JPH06274530A - Syntax analyzed result display system - Google Patents

Abstract

(57) [Summary] [Purpose] When displaying the results of parsing, it is easy to understand the correspondence with the original sentence while preserving the order of the original sentence, and the display efficiency is good even for long sentences. A parsing result display method that makes it easy to understand the parsing result even when taking a down-level understanding method. [Structure] In a syntax analysis result display method for displaying a hierarchical structure between constituent elements as a result of syntax analysis, a substructure of each constituent element is recursively placed in a pair of parentheses for display.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention effectively includes a syntax analysis result display method, for example, a syntax analysis process, and a mechanism for presenting the analysis result for the purpose of confirming or changing the analysis result by a human. The present invention relates to a syntax analysis result display method suitable for application to a natural language processing device, such as a machine translation system or a sentence understanding support device.

[0002]

2. Description of the Related Art In a device including a syntactic analysis process of a natural language sentence, it is effective to display the syntactic analysis result when determining the validity of the syntactic analysis result. In addition, it is effective to display the syntactic analysis result in order to help understanding of the sentence. As the simplest display method, for example, regarding the original sentence, “delivering a message to a designated partner at a designated time”, a dependency relation as shown in FIG. 21 and a syntax analysis tree as shown in FIG. Alternatively, FIG.
There are methods such as semantic network representation as shown in.

Natural language originally has an ambiguity that cannot be correctly judged only by using grammar and logic. Therefore, it is extremely difficult to perform a complete processing only by a device, and in many cases, some kind of processing is required. Requires human decision or calibration in form. For example, for proofreading in machine translation,
There are pre-editing performed on the original text before translation, inter-editing performed during the translation process, and post-editing performed on the generated translation.

As a kind of inter-edit, there is a method of proofreading the syntax analysis result when the syntax analysis is completed.
This method is expected to give good results with a relatively small amount of calibration for modern devices, where the parsing capabilities of the device are significantly improved. In addition, this method is also advantageous in that proofreading can be performed if only knowledge of the original language is available. In order to realize this method, a method of displaying the analysis result in the shape of an analysis tree and calibrating it has already been proposed. However, the display by the parse tree has a merit that the hierarchical structure of the syntax can be easily grasped as a whole, but it requires a large space on the screen,
Therefore, it may not fit on the screen, and in the case of a long sentence extending over a plurality of lines, it is inconvenient in that the whole sentence cannot be seen at once, and there is a drawback that the efficiency of calibration is not good. (See Japanese Patent Laid-Open No. 61-278970).

In order to solve the above-mentioned drawbacks, for example, Japanese Patent Application Laid-Open No. 61-278970 proposes a "control method for displaying and proofreading a syntactic analysis result in a natural language processing device". In this publication,
By displaying each node (eg, phrase, clause) as a result of parsing in at least one stage of parsing as an identifying expression (eg, underline) on the text, a large space required for the parse tree display. Is eliminated, and the stage of the displayed syntactic analysis result is changed according to the operator's request, so that the syntactic analysis result at various stages is displayed as necessary.

The display method shown in FIGS. 21 to 23 makes it easy to grasp the structure intuitively. For example, Japanese Patent Laid-Open No. 61-27897.
As pointed out in Japanese Patent No. 0, if a long sentence extends over a plurality of lines, the display width cannot be easily understood due to the limitation of the left and right display widths. In addition, the parse tree display of FIG. 22 has many syntactic intermediate nodes,
There is a problem in that it becomes complicated and it is difficult to grasp the order relation with the original sentence in the semantic network expression of FIG. Furthermore, such graphical representations, in particular FIG. 22 and FIG.
With 3, it is very difficult to arrange the expression. JP-A-61-
In the method described in Japanese Patent No. 278970, in order to solve the problem of long sentences and the problem that a wide display range is required as described above, a node in one hierarchy of the sentence hierarchical structure is added to the sentence text as shown in FIG. As shown in FIG. 25, an underlined identification expression is used for display.

In the displays shown in FIGS. 21 to 23, since the nodes for each clause are also displayed, when humans take a top-down way of understanding, the problem is that the detailed (terminal) nodes interfere. There is. In addition, there is a problem that it is necessary to follow the hierarchy of details. For this problem,
In the method disclosed in Japanese Patent Laid-Open No. 61-278970, the level of the displayed hierarchy can be changed (FIG. 24:
By displaying a plurality of layers at the same time (bottom-up shift, top-down shift) (FIG. 25: a plurality of underlines), it is possible to follow up without being restricted by the detailed layers. However, this method does not require a large area,
One underline can display only one hierarchical section.

In order to solve this point, for example, in Japanese Patent Laid-Open No. 4-348645, a "syntax analysis result display system" has been proposed. This Japanese Patent Laid-Open No. 4-34846
In the method described in Japanese Patent Publication No. 4, in order to efficiently display the syntactic analysis result even in a long sentence, for example, the original sentence “American Telegraph and Telephone Company applied for a patent for an epoch-making analysis method, It has been clarified that the patents have been granted. ", Suggesting a display method of a dependency structure as shown in FIG. By this display, the order of the original sentences is preserved, the correspondence with the original sentences can be easily grasped, and the display efficiency is shown even for long sentences. Furthermore, the above-mentioned JP-A-4-348.
In Japanese Patent No. 464 publication, when a person takes a top-down way of understanding, as shown in FIG. 27, a display in which a substructure is not displayed depending on the depth of a hierarchy so as not to obstruct the nodes of details (ends) The method, that is, the hierarchical display of the hierarchical depth 2 is shown. However, in the display as shown in FIG. 27, it is not possible to show the layers below the component of depth 2 of the layer.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above situation, and when displaying a syntactic analysis result, the order of original sentences is preserved and the correspondence with the original sentence is grasped. It is an object of the present invention to provide a syntactic analysis result display method that is easy and has a high display efficiency even for long sentences, and that allows a human to take a top-down way of understanding to easily understand the syntactic analysis result. In particular, the inventions of claims 5 and 6 are
In the hierarchical display described in the above-mentioned Japanese Patent Application Laid-Open No. 4-348464, the drawback is that the detailed hierarchy cannot be shown.

