Community Question Answering (CQA) services have evolved into a popular way of online information seeking, where users can interact and exchange knowledge in the form of questions and answers. In this paper, we study the problem of finding historical questions that are semantically equivalent to the queried ones, assuming that the answers to the similar questions should also answer the new ones. The major challenge of question retrieval is the word mismatch problem between questions, as users can formulate the same question using different wording. Most existing methods measure the similarity between questions based on the bag-of-words (BOWs) representation capturing no semantics between words. Therefore, this study proposes to use word embeddings, which can capture semantic and syntactic information from contexts, to vectorize the questions. The questions are clustered using Kmeans to speed up the search and ranking tasks. The similarity between the questions is measured using cosine similarity based on their weighted continuous valued vectors. We run our experiments on real world data set from Yahoo! Answers in English and Arabic to show the efficiency and generality of our proposed method.

With the rapid growth of mobile applications, the user is increasingly confronted with a lot of information and tend to reject notifications sent by applications installed within his/her mobile device. This rejection affects the performance of many systems, especially proactive recommender systems. Therefore, it is no longer enough for a recommender system to determine what to recommend according to users' needs, but it also has to deal with the risk of disturbing the user during the recommendation process. We believe that the several embedded applications within the user's device along with other parameters could help understand and assess the user's interruptibility in some situations. In this paper, we address intrusiveness within a proactive recommendation approach that makes use of the user's context and the applications embedded within the user's mobile device in order to assess the intrusiveness level of a given situation before recommending.

The overwhelming advances in mobile technologies allow recommender systems to be highly contextualized and able to deliver recommendation without an explicit request. However, it is no longer enough for a recommender system to determine what to recommend according to the users' needs, but it also has to deal with the risk of disturbing the user during recommendation. We believe that mobile technologies along with contextual information may help alleviate this issue. In this paper, we address intrusiveness as a probabilistic approach that makes use of the several embedded applications within the user's device and the user's contextual information in order to figure out intrusive recommendations that are subject to rejection. The experiments that we conducted have shown that the proposed approach yields promising results.

Due to the emergence and the prevalence of social networks, social interactions are found beneficial for Recommender Systems. Obviously, users can now rate items, comment and suggest them to friends through social networks. Therefore, these users behaviors must be integrated to predict her preferences. However, most of the works proposed in the literature integrate only users ratings in recommendation process and ignore other behaviors made by users while/after seeing an item. In this paper, we propose a new approach that integrates in a generic way all the user behaviors in order to predict her interests. We conduct a comprehensive effectiveness evaluation on real dataset crawled from Pinhole platform. We consider several social behaviors such as comment, time spent, recommendations and shares. We evaluate the impact of each behavior in the prediction accuracy. Experimental results demonstrate the importance of all social behaviors and the effectiveness of our approach compared to collaborative filtering rating-based and time-spent-based approaches.

In this paper, we tackle the task of automatic grouping of writing styles, also called author clustering. This task deals with identifying authorship links and single-authored groups of documents [3]. This task is very important for other domains like plagiarism detection and author diarization. In this paper, we present a hybrid method that combines lexical and syntactic features for multilingual automatic grouping of writing style. The proposed method has been evaluated on two publicly available corpora. The obtained results outperform the ones obtained by the best state-of-the-art methods.

While public conversations in Twitter have gained increasing interest in the marketing sector, relatively very little data-mining research have been conducted in this area. In this paper, we empirically evaluate whether employing reply links in public conversations can enhance the product feature extraction from tweets. We introduce a conversation-based method that considers a conversation as a reply tree and employs anaphora resolution in a backtracking mechanism to effectively extract the product features involved in the messages. We also develop a conversation filtering process based on a set of filtering measures including content relevance and social metrics. We conducted our experiments using a manually annotated Twitter corpus involving smartphones and other electronics products. The experimental results show the effectiveness of our proposed method.

