WladimirSidorenko / PotTS

The Potsdam Twitter Sentiment Corpus

Description

This directory contains the data of the Potsdam Twitter Sentiment Corpus (ISLRN 714-621-985-491-3). To open the files of this corpus, you need to download and launch MMAX2—a freely distributed annotation tool—and then select one of the *.mmax projects from the directories corpus/annotator-1/ or corpus/annotator-2/.

Folder Structure

The folders of this project are structured as follows:

• corpus/ – directory containing corpus files;

  • annotator1/ – directory containing MMAX projects for the first annotator;
    • markables/ – directory containing annotation files for the first annotator;
  • annotator2/ – directory containing MMAX projects for the second annotator;
    • markables/ – directory containing annotation files for the second annotator;
  • basedata/ and source/ – original corpus tokenization;
  • custom/, scheme/, and style/ – auxiliary MMAX2 data;

• docs/ – directory containing annotation guidelines and other accompanying documents;

• scripts/ – directory containing scripts that were used to process corpus data;

  • examples/ – directory containing examples of input files for the scripts;
  • align.py – auxiliary module used for annotation alignment;
  • alt_fio.py – auxiliary module for AWK-like input/output operations;
  • conll.py – auxiliary module for handling CONLL sentences;
  • measure_corpus_agreement.py – script for measuring corpus agreement;
  • merge_conll_mmax.py – script for aligning annotation from the corpus with the automatically processed CONLL data;

You can see the examples of invocations in the script files or by just typing --help to see their usage.

Note

I strongly recommend using the annotation of annotator-2 on the branch eexpression-revision (run git checkout eexpression-revision after cloning this project).