METHOD AND SYSTEM FOR RECOMMENDING CONTENT BASED ON UNSTRUCTURED TEXT, STRUCTURED DATA AND USER INTERACTIONS WITH CONTENTS

A software system for recommending content that recommends combining two known filtering methods - content and collaborative with a third filtering - based on unstructured text, using the text in its entirety, as opposed to extracting keywords or automatically classifying texts by topic. Content fil...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: RAC, TIBOR, RASTOVIĆ, BORKO, OSTROGONAC, STEVAN, SANTRAČ, BORIS
Format: Patent
Sprache:eng ; srp
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:A software system for recommending content that recommends combining two known filtering methods - content and collaborative with a third filtering - based on unstructured text, using the text in its entirety, as opposed to extracting keywords or automatically classifying texts by topic. Content filtering based on unstructured text is performed using a combination of two models for performing vector representations of text - for sparse and compact representation. In this way, the mutual similarity of texts in the system is determined very precisely, and thus the relevance of individual content for a specific end user of the system, which is a set of content with which that user has previously interacted. Softverski sistem za preporučivanje sadržaja - Recommender Systems, koji karakteriše kombinovanje dva poznata načina filtriranja, sadržajno i kolaborativno, sa trećim filtriranjem - na osnovu nestrukturiranog teksta, pri čemu se tekst koristi u celosti, nasuprot izdvajanju ključnih reči ili automatskoj klasifikaciji tekstova prema temama. Filtriranje sadržaja na osnovu nestrukturiranog teksta se vrši korišćenjem kombinacije dva modela za izvođenje vektorskih reprezentacija teksta - za retku i kompaktnu reprezentaciju. Na ovaj način se veoma precizno određuje međusobna sličnost tekstova u sistemu, a samim tim i relevantnost pojedinačnih sadržaja za konkretnog krajnjeg korisnika sistema, kojeg predstavlja skup sadržaja sa kojima je taj korisnik ranije ostvario interakcije.