Multilingual Data & Knowledge
SALLD-2 @ LREC 2022:
Reinforcing synergies between Sentiment Analysis and Linguistic Linked Data
The second workshop on Sentiment Analysis and Linguistic Linked Data (SALLD-2) was held in conjunction with LREC 2022 in Marseille, France, on June 24, 2022, aimed to foster research and discussion about the interoperability of Linguistic Linked Data (LLD) principles and Sentiment Analysis (SA). More specifically, the half-day workshop provided a platform to explore relevant methodologies, resources, tools, and applications underpinning this connection, as well as to understand the primary approaches – with their respective advantages and limitations – and case studies.
Following on SALLD-1, co-located with LDK 2021 – 3rd Conference on Language, Data and Knowledge, held in Zaragoza, Spain, on September 1, 2021 – the SALLD series was initiated in the framework of NexusLinguarum COST Action (European network for Web-centred linguistic data science, CA 18209), and with its support.
SALLD-2 featured an invited talk by Anna Fensel, Associate Professor at Wageningen University and Research and at the University of Innsbruck, entitled ‘From Data to Meaning in Representation of Emotions’. The talk focused on the research by Fensel and her team regarding emotion analysis based on knowledge graphs.
Five peer-reviewed papers were accepted to the workshop: three related to sentiment lexicons, one to the harmonization of language resources, and one to the exploitation of information available in the Linked Open Data cloud for sentiment analysis.
In the first category, the paper ‘Movie Rating Prediction using Sentiment Features’, by Apóstolo et al., describes a new sentiment lexicon, Expanded OntoSenticNet (EON), which combines OntoSenticNet with SentiWordNet and is used to predict movie ratings. Likewise, ‘Sentiment Analysis of Serbian Old Novels’, by Stanković et al., describes a polarity lexicon in Serbian based on three existing lexicons (NRC, Affin, and Bing), the Ontolex-lemon model, and the sentiment vocabulary Marl. Finally, ‘Evaluating a New Danish Sentiment Resource: the Danish Sentiment Lexicon, DSL’, by Schneidermann and Pedersen, describes an evaluation of Danish sentiment lexicons. We are pleased to highlight the importance of publishing non-English sentiment lexicons for resource-scarce languages.
The second category has been addressed in the paper ‘O-Dang! The Ontology of Dangerous Speech Messages’, by Stranisci et al., which defines an interoperable knowledge graph to link linguistic resources related to dangerous speech. The graph was populated with eight Italian language resources on dangerous speech, demonstrating the considerable potential of linked data technology to interlink language resources.
Finally, the use of open-linked data for sentiment analysis is exploited in ‘Correlating Facts and Social Media Trends on Environmental Quantities Leveraging Commonsense Reasoning and Human Sentiments’, by McNamee et al. This paper analyzed how external information (e.g. temperature or pollution) can be combined with the insights of sentiment analysis.
While SALLD is gradually evolving into a platform where different communities (computer science, linguistics, information science, knowledge engineering, language data science) interact and cooperate within the scope of SA and LLD, we note that the interoperability of these two factors is still rare in most of the submissions we receive. It is hoped that future editions of this workshop will contribute to further strengthening the relevant collaboration between these two fields.
website // publication // video recordings
Lexicala by K Dictionaries
Universidade de Aveiro
Università degli Studi di Parma
Carlos A. Iglesias
Universidad Politécnica de Madrid