[0010]

In order to solve the above problems, the present invention provides a syntactic analysis result display system for displaying a hierarchical structure between constituent elements of a natural language sentence as a syntactic analysis result of a natural language sentence, (1) Having a hierarchical display means for displaying the substructures of each component recursively within the parenthesis pair, or (2) Arranging a plurality of substructures of the same hierarchy of each component in the parenthesis pair. Or (3) a constituent element instruction receiving means for receiving the designation of the constituent element, and a constituent element in a structure lower than the constituent element designated by the constituent element instruction receiving means. Having a hierarchical means for suppressing or substituting the display, or (4) depth instruction receiving means for receiving designation of the depth of the hierarchy in the hierarchical structure, and the depth instruction. It is characterized by having the hierarchical display means for suppressing or substituting the display of the constituent elements in the structure lower than the constituent element of the depth designated by the attaching means, or (5) the hierarchical relationship between the constituent elements is tree. A tree structure display means for displaying a structure, the hierarchy display means according to any one of (1) to (4) (for the hierarchy display means according to (3), the component instruction acceptance means, the ( 4) The hierarchical display means described above includes the depth instruction receiving means) and a component instruction receiving means 2 for receiving the designation of a component, and the tree structure displaying means includes the component instruction receiving means. 2 displays the structure below the constituent element designated by 2 by the hierarchical display means, or (6) a tree structure display means for displaying a hierarchical relationship between the constituent elements in a tree structure; ) To (4)
Any one of the hierarchical display means (for the hierarchical display means according to claim 3, the constituent element instruction receiving means, for the hierarchical display means according to (4), the depth instruction receiving means) , And depth instruction receiving means 2 for receiving the designation of the depth of the hierarchy in the hierarchical structure,
In the tree structure display means, the depth instruction receiving means 2
Is displayed by the hierarchical display means.

[0011]

In the syntax analysis result display method for displaying the hierarchical structure between the constituent elements of the natural language sentence as the result of the syntax analysis of the natural language sentence, (1) a hierarchy in which the substructure of each constituent element is recursively placed in a pair of parentheses. By displaying means or (2) hierarchical display means for displaying a plurality of substructures of the same hierarchy of the respective constituent elements side by side in parenthesis pairs, the syntactic structure is shown as a display of the syntactic analysis result. Further, (3) a hierarchical display means for suppressing or substituting the display of the constituent elements in the structure lower than the constituent element designated by the constituent element instruction receiving means for accepting the designation of the constituent element, or (4)
The structure specified by the hierarchy display means for suppressing or substituting the display of the component in the structure lower than the component having the depth specified by the depth instruction receiving means for receiving the designation of the depth of the hierarchy in the hierarchical structure Make the underlying structure of an element unobtrusive. Furthermore, in the tree structure display means for displaying the hierarchical relationship between the constituent elements in a tree structure, (5) the structures below the constituent elements designated by the constituent element instruction receiving means 2 are described in (1) to (4) above. Or (6) the structure below the component of the depth designated by the depth designation receiving means 2 is displayed.
By displaying by the hierarchical display means described in (4) to (4), in addition to the tree structure display, the subordinate configuration of the instructed component is made inconspicuous.

[0012]

EXAMPLES Examples of the present invention will be described below with reference to Japanese example sentences. FIG. 11 shows the Japanese example sentence, that is,
In the processing of a sentence in an agglutinative language that does not leave breaks between words like Japanese, morphological analysis is the first barrier. " Taking the hierarchical structure of the dependency relations between Japanese phrases as an example, assume that a list of dependency relations between phrases as shown in Table 1 is obtained for the above example sentence. In the following, the relation name between clauses is omitted for simplicity.

[0013]

[Table 1]

Each data in Table 1 includes a clause number, a clause notation,
It is in the form of the modified destination clause number. Based on this, according to the description in the above-mentioned Japanese Patent Laid-Open No. 4-348464, the tree structure display as shown in FIGS. 12 and 13 can be obtained. Here, the bunsetsu is a unit as a constituent element.
Further, the depth of the hierarchy in the tree structure display is the depth in the tree structure of each constituent element when the constituent element of the head, which is the highest part in the structure, is set to the depth 0 of the hierarchy. , "Morphological analysis" and "Gateway" are 1,
"Sentence" is 2, "like Japanese" and "between words" are 5
Becomes

FIG. 1 is a display example for explaining the invention described in claim 1. As shown in the drawing, the substructures of the respective constituent elements are recursively placed in parentheses. That is, the substructure of the constituent element of the head of the example sentence, which is "is", "in the processing of a sentence in an agglutinated language that does not leave a break between words such as Japanese,"
“Morphological analysis” and “first barrier” are placed in parentheses, and further, the substructures of the constituent elements of the head of each substructure are recursively placed in parentheses. In "Daiichi Kanmon", the substructure "Daiichi" of the head "Kanmon" is placed in parentheses, and "In the processing of agglomerative language sentences with no breaks between words like Japanese," , The substructure of the head "in processing", "of a sentence in an agglutinative language with no breaks between words such as Japanese", is placed in parentheses, and the substructures are recursively placed in parentheses below. . FIG. 1 is a display example using "[", "]" as a pair of parentheses.

FIG. 2 is a display example for explaining the invention described in claim 2. In the invention of claim 1, when there are a plurality of substructures arranged in parallel with each other, they are placed in parentheses. However, in the invention of claim 2, a plurality of parallel substructures are arranged in parentheses. Three parallel substructures of the head of the example sentence, "is", "in the processing of a sentence in an agglutinative language without breaks between words such as Japanese,""morphologicalanalysis," and "first barrier." ”Are placed within a single bracket. Even in the substructure “without breaks between words like Japanese”, three parallel substructures “like Japanese”, “between words”, and “breaks” of the head “do not put” , Put them in one bracket. FIG. 2 is a display example in which “|” is used as a separator when arranging the substructures.

