New method classifies sentence time without pretrained text embeddings
Classifying Dominant Temporal Orientation without Pretrained Text Embeddings: A Novel Morphosyntactic Inventory Vector Approach
Computation and Language
Summary
Determining whether a sentence refers to the past, present, or future is challenging, especially when sentences are complex. The authors propose a way to figure out the main time frame of a sentence by counting parts of speech, relationships between words, and specific future-related patterns without relying on existing language models. They tested their method on nearly 1,800 complex English sentences and found it correctly identified the time frame 92% of the time. This approach is simpler and doesn’t use large, pretrained text models.
What this means in practice
- •For natural language processing engineers: Improve time frame detection in complex sentences using a lightweight approach that avoids pretrained embeddings and large models.
- •For language technology product teams: Create faster text processing tools that identify sentence tense accurately without requiring large pretrained models, lowering computational costs.$Commercial implications: Enables development of efficient language processing features in commercial applications by avoiding costly pretrained embedding dependencies.
Authors
Jonathan Cleveland, Peter S. Bearman
Abstract
Computational methods have consistently struggled to determine the dominant temporal orientation of a sentence. This difficulty is especially pronounced when a sentence contains multiple embedded clauses with competing tense and aspectual information. To address this difficulty, we propose an alternative approach for identifying a sentence's past, present, or future global reference interval. Our method does not use any form of pretrained embeddings. We rather encode sentences using fixed-length inventory vectors that are comprised of part-of-speech counts, dependency relation counts, and explicit futurate pattern counts. We term this inventory vector of a sentence a "Morphosyntacton". The method does not use any padding, sequence models, or large language models. Evaluation on 1,799 syntactically complex English sentences, annotated as past, present or future, shows balanced and high accuracy multiclass classification, achieving an overall multi-class accuracy of 92%.