With Learning-To-Rank Topic Modeling

Topic-based ranking of authors, papers and journals can serve as a vital tool for identifying authorities of a given topic within a particular domain. Existing methods that measure topic-based scholarly output are limited to homogeneous networks. This study proposes a new informative metric called Topic-based Heterogeneous Rank (TH Rank) which measures the impact of a scholarly entity with respect to a given topic in a heterogeneous scholarly network containing authors, papers and journals. TH Rank calculates topic-dependent ranks for authors by considering the combined impact of the multiple factors which contribute to an author's level of prestige. Information retrieval serves as the test field and articles about information retrieval published between 1956 and 2014 were extracted from web of science. Initial results show that TH Rank can effectively identify the most prestigious authors, papers and journals related to a specific topic.

This paper investigates the problem of finding author interest in coauthor network through topic modeling with providing several performance evaluation measures. Intuitively, there are two types of explicit grouping exists in research papers (1) authors who have co-authored with author A in one document (subgroup) and (2) authors who have co-authored with author A in all documents (group). Traditional methods use graph-link structure by using keywords based matching and ignored semantics-based information, while topic modeling considered semantics-based information but ignored both types of explicit grouping e.g. State-of-the-art Author-Topic model used only one kind of explicit grouping single document (subgroup) for finding author interest. In this paper, we introduce Group-Author-Topic (GAT) modeling which exploits both types of grouping simultaneously. We compare four different topic modeling methods for same task on large DBLP dataset. We provide three performance measures for method evaluation from different domains which are; perplexity, entropy, and prediction ranking accuracy. We show the trade of between these performance evaluation measures. Experimental results demonstrate that our proposed method significantly outperformed the baselines in finding author interest. The trade of between used evaluation measures shows that they are equally useful for evaluating topic modeling methods.

Online bibliographic databases are powerful resources for research in data mining and social network analysis especially Co-author Networks. Predicting future rising stars is to find brilliant scholars/researchers in Co-author Networks. In this paper, we propose a solution for rising star prediction by applying machine learning techniques. For classification task, discriminative and generative modeling techniques are considered and two algorithms are chosen for each category. The author, co-authorship and venue based information are incorporated, resulting in eleven features with their mathematical formulations. Extensive experiments are performed to analyze the impact of individual feature, category wise and their combination w.r.t classification accuracy. Then, two ranking lists for top 30 scholars are presented from predicted rising stars. In addition, this concept is demonstrated for prediction of rising stars in database domain. Data from DBLP and Arnetminer databases (1996-2000 for wide disciplines) are used for algorithms' experimental analysis.

2008

In this work, we address the problem of joint modeling of text and citations in the topic modeling framework. We present two different models called the Pairwise-Link-LDA and the Link-PLSA-LDA models.

Scientific articles are not born equal. Some generate an entire discipline while others make relatively fewer contributions. When reviewing scientific literatures, it would be useful to identify those important articles and understand how they influence others. In this paper, we introduce J-Index, a quantitative metric modeling topic-level academic influence. J-Index is calculated based on the novelty of each article as well as its contributions to the articles where it is cited. We devise a gen-erative model named Reference Topic Model (RefTM) which jointly utilizes the textual content and citation information in scientific literatures. We show how to learn RefTM to discover both the novelty of each paper and the strength of each citation. Experiments on a collection of more than 420,000 research papers demonstrate that RefTM outperforms the state-of-the-art approaches in terms of topic coherence as well as prediction performance, and validate J-Index's effectiveness of capturing topic-level academic influence in scientific liter-atures.

2020

Modeling the interests of researchers in academic social networks is a crucial step in a process of recommending scientific articles, linked to their areas of competence and expertise. In this context, a researcher profile constructed from non-observable variables on the basis of articles which interests him by the LDA (Latent Dirichlet Allocation) topic modeling technique allows the system to capture knowledge about his area of competence and skills, in order to predict these needs in terms of relevant research articles. In this article we are interested in the results produced by two different implementations of LDA Gensim and Mallet on the basis of information provided by the researchers (explicit information), in order to compare their interpretability and checked if they are reliable sources for model the areas of competence and expertise of scientists. Keywords-1 st Topic Modeling, 2 nd LDA , 3 rd Mallet, 4 th Researcher Profil.

Conference mining and expert finding are useful academic knowledge discovery problems from an academic recommendation point of view. Group level (GL) topic modeling can provide us with richer text semantics and relationships, which results in denser topics. And denser topics are more useful for academic discovery issues in contrast to Element level (EL) or Document level (DL) topic modeling, which produces sparser topics. Previous methods performed academic knowledge discovery by using network connectivity (only links not text of documents), keywordsbased matching (no semantics) or by using semantics-based intrinsic structure of the words presented between documents (semantics at DL), while ignoring semantics-based intrinsic structure of the words and relationships between conferences (semantics at GL). In this paper, we consider semantics-based intrinsic structure of words and relationships presented in conferences (richer text semantics and relationships) by modeling from GL. We propose group topic modeling methods based on Latent Dirichlet Allocation (LDA). Detailed empirical evaluation shows that our proposed GL methods significantly outperformed DL methods for conference mining and expert finding problems.

Proceedings of the 29th Annual ACM Symposium on Applied Computing, 2014

The problem of searching for experts in a given academic field is hugely important in both industry and academia. We study exactly this issue with respect to a database of authors and their publications. The idea is to use Latent Semantic Indexing (LSI) and Latent Dirichlet Allocation (LDA) to perform topic modelling in order to find authors who have worked in a query field. We then construct a coauthorship graph and motivate the use of influence maximisation and a variety of graph centrality measures to obtain a ranked list of experts. The ranked lists are further improved using a Markov Chainbased rank aggregation approach. The complete method is readily scalable to large datasets. To demonstrate the efficacy of the approach we report on an extensive set of computational simulations using the Arnetminer dataset. An improvement in mean average precision is demonstrated over the baseline case of simply using the order of authors found by the topic models.

2019

This study presents a method to analyze textual data and applying it to the field of Library and Information Science. This paper subsumes a special case of Latent Dirichlet Allocation and Author-Topic models where each article has one unique author and each author has one unique topic. Topic Modeling Toolkit is used to perform the author-topic modeling. The study further which considers topics and their changes over time by taking into account both the word co-occurrence pattern and time. 393 full-text articles were downloaded from DESIDOC Journal of Library and Information Technology and were analyzed accordingly. 16 core topics have been identified throughout the period of ten years. These core topics can be considered as the core area of research in the journal from 2008 to 2017. This paper further identifies top five authors associated with the representative articles for each studied year. These authors can be treated as the subject-experts for the modeled topics as indicated. ...

2010

Abstract In this paper, we propose a multiword-enhanced author topic model that clusters authors with similar interests and expertise, and apply it to an information retrieval system that returns a ranked list of authors related to a keyword. For example, we can retrieve Eugene Charniak via search for statistical parsing.