Extrapolative search techniques

A method implemented by a computing device comprising:

• identifying, by the computing device, text of a document to utilize for theme derivation in connection with insertion of a stock image at a position within the document that corresponds to the text;
• forming, by the computing device, a phrase array for the text by separating the text into constituent phrases;
• generating, by the computing device, a hashmap of words contained in the text that maps different words to relevance values derived for the different words;
• recognizing, by the computing device, a set of documents that are similar to the document based on comparison of the phrase array and hashmap for the document to phrase arrays and hashmaps computed for documents included in an established source collection;
• producing, by the computing device, an array of dominant words that are similar across the document and the set of documents by:
• sorting, by the computing device, elements of the hashmap for the document and the hashmaps computed for the documents included in the established source collection according to the relevance values;
• and
• populating, by the computing device, the array of dominant words with top-ranked elements of the hashmap and the hashmaps according to the sorting until a defined threshold number of words have been added to the array of dominant words;
• establishing, by the computing device, dominant phrases having words included in the array of dominant words as theme phrase candidates for the text;
• and
• utilizing, by the computing device, one or more of the dominant phrases as theme phrases for a search to obtain suggested images for insertion at the position within the document.

One or more computer-readable storage media comprising instructions stored thereon that, responsive to execution by a computing device, cause the computing device to derive theme phrases corresponding to selected text of a document for use in a search including:

• identifying, by the computing device, text of the document to utilize for theme derivation in connection with insertion of a stock image at a position within the document that corresponds to the text;
• forming, by the computing device, a phrase array for the text by separating the text into constituent phrases;
• generating, by the computing device, a hashmap of words contained in the text that maps different words to relevance values derived for the different words;
• recognizing, by the computing device, a set of documents that are similar to the document based on comparison of the phrase array and hashmap for the document to phrase arrays and hashmaps computed for documents included in an established source collection;
• producing, by the computing device, an array of dominant words that are similar across the document and the set of documents by:
• sorting, by the computing device, elements of the hashmap for the document and the hashmaps computed for the documents included in the established source collection according to the relevance values;
• and
• populating, by the computing device, the array of dominant words with top-ranked elements of the hashmap and the hashmaps according to the sorting until a defined threshold number of words have been added to the array of dominant words;
• establishing, by the computing device, dominant phrases having words included in the array of dominant words as theme phrase candidates for the text;
• and
• utilizing, by the computing device, one or more of the dominant phrases as theme phrases for a search to obtain suggested images for insertion at the position within the document.