FIG. 3 and FIG. 4 are display examples for explaining the invention described in claim 3, which is a structure lower than the designated component in addition to the display of the invention described in claim 1 or 2. Suppress or substitute the display of the components of. For example, if "glue language" and "gateway" are specified among the constituent elements, the substructures "there is no break between words like Japanese" and "first" are suppressed or displayed. Will be replaced. Although various methods of suppressing or displaying alternatives can be considered, for example, in FIG.
4 shows an example of a display suppressed by, and FIG. 4 shows an example of a substitute display with reduced characters.

5 and 6 are display examples for explaining the invention described in claim 4, which is lower than the depth of the designated hierarchy in addition to the display of the invention described in claim 1 or 2. Suppress or replace the display of structural components. For example, if "1" is specified as the depth of the hierarchy,
The components of the depth 1 of the hierarchy are "In processing,""Morphologicalanalysis," and "Kanmon," which are substructures such as "of a stale language sentence that does not break between words like Japanese" and "No. The display of "one" is suppressed or replaced. According to the invention of claim 4, instead of designating the component of claim 3, the depth of the hierarchy of the component is designated. Note that FIG. 5 is a display example in which constituent elements are suppressed and displayed by “..”, and FIG.

FIG. 7 and FIG. 8 are display examples for explaining the invention described in claim 5. In the present invention, in the tree structure display, the structure lower than the designated component is
Instead of the tree structure display, the hierarchical display described in claims 1 to 4 is performed. FIG. 7 shows a tree structure display.
This is a display example in which the constituent elements "sentence" and "gateway" are designated, and the structure lower than each constituent element is displayed by the hierarchical display according to claim 2. FIG. 8 shows the hierarchical display of FIG.
4. Designating the constituents "glued language" and "gateway", and claim 3
This is a display example in which the hierarchical display described in (1) is performed and is suppressed by ".

FIGS. 9 and 10 are display examples for explaining the invention described in claim 6. In the present invention, in the tree structure display, the tree structure is displayed for the structure lower than the depth of the designated hierarchy. Instead of the display, the hierarchical display according to claims 1 to 4 is performed. This is described in the above-mentioned JP-A-4-3484.
Although it is similar to the tree structure of FIG. 13 in Japanese Patent Laid-Open No. 64, the structure below the designated depth is not a solid display as in FIG. 13 but a hierarchical display. Figure 9
It is a display example by the hierarchy display according to claim 2, wherein the depth of the hierarchy in the tree structure display is 1. FIG. 10 shows a display example by the hierarchy display of claim 4 (FIG. 5), where the hierarchy depth in the hierarchy display is 1.

A method for realizing the display of the above configuration analysis result will be described below. Table 1 shows an example of the dependency parsing result between clauses, but in the following, the data of each component will generally take the form of a number, a notation fragment, or a modified number.

(1) In the invention of claim 1, processing when the number of head component data is n
(N) 'is shown below. The display of FIG. 1 can be performed by displaying the return value of the process'process 1 (n) '. [Process 1 (n)] After the work area (n) is initialized, the component data are sequentially taken out one by one from the number 1, and if the number of the component data is n, the notation fragment of the component data If the modification destination number is n, the return value of'process 1 (i) 'for the number i of the component data of the modification destination number n is enclosed in "[", "]" in the work area (n). Add and finally return work area (n).

The process of obtaining the display of FIG. 1 will be described below with respect to the invention of claim 1. For example, in the above-mentioned example sentence, when the number 11 of the component data is used as the head, it is performed by "Process 1 (11)". In "Processing 1 (11)",
After the work area (11) is initialized, the constituent element data is sequentially fetched from the number 1, and at the time of the number 7, the modified number is 11, so the return value of the “process 1 (7)” is “[ The value enclosed by "" and "]" is added to the work area (11). In'process 1 (7) ', after initializing the work area (7) and extracting the component data in order from the number 1, the modified destination number is 7 at the time of the number 6, so the'process 1 ( 6) 'The return value enclosed in "[" and "]" is added to the work area (7).

In the'process 1 (6) ', after the work area (6) is initialized, the component element data is taken out in order from the number 1, and when the number is 5, the modified number is 6,
A value enclosed by "[" and "]" is added to the work area (6) as the return value of "process 1 (5)". In'process 1 (5) ', after initializing the work area (5) and extracting the component data sequentially from the number 1, the modified number is 5 at the time of the number 4, so the'process 1 ( 4) 'return value is "["
And the value enclosed in "]" is added to the work area (5). In'Process 1 (4) ', after initializing the work area (4) and extracting the component data in order from the number 1, the modification target number is 4 at the time of the number 1, so the'process 1 ( The return value of 1) 'is enclosed in "[" and "]" and added to the work area (4).

In the "process 1 (1)", after the work area (1) is initialized, the constituent element data is sequentially taken out from the number 1, and when the number 1 is the own number, the notation fragment " Add "like Japanese" to work area (1). In the following, in “Processing 1 (1)”, there is no constituent element data having the number 1 as the modification destination number, so the value “like Japanese” is returned. As a result, in'Processing 1 (4) ',
Add "[like Japanese]" to work area (4).

Further, in the "process 1 (4)", when the number is 2, the modification number is 4, so the value enclosed in "[" and "]" is the return value of the "process 1 (2)". Add to work area (4). In "process 1 (2)", "between words" is returned and "[between words]" is added to the work area (4), as in "process 1 (1)". Furthermore, in'Process 1 (4) ',
When the number is 3, the modified number is 4, so'Process 1
A value enclosed by "[" and "]" for the return value of (3) 'is added to the work area (4). In "Process 1 (3)", "Process 1"
Similar to (1) ', "break" is returned and "[break]" is added to the work area (4). In addition, 'Process 1
In (4) ', when the number is 4, since it is its own number, the notation fragment "not put" is added to the work area (4). In the following, in "Processing 1 (4)", there is no component data whose modification number is number 4, so "[like Japanese] [between words] [breaks] in work area (4) is placed. Returns the value "not".

