PATSTAT: an assessment on inventors' addresses quality

When trying to consolidate data on inventors in PATSTAT, a part from name, we need some toponomastic data.
An innovator named "J. Smith" would be difficult to connect either with "John Smith" or with "James Smith", even if we have some further data like address, city...

So I investigated with some queries on patstat ediction 10/2009 in what percentuage the fields address, city, zip code and country are filled, by patent office.

I used tables TLS201_APPLN for linking application authority, and TLS206_ASCII (that is the ascii/parsed version of TLS206, included in patstat) for person ids and data.

For selecting distinct person id for inventors (A_I_FLAG = 'I') I used this SQL

Select
  T1.APPLN_AUTH, Count(Distinct t6.PERSON_ID)
From
  tls201_appln T1 Inner Join tls206_ascii t6 
  On t6.APPLN_ID = T1.APPLN_ID
Where   t6.A_I_FLAG = 'I';

Here below I'm pasting the  resulting table for the TOP 20 by inventor count, the full table can be downloaded @ this link.

APPLN_AUTH	inventors	no state	no zip	no country	no address	no city
US	5960856	86%	98%	21%	97%	25%
EP	3705123	100%	100%	0%	1%	1%
DE	2750079	100%	100%	33%	100%	100%
JP	1798271	100%	100%	98%	99%	100%
CN	1537587	100%	100%	2%	100%	100%
CA	1120490	100%	100%	45%	100%	100%
AU	1087573	100%	100%	98%	100%	100%
SU	968915	100%	100%	41%	100%	100%
AT	653048	100%	100%	29%	100%	100%
KR	637296	100%	100%	14%	100%	100%
FR	565254	100%	100%	98%	99%	100%
GB	531087	100%	100%	70%	65%	100%
RU	394691	100%	100%	29%	100%	100%
CH	338739	100%	100%	11%	100%	100%
BR	292047	100%	100%	89%	100%	100%
SE	256248	100%	100%	85%	98%	100%
FI	212722	100%	100%	11%	43%	100%
IT	192460	100%	100%	74%	100%	100%
ES	133471	100%	100%	17%	100%	100%
DD	129845	100%	100%	7%	97%	100%

It might be argued that such a count could have more sense if done by publication authority instead of application authority, since data are taken from search report and it can be noticed that FI data applied @ EPO and published @ WIPO/PCT differ in quality from those applied and published @ EPO.
Maybe it could be topic of a further post...

KITeS Patstat based reports

I've been vacating this blog for the majority of june because I implemented a set of reports based on Patstat EPO data.

At web address db.kites.unibocconi.it you will be able to find the following reports


Patent count by inventor country / priority year
Patent count by applicant country / priority year
Patent count by inventor region / priority year
Patent count by applicant region / priority year
Patent count by inventor nuts3 / priority year
Patent count by applicant nuts3 / priority year
Patent count by applicant name / priority year
Patent count by main IPC - first 4 digits
Patent count by main IPC class reclassified on OST30
Patent count by applicant, priority year, OST30 IPC reclassification
Patent count by applicant country, county, region, OST30 IPC reclassification, priority year
Citations count by applicant name, priority year
Citations count by applicant country, priority year
Citations count by inventor country, priority year
Copatenting by inventor country, priority year
Copatenting by applicant country, priority year
Applicants by IPC - first 4 digits
Inventors by IPC - first 4 digits

Please remember this is still a beta release; feel free to report me any problem or suggestion...

Eventually you will find some other reports password protected; such reports contain detail data and cannot be distributed since PATSTAT data are property of EPO.

Inventors data in patstat: Epo vs uspto

Patstat stores data about applicants and inventors inside a table with the prefix TLS206, indexed using a field named PERSON_ID that can be linked to the applications via table TLS207_PERS_APPLN linking each person_id to an application id (appln_id)Patstat DVD provides two versions of TLS206: the first TLS206_PERSON is a comma separated value file containing mainly the fields name, address and country code.

TLS206_ASCII instead contains the same information already parsed (and also something more, see below for fields list): name is splitted into last, first and middle name; address into street, city, state and zip code.

Data origins for person names and addresses:


1) EPO Register for EP patent applications; details are those that were the most recent in the EP Register at the time of extraction of the data.

2) OECD patents database for US data post 1976-01-01 up to and including November 15th 2005 for Published Grants.

3) PATSTAT weekly file extracts from USPTO website for Published Grants from November 22nd 2005 until today;  Published Applications  from September 29th 2005 to today inclusive.

4) Inventor & Applicant names for US PTO Published Applications from March 1st 2001 to September 22nd 2005 from DOCDB ,  data-format="docdba".

5) all other names  from DOCDB , data-format="docdba" (US data for names and addresses for patents published before 1976-01-01 is taken from the EPO's DOCDB Database)

TLS206_ASCII FIELD LIST                           

prof-last-name
prof-first-name
prof-middle-names
prof-street
prof-city
prof-state
prof-zip-code

 

TLS206_person field list                      

Person_id                         

Person_name                 

Person_address                             

Person_ctry_code         

patstat: counting applications by patent office

This is an answer to a very simple question: how many patent offices are listed in patstat and how many applications from each of them?
By aggregating the data we get the table I'm posting here below, containing 174 application authorities (data from 10/2009 patstat version). When considering instead only those who have more than 1000 applications, the figure decreases to 78 patent offices.
Some will notice that among the 174 the achronim WO (WIPO/PCT patent office) does not appear. This is because WO pubblications always result from an application to a local patent office (domestic or EP) where the PCT procedure is applied. The extent of this practice will be subject of one further post.

