[E-trademarks] Regex search - Proximity Search
Ken Boone
boondogles at hotmail.com
Sun Mar 17 21:50:02 EDT 2024
> ... I also mentioned that I haven't found a way to do a proximity search in regex ...
From the online help (Advanced logical operators):
* Use ~ for proximity searching with quoted search terms.
Allows words in quotes to appear with other words in-between or in a different order. Counts number words that are inserted, deleted, substituted, or transposed (called "positions"). For example, CM:"rain shine"~5 retrieves:but doesn’t retrieve:
* rain shine (no insertions)
* shine rain (one transposition)
* rain and it might shine (three insertions) and
* shine on rain (one transposition and one insertion)
* Shine your car to protect from rain (one transposition and five insertions)
For the earlier GS:"baby bonnet" example, I used the GS2 index instead {to get the plurals or whatever they do for variations of the terms} and added variations of the ~# operator to demonstrate the impacts in the brief search session below.
Id
Query
ResultCount
Mode
1
GS:"baby bonnets" AND LD:true
9
Expert
2
GS2:"baby bonnets" AND LD:true
10
Expert
3
GS2:("baby bonnets" ~2) AND LD:true
55
Expert
4
GS2:("baby bonnets" ~3) AND LD:true
91
Expert
5
GS2:("baby bonnets" ~4) AND LD:true
110
Expert
6
( 3 ) NOT ( 2 )
45
Expert
7
( 4 ) NOT ( 3 )
36
Expert
( 2 ) NOT ( 1 ) returned the single hit bodysuits for babies, bonnets, assuming I found the relevant GS text. (Where's the full document display with highlighting of the search terms?)
( 3 ) NOT ( 2 ) returned hits like:
Baby tops; Bonnets;
Baby bodysuits; Bonnets;
baby caps, bonnets, and baby booties;
where some appear to be relevant and some not.
( 4 ) NOT ( 3 ) returned hits like:
baby shirts and pants, bonnets and socks
Bonnets; Frocks; Baby bodysuits; Baby bottoms;
Baby tops; Bathing costumes; Bonnets;
Baby tops; Bathing costumes; Bonnets; Bralettes;
where relevancy is more complicated.
The online help could have included more examples of ~# proximity searching.
Happy regexing,
Ken Boone
________________________________
From: E-trademarks <e-trademarks-bounces at oppedahl-lists.com> on behalf of Pamela Chestek via E-trademarks <e-trademarks at oppedahl-lists.com>
Sent: Sunday, March 17, 2024 10:56 AM
To: e-trademarks at oppedahl-lists.com <e-trademarks at oppedahl-lists.com>
Cc: Pamela Chestek <pamela at chesteklegal.com>
Subject: Re: [E-trademarks] Regex search
For one thing, it means that whenever you do a search you have to make a decision about whether to use the Advanced syntax or the regex syntax. And it appears there are situations where you must use one or the other, such as my t-shirt example, which means you have to be highly skilled with both. Case in point - use of capitals:
Id Query ResultCount
1 GS2:DVD 98307
2 GS2:/DVD/ 0
3 GS2:/dvd/ 98307
If you don't remember that the search string in regex must be lowercase, and enter the term as it actually appears in the ID, you may have just screwed the pooch. It's just an invitation to disaster because unless you are a master of both you won't realize that your query was malformed.
I also mentioned that I haven't found a way to do a proximity search in regex. But you can only do pattern matching in regex. So to pattern match on a two-word unitary phrase, I guess you have to include two different searches, e.g., CM:("tick tock") AND CM:(/ti[ckqx]/) AND CM:(/to[ckqx]/) to find the homonyms for "tick tock"?
There are so many examples of unreliable queries - forgetting a colon after the field tag, or failing to use all caps for connectors, also give you some results and you may not realize they are unreliable. I have submitted a request that a malformed query be flagged. It might not be possible with the t-shirt example, but it should be possible to at least identify syntactical errors.
Pam
Pamela S. Chestek
Chestek Legal
300 Fayetteville Street
Unit 2492
Raleigh, NC 27602
+1 919-800-8033
pamela at chesteklegal.com<mailto:pamela at chesteklegal.com>
www.chesteklegal.com<http://www.chesteklegal.com/>
On 3/16/2024 11:44 AM, Sam Castree via E-trademarks wrote:
Just curious, but is there a reason why that particular search would need to use regex at all? Maybe I'm way off (I definitely don't claim to have mastered this system yet), but it seems like a..."normal" expert search would work just fine. GS:"baby bonnet" or whatever.