As a result, in the "process 1 (5)",
Add to [[[like Japanese]] [no word] [no break]] work area (5). In addition, 'Process 1
In (5) ', there is no component data whose modification number is number 4, so "[[like Japanese]] in work area (4)
A value of "no gap between words" and "no breaks" is returned in the stale language. In the same manner, "process 1 (11)"
Then, at the time of the number 7, "[[[[[like Japanese]]
In the processing of [spacing], [no breaks] [spacing], []] is added to the work area (7).

Further, in "Processing 1 (11)", the number 8
, The modified number is 11, so'Process 1
A value enclosed by "[" and "]" is added to the work area (11) as the return value of (8) '. In "Process 1 (8)", "for morphological analysis" is returned and "[for morphological analysis]" is added to the work area (11). Furthermore, in the “process 1 (11)”, when the number is 10, the modification number is 11, so
A value enclosed by "[" and "]" is added to the work area (11) as the return value of the "process 1 (10)". In'Process 1 (10) ', after the work area (10) is initialized, the constituent element data is sequentially fetched from number 1, and at the time of number 9, the modified number is 10, so'process 1 ( The return value of 9) 'is enclosed in "[" and "]" and added to the work area (10). In "Process 1 (9)", "First" is returned,
Add "[first]" to work area (9).

Further, in the "process 1 (10)", the number 1
When it is 0, since it is its own number, the notation fragment "barrier" is added to the work area (10). In the following, in "Processing 1 (10)", there is no component data whose modification number is number 10, so the value "[first] barrier" of the work area (10) is returned. As a result, in "Processing 1 (11)",
"[[First] barrier]" is added to the work area (11).
Further, in the "processing 1 (11)", when the number is 11, since it is the own number, the notation fragment "is" is the work area (1
Add to 1). In the following, in "Processing 1 (11)", there is no component data having the number 11 as the modification destination number, so the value of the work area (11) as shown in FIG. 1 is returned.

(2) The invention of claim 2 shows a case where a plurality of parallel substructures are arranged with "|" as a delimiter. In the example described in (1) above, as shown in FIG.
It can be immediately obtained by replacing the character string of "] [" with "|" in the character string of the return value of the process "process 1 (n)".

(3) With respect to the invention of claim 3, the designated component, that is, the component designated for the purpose of suppressing or substituting the display of the component in the structure lower than the designated component Assuming that the list of the numbers is given by L, the process'process 1X (n, L) 'when the number of the constituent element of the head is n is shown below. By displaying the return value of the process'process 1X (n, L) 'or the value obtained by replacing “] [” of the character string with “|”,
The invention of claim 3 can be displayed. If no component is specified, the number list L is empty.

[Processing 1X (n, L)] The number list L
If is empty, the return value of'Process 1 (n) 'is returned.
If the number list L is not empty, after the work area (n) is initialized, the component data are sequentially taken out one by one from the number 1, and if the component data number is n, whether L is non-negative or not. If the modification destination number of the constituent element data is n, the notation fragment or the suppression of the notation fragment or the alternative expression value is'process 1 for the number i of the constituent element data whose modification destination number is n.
A value in which the return value of X (i, L1) 'is enclosed in "[", "]" is added to the work area (n), and finally the work area (n) is returned. However, the L1 is "-" depending on whether n is included in the number list L or L is "-1".
1 "or L. Here, the above" inhibition of notation fragment or alternative expression value "means". ． "Display and reduced display of expression fragments.

With respect to the description of claim 3, the process of obtaining the display of FIG. 3 or 4 will be described below. For example, if you specify the components of "glued language" and "gateway",
The number list becomes “[5,10]”, and “Process 1
X (11, [5,10]) 'will be executed. The process itself is almost the same as the case of'process 1 (11) ', but in'process 1X (5, [5,10])', 'process 1X (4, -1)' is executed. 'Process 1X
(1, -1) ',' Process 1X (2, -1) ',' Process 1
Since X (3, -1) 'is executed, the numbers 1, 2, 3,
The notation fragment portion of 4 replaces the suppression or alternative expression value. Also, in the “process 1X (10, [5,10])”, the “process 1X (9, −1)” is executed, so that the notation fragment portion of the number 9 replaces the suppression or alternative expression value. Therefore, the display is as shown in FIG. 3 or FIG. However, FIG. 3 or FIG. 4 shows an example in which they are arranged by “|” described in (2) above.

(4) With respect to the invention of claim 4, it is possible to suppress or substitute the display of the component in the structure lower than the component of the designated layer depth, that is, the component of the designated layer depth. Assuming that the depth of the designated hierarchy is given, the processing where the number of the constituent element of the head is n and the depth of the designated hierarchy is D'Process 1Y
(N, D) 'is shown below. First, as a preparation, the depth of the hierarchy from the head is obtained for each component data of the syntax analysis result. The process “process A” and its related processes “process A1 (i)” and “process A2 (j)” for this purpose are shown below.

[Processing A] The constituent element data is sequentially taken out one by one, and when the depth of the hierarchy is not set, the following processing 1'processing A1 (i) 'is performed for the number i. [Processing A1 (i)] Regarding the component data of number i,
If the modification destination number of the component data is not set,
If the modification destination number of the component data is set to “0”, a value obtained by adding 1 to the return value of the process “process 2A (j)” described below for the modification destination number j is added to the depth of the hierarchy. Set it as S and return it. [Processing A2 (j)] Regarding the component data of number j,
If the layer depth of the component data is set, the value is set, and if the layer depth of the component data is not set, the return value of the process'process A1 (j) 'is returned. return.

The process of the process'process A'in the example sentence will be described below. For example, in the'process A ', if the component data is sequentially fetched from the number 1, first, the'process A1
(1) 'is executed. In "Processing A1 (1)", number 1
Since the modification number of the data of is "4,"
The value obtained by adding 1 to the return value of 2 (4) 'is set as the depth of the layer of number 1, and that value is returned. In “Process A2 (4)”, the depth of the hierarchy of the data of number 4 has not been set, so “Process A1
(4) 'is executed and the returned value is returned as it is.
In the "process A1 (4)", the modification number of the data of the number 4 is 5, so the value obtained by adding 1 to the return value of the "process A2 (5)" is defined as the depth of the layer of the number 1, and Returns the value.

