TY - JOUR
T1 - SoulMate: Short-Text Author Linking through Multi-Aspect Temporal-Textual Embedding
AU - Najafipour, Saeed
AU - Hosseini, Saeid
AU - Hua, Wen
AU - Kangavari, Mohammad Reza
AU - Zhou, Xiaofang
N1 - Funding Information:
In this paper, we propose a novel framework that processes short-text contents (e.g. tweets) to exploit subgraphs with highly correlated authors. To this end, we first need to link authors through computing the similarity weights, which results in the authors’ weighted graph. Primarily, the time-aware word embedding model considers temporal-textual evidence to infer the similarity rate between temporal splits in multiple dimensions (e.g. Monday and Tuesday in day dimension) and collectively computes the word vector representations. Subsequently, we obtain short-text vectors and author content vectors through combining the word vectors. Similarly, author concept vectors represent how every author is relevant to each of the short-text clusters. We notice that compared to DBSCAN, the k-medoids clustering can better discover the concepts from tweet contents. We then fuse the content-based and conceptual author similarities to calculate the correlation weight between each pair of authors. Consequently, given the authors’ weighted graph, the stack-wise graph-cutting component in our framework can extract the maximum spanning trees, establishing the subgraphs with highly correlated authors. The result of the extensive experiments on a real-world microblog dataset proves the superiority of our proposed temporal-textual framework in short text author linking. To conclude, the short-texts differ insignificance. Hence, to nominate the concepts from short-text clusters, we should not only consider the relevance between the tweet but also grant higher importance to the concepts of those with higher popularity. We leave this task for future work. 7 ACKNOWLEDGMENT This work was supported by both ST Electronics and the National Research Foundation(NRF), Prime Minister’s Office, Singapore under Corporate Laboratory @ University Scheme (Programme Title: STEE Infosec - SUTD Corporate Laboratory). REFERENCES [1]
Publisher Copyright:
© 1989-2012 IEEE.
PY - 2022/1/1
Y1 - 2022/1/1
N2 - Linking authors of short-text contents has important usages in many applications, including Named Entity Recognition (NER) and human community detection. However, certain challenges lie ahead. First, the input short-text contents are noisy, ambiguous, and do not follow the grammatical rules. Second, traditional text mining methods fail to effectively extract concepts through words and phrases. Third, the textual contents are temporally skewed, which can affect the semantic understanding by multiple time facets. Finally, using knowledge-bases can make the results biased to the content of the external database and deviate the meaning from the input short text corpus. To overcome these challenges, we devise a neural network-based temporal-textual framework that generates the subgraphs with highly correlated authors from short-text contents. Our approach, on the one hand, computes the relevance score (edge weight) between the authors through considering a portmanteau of contents and concepts, and on the other hand, employs a stack-wise graph cutting algorithm to extract the communities of the related authors. Experimental results show that compared to other knowledge-centered competitors, our multi-aspect vector space model can achieve a higher performance in linking short-text authors. In addition, given the author linking task, the more comprehensive the dataset is, the higher the significance of the extracted concepts will be.
AB - Linking authors of short-text contents has important usages in many applications, including Named Entity Recognition (NER) and human community detection. However, certain challenges lie ahead. First, the input short-text contents are noisy, ambiguous, and do not follow the grammatical rules. Second, traditional text mining methods fail to effectively extract concepts through words and phrases. Third, the textual contents are temporally skewed, which can affect the semantic understanding by multiple time facets. Finally, using knowledge-bases can make the results biased to the content of the external database and deviate the meaning from the input short text corpus. To overcome these challenges, we devise a neural network-based temporal-textual framework that generates the subgraphs with highly correlated authors from short-text contents. Our approach, on the one hand, computes the relevance score (edge weight) between the authors through considering a portmanteau of contents and concepts, and on the other hand, employs a stack-wise graph cutting algorithm to extract the communities of the related authors. Experimental results show that compared to other knowledge-centered competitors, our multi-aspect vector space model can achieve a higher performance in linking short-text authors. In addition, given the author linking task, the more comprehensive the dataset is, the higher the significance of the extracted concepts will be.
KW - Author linking
KW - Semantic understanding
KW - Short text inference
KW - Temporally multifaceted
KW - Word2Vec
UR - http://www.scopus.com/inward/record.url?scp=85097274499&partnerID=8YFLogxK
U2 - 10.1109/TKDE.2020.2982148
DO - 10.1109/TKDE.2020.2982148
M3 - Journal article
AN - SCOPUS:85097274499
SN - 1041-4347
VL - 34
SP - 448
EP - 461
JO - IEEE Transactions on Knowledge and Data Engineering
JF - IEEE Transactions on Knowledge and Data Engineering
IS - 1
ER -