Bisociative Literature-Based Discovery: Lessons Learned and New Word Embedding Approach

Lavrač, Nada; Martinc, Matej; Pollak, Senja; Pompe-Novak, Maruša; Cestnik, Bojan

doi:10.1007/s00354-020-00108-w

Cited by 7 publications

(3 citation statements)

References 33 publications

(62 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Broadly, however, we aim for annotators to rate structural/relational similarities between the candidate and query higher (3-2 ratings) than attribute/feature based similarities (1-0 ratings). This draws on motivations from a range of literature highlighting the value of structural similarities to creative activities like scientific research -a focus of this work [8,15,30,33]. While we refer readers to our annotator guidelines for detailed instructions for the relevance grades, we present a summary of the highest grade by facet as an illustration here: background: +3 implies that both papers are trying to accomplish the same specific goal, are solving the same specific modeling problem or the same machine learning task, or are motivated in a specific similar manner.…”

Section: Relevance Ratingsmentioning

confidence: 99%

CSFCube -- A Test Collection of Computer Science Research Articles for Faceted Query by Example

Mysore¹,

O'Gorman²,

McCallum³

et al. 2021

Preprint

View full text Add to dashboard Cite

Query by Example is a well-known information retrieval task in which a document is chosen by the user as the search query and the goal is to retrieve relevant documents from a large collection. However, a document often covers multiple aspects of a topic. To address this scenario we introduce the task of faceted Query by Example in which users can also specify a finer grained aspect in addition to the input query document. We focus on the application of this task in scientific literature search. We envision models which are able to retrieve scientific papers analogous to a query scientific paper along specifically chosen rhetorical structure elements as one solution to this problem. In this work, the rhetorical structure elements, which we refer to as facets, indicate "background", "method", or "result" aspects of a scientific paper. We introduce and describe an expert annotated test collection to evaluate models trained to perform this task. Our test collection consists of a diverse set of 50 query documents, drawn from computational linguistics and machine learning venues. We carefully followed the annotation guideline used by TREC for depth-k pooling (k = 100 or 250) and the resulting data collection consists of graded relevance scores with high annotation agreement. The data is freely available for research purposes.

show abstract

Section: Relevance Ratingsmentioning

confidence: 99%

CSFCube -- A Test Collection of Computer Science Research Articles for Faceted Query by Example

Mysore¹,

O'Gorman²,

McCallum³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Arthur Koestler considered bisociation to be the foundation of human creativity in science and art. He also thought that bisocial thinking occurs when a problem is perceived simultaneously in two or more different domains [7]. Figure 7 presents an example of bisociative network.…”

Section: Bisociative Information Networkmentioning

confidence: 99%

Information networks – concept, classification and application

Salihović

Džubur

2021

Sci. Eng. Technol.

View full text Add to dashboard Cite

An information network is a structure used for transmitting various forms and types of information. In basic structure, it consists of branches that connect certain nodes. Many scientists and researchers have dealt with the problem of defining an information network, depending on the functional organization and data transmission, and also with the classification of information networks. Information networks have a very wide application in almost all scientific disciplines. A large number of researches are carried out on the application of information networks (e.g. bisociative, deep information network, heterogeneous information network, and space information network) in the field of medicine for easier detection of diseases, drug development, etc, and other needs to support real-time communication, massive data transmission, and data processing. In accordance with the above mentioned, the aim of this paper is to offer different approaches in defining and classifying general forms of information networks and to notice their wide application in different research disciplines.

show abstract

“…A study by Tshitoyan et al (2019) showed that latent knowledge regarding future discoveries is to a large extent embedded in past publications by retrieving information from the scientific literature with the usage of word2vec embeddings (Mikolov et al, 2013). The recent approach by Lavrač et al (2020) explored how word embeddings (Joulin et al, 2016) can be used for identification of novel bridging terms in the field of plant biology. A similar approach was also explored in the context of COVID-19-related biomarker discovery (Martinc et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

PubMed-Scale Chemical Concept Embeddings Reconstruct Physical Protein Interaction Networks

Škrlj

Kokalj

Lavrač

2021

Front. Res. Metr. Anal.

Self Cite

View full text Add to dashboard Cite

PubMed is the largest resource of curated biomedical knowledge to date, entailing more than 25 million documents. Large quantities of novel literature prevent a single expert from keeping track of all potentially relevant papers, resulting in knowledge gaps. In this article, we present CHEMMESHNET, a newly developed PubMed-based network comprising more than 10,000,000 associations, constructed from expert-curated MeSH annotations of chemicals based on all currently available PubMed articles. By learning latent representations of concepts in the obtained network, we demonstrate in a proof of concept study that purely literature-based representations are sufficient for the reconstruction of a large part of the currently known network of physical, empirically determined protein–protein interactions. We demonstrate that simple linear embeddings of node pairs, when coupled with a neural network–based classifier, reliably reconstruct the existing collection of empirically confirmed protein–protein interactions. Furthermore, we demonstrate how pairs of learned representations can be used to prioritize potentially interesting novel interactions based on the common chemical context. Highly ranked interactions are qualitatively inspected in terms of potential complex formation at the structural level and represent potentially interesting new knowledge. We demonstrate that two protein–protein interactions, prioritized by structure-based approaches, also emerge as probable with regard to the trained machine-learning model.

show abstract

Bisociative Literature-Based Discovery: Lessons Learned and New Word Embedding Approach

Cited by 7 publications

References 33 publications

CSFCube -- A Test Collection of Computer Science Research Articles for Faceted Query by Example

CSFCube -- A Test Collection of Computer Science Research Articles for Faceted Query by Example

Information networks – concept, classification and application

PubMed-Scale Chemical Concept Embeddings Reconstruct Physical Protein Interaction Networks

Contact Info

Product

Resources

About