Google returning false c-index tool results

[Marshall Clark]

Thought I'd give a heads up to everyone regarding the c-index tool at http://graphnical.com/cindex. Google is currently passing false search result number data to this tool which is resulting in wildly inaccurate c-index values.

Don't waste a morning running data like I did!
:rolleyes:

/-Marshall

[orion]

Actually back in the summer I warned users about this tool in the C-index thread (Keywords Co-Occurrence and Semantic Connectivity).

1. The tool calculates pairwise c-indices, only. It does not work for queries greater than two keywords/phrases.

2. Many users does not understand how to reduce an N-term query to a pairwise case, thus false c-indices are obtained when phrases are used.

3. Perhaps the Google API is also returning "dancing" data, which amounts to more noise.

If one does not want to write a script, the simplest way to compute c-indices is by using an EXCEL spreadsheet and by querying directly the search engine in question and properly using the EXACT and FINDALL query modes.

Orion

[Marshall Clark]

Hi Orion,

This then is an additional problem over and above what you've mentioned. Google is actually returning false search total results to Graphnical - when I go search the results myself they are often 2-3x larger than the values reported to the tool. Seems that Google has flagged their IP.

I've heard you mention your own c-index calculator several times. Is it publicly available?

[orion]

No. That "calculator" is part of a large scale project. Any funding is welcome.

I'll be presenting on c-indices and EF-ratios at SES NY. If you can make it there perhaps we can talk a bit more.

Orion

[NuevoJefe]

No. That "calculator" is part of a large scale project. Any funding is welcome.

Care to expand on that?

[orion]

No longer needed. Sorry.

Orion

[NuevoJefe]

NP, I know it was a long time ago. Thanks for the reply though.