Post on 27-Mar-2015
transcript
Use of Classification at the EPO
Pasquale FogliaDG1 Director, EPO
WIPO, IPC Workshop 5 February 2008
Outline
• Introduction
• IPC at the EPO
• Classification Systems available at EPO
• EPO Search: general methodology and special cases
• Documentation and Citation Statistics
• Wish List and some Inconvenient Truths
• Conclusion
Introduction - myself
• Pasquale Foglia• EPO Examiner and Classifier (15y)• EPO Classification Board Electricity (2000-2006)
• DG1 (Operation) Director in AVM Cluster
Intro (1) - Can one tool do it all ?
Introduction
Intro (2) - One tool ?
Introduction Image: courtesy Wenger
Intro (3) - For best results use specialised tools !
Introduction
IPC at the EPO
• DG1 Structure (Clusters, Directorates)
• Internal distribution of patent applications to Directorates, then technical Teams (using "preclassification", together w ECLA)
• "A2" publications (18 months after PR) of EP applications
• Base for ECLA
• Statistics, Forecasts, Planning
• Search (!)
IPC at EPO
Classification systems available at the EPO
IPC-2006: Core level, Advanced level
IPC (editions 1-7)
ECLA (+ ICO, KW)
US Patent Classification
FI, FTerms
Classification at EPO
EPO Search (1): How do we do it?
• Classification is used in the vast majority of the technical fields (essential for e.g. searching concepts or processes)
• ECLA is often used in combination with other classifications, e.g. FI/FT (UCLA less used)
• The best mix of classification tools is quite variable, and field-dependent, e.g. specialised databases
• IPC is a necessary tool for the residual documentation
Search at EPO
EPO Search (2): Using classification
• First search in a superset defined by using your most precise classifications, e.g. ECLA
• Then search in the relevant residual IPC superset (i.e. IPC set minus (ECLA set, 'wrong' IPC set*) "déjà vu" functionality)
* we'll see that later
ECLA, FI/FT
IPC
Search at EPO
EPO Search (3): Relevance of FI/FT classification
16%
42%
42%
Families classified in ECLA (4,9m)
Families classified in FI/FT (4,8m)
Families classified in ECLA and FI/FT (1,8m)
EPO SR citing JP = ~ 17 % (~ 25.000 SR/Y steady )
Search at EPO
A detailed study (1): FI/FT consultation stats
• Section B* (Mechanics)• About 40 subclasses investigated• For each of them, a quantitative analysis was carried out to
establish the ratio between:– the % of EP Search Reports citing JP docs– the % of JP documents classified in that subclass
• Results:– few ratios below 0,8– most ratios around 1 or more– most ratios stable or increasing over 2004-2006 period
• Interpretation:– effective usage of FI/FT together with ECLA in "deep
indexing"-intensive fields
* additionally, also a few tens of subclasses in A, C and D were involved
Search at EPO
A detailed study (2): Stats on CN and KR citations
• The previous analysis was extended to CN and KR patents• NB:
– whereas JP was in most subclasses > 20% (up to 50%)– CN or KR docs was in most cases 1% to 5%– in most of those fields indexing is important
• Result:– for both CN and KR the ratio is consistently well below 1
• Interpretation:– (in the investigated fields) the EPO cannot access better
'added value' information* on CN and KR documentation– some years ago, it was the same with JP doc
* does it exist?Search at EPO
EPO Search (5): Classification is not used...
• ... in part of Organic Chemistry:
• C07C• C07D• C07H• C07J• A61K31
• used instead: CAS, Beilstein (with graphical user interfaces for defining molecular structures)
• T049: ECLA simplified to IPC AL
T049
Search at EPO
EPO Search Reports in Organic Chemistry (last 5y)
93458; 12%
143596; 19%
630866; 82%
50138; 6%
C section Organic chemistry (35% of C)
All SR ~= 775.000
Search at EPO: stats
Biochemistry: a(nother) special case
• C12Q1• G01N33/50-98• C12N• C07K• A61K38, 39, 48• A01K67/027, 033
• ~46.000 SR in the last 5y• (similar split as previous graph)
• Search: Sequence listing + ECLA/ICO Classification (often neither of the two is enough on its own)
Search at EPO
EPO Search Reports: What do we cite?
