Distantly Supervised Relation Extraction in Federated Settings

Abstract

In relation extraction, distant supervision is widely used to automatically label a largescale training dataset by aligning a knowledge base with unstructured text. Most existing studies in this field have assumed there is a great deal of centralized unstructured text. However, in practice, texts are usually distributed on different platforms and cannot be centralized due to privacy restrictions. Therefore, it is worthwhile to investigate distant supervision in the federated learning paradigm, which decouples the training of the model from the need for direct access to raw texts. However, overcoming label noise of distant supervision becomes more difficult in federated settings, because texts containing the same entity pair scatter around different platforms. In this paper, we propose a federated denoising framework to suppress label noise in federated settings. The key of this framework is a multiple instance learning based denoising method that is able to select reliable sentences via cross-platform collaboration. Various experiments on New York Times dataset and miRNA gene regulation relation dataset demonstrate the effectiveness of the proposed method. 1

Contrast with Major Classifier Vectors for Federated Medical Relation Extraction with Heterogeneous Label Distribution

“…FL is an emerging learning paradigm seeking to address the problem of data governance and privacy by training algorithms collaboratively without exchanging the data itself [2]. FedED [16] and Lazy MIL [17] have studied supervised and distant federated relation extraction, respectively. The vanilla FL algorithm, FedAvg, periodically aggregates the local models in the server and updates the local model with its individual data.…”

CLART: A cascaded lattice-and-radical transformer network for Chinese medical named entity recognition

“…Although the aforementioned models have demonstrated promising results in the general domain [39] , [28] , [17] , [10] , Chinese NER faces significant challenges when applied in the medical domain. This can be attributed primarily to the extensive range of medical terms involved, which requires a profound understanding of medical knowledge for accurate recognition [40] .…”

“…Due to the broad range of applications associated with the extracted medical entities, such as medical knowledge graph construction [24] , medical relation extraction [38] , and medical question answering [15] , Chinese medical NER has attracted considerable attention [40] , [18] , [27] , [34] . Although Chinese NER has shown promising results in the general domain [39] , [28] , [17] , [10] , it encounters significant challenges when applied in the medical domain. This is primarily due to the extensive range of medical terms involved in Chinese medical NER, which usually requires a deep understanding of medical texts to ensure accurate recognition [40] .…”

Extracting Chinese events with a joint label space model

“…In this work, we introduce a novel M ulti-layer L abel A ttentive framework to improve Chinese E vent E xtraction (MLAEE). For the above first issue, there have been studies showing that integrating lexicon features into character-based networks could lead to better entity recognition performance [ 28 , 29 ]. Inspired by these methods, we propose to perform event extraction based on characters and enhance character representations by introducing the word lexicon, which is presented in Fig 1(b) .…”