The Natural Language Question

The natural language question is an easy way to phrase a search query, where the APICP flag keepnoise has been left off. Type in a question which contains the elements you are looking for in the text. There are really only these few rules to keep in mind.

  • Use only the important words in your query. Metamorph assumes every important concept in your question must be matched to qualify it as a hit.

  • Pronouns are meaningless to Metamorph; rephrase them as the nouns that they actually represent.

  • Idiomatic expressions should be avoided if possible, unless they are legitimately important to the sense of the question. Metamorph may or may not be aware of the true meaning of such a phrase.

An example of a "good" question to ask would be:

     Have there been any corporate takeovers?
An example of a less desirable question would be:
     What information is there on corporate takeovers?
The reason is that in this second question, the concept "information" must be stated in the text along with the concept "corporate" and "takeover" to be considered a match, probably excluding relevant responses; any hit from a news article is implicitly understood to be information without so stating. All you really need are the two important concepts "corporate" and "takeover".

Another "good" question might be:

     Were there any corporate takeovers in Germany?
or
     Have there been any power struggles in the Near East?

Where idiomatic expressions or phrases exist as entries in the Equivalence File, they will be meaningfully processed as a whole, whether so marked or not. The question parser will check for phrase entries in an entered query. If the word grouping is not known to be a phrase, the words will be processed separately, according to their individual associations. If it is important to you that the words of such an expression or phrase be processed as one unit and Metamorph does not recognize it as such, mark it as a phrase by putting it in quotes.

Once your question is entered, the question parser will select the important words and denote them as root words, to be expanded into sets from the known associations in the Equivalence File. These words and all their equivalences are in the main passed to PPM to be searched for simultaneously while determining where answers to your question might lie.

As a rule, unless escaped with a backslash (\), hyphens are stripped from the words on the query line before expanded to include their associations in the Equivalence File and sent to the search engine. Once passed to PPM or SPM occurrences of those words as separated by either hyphens or white space will be located.


Copyright © Thunderstone Software     Last updated: Dec 10 2018
Copyright © 2019 Thunderstone Software LLC. All rights reserved.