The user profile has been considered, in the literature, as the most important contextual element which can improve the accuracy of the search. It is integrated into the process of information retrieval in order to improve the user experience while searching for specific information. In addition, as time factor has gained increasing importance in recent years, the temporal dynamics are introduced to study the user profile evolution that consists mainly of capturing the changes of the user behavior, interests and preferences, and updating the profile accordingly. We propose, in this paper, a generic time-sensitive user profile that is implicitly constructed as a vector of weighted terms in order to find a trade-off by unifying both current and recurrent interests. The popularity of Social Media makes it as an invaluable source of data used by users to express, share and mark as favorite the content that interests them... For this reason, we modeled the user profile according to a set of data collected from Twitter, i.e a social networking and microblogging service. Then, we apply our re-ranking process to a Web search system in order to adapt the user's online interests to the original retrieved results. The results of the experiments proved the significance of adding a temporal feature by comparing our method with baselines models that do not consider the user profile dynamics.

The conversational element of Twitter has recently become of particular interest to the marketing community. However, most studies on mining product features through Twitter, have so far employed simple individual tweets rather than considering the whole conversations. In this paper, we empirically evaluate whether employing user interactions in public conversations can improve the product feature extraction from tweets. We propose a conversation-based method which considers a conversation as a reply tree and employs reply links, to effectively extract the product features involved in the messages. We also develop a conversation filtering process which combines scores measured from different aspects including content relevance and social aspects. We conducted our experiments using a manually annotated Twitter corpus involving smartphones and other electronics products. The experimental results show the effectiveness of our proposed method.

When looking for recently published scientific papers, a researcher usually focuses on the topics related to her/his scientific interests. The task of a recommender system is to provide a list of unseen papers that match these topics. The core idea of this paper is to leverage the latent topics of interest in the publications of the researchers, and to take advantage of the social structure of the researchers (relations among researchers in the same field) as reliable sources of knowledge to improve the recommendation effectiveness. In particular, we introduce a hybrid approach to the task of scientific papers recommendation, which combines content analysis based on probabilistic topic modeling and ideas from collaborative filtering based on a relevance-based language model. We conducted an experimental study on DBLP, which demonstrates that our approach is promising.

Social mediating technologies have engendered radically new ways of information and communication, particularly during events; in case of natural disaster like earthquakes tsunami and American presidential election. Billions of people create trillions of connections through social media each day, but few of us consider how each click and key press builds relationships that, in aggregate, form a vast social network. This paper is based on data obtained from Twitter because of its popularity and sheer data volume. This content can be combined and processed to detect events, entities and popular moods to feed various new large-scale data-analysis applications. On the downside, these content items are very noisy and highly informal, making it difficult to extract sense out of the stream. Taking to account all the difficulties, we propose a new event detection approach combining linguistic features and Twitter features. Finally, we present our event detection system from microblogs that aims (1) detect new events, (2) to recognize temporal markers pattern of an event, (3) and to classify important events according to thematic pertinence, author pertinence and tweet volume.

This paper present a new recommendation algorithm based on contextual analysis and new measurements. Social Network is one of the most popular Web 2.0 applications and related services, like Facebook, have evolved into a practical means for sharing opinions. Consequently, Social Network web sites have since become rich data sources for opinion mining. This paper proposes to introduce external resource from comments posted by users to predict recommendation and relieve the cold start problem. The novelty of the proposed approach is that posts are not simply characterized by an opinion score, as is the case with machine learning-based classifiers, but instead receive an opinion grade for each distinct notion in the post. Our approach has been implemented with Java and Lenskit framework; the study we have conducted on a movie dataset has shown competitive results. We compared our algorithm to SVD and Slope One algorithms. We have obtained an improvement of 8% in precision and recall as well an improvement of 16% in RMSE and nDCG.

In this paper, we focus on the detection of the Arabic learners' mother tongue. The proposed method is based on the automatic classification using some data statistically extracted from a source corpus. We present a hybrid method that combines surface analysis in texts with an automatic learning method. Unlike the few techniques found in the state of the art, the features selection phase allowed improving performances. Therefore, the obtained results outperformed those provided by the best methods used for Arabic native language detection.

