Text-as-Signal: Quantitative Semantic Scoring with Embeddings, Logprobs, and Noise Reduction

arXiv cs.AI / 4/16/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces a “text-as-signal” pipeline that converts a text corpus into quantitative semantic indicators by combining full-document embeddings with logprob-based scoring from a configurable positional dictionary.
In a case study, the authors apply the method to 11,922 Portuguese AI-related news articles using a six-dimension semantic dictionary to create a corpus “identity space” for both document-level and aggregated corpus-level characterization.
The workflow projects signals onto a noise-reduced low-dimensional manifold for structural interpretation, enabling clearer semantic positioning and comparison across documents.
It leverages Qwen embeddings, UMAP, directly model-output-space-derived semantic indicators, and a three-stage anomaly-detection procedure to support practical tasks like corpus inspection and monitoring.
The identity layer is designed to be configurable, allowing the framework to be adapted to different analytical needs rather than relying on a single fixed schema.

Abstract

This paper presents a practical pipeline for turning text corpora into quantitative semantic signals. Each news item is represented as a full-document embedding, scored through logprob-based evaluation over a configurable positional dictionary, and projected onto a noise-reduced low-dimensional manifold for structural interpretation. In the present case study, the dictionary is instantiated as six semantic dimensions and applied to a corpus of 11,922 Portuguese news articles about Artificial Intelligence. The resulting identity space supports both document-level semantic positioning and corpus-level characterization through aggregated profiles. We show how Qwen embeddings, UMAP, semantic indicators derived directly from the model output space, and a three-stage anomaly-detection procedure combine into an operational text-as-signal workflow for AI engineering tasks such as corpus inspection, monitoring, and downstream analytical support. Because the identity layer is configurable, the same framework can be adapted to the requirements of different analytical streams rather than fixed to a universal schema.