The 1st Workshop on Extraction and Evaluation of Knowledge Entities from Scientific Documents (EEKE 2020) was launched at the ACM/IEEE Joint Conference on Digital Libraries (JCDL) on August 1, 2020. The goal of this workshop is to engage the related communities in open problems in the extraction and evaluation of knowledge entities from scientific documents. Participants are encouraged to identify knowledge entities, explore feature of various entities, analyze the relationship between entities, and construct the extraction platform or knowledge base. Results of this workshop are expected to provide scholars, especially early career researchers, with knowledge recommendations and other knowledge entity-based services [1].
This year, 14 papers (including 3 long papers, 6 short papers, 4 posters and 1 demo) were accepted for presentation and inclusion in the proceedings. In addition, the workshop featured two keynote talks in the different EEKE-related fields. All workshop contributions are documented in the workshop website 1 . The following section briefly lists the various contributions.
Two keynotes were presented in EEKE2020.
The first one was given by Ming Song: Entitymetrics 2.0: Measuring the Impact of Entities and Relations Extracted from Scientific Documents. The concept of entitymetrics was first introduced in 2013, entitymetrics [2] has been applied to measure the impact of entities as well as to gauge the knowledge usage and transfer anchored on entities for knowledge discovery. In this talk, the previous studies employing entitiymetrics are summarized and the limitations of the current approaches are discussed. In addition, the future directions of entitymetrics are suggested.
The second keynote was given by Markus Stocker: Building Scholarly Knowledge Bases with Crowdsourcing and Text Mining. Building on the Open Research Knowledge Graph (http://orkg.org) as a concrete research infrastructure, in this talk Dr. Stocker presented how using crowdsourcing and text mining humans and machines can collaboratively build scholarly knowledge bases. He discussed some key challenges that human and technical infrastructures face as well as the possibilities scholarly knowledge bases enable.
The following papers were presented in 5 sessions.
This paper describes an annotation initiative to capture the scholarly contributions in natural language processing (NLP) articles, particularly, for the articles that discuss machine learning (ML) approaches for various information extraction tasks. They attempted to find a systematic set of patterns of subject-predicate-object statements for the semantic structuring of scholarly contributions, and to apply the discovered patterns in the creation of a larger annotated dataset for training machine readers of research contributions.
This paper proposes an adaptable and transferable methodology to extract biomedical entities including diseases, chemicals, genes and genetic variations from literature data. A heterogeneous co-occurrence network is constructed and a semantic adjacency matrix is generated to identify key genes and genetic variants, and capture the emerging disease-gene associations via a link prediction approach.
This research designs and implements an academic search system with two major innovations: 1) proposing a general query language SSL to describe the academic search intention in a unified and standardized way, 2) proposing a user intention recognition method to help improve traditional automatic question-answering systems. The SSL language and intention recognition method is applied to a QA-oriented academic system which is innovative compared with traditional query-based systems.
-Xin An, Jinghong Li, Shuo Xu, Liang Chen and Sainan Pi
In order to measure the similarity among different patents, this study proposes a novel approach on the basis of sequence alignment. The method takes semantic direction of each sequence structure and the word order information of each component into consideration; an algorithm for calculating the global importance of each sequence structure is put forward. Extensive experimental results show that the proposed approach is significantly more accurate and is not sensitive to several core parameters.
-Fang Tan, Siting Yang, Xiaoyan Wu and Jian Xu
In order to analyze the effect of government funding on the promotion of scientific research, and to help the government manage research funds more rationally, this study proposes a framework for analyzing the relationship between entities in the field of medicine and funds.
The results reveal that the field of genetic research is in a period of rapid development and disease research catch NIH's continuous attention. However, the stimulating effect of government funding on the research popularity is decreasing.
-Sahand Vahidnia, Alireza Abbasi and Hussein A. Abbass
In this study, a method is proposed to extract research trends and their temporal evolution, throughout discrete time periods. Adapting contextualized word embedding techniques, the method utilizes published academic documents as knowledge units and clusters them into groups. Various labeling techniques are explored to evaluate the quality of clusters and explore their explain ability. The results show that utilization of neural embedding in conjunction with paragraph-term weights would provide simple and reliable paragraph embedding that can be used for clustering of the textual data.
Session 5: Poster/ Greeting Notes of EEKE2020 -Qikai Liu, Pengcheng Li, Wei Lu and Qikai Cheng
Datasets play an important role in data-driven scientific research. It is important to recognize dataset entities correctly, especially when it comes to unusual long-tail dataset entities. However, it is very difficult to obtain high quality training corpus in named entity recognition. This paper obtained the data based on a distant supervision method along with two data augmentation methods. A BERT-BiLSTM-CRF model is used to predict long-tail dataset entity.
-Xiaole Li, Yuzhuo Wang Assessing Impact of Method Entities in a Special Task Methods play an important role in the research. Identifying and analyzing entities about research methods can help scholars understand methods used in their field and accelerate the efficiency of scientific research. This paper takes named entity recognition (NER) as an example and evaluates the impact of method entities in this domain. They found that conditional random field (CRF) is the most influential algorithms in NER. Deep learning algorithms have developed rapidly in the past 5 years. F-measure, precision and recall are the most widely used indices and measurements. Scholars do not pay enough attention to use tools and they prefer to use classic datasets.
-Chong Chen, Jingying Zhang, Xiaoyu Chu and Jinglin Zheng
This article proposes primary measurement to compare Summary peer reviews with abstracts from readability and semantic function types. The results show that summary peer reviews highlight some distinct function types, and the in peer is not as dense as in abstracts. Summary peer reviews can be complement to abstracts in literature searches, and can help readers understanding papers more thoroughly.
-Wei Shao, Hua Bolin, Qiang Ma, Jiaying Liu, Hongwei He, Keqi Chen
Finding new terminology is a kind of named entity recognition (NER) problem. However, many high performance methods need labeled data. This paper proposes an unsupervised method based on sentence pattern and part of speech. They initialize a few patterns to extract terminologies in certain sentences, and then try to find the same POS sequences in sentences not matched by initial patterns with obtained terminologies' POS sequences. The new patterns and more terminologies are obtained after several iterations.
Currently the EEKE2020 organizers edit the following two Special issues:
-Special Issue on "Extraction and Evaluation of Knowledge Entities from Scientific Documents" in Journal of Data and Information Science (https://mc03.manuscriptcentral.com/jdis).
-Special Issue on "Scientific Documents Mining and Applications" in Data and Information Management (https://www.editorialmanager.com/dim/default.aspx).