2 Pages Report + Coding

Assignment 3: Relation Extraction

In this assignment, you will implement a Relation Extraction (RE) system, making use of the output of the various NLP tools you learned about in the course. The assignment is somewhat open-ended, and you are not restricted as to the methods you are expected to use. You can write either a rule-based system, a machine-learning based system, or a hybrid system combining both approaches.

The assignment is to be submitted in pairs

Relation Extraction (RE)

RE aims at extracting semantic relations between entities from a given text. We will focus on extracting binary relations, i.e. relations between exactly two entities, from sentences. We will denote the relations rel(arg1,arg2) where arg1 and arg2 are the two entities in a predefined relation rel within a sentence S.

Given a sentence, the task is to decide whether the sentence contains a mention of a relation of interest, and if yes -- extract the relation name and the two corresponding entities.

In this assignment we will be looking for the following relations: Work for, e.g. Work_For(Yoav Goldberg, Bar Ilan University) Live in, e.g. Live_In(Ido Dagan, Israel) Note that the order of the entities within a relation is important, e.g. compare Work_For(Yoav Goldberg, Bar Ilan University) and Work_For(Bar Ilan University, Yoav Goldberg).

Requirements

You need to implement:

1) a system that gets as input a text file with multiple sentences, and returns as output a list of extracted relations together with the sentences the relations were extracted from. (60%)

2) an evaluation program that gets your programs's output and the gold-standard data, and outputs the Precision, Recall and F-1 scores. (10%)

You will also submit a report. (30%

Engineering Java PHP Research Writing Software Architecture

Project ID: #5802516

About the project