Question

I need some help in solving this problem.

We have a large amount of documents of a given specified domain. These documents are from differente sources and therefore their structure can be very different too. On the other side I have a table with some specified fields where some figures has to be filled from the extract of the documents.

For example:

Company x had a business volume of
$20mio in 2010. $1,000,000 was the exchange of
company y this year.

The result should something like this

|| Company | Year | Volume  
||  X      | 2010 |  200,000  
||  Y      | 2010 | 1000,000  

Can you point me please to some links or topics, where I can find further informations how to solve such a problem.

I know that there is no out of the box solution for this, but where should i start to look for.

Thanks in advance.

Answer 1

Ok. There are entire computer science labs devoted to that sort of stuff!
Maybe start by looking a tool called RapidMiner

Also here are a couple of research paper titles I have as PDF’s (which I don’t have links for anymore sadly):

1. Automated Understanding of Financial Statements
Using Neural Networks and Semantic Grammars

James Markovitch
Dun & Bradstreet, Search Technologies
April 1995
Email: [e-mail]
Copyright  1995 James Markovitch

2. An Integrated Approach for Automatic Semantic Structure Extraction in Document Images

Margherita Berardi, Michele Lapi, and Donato Malerba
Dipartimento di Informatica – Università degli Studi di Bari
via Orabona 4 – 70126 Bari
{berardi,lapi,malerba}@di.uniba.it

I think the first one would be of greatest interest in terms of what you are after. Not quite sure how much value it will be though 🙂