Coreference Resolution in Research Papers from Multiple Domains

Analysing the requirements for an Open Research Knowledge Graph: use cases, quality requirements, and construction strategies

“…Pajura and Singh [88] give an overview of the involved tasks for KG population: (a) information extraction to extract a graph from text with entity extraction and relation extraction, and (b) graph construction to clean and complete the extracted graph, as it is usually ambiguous, incomplete and inconsistent. Coreference resolution [17,71] clusters different mentions of the same entity in text and entity linking [63] maps mentions in text to entities in the KG. Entity resolution [104] identifies objects in the KG that refer to the same underlying entity.…”

“…Kannan et al [58] create a multimodal KG for deep learning papers from text and images and the corresponding source code. Brack et al [17] generate a KG for 10 different science domains with the concept types material, method, process, and data. Zhang et al [115] suggest a rule-based approach to mine research problems and proposed solutions from research papers.…”

“…More recent corpora, annotated at phrasal level, aim at constructing a fine-grained KG from scholarly abstracts with the tasks of concept extraction [4,15,44,71,89], binary relation extraction [4,45,71], n-ary relation extraction [55,57,59], and coreference resolution [17,25,71]. They cover several domains, e.g.…”

“…They cover several domains, e.g. material sciences [44]; computational linguistics [45,89]; computer science, material sciences, and physics [4]; machine learning [71]; biomedicine [25,57,64]; or a set of ten scientific, technical and medical domains [15,17,37]. The datasets differentiate between four to seven concept classes (like task, method, tool) and between two to seven binary relation types (like used-for, part-of, evaluatefor).…”

“…Approaches for concept extraction achieve F1 scores ranging from 56.6 to 96.9% (see Table 4), for coreference resolution F1 scores range from 46.0 to 61.4% [17,25,71], and for binary relation extraction from 28.0 to 83.6% (see Table 3). The task of n-ary relation extraction with an F1 score from 28.7 to 56.4% [57,59] is especially challenging, since such relationships usually span beyond sentences or even sections and thus, machine learning models require an understanding of the whole document.…”