In the "process A2 (5)", since the depth of the hierarchy of the data of number 5 is not set, the "process A1 (5)" is executed and the return value is returned as it is. 'Process A1
In (5) ', the modification number of the data of number 5 is 6, so the value obtained by adding 1 to the return value of'process A2 (6)' is taken as the depth of the layer of number 1, and that value is returned. To do. In the "process A2 (6)", the depth of the hierarchy of the data of number 6 is not set, so the "process A1 (6)" is executed and the return value is returned as it is. In the "process A1 (6)", the modification number of the data of the number 6 is 7, so the "process A2"
The value obtained by adding 1 to the return value of (7) 'is set as the depth of the layer of number 1, and the value is returned.

In the "process A2 (7)", since the depth of the hierarchy of the data of number 7 is not set, the "process A1 (7)" is executed and the return value is returned as it is. 'Process A1
In (7) ', the modification number of the data of number 7 is 11, so the value obtained by adding 1 to the return value of'Process A2 (11)' is taken as the depth of the layer of number 1, and that value is returned. To do.
In “Process A2 (11)”, the depth of the hierarchy of the data of number 11 is not set, so “Process A1 (11)” is executed,
The returned value is returned as it is. "Process A1 (11)"
Then, since the modification number of the data of number 11 is not set,
The value 0 is set as the depth of the hierarchy, the value 0 is added to the return value, and the value is set as the depth of the hierarchy of number 1, and the value is returned. Therefore, 'Processing A2 (11)' returns 0.

As a result, in the "process A1 (7)",
The value 1 obtained by adding 1 to the return value 0 of'process A2 (11) 'is taken as the depth of the hierarchy, and the return value 1 is received, and'process A2
(7) 'returns 1. As a result, 'Process A1
In (6) ', 2 is obtained by adding 1 to the return value 1 of'Process A2 (7)' is the depth of the hierarchy, and upon receipt of the return value 2, 'Process A2 (6)' returns 2. As a result, 'Process A
In 1 (5) ', the return value 2 of'process A2 (6)' plus 1 is set to 3 as the depth of the hierarchy, and the return value 3 is received,
'Process A2 (5)' returns 3.

As a result, in the "process A1 (4)",
The value of 4 obtained by adding 1 to the return value 3 of'process A2 (5) 'is set as the depth of the hierarchy, and the return value 4 is received, and'process A2 (4)' is received.
Returns 4. As a result, in the “process A1 (1)”, 5 is obtained by adding 1 to the return value 4 of the “process A2 (4)” and is set as the layer depth. If the depth of the hierarchy is set in the fourth column of the component data by the following, the following data will be obtained.

[0041]

[Table 2]

Next, the number 2 of the component data is taken out and the "process A1 (2)" is executed. 'Process A1
In (2) ', the modification number of the data of number 2 is 4, so the value obtained by adding 1 to the return value of'process A2 (4)' is taken as the depth of the layer of number 1, but In A2 (4) ', the depth of the hierarchy of the data of number 4 has already been set to 4, so 4 is returned, and accordingly,' processing A1
In (2) ′, 5 obtained by adding 1 to it is set as the depth of the hierarchy. As a result, the data of number 2 is updated as follows.

[0043]

[Table 3]

Next, the process'process 1Y (n, D) 'when the depth D of the designated layer is given and the number of the constituent element of the head is n is shown below. The display according to claim 4 can be performed by displaying the return value of the process'process 1Y (n, D) 'or the value obtained by replacing “] [” in the character string with “|”. . If the layer depth is not designated, the designated layer depth D is empty.

[Processing 1Y (n, D)] Depth D of the layer
If is empty, the return value of'Process 1 (n) 'is returned.
If the depth D of the hierarchy is not empty, the work area (n) is initialized, and then the component data are sequentially taken out one by one from number 1. If the number of the component data is n, D is non-negative. If the modification destination number of the constituent element data is n, then the suppression or alternative expression value of the notation fragment is processed for the number i of the constituent element data whose modification destination number is n.
A value obtained by enclosing the return value of (i, D1) 'in "[", "]" is added to the work area (n), and finally the work area (n) is returned. However, the D1 is set to D or "-1" depending on whether or not the hierarchical depth of the component data having the number i is D.
Here, as in the case of (3), the "inhibition of expression fragment or alternative expression value" is, for example, ".." Display or reduced display of expression fragment.

The process of obtaining the display of FIG. 5 or 6 will be described below with respect to the invention of claim 4. For example, if "1" is specified as the depth of the hierarchy, 'Process 1Y
(11,1) 'will be executed. The process itself is
Almost the same as the case of'Processing 1 (11) ',
In "Processing 1Y (7,1)", "Processing 1Y (6,-
1) 'is executed, then'Process 1Y (5, -1)' is executed, then'Process 1Y (4, -1) 'is executed, and further,' Process 1Y (1, -1) 'is executed. ',' Processing 1Y
Since (2, -1) 'and'Process 1Y (3, -1)' are executed, the notation fragment part of the numbers 1, 2, 3, 4, 5, 6 replaces the suppression or alternative expression value. . In addition, even in “Processing 1Y (10,1)”, “Processing 1Y (9, −1)”
Is executed, the notation fragment portion of number 9 replaces the suppression or alternative representation value. Therefore, the display is as shown in FIG. 5 or FIG. However, an example is shown in which the lines are arranged by "|" described in (2) above.

(5) Next, in describing an embodiment of the invention of claim 5, the outline thereof will be described first. According to the invention of claim 5, based on the tree structure display,
The hierarchical display described in 4 is partially used. Therefore, first, in preparation for processing of displaying a tree structure, in each component data table 3 of the syntax analysis result to which the depth of the hierarchy is added,
The process'process B'for further adding the component number which directly modifies from the frontmost and the rearmost is shown. Next, the processing “processing 2” for displaying the tree structure will be described. Next, in the tree structure, a process for displaying the structure below the designated component by the hierarchical display means according to any one of claims 1 to 4, that is, 'process 2X (L)' and its related process. I will show you.

