Hi,

I am looking for a service (accessible via an API) that can compare two
company names and return a % match based on their true equivalence. For
example, in system1 we have a record that "London Borough of Islington" and in
system2 it's referred to as "Islington Council" - these relate to the same
business entity (easy to determine if familiar with the data as a human) but
of course are quite different labels. This seems to need to rely on both word
matching as well as, potentially, records of business entities.

Beyond fuzzy matching, any suggestions / been down this road before?

--
jeremy

On Thursday 7 December 2023 at 08:17:44 UTC, jeremy wrote:
> Hi,
>
> I am looking for a service (accessible via an API) that can compare two
> company names and return a % match based on their true equivalence. For
> example, in system1 we have a record that "London Borough of Islington" and in
> system2 it's referred to as "Islington Council" - these relate to the same
> business entity (easy to determine if familiar with the data as a human) but
> of course are quite different labels. This seems to need to rely on both word
> matching as well as, potentially, records of business entities.
>
> Beyond fuzzy matching, any suggestions / been down this road before?

In the good old days[1] I recall a system that did 'speech to text' (I think Simon Wilson
knows of this one too). Maybe called 'Spinvox'. You'd divert your voicemail to a number
that would then use 'computer voice to text technology' to convert the voicemail
message to a text message. Amazing stuff at the time. Except it wasn't, they had
desks full of people somewhere doing manual transcription. Ended up being quite
a stink where I worked because this had been missed.

Maybe a similar system would work here...

[1] Maybe early-mid 00s?

On 7 Dec 2023 at 09:59:16 GMT, "Stephen Packer" <stephen.packer@gmail.com>
wrote:

> On Thursday 7 December 2023 at 08:17:44 UTC, jeremy wrote:
>> Hi,
>>
>> I am looking for a service (accessible via an API) that can compare two
>> company names and return a % match based on their true equivalence. For
>> example, in system1 we have a record that "London Borough of Islington" and in
>> system2 it's referred to as "Islington Council" - these relate to the same
>> business entity (easy to determine if familiar with the data as a human) but
>> of course are quite different labels. This seems to need to rely on both word
>> matching as well as, potentially, records of business entities.
>>
>> Beyond fuzzy matching, any suggestions / been down this road before?
>
> In the good old days[1] I recall a system that did 'speech to text' (I think
> Simon Wilson
> knows of this one too). Maybe called 'Spinvox'. You'd divert your voicemail to
> a number
> that would then use 'computer voice to text technology' to convert the
> voicemail
> message to a text message. Amazing stuff at the time.

Yes - there was a guy here on ukrm worked for Spinvox - Chris somebody?
--
jeremy

On 07/12/2023 08:17, jeremy wrote:
> Hi,
>
> I am looking for a service (accessible via an API) that can compare two
> company names and return a % match based on their true equivalence. For
> example, in system1 we have a record that "London Borough of Islington" and in
> system2 it's referred to as "Islington Council" - these relate to the same
> business entity (easy to determine if familiar with the data as a human) but
> of course are quite different labels. This seems to need to rely on both word
> matching as well as, potentially, records of business entities.
>
> Beyond fuzzy matching, any suggestions / been down this road before?
>

Both should have the same registration number at companies house if they
are the same legal entity.

Tim <nobody@home.co.uk> wrote in news:uks8qg$183bt$1@dont-email.me:

> On 07/12/2023 08:17, jeremy wrote:
>> Hi,
>>
>> I am looking for a service (accessible via an API) that can compare
>> two company names and return a % match based on their true
>> equivalence. For example, in system1 we have a record that "London
>> Borough of Islington" and in system2 it's referred to as "Islington
>> Council" - these relate to the same business entity (easy to
>> determine if familiar with the data as a human) but of course are
>> quite different labels. This seems to need to rely on both word
>> matching as well as, potentially, records of business entities.
>>
>> Beyond fuzzy matching, any suggestions / been down this road before?
>>
>
> Both should have the same registration number at companies house if
> they are the same legal entity.
>

and VAT number which could apply to sole traders, councils, charities etc
not registered at CH

On 7 Dec 2023 at 11:03:44 GMT, "Tim" <nobody@home.co.uk> wrote:

> On 07/12/2023 08:17, jeremy wrote:
>> Hi,
>>
>> I am looking for a service (accessible via an API) that can compare two
>> company names and return a % match based on their true equivalence. For
>> example, in system1 we have a record that "London Borough of Islington" and in
>> system2 it's referred to as "Islington Council" - these relate to the same
>> business entity (easy to determine if familiar with the data as a human) but
>> of course are quite different labels. This seems to need to rely on both word
>> matching as well as, potentially, records of business entities.
>>
>> Beyond fuzzy matching, any suggestions / been down this road before?
>>
>
> Both should have the same registration number at companies house if they
> are the same legal entity.

Yes, makes sense - however the issue is slightly before that - as clearly the
company will be registered with a name - but in CRM systems for example, they
might simply record "Islington" as the customer name - they know what they
mean - but another system with related information refers to it differently.
The challenge is to compare the two items of data and calculate a % confidence
in them relating to the same entity.

