README

##Description## ExtrHech is an Open Information Extraction system for Spanish language based on matching sequences of Part-of-Speech tags. It has been created as a part of my PhD thesis in the Center for Computing Research of Instituto Politécnico Nacional (CIC-IPN) in Mexico City, Mexico. "Orgullosamente politécnico."

Essentially, it is a system for detection and extraction of potentially infromative verb-based triples, i.e., relations along with their arguments, from arbitrary texts in Spanish language. Of course, the text must be pre-processed correspondingly (see below).

You can read more about the details and theoretical background in my paper Open Information Extraction for Spanish Language based on Syntactic Constraints co-authored with my advisor Prof. Alexander Gelbukh

Currently, ExtrHech definitely lacks industrial strength but it works... or more precisely, worked a few times.

###Keywords### Natural Language Processing, Open Information Extraction, Part-of-Speech tags, Relation Extraction, PhD thesis

What do we have here?

The ./src folder contains all files ever related to the project, a lot of which are outdated and not necessary but I keep them all as a history and push them diligently.

In the ./data folder there are couple files for testing:

• news-v6.es-en.300.es - 300 sentences from Reuters news corpus
• news-v6.es-en.300.es.pos - same 300 sentences POS-tagged according to EAGLES POS tag set. POS-tagging was done using Freeling-2.2 POS-tagger (see below)

In the ./output folder:

• output.extr - what to expect as output

How to run it?

If you happen to have your texts POS-tagged with EAGLES POS-tages and formatted one sentence per line then all you need is to run:

> python ./src/fact_extr_regexp4.py  ./src/facts_extr.config your_input.pos > your_output.extr

To try it out, you can just blindly run

python ./src/fact_extr_regexp4.py ./src/facts_extr.config  ./data/news-v6.es-en.300.es.pos > ./output/my_output.extr

which is equivalent to the ExtrHech.bat file.

However, if you do need to POS-tag your dataset first:

• download Freeling
• to get the idea how to proceed, check ./extraction_pipeline.bat ... yes, it's .bat.

Who to talk to?

• talk to me