IPSJ-shinmori final.dvi

Size: px

Start display at page:

Download "IPSJ-shinmori final.dvi"

ぜんすけよしくに
5 years ago
Views:

1 Vol. 45 No. 3 Mar NTCIR3 Structure Analysis of Japanese Patent Claims using Cue Phrases Akihiro Shinmori, 1 Manabu Okumura, 2 Yuzo Marukawa 2 and Makoto Iwayama 4 The most important part of patent specification documents is where the claims are written. It is common that claims written in Japanese are described in one sentence with peculiar style and are difficult to understand for ordinary people. In this paper, we propose a method to analyze the structure of patent claims for the purpose of improving readability. First, based on the result of literature survey and preliminary study, we propose a framework to represent the structure of patent claims. Next, based on this framework we propose a method to automatically analyze the structure of patent claims. This method utilizes several cue phrases which are used in patent claim descriptions. By using the NTCIR3 patent data collection, we show the effectiveness of our proposal through indirect and direct evaluation INTEC Web and Genome Informatics Corp. 2 Precision and Intelligence Laboratory, Tokyo Institute of Technology 3 CREST Core Research for Evolutional Science and Technology, Japan Science and Technology Agency 4 Precision and Intelligence Laboratory, Tokyo Institute of Technology / Hitachi, Ltd. 40 1) 5) 1

2 2 Mar NTCIR3 2) (1) (2) (3) NTCIR ,968 18) IPA Fig. 1 A sample Japanese patent (Publication Number= ) (259 characters) Fig. 2 A sample Japanese patent (Publication Number= ) (304 characters) ) ),5) 6)

3 Vol. 45 No ) 3. 4) 5 4) Jepson 4) Jepson Jepson 6)7) 6) the Japanese claims drafter seldom intends a true Jepson claim, and the Jepson claims are indeed very rare in US patents 8)9) Markush 1 9) product-by-process 8) means-plus-function function means 9) KNP 10) KNP a) b) a) b)

4 4 Mar ) RST: Rhetorical Structure Theory 11) ) 14) Tcl/Tk RSTTool 15) rhetorical structure analysis segment span rhetorical structure tree rhetorical relation nucleus satellite multi-nuclearmononuclear 5) 4) [ ] XYZ ELABORATION ELABORATION 5) RSTTool 4 4

5 Vol. 45 No Table 1 Structural relations for patent claim PROCEDURE [][][]X COMPONENT [][][] X PRECONDITION [X ][ Y] COMPOSE 1 [][]X FEATURE ( ) 1 [ X][] ELABORATION [X ][Y Z] 3 RSTTool v2.7 Fig. 3 A result of structure analysis of patent claim

6 6 Mar Fig. 4 An example of newline insertion and indentation HTML 16) 5 6 <nl> <BR> <nl> <nl> <nl> Fig. 5 An example of claim which newlines are explicitly inserted <nl> <nl> <nl> Fig. 6 An example of claim which newlines are explicitly inserted 4)5) NTCIR3 48.5% % 2 13) cue phrase

7 Vol. 45 No Table 2 Expression pattern just before newline of the claims in which newlines are explicitly inserted No. 1 ( ) ( ) 46.1% 46.1% 2 ( )( ) 17.5% 63.6% 3 ( )( ()?)( ) 16.5% 80.1% 4 ( ) ( ) 7.2% 87.3% Perl (1) 18) -j (2) 3 JEPSON CUE, FEATURE CUE, COMPOSE CUE, NOUN, POSTP TO, PUNCT TOUTEN, VERB RENYOU, VERB KIHON 2 10 POSTP NO WORD 9 (3) Bison 19) LALR(1) shift/reduce conflict reduce/reduce conflict JEPSON CUECOMPOSE CUE VERB KIHONFEATURE CUE PRECONDITION, COMPO- NENT, PROCEDURE, FEATURE COM- POSE CUE COM- POSE JEPSON CUE ELABORATION 4.3

8 8 Mar Table 3 ( ) ()? ( ) ( ) ( ) ()? ( ) ( )( )? ( ) ( )? ( )?( ) ( ) ( )? ( )? ( ) ( )? ( ) ( ) ( )? ( )? ( ) ( )? ( )? ( ) ( )? ( )? ( ) ( )? ( )( )?( ) ( )? ( )( )?( ) ( )? ( )? ( ) ( )? ( )? ( ) ( )? COMPOSE CUE ( ) ( ) 3 Cue phrases which can be used to analyze patent claims ( ) ( )( ) ( ) ( ) JEPSON CUE FEATURE CUE COMPOSE CUE ( )NOUN POSTP TO ( )PUNCT TOUTEN ( ) ( ) VERB RENYOU ( ) ( ) PUNCT TOUTEN ( ) VERB KIHON (1) 3 JEPSON CUE JEPSON CUE 2 (2) 3 FEATURE CUE FEATURE CUE (3) NOUN POSTP NO ( 4 ) JEPSON CUEFEATURE CUE NOUN POSTP NO (5) 1 (a) ( )( ) NOUN ( b ) COMPOSE CUE ( 6 ) (a) VERB KIHON PUNCT TOUTEN ( )( )( ) ( ) VERB RENYOU PUNCT TOUTEN ( ) VERB RENYOU ( ) PUNCT TOUTEN

