Web personalization involves automated content analysis of text, and modern technologies of semantic analysis of text rely on a number of scales. Among them is the abstractness of meaning, which is not captured by more traditional measures of sentiment, such as valence, arousal and dominance. The present work introduces a physics-inspired approach to constructing the abstractness scale based on databases of hypernym-hyponym relations, e.g., WordNet 3.0. The idea is to define an energy as a function of word coordinates that are distributed in one dimension, and then to find a global minimum of this energy function by relocating words in this dimension. The result is a one-dimensional distribution that assigns "abstractness" values to words. While positions of individual words on this scale are subject to noise, the entire distribution globally defines the universal semantic dimension associated with the notion of hypernym-hyponym relations, called here "abstractness".

Sentiment Classification (SC) is about assigning a positive, negative or neutral label to a piece of text based on its overall opinion. This paper describes our in-progress work on extracting the meaning of words for SC. In particular, we investigate the utility of sense-level polarity information for SC. We first show that methods based on common classification features are not robust and their performance varies widely across different domains. We then show that sense-level polarity information features can significantly improve the performance of SC. We use datasets in different domains to study the robustness of the designated features. Our preliminary results show that the most common sense of the words result in the most robust results across different domains. In addition our observation shows that the sense-level polarity information is useful for producing a set of high-quality seed words which can be used for further improvement of SC task.

Homophily is the tendency of individuals in a social system to link to others who are similar to them and understanding homophily can help us build better user models for personalization and recommender systems. Many studies have verified homophily along demographic dimensions, such as age, location, occupation, etc., not only in real-world social networks but also online. However, there is limited research showing that homophily also exists when similarity is judged by topics of expertise or interests. We demonstrate the existence of topical homophily on Twitter using a novel source of evidence provided by Twitter lists. In this paper, we use LDA to extract topics from Twitter lists (a collection of user accounts created by some user that others can follow) and measure similarity between listed users based on the learned topics. We show that topically similar users are more likely to be linked via a follow relationship than less similar users.

Reading groups domain is a new domain for group recommenders. In this paper we propose a web based group recommender system which is called BoRGo: Book Recommender for Reading Groups, for reading groups domain. BoRGo uses a new information filtering technique which uses the difference between positive and negative feedbacks about a feature of a user profile and also presents an interface for after recommendation processes like achieving a consensus on the reading list.

The importance of URLs in the representation of a document cannot be overstated. Shorthand mnemonics such as ``wiki'' or ``blog'' are often embedded in a URL to convey its functional purpose or genre. Other mnemonics have evolved from use (e.g., a Wordpress particle is strongly suggestive of blogs). Can we leverage from this predictive power to induce the genre of a document from the representation of a URL? This paper presents a methodology for webpage genre classification from URLs which, to our knowledge, has not been previously attempted. Experiments using machine learning techniques to evaluate this claim show promising results and a novel algorithm for character n-gram decomposition is provided. Such a capability could be useful to improve personalized search results, disambiguate content, efficiently crawl the Web in search of relevant documents, and construct behavioral profiles from clickstream data without parsing the entire document.

This poster proposes a human computation approach to collecting representative pictures for words so that the collected pictures can efficiently and effectively convey the meaning of the words or concepts. A large collection of representative pictures can be used in text-to-picture communication systems, and may also be used to teach computers to learn what representative pictures are. We have developed a web application to help players of Draw Something, a popular social mobile game, search pictures for drawing inspiration while at the same time they implicitly help us collect representative pictures for words. Our preliminary result shows that the proposed approach has the potential to harvest Draw Something players for collecting desired data.

A natural, common way of doing this is by crowdsourcing this stage as well, and specifically Human computation is a fast-growing field that seeks to harness letting people vote over different proposals that were the relative strengths of humans to solve problems that submitted by their peers. For example, in EteRNA thousands are difficult for computers to solve alone. The field has recently of designs are submitted each month, but only a small number been gaining traction within the AI community, as k of them can be synthesized in the lab (as of late 2011, increasingly more deep connections between AI and human k 8). To single out k designs to be synthesized, players computation are uncovered (Dai, Mausam, and Weld 2010; vote by reporting their k favorite designs, each of which is Shahaf and Horvitz 2010).

Natural language programming has proven to be a very challenging task. We present a novel idea which suggests using crowdsourcing to do natural language programming. Our approach asks non-expert workers to provide input/output examples for a task defined in natural language form. We then use a Programming by Example system to induce the intended program from the input/output examples. Our early results are promising, encouraging further research in this area.

To ensure quality results from unreliable crowdsourced workers, task designers often construct complex workflows and aggregate worker responses from redundant runs. Frequently, they create several alternative workflows to accomplish the task, and choose a single workflow to deploy (perhaps the one that achieves the best performance during early experiments). However, this seemingly natural design paradigm does not achieve the full potential of crowdsourcing. In particular, using a single workflow (even the best) to accomplish a task is suboptimal. We show that alternative workflows can compose synergistically to yield a much higher quality output. We formalize the insight with a novel probabilistic graphical model, design and implement AgentHunt, a POMDP-based controller that dynamically switches between these workflows to achieve higher returns on investment, and design offline and online methods for learning model parameters. Live experiments on Amazon Mechanical Turk demonstrate the superiority of AgentHunt for the practical task of generating NLP training data, yielding up to 50% error reduction and greater net utility compared to previous methods.

People complete tasks more quickly when they have concrete plans, especially for open-ended, creative tasks. However, people often fail to create such action plans. (How) can systems provide people with these concrete steps automatically? To scalably provide personalized action plans, this paper introduces and evaluates crowdsourcing and peer approaches for creating plans, and NLP techniques for reusing them. We evaluated the effects of action plans on different types of tasks. A between-subjects experiment found that people who received crowd-created plans completed more tasks than people asked to self-create plans and than a control group without action plans. We found that crowd-created action plans are especially effective for lingering and high-level tasks. A second experiment found that peer-provided plans led to more completed tasks than no plans. A third experiment found that participants who received reused action plans also completed more tasks than a control group without action plans. We have incorporated these principles into TaskGenies: a crowd-powered task management system.