Ralf Krestel

Latent dirichlet allocation for tag recommendation

Tagging systems have become major infrastructures on the Web. They allow users to create tags that annotate and categorize content and share them with other users, very helpful in particular for searching multimedia content. However, as tagging is not constrained by a controlled vocabulary and annotation guidelines, tags tend to be noisy and sparse. Especially new resources annotated by only a few users have often rather idiosyncratic tags that do not reflect a common perspective useful for search. In this paper we introduce an approach based on Latent Dirichlet Allocation (LDA) for recommending tags of resources in order to improve search. Resources annotated by many users and thus equipped with a fairly stable and complete tag set are used to elicit latent topics to which new resources with only a few tags are mapped. Based on this, other tags belonging to a topic can be recommended for the new resource. Our evaluation shows that the approach achieves significantly better precision and recall than the use of association rules, suggested in previous work, and also recommends more specific tags. Moreover, extending resources with these recommended tags significantly improves search for new resources.

Personalized topic-based tag recommendation

More and more content on the Web is generated by users. To organize this information and make it accessible via current search technology, tagging systems have gained tremendous popularity. Especially for multimedia content they allow to annotate resources with keywords (tags) which opens the door for classic text-based information retrieval. To support the user in choosing the right keywords, tag recommendation algorithms have emerged. In this setting, not only the content is decisive for recommending relevant tags but also the users preferences. In this paper we introduce an approach to personalized tag recommendation that combines a probabilistic model of tags from the resource with tags from the user. As models we investigate simple language models as well as Latent Dirichlet Allocation. Extensive experiments on a real world dataset crawled from a big tagging system show that personalization improves tag recommendation, and our approach significantly outperforms state-of-the-art approaches.

Language Models and Topic Models for Personalizing Tag Recommendation

More and more content on the Web is generated by users. To organize this information and make it accessible via current search technology, tagging systems have gained tremendous popularity. Especially for multimedia content they allow to annotate resources with keywords (tags) which opens the door for classic text-based information retrieval. To support the user in choosing the right keywords, tag recommendation algorithms have emerged. In this setting, not only the content is decisive for recommending relevant tags but also the users preferences. In this paper we introduce an approach to personalized tag recommendation that combines a probabilistic model of tags from the resource with tags from the user. As models we investigate simple language models as well as Latent Dirichlet Allocation. Extensive experiments on a real world dataset crawled from a big tagging system show that personalization improves tag recommendation, and our approach significantly outperforms state-of-the-art approaches.

Diversifying Product Review Rankings: Getting the Full Picture

E-commerce Web sites owe much of their popularity to consumer reviews provided together with product descriptions. On-line customers spend hours and hours going through heaps of textual reviews to build confidence in products they are planning to buy. At the same time, popular products have thousands of user-generated reviews. Current approaches to present them to the user or recommend an individual review for a product are based on the helpfulness or usefulness of each review. In this paper we look at the top-k reviews in a ranking to give a good summary to the user with each review complementing the others. To this end we use Latent Dirichlet Allocation to detect latent topics within reviews and make use of the assigned star rating for the product as an indicator of the polarity expressed towards the product and the latent topics within the review. We present a framework to cover different ranking strategies based on theusers need: Summarizing all reviews, focus on a particular latent topic, or focus on positive, negative or neutral aspects. We evaluated the system using manually annotated review data from a commercial review Web site.

Converting a Historical Architecture Encyclopedia into a Semantic Knowledge Base

Digitizing a historical document using ontologies and natural language processing techniques can transform it from arcane text to a useful knowledge base.The Handbook on Architecture (Handbuch der Architektur) was perhaps one of the most ambitious publishing projects ever. Like a 19thcentury Wikipedia, it attempted nothing less than a full account of all architectural knowledge available at the time, both past and present. It covers topics from Greek temples to contemporary hospitals and universities; from the design of individual construction elements such as window sills to large-scale town planning; from physics to design; from planning to construction. It also discusses architectural history and styles and a multitude of other topics, such as building conception, statics, and interior design.Not surprisingly, this project took longer than planned. The encyclopedias first volume was partly published in 1880, and over the next 63 years more than 100 architects worked on what would become more than 140 individual publications with over 25,000 pages. One important insight of our work is that targeted text analysis support, already available today, can easily be integrated into common desktop tools to support users for their task at hand. While NLP techniques are far from perfect or comprehensive, they can already deliver knowledge discovery support that goes significantly beyond the currently used approach of full-text search and information retrieval.

Which Answer is Best?: Predicting Accepted Answers in MOOC Forums

Massive Open Online Courses (MOOCs) have grown in reach and importance over the last few years, enabling a vast userbase to enroll in online courses. Besides watching videos, user participate in discussion forums to further their understanding of the course material. As in other community-based question-answering communities, in many MOOC forums a user posting a question can mark the answer they are most satisfied with. In this paper, we present a machine learning model that predicts this accepted answer to a forum question using historical forum data.

Tweet-Recommender: Finding Relevant Tweets for News Articles

Twitter has become a prime source for disseminating news and opinions. However, the length of tweets prohibits detailed descriptions; instead, tweets sometimes contain URLs that link to detailed news articles. In this paper, we devise generic techniques for recommending tweets for any given news article. To evaluate and compare the different techniques, we collected tens of thousands of tweets and news articles and conducted a user study on the relevance of recommendations.

Diversifying customer review rankings

E-commerce Web sites owe much of their popularity to consumer reviews accompanying product descriptions. On-line customers spend hours and hours going through heaps of textual reviews to decide which products to buy. At the same time, each popular product has thousands of user-generated reviews, making it impossible for a buyer to read everything. Current approaches to display reviews to users or recommend an individual review for a product are based on the recency or helpfulness of each review. In this paper, we present a framework to rank product reviews by optimizing the coverage of the ranking with respect to sentiment or aspects, or by summarizing all reviews with the top-K reviews in the ranking. To accomplish this, we make use of the assigned star rating for a product as an indicator for a reviews sentiment polarity and compare bag-of-words (language model) with topic models (latent Dirichlet allocation) as a mean to represent aspects. Our evaluation on manually annotated review data from a commercial review Web site demonstrates the effectiveness of our approach, outperforming plain recency ranking by 30% and obtaining best results by combining language and topic model representations.

Recommending patents based on latent topics

The availability of large volumes of granted patents and applications, all publicly available on the Web, enables the use of sophisticated text mining and information retrieval methods to facilitate access and analysis of patents. In this paper we investigate techniques to automatically recommend patents given a query patent. This task is critical for a variety of patent-related analysis problems such as finding relevant citations, research of relevant prior art, and infringement analysis. We investigate the use of latent Dirichlet allocation and Dirichlet multinomial regression to represent patent documents and to compute similarity scores. We compare our methods with state-of-the-art document representations and retrieval techniques and demonstrate the effectiveness of our approach on a collection of US patent publications.

Learning Temporal Tagging Behaviour

Social networking services, such as Facebook, Google+, and Twitter are commonly used to share relevant Web documents with a peer group. By sharing a document with her peers, a user recommends the content for others and annotates it with a short description text. This short description yield many chances for text summarization and categorization. Because todays social networking platforms are real-time media, the sharing behaviour is subject to many temporal effects, i.e., current events, breaking news, and trending topics. In this paper, we focus on time-dependent hashtag usage of the Twitter community to annotate shared Web-text documents. We introduce a framework for time-dependent hashtag recommendation models and introduce two content-based models. Finally, we evaluate the introduced models with respect to recommendation quality based on a Twitter-dataset consisting of links to Web documents that were aligned with hashtags.