2015 IR Workshop

[Chinese]

Workshop Topics：	New Trends and Technologies of Cross-discipline NLP and IR
Date:	2015-12-18 (Friday)
Venue：	Auditorium 106 at Institute of Information Science , Academia Sinica
Sponsors：	Institute of Information Science, Academia Sinica Department of Computer Science and Information Engineering, National Taiwan University The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)
Workshop Chairs:	Pu-Jen Cheng (National Taiwan University) Lun-Wei Ku (Academia Sinica)

8:50-9:10	Registration
9:10-9:20	Opening	Pu-Jen Cheng Lun-Wei Ku	National Taiwan University Academia Sinica
9:20-10:20	Keynote Speech: Jointly Modeling Topics, Events and User Interests on Twitter	Jing Jiang	Singapore Management University
Talk Title: Jointly Modeling Topics, Events and User Interests on Twitter Abstract: With the rapid growth of social media, Twitter has become one of the most widely adopted platforms for people to post short and instant messages. Tweets reflect both individual users’ personal interests and major events that have attracted the public’s attention. In this talk, I will present our recent work on modeling topics, events and user interests on Twitter. Our work is based on probabilistic topic models, which have principled theoretical foundations. Based on our observations with Twitter, we introduced several important modifications to standard models in order to capture Twitter’s special properties. Our experiments showed that our models can effectively identify meaningful events, group events by topics, and potentially recommend events to users based on their topical interests.
10:20-10:40	Coffee Break
10:40-11:30	Invited Talk 1: A Community-based Method for Valence-Arousal Prediction of Affective Words	Liang-Chih Yu	Yuan Ze University
Talk Title: A Community-based Method for Valence-Arousal Prediction of Affective Words Abstract: Compared to the categorical approach that represents affective states as several discrete classes (e.g., positive and negative), the dimensional approach represents affective states as continuous numerical values in multiple dimensions, such as the valence-arousal (VA) space, thus allowing for more fine-grained sentiment analysis. In building dimensional sentiment applications, affective lexicons with valence-arousal ratings are useful resources but are still very rare. Several semi-supervised methods such as the kernel method, linear regression, and the pagerank algorithm have been investigated to automatically determine the VA ratings of affective words from a set of semantically similar seed words. These methods suffer from two major limitations. First, they apply an equal weight to all seeds similar to an unseen word in predicting its VA ratings. Second, even similar seeds may have quite different ratings (or an inverse polarity) of valence/arousal to the unseen word, thus reducing prediction performance. To overcome these limitations, this study proposes a community-based weighted graph model that can select seeds which are both similar to and have similar ratings (or the same polarity) with each unseen word to form a community (sub-graph) so that its VA ratings can be estimated from such high-quality seeds using a weighted propagation scheme. That is, seeds more similar to unseen words contribute more to the estimation process. Experimental results show that the proposed method yields better prediction performance for both English and Chinese datasets.
11:30-12:20	Invited Talk 2 Active Learning by Learning	Hsuan-Tien Lin	National Taiwan University
Talk Title: Active Learning by Learning Abstract: Active learning is an important technique that helps reduce labeling efforts in machine learning applications. Currently, most active learning strategies are constructed based on some human-designed philosophy; that is, they reflect what human beings assume to be "good labeling questions." However, given that a single human-designed philosophy is unlikely to work on all scenarios, choosing and blending those strategies under different scenarios is an important but challenging practical task. This paper tackles this task by letting the machines adaptively "learn" from the performance of a set of given strategies on a particular data set. More specifically, we design a learning algorithm that connects active learning with the well-known multi-armed bandit problem. Further, we postulate that, given an appropriate choice for the multi-armed bandit learner, it is possible to estimate the performance of different strategies on the fly. Extensive empirical studies of the resulting algorithm confirm that it performs better than strategies that are based on human-designed philosophy.
12:20-13:50	Lunch
13:50-14:50	Invited Talk 3 Meet Industry People: Capture the great moment of social network efficiently and effectively	Shih-En Chou	QSearch, Co-Founder (ASIA BEAT 創業競賽冠軍)
Talk Title: Capture the great moment of social network efficiently and effectively Abstract: Social media has greatly changed the way we communicate and huge amount of social behavior data is thus recorded and accumulated simultaneously. The data is now widely applied to many emerging research issues in combination with social behavior analysis. More recently, time domain analysis is especially popular on conducting behavior change investigation, in which people take snapshots on a particular subject of network on regular intervals, and hot messages (posts) are in urgent need of snapshot so as to precisely learn about user’s behavior as time moves. Scraping social networking sites such as Twitter, Facebook, etc. is not an easy task for data acquisition departments of most institutions since these sites often have complex structures and also restrict the amount and frequency of the data that they let out to common crawlers. In this talk, I will introduce the proposed crawling algorithm and address the challenges of large-scale dynamic content crawling.
14:50-15:10	Coffee Break
15:10-16:00	Invited Talk 4 A Novel Social Influence Model based on Multiple States and Negative Social Influences	Jen-Wei Huang	National Cheng Kung University
Talk Title: A Novel Social Influence Model based on Multiple States and Negative Social Influences Abstract: Social influence has been a significant and popular topic in the social network analysis. People usually rely on the social influence model to predict and learn the influence diffusion process in real world. We studied the most popular influence models nowadays and found out some limitations of the traditional influence models. Traditional models only categorize nodes into two types of states, say active and inactive states. In addition, most previous models have only taken positive influences into account. We try to break the above limitations and propose a novel social influence model based on multiple states and negative social influence. According to the new propagation method, the strength of the social influence may be reduced over time. The experimental results show that the proposed model outperforms previous models in precisions of prediction.
16:00-16:50	Invited Talk 5 Using Non-Verbal Information to Augment Designs of Language-based Interactions	Hao-Chuan Wang	National Tsing Hua University
Talk Title: Using Non-Verbal Information to Augment Designs of Language-based Interactions Abstract: Language plays an important role in communications and interactions, but communications don’t necessarily have to stay verbal to be the best. In human-human communication, for example, verbal messages can be supplemented in non-verbal ways, such as facial expressions, gestures or emoticons/emojis in online communication today. In human-machine communication like user interface design, designers also use graphic icons to communicate with users what are the intended functions of specific interface features. Because non-verbal information can inform the users what language refers, intends to mean, or tries to emphasize, we see opportunities in using non-verbal information to functionally augment language-based interactions when the problems in expression and comprehension are severe. In this talk, I will present a view on how non-verbal and verbal information might be coordinated and integrated in interaction designs. I’ll showcase a number of examples that we pursued in the Collaborative and Social Computing Lab at National Tsing Hua University to illustrate the benefits and constraints of injecting non-verbal information to verbal interactions as a general design pattern.
16:50-17:00	Closing	Pu-Jen Cheng Lun-Wei Ku	National Taiwan University Academia Sinica