the inside of the arbitrary quotation part is analyzed. When the analysis of the quotation part is completed, the analysis is conducted by assuming that the grammatical code indicating the grammatical role obtained as a result of the analysis is the grammatical role that the quotation part plays for this sentence. In other words, the optional quotation section has the meaning of delimiting the range of analysis of a sentence. Also,
A title quotation is a quotation without punctuation marks (periods, commas) before and after the quotation mark that closes the quotation, and it is assumed that the title (noun phrase) is inside the quotation.

Proceed with the analysis in the same way as with arbitrary citations. The difference is that the sentence is not analyzed using the analysis result of the title quotation part as is, but the sentence is analyzed by assigning the role of the noun phrase, no matter what the analysis result is.

This will be explained below using an example.

a、All you "need" is 1ove
.

b, He 5ays,”all you need
is 1ove,”c,”All you need
is l0Ve"ars the words of
The Beatles' song.

First, these three sentences a, b, and c are subjected to citation recognition processing by the morphological analysis unit. Count the number of quotation marks. If there is an even number (usually two, but sometimes four), count the number of words in the quotation. If there is only one, it is recognized as an emphasized quotation. Sentence a is an example of an emphatic quotation. In other cases, check for punctuation marks (periods, commas) before and after the quotation mark that closes the quotation mark. B is recognized as an arbitrary citation because it exists, and C is recognized as a title citation because there is neither. Next, proceed to the syntax parsing section, which is described in step a above.

For example, suppose the following grammar rules are used to analyze the English sentences b and c.

A1 Part of speech classification code prn (nominative pronoun) n0u (noun) no2 (
(noun possessive case) dat (definite article) vil (transitive phrase as object) vt2 (transitive that clause as object) vil (transitive complement) pre (preposition) cma (comma) prd (period) B0 grammar Code NP (noun phrase) SN (nominative noun phrase) CN (complement noun phrase) ON (object noun phrase) DP (prepositional phrase)
CN (comma) PD (period) VC (predicate)
QC (thati that can be omitted) WC (predicate without object) SG (sentence without sentence final symbol)
SE (sentence including sentence-final symbol) C0 grammar rule 1, SE → SG PD 2, 5G -> SN V (1°3.0C
→ SG 4, VC-) vil CN 5, VC-+ vtl CN 6, VC+vt2 (CN) QC7, WC-
+ vtl 8.5N −+ NP 9, SN −+ dat NPl, 0.
SN-+prn (SN WC)11, CN
4 Nl' 12, CN −+ det NPl3, ON
4 NP 14, ON -) det NPl5, N
P → nou (DP)16, NP −*
no2 NPl7, DP → pre 0
N IL CN -) cma 19, yen) → prd Note that the elements in parentheses may or may not be present.

Also, the number at the beginning of the line indicates the rule number.

Analyze sentence a starting from the end of the sentence. For simplicity, assume that each word only has a part-of-speech classification that leads to the correct answer. In case a, the analysis is the same as when there is no cited part.

a, all(prn) you(prn) n
eed(vtl) 1s(via) 10νe(nou
) , (prd) 1, (1!]) PD −+ prd (,)2
, (15) NP + nou (Love)3
, (11) CN-+NP (love)4,
(4) VC-+ vil CN (is 1ov
e) 5, (7) WC-*vtl (ne
ed)6, (1,0) SN → prn (yo
u)7, (10) SN-+prn SN WC(al
you need)8, (2) SG -* S
N VC (all you need is 1ove)
)9, (1) SE −+ SG PD (all
(you need is 1ove,) However, the beginning of the line is the serial number, and the number in parentheses is the rule number.

Sentence b is “all you need is 1ove”
” is an arbitrary quotation. The analysis of this part is the same as 2 to 7 in a. Also, the quotation mark itself is not subject to analysis.

b, he(prn) 5ays(vt2),
(cma) all(prn) you(prn)
need(vtl) 1s(vil) 1ove(
nou), (prd) 1, (19) PD −+ prd (,
)2, (15) NP-+ nou (l
ove)8, (2) SG-+SN VC(all
you need is 1ove)9, (3)
QC-) SG (all you need
is 1ove) 10, (17) CN −+
crma (1)11, (6) VC-) v
t2 CN QC(says, all you ne
edis 1ove) 12, (10) SN → prn (he))
3. (2) SG −+ SN VC (ha 5a
ys, all you need Love) 14, (1) SE −+ SG PD (he
5ays, all you need 1ov
e,) The analysis is proceeding by using the SG (sentence without period), which is the analysis result of all you need is Love, as it is.