First, the process "process B" as the preparation described above is shown below. [Processing B] The constituent data is taken out one by one from the number 1, and if the modification destination number M of the constituent data of the number N is empty, nothing is done. If the modified number M is not empty,
When the number N is smaller than the modification destination number M, and the most forward modification destination number of the component data of the number M is not set, N is set to the most advanced modification source number of the component data of the number M. However, when the number N is larger than the modification destination number M and the rearmost modification source number of the component data of the number N is not set, N is added to the rearmost modification source number of the component data of the number M. Set. However, in the above case, if the frontmost or rearmost modification source number is set, nothing is done.

Hereinafter, the process “process B” in the above example sentence
Shows the process of. For example, in the case of extracting the component data from the number 1 in'Process B ', first, the number 1 is smaller than the modification destination number 4, and the most front modification source number of the component of the number 4 is not set. Therefore, 1 is set to the most forward modification source number of the component number 4. Next, the number 2
Although it is smaller than the modification destination number 4, the modification source number at the forefront of the constituent element of number 4 has already been set, so nothing is done. Similarly, in case of the number 3, nothing is done. Next, the number 4 is smaller than the modification destination number 5, and the foremost modification source number of the component of number 5 is not set, so 4 is set for the most advanced modification source number of the component of number 5. To do. Thereafter, in the same manner, the following data is obtained. However, the frontmost modification source number is set in the fifth column, and the rearmost modification source number is set in the sixth column. If there is no corresponding modification source, empty shall be set.

[0050]

[Table 4]

Note that, in the dependency relation between Japanese phrases, there is only a backward modification, and therefore there is no backward modification source as shown in Table 4.

The preparation for obtaining the modification source range has been described above.
Next, the above-mentioned tree structure display processing'processing 2'is shown. [Processing 2] Column arrangement [1] of depth N of maximum hierarchy
After initializing [N] with "" (blank), the constituent element data is sequentially taken out one by one, and the processing is performed by the following procedure. Step 1. The column arrays [1] to [d-1] are sequentially output for the depth d of the hierarchy of the component data. However, when d is "1", nothing is output. Step 2. If the number of the component data is smaller or larger than the modification number of the component data, the following processing is performed. 1. In case of small If the column array [d] is not “┃”, “┏” is output and “┃” is set in the column array [d]. If the column array [d] is “┃”, “┣” is output. 2. In the case of a large number, if it is smaller than the rearmost modification source number, "┣" is output,
Set “┃” in the column array [d]. If it is the same as the rearmost modification source number, “┗” is output and “” is set in the column array [d]. Step 3. The notation fragment of the component data is output. Step 4. Further, the column arrays [d + 1] to [N] are replaced with "".
To re-initialize with. However, in the above procedure, if d is "0", steps 1 and 2 are not performed.

The process of obtaining the display of FIG. 19 will be described below with respect to the invention of claim 5. For example, in the data in Table 4 above, after the column array is initialized, the component data of number 1 is sequentially extracted, and the depth of the layer is 5, so
Output the first 4 column arrays (4 blanks). Number 1
Is smaller than the modification number 4 and the column array [5] is blank, so “┏” is output and “┃” is set in the column array [5]. Furthermore, the notation fragment “like Japanese” is output.
As described above, the display on the first line is obtained as shown in FIG. In the following component data of number 2, since the depth of the layer is 5, the first 4 column arrays (4 blanks) are output. The number 2 is smaller than the modification number 4, and the column array [5] is "┃", so "┣" is output. Further, the notation fragment “between words” is output. Further, the component data of number 3 is processed in the same manner. By the above, the display up to the third line is obtained as shown in FIG.

Next, in the component data of number 4, since the depth of the layer is 4, the first three column arrays (three blanks) are output. Number 4 is less than its modification number 5,
Since the column array [4] is blank, “┏” is output and “┃” is set in the column array [4]. Further, the notation fragment “not placed” is output, and the column array [5] is set to blank. By the above, the display on the fourth line is obtained, and similarly, the display on the fifth and sixth lines is obtained. As described above, as shown in FIG. 17, the display up to the sixth line can be obtained.

In the following component data of No. 7, since the depth of the layer is 1, no column array is output first.
The number 7 is smaller than the modification number 11, and the column sequence [1]
Is blank, so output “┏” and set “┃” in the column array [1]. Furthermore, the notation fragment “in processing” is output, and the column arrays [2] to [5] are set to blanks. By the above, the display of the 7th line is obtained. The same applies to the component data of number 8, but the column array [1]
Is “┃”, “┣” is output and “morphological analysis is” is output, as shown in FIG.

Further, in the component element data of number 9, since the depth of the hierarchy is 2, first, "┃" of the column array [1]
Is output. Since the number 9 is smaller than the modification number 10 and the column array [2] is blank, “┏” is output and “┃” is set in the column array [2]. Furthermore, the notation fragment “first” is output, and the column arrays [3] to [5] are set to blanks. As described above, the display on the ninth line is obtained. The 10th line by the number 10 is obtained in the same manner as the 8th line, and the 11th line by the number 11 immediately outputs the notation fragment "is", and finally the entire tree structure as shown in FIG. Can be obtained.

Next, in the tree structure described above, a process for displaying by the hierarchy display means, that is,
The'process 2X (L) 'and its related process are shown below. In the tree structure, a component for displaying a structure below the designated component by the hierarchical display means is designated. Here, it is assumed that the list of the numbers of the designated components is given by L.

