Abstract
This work demonstrates how neural network models (NNs) can be exploited towards resolving citation links in the scientific literature, which involves locating passages in the source paper the author had intended when citing the paper. We look at two kinds of models: triplet and binary. The triplet network model works by ranking potential candidates, using what is generally known as the triplet loss, while the binary model tackles the issue by turning it into a binary decision problem, i.e., by labeling a candidate as true or false, depending on how likely a target it is. Experiments are conducted using three datasets developed by the CL-SciSumm project from a large repository of scientific papers in the Association for Computational Linguistics (ACL) repository. The results find that NNs are extremely susceptible to how the input is represented: they perform better on inputs expressed in binary format than on those encoded using the TFIDF metric or neural embeddings. Furthermore, in response to a difficulty NNs and baselines had in predicting the exact location of a target, we introduce the idea of approximately correct targets (ACTs) where the goal is to find a region which likely contains a true target rather than its exact location. We show that with the ACTs, NNs consistently outperform Ranking SVM and TFIDF on the aforementioned datasets.
Highlights
The work described in this paper owes its birth to recent efforts at CL-SciSumm Shared Task Project (Jaidka et al, 2016) to develop a systematic approach to relating citing snippets to their sources in the paper they refer to
To get a broad picture of how the models perform, we introduce an idea we call approximately correct targets (ACTs), where we are interested in finding out whether they pick up exact sentences humans labeled as true targets, and finding out how close predictions are to the true targets
We have presented approaches to linking citation and reference that draw upon neural networks (NNs), and described in detail what machinery is involved and what we found in experiments with the three datasets, TAC2014, DSA2016, and SRD2016
Summary
The work described in this paper owes its birth to recent efforts at CL-SciSumm Shared Task Project (Jaidka et al, 2016) to develop a systematic approach to relating citing snippets to their sources in the paper they refer to. One particular (much publicized) feature of NNs is that they are an end-to-end system, meaning that they are designed to learn whatever features they need by themselves, freeing humans of the drudgery of making them up. This is something that has not been explored in the previous CL-SciSumm literature, with an exception of Nomoto (2016), who presented a preliminary attempt to leverage neural network to address resolving citation links, from which the current work descends.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