The sentence in C is “all you need is 1ov”
e" is a title quotation. The analysis of this part is the same as 2 to 7 in a. Also, the quotation mark itself is not subject to analysis.

b, ali(prn) you(prn) n
eed(vtl) 1s(vil) love(nou
) are(vil) the(det) word
s(nou) of(pre) the(det)
Beatles(no2) song(nou)
, (prd)1, (19) PD −+ prd
(,)2, (15) NP −+ nou
(song)3, (16) NP −+ no2
NP (Beatles' song) 4, (14)
ON →det NP (the Beatles'
song) 5, (17) Kano-+ pre ON (
of the Beatles' song)6, (
15) NP −* nou DP (words o
f the Beatles' song) 7, (12) CN −+ det NP (the
words of the Beatles' son
g) 8, (4) VC→vil CN (ars th
e words of the13eat1. es'
song) 9, (15) NP −+ nou
(love)15, (2) SG 4 SN
VC(all you need is 1ove)1
6. (0) NP→SG Give SG the role of NP (noun phrase) 17. (8) SN→NP (all you
need is 1ove) 18, (2) SG
4 SN VC (all you need is
1overare the words of
the Beatlcs' song) 19, (1) SE -* SG PD (al
l you need is 1overare th
e tmods of the Beatles
' song, ) The SG grammar code was obtained on the 15th line. Even if you continue parsing using this S(i, you will not reach SH in the end because there is no parsing rule. Therefore, in line 16, the grammar code NP is added as a new role for this title quotation. This is not a grammar rule, but it can be thought of as labeling the SG obtained in line 15 as NP in line 16. Through this process, the sentence in C can be successfully analyzed.

1-As is clear from the above explanation, according to claim (1),
You can now distinguish between highlighted quotations and other quotations, which you did not previously distinguish between.

According to claim (2), any matter that has not been particularly distinguished so far:
You can now distinguish between quotations and title quotations.

According to claim (3), it is now possible to perform analysis without specifying an ineffective range by recognizing one emphasized citation.

According to claim (4), more accurate, faster, and more efficient analysis can be performed not only in arbitrary citations but also in title citations.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of a translation device equipped with a dictionary lookup method according to the present invention, Fig. 2 is a diagram showing the flow of the main body of translation, and Fig. 3 is a diagram showing citation recognition in the morphological analysis section. FIG. 4 is a diagram showing the flow of syntax analysis. 1...CRT, 2...Keyboard, 3...OCR
, 4... Input sentence trace, 5... Spelling, error part, 6
...Previous editorial department. 7... Translation body part, 8... Post-editing part, 9... Dictionary, 10... Grammar rules, 11... Output document, 12...
printer.

Claims (1)

[Claims] 1. In the morphological analysis unit of a natural language analysis system such as machine translation, the presence or absence of a quotation part in the input text is checked, and the number of words in the quotation part is counted for sentences in which the quotation part exists. , a quotation part recognition method characterized in that if the number of words is one, the quotation part is recognized as an emphasized quotation. 2. The morphological analysis unit of a natural language analysis system such as machine translation checks the presence or absence of quotations in the input text, and for sentences with quotations, punctuation marks immediately before and after the quotation mark that closes the quotation are detected. This quotation recognition method is characterized by checking the presence or absence of punctuation marks and recognizing the quotation parts with punctuation marks as optional quotation parts and the quotation parts without punctuation marks as title quotation parts. 3. In a natural language analysis system such as machine translation, the syntactic analysis unit, which is the next processing step after the morphological analysis unit, is equipped with syntactic analysis grammar rules for analysis, and the emphasis certified in claim (1) A quotation part analysis method characterized in that a sentence in which a highlighted quotation part exists is analyzed on the assumption that there is no quotation mark in the quotation part. 4. In a natural language analysis system such as machine translation, the syntactic analysis unit, which is the next processing step after the morphological analysis unit,
Claim (2) is provided with parsing grammar rules for analysis.
) For the two types of quoted parts that have been certified in The sentence is analyzed assuming that the part plays a required syntactic role in the sentence, and the title quotation section analyzes the sentence as if the quoted part has a required syntactic role in the sentence, even if the analysis result does not become a noun phrase. A quotation partial analysis method characterized by analyzing the sentence as a noun phrase.