Cheers,
Sam Castree, III
Sam Castree Law, LLC
3421 W. Elm St.
McHenry, IL 60050
(815) 344-6300
On Fri, Mar 15, 2024 at 7:56 PM Pamela Chestek via E-trademarks <e-trademarks at oppedahl-lists.com<mailto:e-trademarks at oppedahl-lists.com>> wrote:
Indeed, and the help materials couldn't be more terse and the
information more difficult to find, other than not to exist at all. I
have yet to figure out how (if possible) to search for a two-word
expression, like "baby bonnet," using regex.
Id Query ResultCount
3 GS2:(/baby/ AND /bonnet/) 2484
2 GS2:/baby bonnet/ 0
1 GS2:"baby bonnet" 43
So you're left with trying to remember two similar, but not identical,
search systems, including how they treat quotation marks, reserved
characters, capitalization, etc.
Pam
Pamela S. Chestek
Chestek Legal
300 Fayetteville Street
Unit 2492
Raleigh, NC 27602
+1 919-800-8033
pamela at chesteklegal.com<mailto:pamela at chesteklegal.com>
www.chesteklegal.com<http://www.chesteklegal.com/>
On 3/15/2024 8:40 PM, Richard Schafer via E-trademarks wrote:
> The fact that we have to make guesses about how the indexes work shows how poorly the PTO has implemented and documented this system. I wonder if examining attorneys have the same poor explanation about how the indexes work. If they do, that's really bad. If they have a fuller explanation, what possible reason would the PTO have for not making that explanation public?
>
> Best regards,
> Richard A. Schafer | Schafer IP Law
> P.O. Box 230081 | Houston, TX 77223
> M: 832.283.6564 | richard at schafer-ip.com<mailto:richard at schafer-ip.com>
>
> -----Original Message-----
> From: E-trademarks <e-trademarks-bounces at oppedahl-lists.com<mailto:e-trademarks-bounces at oppedahl-lists.com>> On Behalf Of Neil R. Ormos via E-trademarks
> Sent: Friday, March 15, 2024 6:19 PM
> To: E-trademarks Mailing List <e-trademarks at oppedahl-lists.com<mailto:e-trademarks at oppedahl-lists.com>>
> Cc: Neil R. Ormos <ormos-lists at ormos.org<mailto:ormos-lists at ormos.org>>
> Subject: Re: [E-trademarks] Regex search
>
> Pamela Chestek via E-trademarks wrote:
>
>> Can anyone explain to me why the first two search queries didn't give
>> me the same results as the third query? I was in expert mode.
>
>> GS:/t\-shirt/ 0
>> GS:/t[-]shirt/ 0
>> GS:t-shirt 511200
> When the GS index is used with a regular expression search, the index appears to contain each of the words of the goods and services field in isolation. Spaces and hyphens separate words, but some other punctuation does not. Your regexps do not match because t-shirt does not appear as a single word in the index.
>
> The GS index behaves differently when used with other types of search.
>
> <https://oppedahl-lists.com/pipermail/e-trademarks_oppedahl-lists.com/2024-January/000554.html>
>
> And the GS2 index also behaves differently.
>
> GS2:/t[-]shirt/
>
> returns 61,341 records.
>
> I know GS2 does some sort of stemming, but I haven't been able to synthesize a complete explanation. If someone has GS2 figured it out, or knows of a document that describes its behavior, I hope they will explain.
>
>
> --
> E-trademarks mailing list
> E-trademarks at oppedahl-lists.com<mailto:E-trademarks at oppedahl-lists.com>
> http://oppedahl-lists.com/mailman/listinfo/e-trademarks_oppedahl-lists.com
>
--
E-trademarks mailing list
E-trademarks at oppedahl-lists.com<mailto:E-trademarks at oppedahl-lists.com>
http://oppedahl-lists.com/mailman/listinfo/e-trademarks_oppedahl-lists.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oppedahl-lists.com/pipermail/e-trademarks_oppedahl-lists.com/attachments/20240318/717e2632/attachment.htm>
More information about the E-trademarks
mailing list