From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space

TL;DR

This paper introduces TESS, a novel approach that creates an interpretable semantic space to effectively translate textual descriptions into numerical signals for improved time-series forecasting, addressing the modality gap.

Contribution

The paper proposes TESS, a new framework that bridges the modality gap by extracting interpretable temporal primitives from text to enhance forecasting accuracy.

Findings

Up to 29% reduction in forecasting error.

Effective translation of textual semantics into numerical cues.

Robust performance across four real-world datasets.

Abstract

Incorporating textual information into time-series forecasting holds promise for addressing event-driven non-stationarity; however, a fundamental modality gap hinders effective fusion: textual descriptions express temporal impacts implicitly and qualitatively, whereas forecasting models rely on explicit and quantitative signals. Through controlled semi-synthetic experiments, we show that existing methods over-attend to redundant tokens and struggle to reliably translate textual semantics into usable numerical cues. To bridge this gap, we propose TESS, which introduces a Temporal Evolution Semantic Space as an intermediate bottleneck between modalities. This space consists of interpretable, numerically grounded temporal primitives (mean shift, volatility, shape, and lag) extracted from text by an LLM via structured prompting and filtered through confidence-aware gating. Experiments on four real-world datasets demonstrate up to a 29 percent reduction in forecasting error compared to state-of-the-art unimodal and multimodal baselines. The code will be released after acceptance.