--
jeremy

On 07/12/2023 12:00, jeremy wrote:
> On 7 Dec 2023 at 11:03:44 GMT, "Tim" <nobody@home.co.uk> wrote:
>
>> On 07/12/2023 08:17, jeremy wrote:
>>> Hi,
>>>
>>> I am looking for a service (accessible via an API) that can compare two
>>> company names and return a % match based on their true equivalence. For
>>> example, in system1 we have a record that "London Borough of Islington" and in
>>> system2 it's referred to as "Islington Council" - these relate to the same
>>> business entity (easy to determine if familiar with the data as a human) but
>>> of course are quite different labels. This seems to need to rely on both word
>>> matching as well as, potentially, records of business entities.
>>>
>>> Beyond fuzzy matching, any suggestions / been down this road before?
>>>
>>
>> Both should have the same registration number at companies house if they
>> are the same legal entity.
>
> Yes, makes sense - however the issue is slightly before that - as clearly the
> company will be registered with a name - but in CRM systems for example, they
> might simply record "Islington" as the customer name - they know what they
> mean - but another system with related information refers to it differently.
> The challenge is to compare the two items of data and calculate a % confidence
> in them relating to the same entity.
>
>

Companies house BETA has a REST API that might assist?

https://developer.company-information.service.gov.uk/overview

Might need an advanced search to reduce the otherwise massive number of
hits - say on Standard Industrial Classification (SIC) code to search
by nature of business?

582 hits for 'Islington'

One hit for "Islington" SIC 84110 , though the CRM or whatever probably
won't hold that data.

https://find-and-update.company-information.service.gov.uk/advanced-search/get-results?companyNameIncludes=islington&companyNameExcludes=&registeredOfficeAddress=&incorporationFromDay=&incorporationFromMonth=&incorporationFromYear=&incorporationToDay=&incorporationToMonth=&incorporationToYear=&sicCodes=84110&dissolvedFromDay=&dissolvedFromMonth=&dissolvedFromYear=&dissolvedToDay=&dissolvedToMonth=&dissolvedToYear=

--
Moto Morini 2C/375
Gilera 175 Sport, Husqvarna 401 Svartpilen
Honda CB250RS (Not Waynetta!)
"Do not adjust your mind, there is a fault in the reality"

On 07/12/2023 08:17, jeremy wrote:
> Hi,
>
> I am looking for a service (accessible via an API) that can compare two
> company names and return a % match based on their true equivalence. For
> example, in system1 we have a record that "London Borough of Islington" and in
> system2 it's referred to as "Islington Council" - these relate to the same
> business entity (easy to determine if familiar with the data as a human) but
> of course are quite different labels. This seems to need to rely on both word
> matching as well as, potentially, records of business entities.
>
> Beyond fuzzy matching, any suggestions / been down this road before?
>

Paid for?

Bureau Van Dijk <https://www.bvdinfo.com/en-gb/>

--
Bruce Horrocks
FJR1300AS

On 07/12/2023 12:00, jeremy wrote:
> On 7 Dec 2023 at 11:03:44 GMT, "Tim" <nobody@home.co.uk> wrote:
>
>> On 07/12/2023 08:17, jeremy wrote:
>>> Hi,
>>>
>>> I am looking for a service (accessible via an API) that can compare two
>>> company names and return a % match based on their true equivalence. For
>>> example, in system1 we have a record that "London Borough of Islington" and in
>>> system2 it's referred to as "Islington Council" - these relate to the same
>>> business entity (easy to determine if familiar with the data as a human) but
>>> of course are quite different labels. This seems to need to rely on both word
>>> matching as well as, potentially, records of business entities.
>>>
>>> Beyond fuzzy matching, any suggestions / been down this road before?
>>>
>>
>> Both should have the same registration number at companies house if they
>> are the same legal entity.
>
> Yes, makes sense - however the issue is slightly before that - as clearly the
> company will be registered with a name - but in CRM systems for example, they
> might simply record "Islington" as the customer name - they know what they
> mean - but another system with related information refers to it differently.
> The challenge is to compare the two items of data and calculate a % confidence
> in them relating to the same entity.
>
>
You should also be able to get all of the postcodes belonging to each
registered entity. My last employer had a number of offices across the
UK and trading under different names in the group of companies, but it
was simple to find the parent company and therefore all the postcodes in
the group and thus whether two dissimilar names were related.

On 07/12/2023 09:59, Stephen Packer wrote:
> On Thursday 7 December 2023 at 08:17:44 UTC, jeremy wrote:
>> Hi,
>>
>> I am looking for a service (accessible via an API) that can compare two
>> company names and return a % match based on their true equivalence. For
>> example, in system1 we have a record that "London Borough of Islington" and in
>> system2 it's referred to as "Islington Council" - these relate to the same
>> business entity (easy to determine if familiar with the data as a human) but
>> of course are quite different labels. This seems to need to rely on both word
>> matching as well as, potentially, records of business entities.
>>
>> Beyond fuzzy matching, any suggestions / been down this road before?
>
> In the good old days[1] I recall a system that did 'speech to text' (I think Simon Wilson
> knows of this one too). Maybe called 'Spinvox'. You'd divert your voicemail to a number
> that would then use 'computer voice to text technology' to convert the voicemail
> message to a text message. Amazing stuff at the time. Except it wasn't, they had
> desks full of people somewhere doing manual transcription. Ended up being quite
> a stink where I worked because this had been missed.
>
> Maybe a similar system would work here...
>
> [1] Maybe early-mid 00s?

I ain't saying nuffink. Except it was Catman, formerly of this parish
that got me involved in that.

Amazon has (used to have?) a mechanical turk[1] service.

[1] if one is still allowed to use such an expression.

--
/Simon