Temporal Information Extraction from Textual Data using Long Short Term Memory Recurrent Neural Network

Abstract

Temporal information extraction from raw text is always challenging. It is time consuming and sometimes difficult to extract temporal expression manually. For this reason, an automatic system is a demand to find the temporal expressions from the textual data automatically. In this paper, we have developed a temporal information extraction system using Long Short Term Memory (LSTM) recurrent neural network (RNN) along with word embedding where temporal expressions are extracted from TempEval-2 dataset. Performance of the proposed LSTM RNN based system is highly comparable with the other entries of TempEval-2 challenge. As LSTM RNN can handle both long and short term dependencies, the proposed system shows robust result than other renowned existing systems.

Keywords: Temporal information; LSTM RNN; TempEval-2; Word embedding, TIMEX3