Tingting Liang

Meta-Path Based Service Recommendation in Heterogeneous Information Networks

In the scenario of service recommendation, there are multiple object types (e.g. services, mashups, categories, contents and providers) and rich relationships among these objects, which naturally constitute a heterogeneous information network (HIN). In this paper, we propose to recommend services for mashup creation by exploiting different types of relationships in service related HIN. Specifically, we first introduce meta-path based measure for similarity estimation between mashups along different types of paths in HIN. We then design a recommendation model based on collaborative filtering and meta-path based similarities, and employ Bayesian ranking based optimization algorithm for model learning. Comprehensive experiments based on real data demonstrate the effectiveness of the HIN based service recommendation approach.

Co-Clustering WSDL Documents to Bootstrap Service Discovery

With the increasing popularity of web service, it is indispensable to efficiently locate the desired service. Utilizing WSDL documents to cluster web services into functionally similar service groups is becoming mainstream in recent years. However, most existing algorithms cluster WSDL documents solely and ignore the distribution of words rather than cluster them simultaneously. Different from the traditional clustering algorithms that are on one-way clustering, this paper proposes a novel approach named WCCluster to simultaneously cluster WSDL documents and the words extracted from them to improve the accuracy of clustering. WCCluster poses co-clustering as a bipartite graph partitioning problem, and uses a spectral graph algorithm in which proper singular vectors are utilized as a real relaxation to the NP-complete graph partitioning problem. To evaluate the proposed approach, we design comprehensive experiments based on a real-world data set, and the results demonstrate the effectiveness of WCCluster.

Data Augmented Maximum Margin Matrix Factorization for Flickr Group Recommendation

User groups on photo sharing websites, such as Flickr, are self-organized communities to share photos and conversations with similar interest and have gained massive popularity. However, the huge volume of groups brings troubles for users to decide which group to choose. Further, directly applying collaborative filtering techniques to group recommendation will suffer from cold start problem since many users do not affiliate to any group. In this paper, we propose a hybrid recommendation approach named Data Augmented Maximum Margin Matrix Factorization (DAM3F), by integrating collaborative user-group information and user similarity graph. Specifically, Maximum Margin Matrix Factorization (MMMF) is employed for the collaborative recommendation, while the user similarity graph obtained from the user uploaded images and annotated tags is used as an complementary part to handle the cold start problem and to improve the performance of MMMF. The experiments conducted on our crawled dataset with 2196 users, 985 groups and 334467 images from Flickr demonstrate the effectiveness of the proposed approach.

Incorporating Heterogeneous Information for Mashup Discovery with Consistent Regularization

With the development of service oriented computing, web mashups which provide composite services are increasing rapidly in recent years, posing a challenge for the searching of appropriate mashups for a given query. To the best of our knowledge, most approaches on service discovery are mainly based on the semantic information of services, and the services are ranked by their QoS values. However, these methods can’t be applied to mashup discovery seamlessly, since they merely rely on the description of mashups, but neglecting the information of service components. Besides, those semantic based techniques do not consider the compositive structure of mashups and their components. In this paper, we propose an efficient consistent regularization framework to enhance mashup discovery by leveraging heterogeneous information network between mashups and their components. Our model also integrates mashup discovery and ranking properly. Comprehensive experiments have been conducted on a real-world ProgrammableWeb.com (http://www.programmableweb.com) dataset with mashups and APIs (In ProgrammableWeb.com, APIs are the service components of mashups. Our model verified on the ProgrammableWeb.com dataset could also be applied to other compositive service discovery scenarios.). Experimental results show that our model achieves a better performance compared with ProgrammableWeb.com search engine and a state-of-the-art semantic based model.

EARec: leveraging expertise and authority for pull-request reviewer recommendation in GitHub

Pull-Request (PR) is a primary way of code contribution from developers to improve quality of software projects in GitHub. For a popular GitHub project, tens of PR are submitted daily, while only a small number of developers, i.e core developers, have the grant to judge whether to merge these changes into the main branches or not. Due to the time-consumption of PR review and the diversity of PR aspects, it is becoming a big challenge for core developers to quickly discover the useful PR. Currently, recommending appropriate reviewers (developers) for incoming PR to quickly collect meaningful comments, is treated as an effective and crowdsourced way to help core developers to make decisions and thus accelerate project development. In this paper, we propose a reviewer recommendation approach (EARec) which simultaneously considers developer expertise and authority. Specifically, we first construct a graph of incoming PR and possible reviewers, and then take advantage of text similarity of PR and social relations of reviewers to find the appropriate reviewers. The experimental analysis on MSR Mining Challenge Dataset\footnote{http://ghtorrent.org/msr14.html} provides good evaluation for our approach in terms of precision and recall.

Exploiting Heterogeneous Information for Tag Recommendation in API Management.