[Processing 2X (L)] Initialize the column arrays [1] to [N] for the depth N of the maximum hierarchy with "" (blank), and then sequentially take out the component data one by one. For the number i of each component data, if the following process'process 3X (i, L) 'is "0", the process is performed according to the following procedure, and if "1", nothing is output. . Step 1. The column arrays [1] to [d-1] are sequentially output for the depth d of the hierarchy of the component data. However, when d is "1", nothing is output. Step 2. If the number of the component data is smaller or larger than the modification number of the component data, the following processing is performed. 1. In case of small If the column array [d] is not “┃”, “┏” is output and “┃” is set in the column array [d]. If the column array [d] is “┃”, “┣” is output. 2. In the case of a large number, if it is smaller than the rearmost modification source number, "┣" is output,
Set “┃” in the column array [d]. If it is the same as the rearmost modification source number, “┗” is output and “” is set in the column array [d].

Step 3. The return value of the process'process 1 (i) 'is output. Step 4. Further, the column arrays [d + 1] to [N] are replaced with "".
To re-initialize with. However, in the above procedure, if d is "0", steps 1 and 2 are not performed. [Processing 3X (i, L)] If j is empty or equal to i according to the modification destination number j of the component data of number i,
"0", "1" if j is in the number list L
Otherwise, the return value of'Process 3X (j, L) 'is returned. The “process 3X (i, L)” is a process of checking whether or not it is a component of the number in the number list L above the component of the number of the first argument.

The process of obtaining the display of FIGS. 7 to 8 or 20 will be described below with respect to the invention of claim 5. For example, if the constituent elements of "sentence" and "gateway" are specified, the number list becomes "[6,10]", and the "process 2"
X ([6,10]) 'will be executed. In the first numbers 1 to 5, since "Processing 3X (i [6,10])" is "1", nothing is output. In the case of number 6, "Processing 3X (6, [6, 10])" is "0", so that it is output for the first time. It is almost the same as the case of the number 6 in the case of'processing 2 ', but instead of the notation fragment,' processing 1
(6) 'return value "[[[like Japanese | between words |
The output of the sentence "]" in the stuck language] is output.

Numbers 7 and 8 are'process 3X (i, [6, 1
"0])" is "0", "process 1 (7)" and "process 1 (8)" are output in the same manner as "process 2". Number 9
"Process 3X (9, [6,10])" is "1", nothing is output. At number 10, 'Process 3X (1
Since 0, [6, 10]) 'is "0", it is output. Although it is almost the same as the case of the number 10 in the case of the "process 2", the return value "[first] barrier" of the "process 1 (10)" is output instead of the notation fragment. The number 11 is “0” in “Processing 3X (11, [6,10])”.
"Processing 1 (11)" is output in the same manner as "Processing 2". By the above, the display as shown in FIG. 25 is obtained.
Here, an example is shown in which “] [” of the return value of “Processing 1” is replaced with “|”.

Further, in the hierarchical display, when a component is designated as in claim 3, the list of the numbers of the designated components is set to L0, and the above'Process 2X
Instead of (L) ',' Process 2XX (L, L0) '
Instead of the process “process 1 (i)” described in the procedure 3, the process “process 1X (i, L0)” described in (3) is performed.
As an example of this case, in addition to the above example, if "glue language" and "gateway" are designated as the designated components in the hierarchical display, L0 is "[5, 1
0] ”and the display is as shown in FIG.
FIG. 8 shows an example of the suppression display by "..".

Further, in the hierarchical display, when the depth of the hierarchy is specified as in claim 4, the depth of the specified hierarchy is set to D0, and instead of the above-mentioned'Process 2X (L) ', As'process 2XY (L, D0) ', instead of the process'process 1 (i)' described in step 3, the process' process described in (4) is performed for the depth X of the hierarchy of the component number i. Process with 1Y (i, X + D0) '. As an example of this case, if the depth "1" of the hierarchy is specified in the hierarchy display in addition to the above example, the display is as shown in FIG. Note that FIG. 20 shows an example of the suppression display by "..".

(6) In the tree structure, the component of the depth of the designated hierarchy (the processing for displaying the following structure by the hierarchy display means, that is, the'processing 2Y (L) 'and its related processing is Shown below) Here, the procedure of the process'process 2Y (L) 'when the depth of the designated hierarchy is given by D will be shown.

[Processing 2Y (D)] After initializing the column arrays [1] to [N] for the depth N of the maximum hierarchy with "" (blank), the component data are sequentially taken out one by one, With respect to the number i, if "Processing 3Y (i, D)" is "0", processing is performed according to the following procedure, and if "1", nothing is output. Step 1. The column arrays [1] to [d-1] are sequentially output for the depth d of the hierarchy of the component data. Step 2. If the number of the component data is smaller or larger than the modification destination number of the component data, the following processing is performed. 1. In the case of small If the column array [d] is not “┃”, “┏” is output and “┃” is set in the column array [d]. Column arrangement [d] is "┃"
If so, "┣" is output. 2. In the case of a large number, if it is smaller than the rearmost modification source number, "┣" is output,
Set “┃” in the column array [d]. If it is the same as the rearmost modification source number, “┗” is output and “” is set in the column array [d].

Step 3. Process described in (1) above'Process 1
Output the return value of (i) '. Step 4. Further, the column arrays [d + 1] to [N] are replaced with "".
To re-initialize with. However, in the above procedure, if d is "0", steps 1 and 2 are not performed. [Process 3Y (i, D)] Depending on the depth d of the layer of number i, if d is larger than D, "1"
Otherwise, "0" is returned. 'Process 3Y (i,
D) 'is the depth of the hierarchy of the component of the number of the first argument,
This is a process of checking whether the depth is deeper than the designated depth D.

The process of obtaining the display of FIG. 12 or 13 will be described below with respect to the invention of claim 6. The process itself is almost the same as the process described in (5) above. For example, if the depth of the hierarchy is designated as "1", the display as shown in FIG. 12 is obtained. In addition to this, as in the case of (5) above, by using “Processing 1X” or “Processing 1Y” instead of “Processing 1”, a display such as that shown in FIG. 13 can be obtained. it can. As described above, as an example, the dependency analysis result between Japanese phrases was used.
The processing of each claim can be made more efficient by taking advantage of the property of only backward modification, but they are not shown here.

