Natural language processing and information systems : 25th international conference on applications of natural language to information systems, NLDB 2020, Saarbrücken, Germany, June 24–26, 2020, proceedings

Abstract

This volume contains the papers presented at NLDB 2020, the 25th International Conference on Applications of Natural Language to Information Systems held June 24–26, 2020, as a video conference at the German Research Center for Artificial Intelligence in Saarbrücken, Germany. We received 68 submissions for the conference.
Each paper was assigned to three reviewers, taking into account preferences expressed by the Program Committee members as much as possible. After the review deadline, Program Committee members were asked to complete missing reviews. In addition, Organization Committee members and the Program Committee chair acted as
meta-reviewers – they wrote additional reviews for borderline cases and for the papers which received reviews with considerably conflicting assessments. At the end, each paper received at least three reviews. On the basis of these reviews, the Organization Committee members and the Program Committee chair decided to accept papers with an average score around weak acceptance as full papers, papers with a
slightly lower score as short papers. In borderline cases, credit was given to experimentally-oriented papers with novel and ambitious concepts.
The final acceptance rate counting the number of full papers according to NLDB tradition was 22% (15 out of 68), similarly competitive in comparison to the previous
years. In addition, 10 submissions were accepted as short papers, and no posters, since NLDB 2020 had to be a video conference. Full papers were allowed a maximum of 12 and short papers a maximum of 8 pages. Originally, two more short papers were accepted, but the authors preferred to retract their submissions for personal reasons.
Following the trends of previous years, there is more diversification in the topics and specific issues addressed in comparison to a decade ago. Several papers address some languages for which not too rich resources are available – Arabic and Russian. Some currently hot topics are dealt with intensively, including sentiment analysis and chatbots, and successful tools are reused and adapted such as the transformer BERT. Finally, going beyond language proper is examined by several contributions, including visual data, affect, emotions, and personality.
In addition to the reviewed papers, there were three invited talks at NLDB 2020:
– Claire Gardent, LORIA Nancy, France
– Ehud Reiter, University of Aberdeen and Arria, UK
– Josef van Genabith, DFKI Saarbrücken, Germany
The accepted contributions (long and short papers) covered a wide range of topics, which we classified in six topic areas, each covering a section in this volume:
– Semantic Analysis
– Question Answering and Answer Generation
– Classification
– Sentiment Analysis
-- Personality, Affect, and Emotion
– Retrieval, Conversational Agents, and Multimodal Analysis

Semantic Analysis

Two long and two short papers were categorized in this section. The first one incorporates psycholinguistic evidence into subword properties for training vector representations. The next two papers address named-entity recognition, both in non-standard domains. The first one works in the cybersecurity domain in Russian, showing superiority of the BERT model. The second one addresses biomedical data
using a deep neural network (NN) architecture. The final paper in this section features explanations about semantic parsing, in the context of a natural language programming task.

Question Answering and Answer Generation

Three long and two short papers were categorized in this section. The first paper obtains performance increase through query expansion and coreference resolution measures. The next two papers extend question answering by dialog patterns, the first one automates building chatbots including clarification dialogs, the second one organizes answering procedural questions by incrementally following hierarchically organized knowledge. The last two papers address limitations of knowledge; the first one deals with large data in open domains, the second one obtains control over limitations caused by missing knowledge.

Classification

One long and three short papers were categorized in this section. There are two technology-based and two application-oriented contributions. One of the technology-based contributions applies an iterative procedure, exemplified for classifying short texts, the other one introduces systematic selection techniques to increase stability in Latent Dirichlet Allocation. The application-oriented contributions aim at
identifying reports about defects and associated requests for improvements within dedicated reviews and court decisions in the area of housing law, respectively.

Sentiment Analysis

Two long and one short paper were categorized in this section. Each approach features a specific, non-standard perspective. The first approach features the role of positional information attributed to a word contributing to an aspect of interest. The second one studies the role of attention and demonstrates its relevance for assessing
analytical texts. The last one emphasizes the exploitation of sentiment to drive strategies for curriculum learning.

Personality, Affect, and Emotion

Four long and one short paper were categorized in this section. The first two papers in this category address the role of personality for quite diverse purposes, one for the discrimination of honest versus deceptive authors, the other one to model language behavior of literary figures. The next two papers analyze emotions, one in the context of movies, the other one by analyzing low-level linguistic properties in social media
blogs. The last one attempts to infer doubts about a specific disease from analyzing social media dialogs.

Retrieval, Conversational Agents, and Multimodal Analysis

Three long and one short paper were categorized in this section. The first two address extended mechanisms for answer choice taking into account the dialog context and semantic similarity between a new question and already processed ones, respectively.
The next paper features anchoring entities based on textual and visual data, and the final paper describes a compound practical system with occasional human intervention.
The conference organizers are indebted to the reviewers for their engagement in a vigorous submission evaluation process. We would also like to thank, for the organization help, some members of the DFKI GmbH.

Citation

(2020). E. Métais, F. Meziane, H. Horacek, & P. Cimiano (Eds.), Natural language processing and information systems : 25th international conference on applications of natural language to information systems, NLDB 2020, Saarbrücken, Germany, June 24–26, 2020, proceedings. Switzerland: Springer. https://doi.org/10.1007/978-3-030-51310-8

• Book Type: Authored Book
• Online Publication Date: Jun 17, 2020
• Publication Date: Jun 1, 2020
• Deposit Date: Jul 6, 2020
• Series Title: Lecture Notes in Computer Science (LNCS)
• ISBN: 9783030513092-(print);-9783030513108-(online)
• DOI: https://doi.org/10.1007/978-3-030-51310-8
• Publisher URL: https://doi.org/10.1007/978-3-030-51310-8
• Related Public URLs: https://www.springer.com/gp/