OPTA – Computer-based methods of identifying opinion targets and opinion expressions in Polish texts
The project investigated syntactic relations between opinion targets and opinion expressions, using parsing and machine learning to link opinions with their targets in Polish texts.
Lead Investigator: Aleksander Wawer
- Grant Number
- 2012/05/N/ST6/03587
- Start Date
- January 2013
- End Date
- January 2016
Project factsheet
| Polish name: | Automatyczne metody rozpoznawania przedmiotów i wyrażeń opinii w języku polskim |
| Project type: | A National Science Centre grant 2012/05/N/ST6/03587 |
| NCN project info: | https://projekty.ncn.gov.pl/index.php?s=5389 |
Project summary
The project is focused on investigating syntactic relations between opinion targets and opinion expressions. The goal is to use available parsing techniques (eg. shallow and dependency) and machine learning to facilitate automated association of opinion expressions with their targets. The project involves corpus-based analyses based on the National Corpus of Polish (NKJP) and opinion corpora obtained in the NEKST project.
Resources
Treebanks of Opinion Targets in Polish:
- Product review sentences annotated with sentiments and opinion targets: download.
- A subset of Skladnica sentences annotated with sentiments and opinion targets: download.
Dependency Patterns for Extracting Opinion Targets in Polish with pattern matching code: download.
OPFI - Opinion Finder - a hybrid (CRF+patterns) application for extracting opinion targets in Polish: download.
word2vec vectors - trained on various combinations of Wikipedia and the National Corpus of Polish: visit page.