L0GANA new technology
in text processing 
     |  Analog1  |  Analog2  |  magyarul  

With Logana’s text search method you can find incorrectly searched or even incorrectly stored text that normally you would not be able to with other search engines (Google, Yahoo, Bing etc.). This functionality - developed by our own teams - is part of Logana’s highly effective, associative text comparison algorithm, which is independent of language and character sets. You can try the demos (Analog1 and Analog2) for the service at www.logana.com.

The Analog1 (Associative Name Seeker) is an irregular solution for a regular text search task, used to perform a distorted search of more than a thousand international celebrities in the “Sample data” table.

The most crucial parameters of a search engine are generally completeness and speed. Completeness means finding all the objects matching the requested criteria, while speed refers to a reaction time normally expected of a computer, that is, a few seconds.

Traditional text search engines like Google, Yahoo, Bing etc., match the speed requirements; however, the same can very rarely be said of those of completeness, as these search tools cannot adequately manage textual errors in the users’ searches or in the documents stored.

To prove this, let us examine the traditional text search tools and use any of them to search for the name “Veiszäcker”. (Richard Karl Freiherr von Weizsäcker served as President of the Federal Republic of Germany between 1984 and 1994).

Yahoo and Bing return two results (excluding the link to Logana’s website): two comments on a Swedish chat website, where the German politician’s name is referenced as provided (Richard von Veiszäcker). Although Google returns more results, all the search engines share two unfortunate side effects. One of these is that these incorrectly typed websites do not appear when searching for correctly typed names, and the other is that the results do not have links to the correctly spelled names attached to them…

All of this despite that fact that the actual name, Weizsäcker, for which Google returns more than a million results, only differs from the formerly entered name in one correctly spelled (W instead of V) and one swapped character (z and s), meaning their similarity is quite high (80% according to Logana’s text comparison process).

Let us now search with Logana’s method in the “Sample data” table.

In the “Text search” text box, enter “Veiszäcker”, leave the value for the “Similarity” data box at 100%, then click the “Search” button. The “List of results” pane will be empty.

Well, this is what a value of 100% returns. However, let us set the limit for similarity in the “Similarity” data box lower, for example 80%. Click the “Search” button and the required name will appear.

The Analog2 (Associative KeyWord Collector) demonstrates a unique associative text processing method.

Its use is simple. The user enters a text in the “Text to examine” box (the character limit in the demo is set at 32,000 characters) and clicks the “Start collecting” button, after which the most typical words of the text “keywords” in librarian terminology automatically appear in the “Keywords returned” list. Moreover, the text entered cannot only be English; users can enter text in French, German, Hungarian, Italian or Spanish.

Text input into the “Text to examine” box follows a regular Windows method. (For example, highlighting the original text and pasting it to the destination with CTRL+C and CTRL+V). The next step is to select the input language from the “Language” list, then enter the percentage values for the “Filter” and “Reduction” boxes.

Lower filter values return more unique words, while lower reduction values return words more different from each other. Simply put, the higher the value for both parameters, the more keywords will appear in the “Keywords returned” list.

Have fun using the demos provided! Keep in mind that there are several other uses of Logana’s associative text processing technology (phoneme-based speech recognition, finding mutant gene sequences in DNA chains, comparing text documents, processing similiarity data, identifying data conections etc.).

If we have piqued your interest, please contact us at analog@logana.com.