[0068]

As is apparent from the above description, the present invention has the following effects. (1) In the inventions of claims 1 and 2, the syntactic structure can be shown as the display of the syntactic analysis result, and since it is the same one-dimensional display as the original sentence, even a long sentence can be displayed without a problem of display space. (2) In the invention of claim 2, it is expected that the structure of the same layer can be easily seen. (3) In the inventions of claims 3 and 4, in addition to the effects of claims 1 and 2, by making the substructure inconspicuous, the skeleton part of the syntax analysis result can be displayed in an easy-to-understand manner. (4) In the inventions of claims 5 and 6, in addition to the tree structure display that is easy to understand as the syntactic analysis result, the skeleton part of the syntactic analysis result can be easily displayed by the one-dimensional display of claims 1 to 4 for the substructure. .

[Brief description of drawings]

FIG. 1 is a diagram showing a display example for explaining an embodiment of a syntax analysis result display system according to the present invention.

FIG. 2 is a diagram showing a display example of the invention of claim 2;

FIG. 3 is a diagram showing a display example of the invention of claim 3 using a constituent element inhibition display.

FIG. 4 is a diagram showing a display example of the invention of claim 3 using a component alternative display.

FIG. 5 is a diagram showing a display example of the invention of claim 4 using a constituent element inhibition display.

FIG. 6 is a diagram showing a display example of the invention of claim 4 using a component alternative display.

FIG. 7 is a diagram showing a display example of the invention of claim 5 using the hierarchical display according to claim 2;

FIG. 8 is a diagram showing a display example of the invention of claim 5 using the hierarchical display according to the invention of claim 3;

FIG. 9 is a diagram showing a display example of the invention of claim 6 using the hierarchical display according to the invention of claim 2;

FIG. 10 is a diagram showing a display example of the invention of claim 6 using the hierarchical display according to the invention of claim 4;

FIG. 11 is a diagram showing a Japanese original sentence.

FIG. 12 is a diagram showing a display example of a dependency tree structure display.

[Fig. 13] Fig. 13 is a diagram showing a display example of a dependency tree structure display of a depth 1 of a hierarchy.

FIG. 14 is a diagram showing a process of processing for the invention of claim 2;

FIG. 15 is a diagram for explaining a process of processing according to the invention of claim 5;

FIG. 16 is a diagram for explaining a process of processing according to the invention of claim 5;

FIG. 17 is a diagram for explaining a process of processing according to the invention of claim 5;

FIG. 18 is a diagram for explaining a process of processing according to the invention of claim 5;

FIG. 19 is a diagram for explaining a process of processing of the invention of claim 5;

FIG. 20 is a diagram showing a display example of the invention of claim 5 using the hierarchical display according to the invention of claim 4;

FIG. 21 is a diagram showing a dependency relationship.

FIG. 22 is a diagram showing a syntax analysis tree.

FIG. 23 is a diagram showing a semantic network representation.

FIG. 24 is a diagram showing a hierarchical sectional view by underlining.

FIG. 25 is a diagram showing another hierarchical sectional view by underlining.

FIG. 26 is a diagram showing a dependency structure diagram.

[Fig. 27] Fig. 27 is a diagram illustrating a layer display of a layer depth of 2.

Claims (6)

[Claims] 1. A syntactic analysis result display system for displaying a hierarchical structure between constituent elements of a natural language sentence as a syntactic analysis result of a natural language sentence, wherein substructures of each constituent element are recursively placed within a pair of parentheses. A method of displaying a syntactic analysis result, characterized by having hierarchical display means for displaying. 2. A syntactic analysis result display method for displaying a hierarchical structure between constituent elements of a natural language sentence as a syntactic analysis result of a natural language sentence, wherein a plurality of substructures of the same hierarchy of each constituent element are arranged side by side within a pair of parentheses. A method of displaying a syntactic analysis result, characterized in that it has a hierarchical display means for displaying in the. 3. In a syntax analysis result display method for displaying a hierarchical structure between constituent elements of a natural language sentence as a result of syntax analysis of a natural language sentence, constituent element instruction receiving means for receiving designation of said constituent element, and said A method for displaying a syntactic analysis result, comprising hierarchical display means for suppressing or substituting the display of the constituent elements in the structure lower than the constituent element designated by the constituent element instruction receiving means. 4. In a syntax analysis result display method for displaying a hierarchical structure between constituent elements of a natural language sentence as a syntactic analysis result of a natural language sentence, a depth instruction acceptance for accepting designation of a depth of the hierarchy in the hierarchical structure. A syntactic analysis result display system comprising: a means and a hierarchical display means for suppressing or substituting the display of the constituent elements in the structure lower than the constituent element of the depth designated by the depth instruction receiving means. 5. A tree structure display means for displaying a hierarchical relationship between constituent elements in a tree structure in a syntax analysis result display method for displaying a hierarchical structure between constituent elements of a natural language sentence as a result of syntax analysis of a natural language sentence. The hierarchy display means according to any one of claims 1 to 4 (for the hierarchy display means according to claim 3, the component instruction receiving means, and for the hierarchy display means according to claim 4, the depth Instruction receiving means),
And a structure element instruction receiving means for receiving specification of a component, wherein the tree structure display means displays, by the hierarchy display means, a structure below the structure element instructed by the structure element instruction receiving means. Parsing result display method. 6. A syntactic analysis result display method for displaying a hierarchical structure between constituent elements of a natural language sentence as a syntactic analysis result of a natural language sentence, and a tree structure display means for displaying a hierarchical relationship between the constituent elements in a tree structure. The hierarchy display means according to any one of claims 1 to 4 (for the hierarchy display means according to claim 3, the component instruction receiving means, and for the hierarchy display means according to claim 4, the depth Instruction receiving means),
And a depth instruction receiving means for receiving designation of the depth of the hierarchy in the hierarchical structure, wherein the tree structure display means is configured by the depth instruction receiving means to display a structure below a depth component by the hierarchy display means. A parsing result display method characterized by displaying.