APPLN_AUTH	count
	1
AE	36
AF	1
AL	3
AM	147
AN	2
AP	4742
AR	79581
AT	1028408
AU	1553990
AZ	105
BA	344
BB	5
BD	9
BE	642989
BG	53602
BH	1
BI	9
BO	10
BR	496326
BS	2
BW	1
BX	105
BY	761
BZ	1
CA	2580140
CG	3
CH	1055204
CI	2
CL	4453
CM	7
CN	2803867
CO	222
CR	379
CS	166312
CU	2843
CY	2621
CZ	66921
DD	266209
DE	6792342
DK	429570
DM	12
DO	191
DZ	1547
EA	13114
EC	4949
EE	6251
EG	11450
EM	4254
EP	2388584
ER	6
ES	902689
FI	271371
FR	2917043
GA	2
GB	3319126
GC	420
GD	3
GE	186
GH	7
GI	5
GM	10
GN	8
GR	97444
GT	1301
HK	69450
HN	10
HR	11297
HT	7
HU	137061
IB	65369
ID	14768
IE	91248
IL	160970
IN	66238
IQ	14
IR	88
IS	7797
IT	708724
JM	6
JO	22
JP	16282860
KE	1392
KG	19
KP	60
KR	1659443
KZ	477
LB	109
LI	22
LK	152
LR	3
LS	3
LT	3651
LU	68491
LV	4835
LY	2
MA	10092
MC	2789
MD	4586
MG	1
MK	87
MN	246
MO	1
MR	1
MT	566
MU	7
MW	739
MX	162007
MY	11106
NE	3
NG	19
NI	207
NL	607873
NN	4
NO	226939
NP	2
NR	2
NZ	109638
OA	12934
OM	1
PA	2108
PE	431
PH	23261
PK	147
PL	233730
PT	81034
PY	12
RH	40
RO	61095
RS	73
RU	394084
SA	73
SB	1
SC	1
SD	67
SE	831377
SG	51063
SH	1
SI	17619
SK	23261
SL	2
SM	44
SN	10
SO	1
SR	3
ST	1
SU	1249050
SV	692
SY	44
SZ	4
TH	260
TJ	375
TM	2
TN	269
TR	42800
TT	52
TW	369739
TZ	1
UA	50213
US	11376401
UY	6573
UZ	53
VA	6
VE	100
VN	240
WO	5065
XH	1242
XP	16
YE	1
YU	33687
ZA	256542
ZM	2742
ZR	1
ZW	2909

converting patstat text fields into plain ascii

Due to the many different sources of data, the text fields in patstat (especially in table TLS206 containing names and adresses of applicants & inventors) would containg a lot of non-ascii chars;

In case anybody would need, this is a simple version for standardizing to plain ascii characters; be aware that using this table on patents titles or abstracts could lead to unexpected results.

-->

NONASCII	ASCII	DESCRIPTION
Ç	C	Ç CEDILLE
É	E	é
Ö	O	ö
Ü	U	ü
Ä	A	ä
À	A	à
Ú	U	ú
Á	A	á
Î	I	î
Å	A	å
È	E	è
Ã	A	Ã
Â	A	Â
Ë	E	Ë
Ø	O	Ø
Ó	O	Ó
Ñ	N	Ñ N TILDE
Ê	E	Ê
Ô	O	Ô
Æ	AE	Æ
ß	SS	ß
Í	I	I
Ï	I	Ï
Õ	O	Õ
¢	C	¢
Û	U	Û
†	A	Ä
š	U	ü
.O SLASHED.	O	Ø
.ANG.	A	å
{ACUTE OVER (M)}	M	M
{HACEK OVER (S)}	S	Š
{HACEK OVER (C)}	C	Č
{HACEK OVER (Z)}	Z	Ž
{UMLAUT OVER (S)}	S	S
{UMLAUT OVER (C)}	C	C

Obviously there is another problem that is "how non ascii characters are incorporated in the txt fields?"
If we deal with table TLS206_PERSON this problem has a high relevance but we can highlight it by seeking for the char  (FI « = AE) and we may use a table like the one made by Julio Raffo on
http://wiki.epfl.ch/patstat/corrupted

If we use TLS206_ascii the problem is not relevant.

Legal status forthcoming in patstat?

Legal status informations, taken from INPADOC database, describe designated states, extension, examination process and so on; in short refer to the entries and procedural steps occurring during the patent grant procedure and the subsequent life of a patent. These are normally published in the patent gazette of the patent-granting country or organisation concerned.

FI for patent EP 1.000.000 you can find here the legal status of the patent.

Each record is described by a PRS code.
For current code explanations please go to the download section under "Legal status codes in English".
Legal code classifications can be found in the download section under "Classification of recently used PRS codes" .

Nowadays this informations are not included in PATSTAT; this is what Geert Boedt posted on patstat forum.

Following the "Patent statistics for decision makers" conference, the EPO has had many requests to link legal status data to PATSTAT.
Some PATSTAT customers already have experience in linking these datasets and we understand several other users are considering carrying out work in this area.
In order to avoid unnecessary double work, the EPO is looking into possibilities to adapt the INPADOC database such that it could be easily linked to PATSTAT.
It would of course be helpful if you would like to share any experiences or user requests with the PATSTAT user community and the EPO.
If you have any suggestions on how you might like to use PATSTAT in conjunction with INPADOC Legal Status Data, we would like to hear from you.

So there's hope in the next future to see legal data inserted in PATSTAT.