As web-enabled software becomes the standard for business processes, the ways organizations, partners and customers interface with it have become a critical differentiator in the market place, i.e., API Economy. With the rapid proliferation of APIs, it is increasingly important for users to effectively manage objective APIs in kinds of API markets, e.g., ProgramableWeb (PW), Mashape, etc. In this paper, to facilitate the process of API management, we propose a graphbased recommendation approach called ATRec to automatically assign tags to unlabeled APIs by exploiting both graph structure information and semantic similarity. Specifically, ATRec first leverages the multi-type relations (i.e., among APIs, mashups, and mashup assigned tags) to construct a heterogeneous network, in which a Random Walk with Restart (RWR) model is applied to alleviate the total cold start problem where no API has ever been tagged. Furthermore, we apply the recommended API tags in two API management scenarios (API search, API recommendation). Comprehensive experiments based on a real dataset crawled from PW demonstrate the effectiveness of the proposed approach.

SMS: A Framework for Service Discovery by Incorporating Social Media Information

With the explosive growth of services, including Web services, cloud services, APIs and mashups, discovering the appropriate services for consumers is becoming an imperative issue. The traditional service discovery approaches mainly face two challenges: 1) the single source of description documents limits the effectiveness of discovery due to the insufficiency of semantic information; 2) more factors should be considered with the generally increasing functional and nonfunctional requirements of consumers. In this paper, we propose a novel framework, called SMS, for effectively discovering the appropriate services by incorporating social media information. Specifically, we present different methods to measure four social factors (semantic similarity, popularity, activity, decay factor) collected from Twitter. Latent Semantic Indexing (LSI) model is applied to mine semantic information of services from meta-data of Twitter Lists that contains them. In addition, we assume the target query-service matching function as a linear combination of multiple social factors and design a weight learning algorithm to learn an optimal combination of the measured social factors. Comprehensive experiments based on a real-world dataset crawled from Twitter demonstrate the effectiveness of the proposed framework SMS, through some compared approaches.

Crowdsourcing Based API Search via Leveraging Twitter Lists Information

With the rapid growth of open APIs on the Internet, searching appropriate APIs for a given query becomes a challenging problem. General API search systems, such as ProgrammableWeb, usually can not provide satisfactory results of API search due to the simple keywords matching between queries and API information offered by providers (e.g. name and description). In this paper, we propose a crowdsourcing based search approach named CrowdAPS to effectively find the appropriate APIs. Specifically, CrowdAPS leverages Twitter lists, which is a tool used by individual users to organize accounts that interest them on semantics. List meta-data, including list name and description, is generated from collective intelligence and can be used by Latent Semantic Indexing (LSI) model to acquire semantic similarity between APIs and queries. Furthermore, CrowdAPS exploits list number to infer the popularity of APIs. The final search result relies on the integration of semantic similarity and popularity. Comprehensive experiment based on real-world datasets crawled from ProgrammableWeb and Twitter demonstrates the effectiveness of CrowdAPS.

Mobile Application Rating Prediction via Feature-Oriented Matrix Factorization

With the proliferation of mobile application (app) markets (e.g., Google Play, Apple App Store), predicting user preferences on apps becomes a challenging problem. Different from previous work, we assume that a user likes an app because he/she likes certain features of the app (e.g., permission, genre, topic). Based on this assumption, we propose a feature-oriented approach to predict user preferences on apps. Specifically, we transform the original app rating matrix to feature rating data and predict the unknown ratings on the features through a latent factor model, instead of directly predicting ratings on apps. The predicted user ratings on features can be used to generate the ratings on apps. Two integration methods are presented to give different significance for feature preferences. The approach has some obvious advantages: as it integrates feature information to analyze the details of user preference, it can generalize better as the feature rating data is denser, and improve the interpretation of the prediction of app ratings. Experimental results on a real-world dataset demonstrate the effectiveness of the proposed approach.

Time-aware metric embedding with asymmetric projection for successive POI recommendation

Successive Point-of-Interest (POI) recommendation aims to recommend next POIs for a given user based on this user’s current location. Indeed, with the rapid growth of Location-based Social Networks (LBSNs), successive POI recommendation has become an important and challenging task, since it can help to meet users’ dynamic interests based on their recent check-in behaviors. While some efforts have been made for this task, most of them do not capture the following properties: 1) The transition between consecutive POIs in user check-in sequences presents asymmetric property, however existing approaches usually assume the forward and backward transition probabilities between a POI pair are symmetric. 2) Users usually prefer different successive POIs at different time, but most existing studies do not consider this dynamic factor. To this end, in this paper, we propose a time-aware metric embedding approach with asymmetric projection (referred to as MEAP-T) for successive POI recommendation, which takes the above two properties into consideration. In addition, we exploit three latent Euclidean spaces to project the POI-POI, POI-user, and POI-time relationships. Finally, the experimental results on two real-world datasets show MEAP-T outperforms the state-of-the-art methods in terms of both precision and recall.