EPO SR produced in the last 5 years (tot. 775.000)
0
100000
200000
300000
400000
500000
600000
US EP WO DE XP JP GB FR AT IT CA CN RU KR
N°
of
Se
arc
h R
ep
ort
s
CN=2500, RU=2150, KR=1800
EC classified Pat: 95%
NPL: 24%
JP: 17%
1%
EPO Citation Statistics
Patent Publication Statistics (from WPI and EPODOC)
The % of only-IPC classified families is slowly increasing, and the country-of-origin split is rapidly changing: KR+CN share is increasing
Published families (all)
7%
35%
9%
5% 3%
41%
Pat w ECXP w ECJPSU/RUKRCN
Families published until 2002
39%
9%
36%
11%
4%
1%
Families published last 5 years
46%
1%
34%
2%
8%8%
16%
18%
17%
Publication Statistics
Patent Publications (families)
0
100
200
300
400
500
600
700
1999 2000 2001 2002 2003 2004 2005 2006 2007
x100
0
0%10%20%30%40%50%60%70%80%90%
100%
1999 2000 2001 2002 2003 2004 2005 2006 2007
KR patents
CN patents
EC classified patents
KR+CN = ~24% of this patent doc
over last 9 years
Publication Statistics
Theoretical level of CN, KR citations
• JP doc = 35 %• SR with JP cited = 17 % (~2:1 ratio)
• CN+KR doc = 8% (all) ; 16% (last 5y)• →• SR with CN or KR cited = ~ 4% (at least)
• the reality is (next slide)…
Statistics
% of EPO Search Reports citing CN, KR docs
0,00%0,05%0,10%0,15%0,20%0,25%0,30%0,35%0,40%0,45%0,50%
1999
2000
2001
2002
2003
2004
2005
2006
2007
SR citing KR
SR citing CN
800
EPO Citation Statistics
Example: G09G - Displays
• Almost all patents: JP or KR PR• World leaders: Samsung, LG,
Pioneer, Panasonic, ...
• Plasma displays• One (1!) IPC group: G09G3/28• 8.740 families• 28% KR patents not EC or FT classified
• Last 5 years: 511 EPO SR• SR with KR docs: 16• 3%• SR with JP docs: 283• 55%
43%
26%
28%
4%
EC class JP KR CN
EPO Citation Statistics Image: courtesy Sony
Better Patent Search: not only IPC classification!
• In general, added-value systems need improvements:
– share internal classification schemes and doc inventory for search
– better availability (in format and language) of national patent publications, e.g. Utility Models
– better translation engines
– easy availability of references and citations
– (categorised) full-text, controlled keywords, extended abstracts in English
– relevant information (e.g. sequence listings) must be published according to the required standards
EPO Wish List
Reformed IPC: some inconvenient truths
• The Reform has not addressed/overcome some fundamental problems of the IPC
• IPC is rarely used for search (at the EPO…)
• CL: anybody cares?
• Invention Information/Additional Information– not consistently applied (next slide)
– cheer up: ECLA Reform has received a similar lukewarm welcome among EPO classifiers/searchers (~ 12% of subclasses)
• Not harmonised IPC application (next slide)
• Full compulsory Reclassification?
IPC: some inconvenient truths
Facing Reality: Application of "Additional Information" IPC
Patents Published during Reformed IPC (2006/2007 - 87% of total)
0100000
200000300000400000
500000600000700000
800000900000
JP US CN DE WO KR EP RU
% of above Patents bearing "Additional Information" classification
0
2
4
6
8
10
12
JP US CN DE WO KR EP RU
Average: 7%does it matter when doing a search?
IPC: some inconvenient truths
Facing Reality: Not harmonised use of IPC
• Possibly the most serious problem of the IPC (even more than the size of the groups)
• One -expensive- trick to reduce noise used by a few EPO examiners is "negative" classification
• → Harmonisation would be better !
ECLA, FI/FT
IPC
• implemented by using (controlled) keywords• sporadically allocated to documents that are
normally not classified in ECLA (e.g. RU, CN) and stumbled upon during searches
• the KW is composed of an IPC symbol followed by an "X"
• meaning: the document bearing it should not have been classified in that IPC group (according to EPO interpretation...)
Negative Set
IPC: some inconvenient truths
So, do we still need the IPC? For what?
• Paradoxically, due to the rapid increase of % of patents only published with IPC, its importance is growing!
• Though, not (only) for Search• Where is the IPC in the Toolbox analogy?
• The IPC is ... the BOX !
• IPC is the only binding element among a (growing) plethora of unrelated and specialised tools
• As any "universal" language, it's incomplete and imprecise, but it's very much needed for e.g. concordance, navigation, link
IPC: a fundamental question
IPC Community: a message in a bottle
• Face reality: nature (and patent offices, and classifiers, and searchers) follow the path of minimum resistance
• Make pragmatic choices:– Keep what's used / Involve Stakeholders / Innovate– Timely improve where needed: Technology Watch!– Reclassification: some AL projects (H04W) are at a stand-
still: implement what's possible rather than nothing at all
• Other classifications may be locally more precise: acknowledge this fact, coordinate among them and help the user: external links, references, navigation facilities, ...
• IPC CL for NPL (Non-Patent Literature): can WIPO convince publishers?
• Broader cooperation is needed: EPO is ready to help!
Conclusion (1/3)
Every cockroach is beautiful to its mother, but...
Reformed IPC
Conclusion (2/3) Image: courtesy Disney-Pixar
... many bugs make quite a powerful bunch !
Reformed IPC
Thank You!
Conclusion Image: courtesy Disney-Pixar