template-oie

KO | EN

gitlite — search

#react #ai #typescript #android #rust #deep-learning #docker #nodejs #macos #awesome #cli #awesome-list #vue #ios

(★ 17)

Extract templated Open Information Extraction

template-oie Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Templated Open Information Extraction

- [Introduction](#introduction)
- [Prerequisites](#prerequisites)
- [Installation](#installation)
- [Running](#running)
- [Output format](#output-format)
- [Examples](#examples)
- [Users of template-oie](#users-of-template-oie)

## Introduction

Extracts templated Open Information Extraction: allowing for diverse, non-contiguous, multi-word predicates, while keeping the arguments short and useful for downstream applications.

For example, given the sentence:

_Under the agreement with the House and Senate leaders , the minimum wage would rise from the current $ 3.35 an hour to $ 4.25 an hour._

One of the extractions is:

* Under {A0} {A1} would **rise** from {A2} to {A3}

A0:	the agreement
        A1:	the minimum wage
        A2:	$ 3.35 an hour
        A3:	$ 4.25 an hour

Note that that the head of the predicate (**rise**) is also identified.

## Prerequisites
1. python 2.7
2. pip 9.x

## Installation
1. Install required packages:<br>
```bash
pip install -r ./requirements.txt
```
2. Download spaCy English models:<br>
```bash
python -m spacy download en
```

## Running 
```
Usage:
    prop_extraction --in=INPUT_FILE --out=OUTPUT_FILE [--id]

Extract propositions from a given input file, output is produced in separate output file.
If both in and out paramaters are directories, the script will iterate over all *.txt files in the input directory and
output to *.prop files in output directory.

Options:
   --in=INPUT_FILE      The input file, each sentence in a separate line
   --out=OUTPUT_FILE    The output file, Each extraction in a tab separated line, each consisting of original sentence,
   predicate template, lemmatized predicate template,argument name, argument value, ...
   --id                 Indicate that the input file is composed of sentence-id \t sentence, and copy this id in the output.
```

## Output format

Each extraction is presented in a tab separated line, consisting of:
1. Original sentence <br>
Words are separated by a single space, *chunks by double spaces*.
2. Index of the main predicate's chunk
3. Predicate template
4. Lemmatized predicate
6. Argument1 name
7. Argument1 value
8. ...

## Examples

See [the example folder](./examples) for the output over more than 3K sentences from the news domain.

```bash
python ./prop_extraction.py --in=../examples/sentences.txt --out=../examples/sentences.prop
```

## Users of template-oie

The following projects make use of template-oie:

* [Open Knowledge Representation](https://github.com/vered1986/OKR)
* [Chirps](https://github.com/vered1986/Chirps)

template-oie

template-oie Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

template-oie

template-oie Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README