Sequential latent Dirichlet allocation

Du, Lan; Buntine, Wray L.; Jin, Huidong; Chen, Changyou

doi:10.1007/s10115-011-0425-1

Cited by 40 publications

(25 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The LDA method (Blei et al 2003) allowed us to understand the topics of the projects in an efficient and reliable manner. It allowed us to calculate diversity and the degree of multi-disciplinarity, and can also aid future researchers with understanding the topics of innovation projects, in addition to publications or patents (Du et al 2012).…”

Section: Resultsmentioning

confidence: 99%

Multi-disciplinarity breeds diversity: the influence of innovation project characteristics on diversity creation in nanotechnology

et al. 2017

View full text Add to dashboard Cite

Nanotechnology is an emerging and promising field of research. Creating sufficient technological diversity among its alternatives is important for the long-term success of nanotechnologies, as well as for other emerging technologies. Diversity prevents early lock-in, facilitates recombinant innovation, increases resilience, and allows market growth. Creation of new technological alternatives usually takes place in innovation projects in which public and private partners often collaborate. Currently, there is little empirical evidence about which characteristics of innovation projects influence diversity. In this paper we study the influence of characteristics of EU-funded nanotechnology projects on the creation of technological diversity. In addition to actor diversity and the network of the project, we also include novel variables that have a plausible influence on diversity creation: the degree of multi-disciplinarity of the project and the size of the joint knowledge base of project partners. We apply topic modelling (Latent Dirichlet allocation) as a novel method to categorize technological alternatives. Using an ordinal logistic regression model, our results show that the largest contribution to diversity comes from the multidisciplinary nature of a project. The joint knowledge base of project partners and the geographical distance between them were positively associated with technological diversity creation. In contrast, the number and diversity of actors and the degree of clustering showed a negative association with technological diversity creation. These results extend current micro-level explanations of how the diversity of an emerging technology is created. The contribution of this study could also be helpful for policy makers to influence the level of diversity in a technological field, and hence to contribute to survival of emerging technologies.

show abstract

Section: Resultsmentioning

confidence: 99%

Multi-disciplinarity breeds diversity: the influence of innovation project characteristics on diversity creation in nanotechnology

et al. 2017

View full text Add to dashboard Cite

show abstract

“…shows the graphical model for SCNTM. In [23], seqLDA was introduced as novel variant of LDA underlying sequential structure. Hierarchy modeling was applied to documents which were considered as a multiple segments and each segment is associated to its predecessor and subsequent segments.…”

Section: Supervised Citation Network Topic Model (Scntm)mentioning

confidence: 99%

A Survey on Journey of Topic Modeling Techniques from SVD to Deep Learning

Sharma¹,

Kumar²,

Chand³

2017

IJMECS

View full text Add to dashboard Cite

Abstract-Topic modeling techniques have been primarily being used to mine the topics from text corpora. These techniques reveal the hidden thematic structure in a collection of documents and facilitate to build up new ways to browse, search and summarize large archive of texts. A topic is a group of words that frequently occur together. A topic modeling can connect words with similar meanings and make a distinction between uses of words with several meanings. Here we present a survey on journey of topic modeling techniques comprising Latent Dirichlet Allocation (LDA) and non-LDA based techniques and the reason for classify the techniques into LDA and non-LDA is that LDA has ruled the topic modeling techniques since its inception. We have used the three hierarchical classification criteria's for classifying topic models that include LDA and non-LDA based, bag-of-words or sequence-of-words approach and unsupervised or supervised learning for our survey. Purpose of this survey is to explore the topic modeling techniques since Singular Value Decomposition (SVD) topic model to the latest topic models in deep learning. Also, provide the brief summary of current probabilistic topic models as well as a motivation for future research.

show abstract

“…However, this requires recording the number of customers on each table and could be expensive. The other way is to fix a k > 0 and use an adaptive rejection sampler to sample b k 's, as was done by Du et al [20]. We implemented both methods and used the second in these experiments as it produced better training likelihoods.…”

Section: Handling Hyperparametersmentioning

confidence: 99%

“…Sampling is performed on an extended version of LDA with multiple levels. Du et al developed a series of models exhibiting sharing across segments in a document both hierarchically and sequentially [20], [21] that were very competitive against standard LDA. Note that the above works, while hierarchical, do not consider the problem of topic sharing between groups of data sets, nor do they consider correlations among words in the topic.…”

Section: Related Workmentioning

confidence: 99%

Differential Topic Models

Chen

Buntine

Ding

et al. 2015

IEEE Trans. Pattern Anal. Mach. Intell.

Self Cite

View full text Add to dashboard Cite

Abstract-In applications we may want to compare different document collections: they could have shared content but also different and unique aspects in particular collections. This task has been called comparative text mining or cross-collection modeling. We present a differential topic model for this application that models both topic differences and similarities. For this we use hierarchical Bayesian nonparametric models. Moreover, we found it was important to properly model power-law phenomena in topic-word distributions and thus we used the full Pitman-Yor process rather than just a Dirichlet process. Furthermore, we propose the transformed Pitman-Yor process (TPYP) to incorporate prior knowledge such as vocabulary variations in different collections into the model. To deal with the non-conjugate issue between model prior and likelihood in the TPYP, we thus propose an efficient sampling algorithm using a data augmentation technique based on the multinomial theorem. Experimental results show the model discovers interesting aspects of different collections. We also show the proposed MCMC based algorithm achieves a dramatically reduced test perplexity compared to some existing topic models. Finally, we show our model outperforms the state-of-the-art for document classification/ideology prediction on a number of text collections.

show abstract

Sequential latent Dirichlet allocation

Cited by 40 publications

References 28 publications

Multi-disciplinarity breeds diversity: the influence of innovation project characteristics on diversity creation in nanotechnology

Multi-disciplinarity breeds diversity: the influence of innovation project characteristics on diversity creation in nanotechnology

A Survey on Journey of Topic Modeling Techniques from SVD to Deep Learning

Differential Topic Models

Contact Info

Product

Resources

About