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

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—distribution 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% reduction in forecasting error compared to state-of-the-art unimodal and multimodal baselines. Code is available at: https://github.com/olivia3395/TESS.

Real-world forecasts, such as stock prices or electricity demand, often change suddenly after news, weather, or public events. Text can explain these changes, but forecasting models struggle to use it because language is vague while predictions require numerical signals. We propose TESS, which translates text into simple descriptions of future behavior, such as rise, volatility, shape, and delayed effects, and filters uncertain signals with a confidence mechanism. This helps models make more accurate forecasts under event-driven changes, as shown on financial, electricity, and environmental datasets.