Temporal Information Extraction for Question Answering Using Syntactic Dependencies in an LSTM-based Architecture

Abstract

In this paper, we propose to use a set of simple, uniform in architecture LSTMbased models to recover different kinds of temporal relations from text. Using the shortest dependency path between entities as input, the same architecture is implemented to extract intra-sentence, crosssentence, and document creation time relations. A "double-checking" technique reverses entity pairs in classification, boosting the recall of positive cases and reducing misclassifications between opposite classes. An efficient pruning algorithm resolves conflicts globally. Evaluated on QA-TempEval (SemEval2015 Task 5), our proposed technique outperforms state-ofthe-art methods by a large margin. We also conduct intrinsic evaluation and post stateof-the-art results on Timebank-Dense.

Context-Aware Neural Model for Temporal Information Extraction

“…Following Meng et al (2017), we augment the data by flipping all pairs, except for relations involving document creation time (DCT). In other words, if a pair (e i , e j ) exists, we add (e j , e i ) to the dataset with the opposite label (e.g.…”

“…TimeBank-Dense dataset labels three types of pairs: intra-sentence, cross-sentence and document creation time (DCT). For intra-sentence pairs and cross-sentence pairs, we follow Meng et al (2017). The shortest dependency path between the two entities is identified, and the word embeddings from the path to the least common ancestor for each entity are processed by two LSTM branches, with a separate max pooling layer for each branch.…”

“…Recently, there have been some attempts to extract temporal relations with neural network models, particularly with recurrent neural networks (RNN) models (Meng et al, 2017;Cheng and Miyao, 2017;Tourille et al, 2017) and convolutional neural networks (CNN) (Lin et al, 2017). These models predominantly use token embeddings as input, avoiding handcrafted features for each task.…”

“…High-ranking results overwrite low-ranking ones. Meng et al (2017) used a greedy pruning algorithm to remove weak edges from the timegraph until it is coherent.…”

Embedding Time Expressions for Deep Temporal Ordering Models

“…Inferring temporal event order is challenging as it often disagrees with the narrative order in text. Past work on temporal relation extraction has exploited cues such as global constraints on the temporal graph structure (Bramsen et al, 2006;Chambers and Jurafsky, 2008;Ning et al, 2017), world knowledge (Ning et al, 2018b), grouping of events (Tourille et al, 2017), or fusing these cues more effectively with deep models (Meng et al, 2017;Cheng and Miyao, 2017). One key component of temporal understanding is time expressions (timexes) that help anchor events to the time axis, but few recent systems effectively use the knowledge derivable from time expressions in their models.…”