Just-In-Time Recommender Systems involve all systems able to provide recommendations tailored to the preferences and needs of users in order to help them access useful and interesting resources within a large data space. The user does not need to formulate a query, this latter is implicit and corresponds to the resources that match the user's interests at the right time. In this paper, we propose a proactive context-aware recommendation approach for mobile devices that covers many domains. It aims at recommending relevant items that match users' personal interests at the right time without waiting for users to initiate any interaction.

Due to the diversity of alternative programs to watch and the change of viewers' contexts, real-time prediction of viewers' preferences in certain circumstances becomes increasingly hard. However, most existing TV recommender systems used only current time and location in a heuristic way and ignore other contextual information on which viewers' preferences may depend. This paper proposes a probabilistic approach that incorporates contextual information in order to predict the relevance of TV contents. We consider several viewer's current context elements and integrate them into a probabilistic model. We conduct a comprehensive effectiveness evaluation on a real dataset crawled from Pinhole platform. Experimental results demonstrate that our model outperforms the other context-aware models.

We propose a machine learning approach for automatically classifying opinions of Twitter texts written in Modern Standard Arabic (MSA). Tweets are classified as either positive, negative, neutral or non-opinion. Various features for opinion classification have been used which are mainly linguistic and numeric. Our in-house collected and developed training data consists of tweets preserving their specifications such as @usermentions, #hashtags which are used as tweet-particular features. Four machine learning algorithms were applied on our dataset: Support Vector Machine (SVM), Naive Bayes (NB), J48 decision tree and Random forest. The experiments results show that SVM gives the highest F measure (72%), while the j48 classifier gives the highest precision (70,97%). Our experimental results demonstrate that tweet's specific features can significantly improve classification performance in comparison to other features combination.

With the expansion of the Web 2.0, daily huge amount of data is produced everywhere, namely new articles. These contents need to be exploited in order to extract relevant information and to build knowledge databases. In this concern, processing the temporal dimension of language and extracting temporal information from electronic news articles is becoming a prominent task. In this concern, we propose an approach for identifying inter-sentential temporal relations between main events from news articles. Our approach is based on a complete linguistic analysis of texts and supervised learning models.

The increase of mobile computing systems and the growth of mobile Web highly influence the needs for personalized mobile browsers. To this aim, mobile applications take benefits from context-aware computing by adapting search process through user's context information (parameters). However, Contextualized Mobile Information Retrieval still remains a challenging problem. This last is to identify contextual parameters that improve search effectiveness and should therefore be in the user's focus. We investigate in this paper the problem of filtering mobile user's context and we propose a Language model for relevant context fields recognition. The proposed model interprets a context field relevance as a metric measure and estimates it using different features. In particular, we propose to filter as much as possible the mobile user's context to emanate efficient information that help in a personalization access to information.person- alized mobile browsers.

Event extraction is a significant task in information extraction. This importance increases more and more with the explosion of textual data available on the Web, the appearance of Web 2.0 and the tendency towards the Semantic Web. Thus, we propose a generic approach to extract events from text and to analyze them. We propose an event extraction algorithm with a polynomial complexity O(n5), and a new similarity measurement between events. We use this measurement to gather similar events. We also present a semantic map of events, and we validate the first component of our approach by the development of the “EventEC” system.

Previous researches suggested that user resistance is one of the major problems facing Information Technology project effectiveness in developing countries. This user resistance is based partially on the non alignment of the new designed IT project with the organization cultural and social factors characterizing the internal environment of organization in developing countries. On the other hand, technical anomalies are also considered to be effectiveness barriers and direct causes of user resistance. The purpose of our research is to provide technical solutions to reduce the user resistance and raising the effectiveness of the IT through the using of Webservices technology.

In this paper, we present an approach to automatically extract and classify opinions in texts. We propose a similarity measurement calculating semantically distances between a word and predefined subgroups of seed words. We have evaluated our algorithm on the semantic evaluation company “SemEval 2007” corpus, and we obtained the best value of Precision and F1 62% and 61%. As an improvement of 20 % compared to others participants.