9 Vol. 45 No. 3 9 ( 7 ) (b) COMPOSE CUE COMPOSE CUE ( )? ( ) ( ) NOUN POSTP TO PUNCT TOUTEN ( )( ) VERB RENYOU PUNCT TOUTEN (8) NOUN NOUN POSTP NO (9) WORD 5 A rst claim BR RSTTool.rs2 XML 4 7 rst claim RST- Tool 7 rst claim Fig. 7 Positioning of structure analysis program 5. NTCIR3 NTCIR , ,956 59,956 IPCA H 1999 accept % Pentium III 1GHz 512MB

10 10 Mar Linux OS PC 5.1 JEPSON CUE Recall Precision n i c RP F-measure F R = c n P = c i F = 2 R P R + P (1) (2) (3) ( ) ( ) ( )( ) , (1) (2) (3) 2 4 ( )( ) 0.06%

11 Vol. 45 No Table 4 4 Indirect evaluation result on the analysis precision 1 2 R P N/A F-measureF N/A Table 5 5 Direct evaluation result on the analysis precision % 2 2.1% 6 6.3% 5 - (4) Fig. 9 A sample Japanese patent misjudged as Jepsontype (Publication Number= )

12 12 Mar Fig. 8 An example of under-segmentation Fig. 10 Output of newline-insertion and indentation for the claim of KNP 10) 6.2

13 Vol. 45 No % F- measure R F-measure % RSTTool 4) 6) 7. 12)13) 20) 20) 20) 17) 8.

14 14 Mar NTCIR3 21) NTCIR3 1),, :, 7 /, pp , (2001). 2),,, :,, FI , (2001). 3) ,, (1995). 4) :,, (1999). 5) :,, (1999). 6) William Lise: An Investigation of Terminology and Syntax in Japanese and US Patents and the Implications for the Patent Translator. (2002). 7) :,, (1989). 8) : 6,, (1999). 9) :,, (1999). 10) : KNP,, Vol. 41, No. 11, pp , (2000). 11) Bill Mann: An Introduction to Rhetorical Structure Theory (RST), mannb/rst/rintro99.htm, (1999). 12) Daniel Marcu: The Theory and Practice of Discourse Parsing and Summarization, MIT Press, (2000). 13) Daniel Marcu: The Rhetorical Parsing of Unrestricted Texts: A Surface-based Approach, Computational Linguistics, Vol. 26, No. 3, pp , (2000). 14) John Bateman, Thomas Kamps, Jorg Kleinz, Klaus Reichenberger: Toward Constructive Text, Diagram, and Layout Generation for Information Presentation, Computational Linguistics, Vol. 27, No. 3, pp , (2000). 15) Michael O Donnell: RST-Tool: An RST Analysis Tool, Proceedings of the 6th European Workshop on Natural Language Generation, (1997). 16) 2, ) Svelana Sheremetyeva, Sergey Nirenburg: Knowledge Elicitation for Authoring Patent Claims, IEEE Computer, July 1996, pp , (1996). 18),,, Vol.41, No.11, ) Charles Donnelly, Richard Stallman: Bison: The YACC-compatible Parser Generator, Version 1.25, (1995). 20) : QJR,, NL-110-9, pp , (1995). 21) :,, (1988). A.1 NOUN WORD WORD WORD... WORD NOUN NOUN POSTP_TO PUNCT_TOUTEN WORD WORD... WORD NOUN NOUN NOUN POSTP_TO

15 Vol. 45 No PUNCT_TOUTEN WORD... WORD NOUN NOUN POSTP_TO PUNCT_TOUTEN COMPOSE_CUE WORD WORD NOUN JEPSON_CUE WORD... WORD WORD VERB_KIHON NOUN FEATURE_CUE WORD WORD NOUN ( ) ( ) , AAAIACL, oku@pi.titech.ac.jp, oku/ (CREST) () 2000 () () ACLAAAI 1992 () 2000 () ACM

ID 3) 9 4) 5) ID 2 ID 2 ID 2 Bluetooth ID 2 SRCid1 DSTid2 2 id1 id2 ID SRC DST SRC 2 2 ID 2 2 QR 6) 8) 6) QR QR QR QR

ID 3) 9 4) 5) ID 2 ID 2 ID 2 Bluetooth ID 2 SRCid1 DSTid2 2 id1 id2 ID SRC DST SRC 2 2 ID 2 2 QR 6) 8) 6) QR QR QR QR Vol. 51 No. 11 2081 2088 (Nov. 2010) 2 1 1 1 which appended specific characters to the information such as identification to avoid parity check errors, before QR Code encoding with the structured append