No abstract

No abstract

An increasing number of Web applications are allowing users to play more active roles for enriching the source content. The enriched data can be used for various applications such as text summarization, opinion mining and ontology creation. In this paper, we propose a novel Web content summarization method that creates a text summary by exploiting user feedback (comments and tags) in a social bookmarking service. We had manually analyzed user feedback in several representative social services including del.icio.us, Digg, YouTube, and Amazon.com. We found that (1) user comments in each social service have its own characteristics with respect to summarization, and (2) a tag frequency rank does not necessarily represent its usefulness for summarization. Based on these observations, we conjecture that user feedback in social bookmarking services is more suitable for summarization than other type of social services. We implemented prototype system called SSNote that analyzes tags and user comments in del.icio.us, and extracts summaries. Performance evaluations of the system were conducted by comparing its output summary with manual summaries generated by human evaluators. Experimental results show that our approach highlights the potential benefits of user feedback in social bookmarking services.

No abstract

Knowledge-grounded conversation models aim at generating informative responses for the given dialogue context, based on external knowledge. To generate an informative and context-coherent response, it is important to conjugate dialogue context and external knowledge in a balanced manner. However, existing studies have paid less attention to finding appropriate knowledge sentences from external knowledge sources than to generating proper sentences with correct dialogue acts. In this paper, we propose two knowledge selection strategies: 1) Reduce-Match and 2) Match-Reduce and explore several neural knowledge-grounded conversation models based on each strategy. Models based on Reduce-Match strategy first distill the whole dialogue context into a single vector with salient features preserved and then compare this context vector with the representation of knowledge sentences to predict a relevant knowledge sentence. Models based on Match-Reduce strategy first match every turn of the context with knowledge sentences to capture fine-grained interactions and aggregate them while minimizing information loss to predict the knowledge sentence. Experimental results show that conversation models using each of our knowledge selection strategies outperform the competitive baselines not only in terms of knowledge selection accuracy but also in response generation performance. Our best model based on Match-Reduce outperforms the baselines in the comparative studies with the Wizard of Wikipedia dataset. Also, our best model based on Reduce-Match outperforms them with the CMU Document Grounded Conversations dataset.