My research revolves around text analysis and mining and focuses on two components. The first and main focus is to enhance models/algorithms related to opinion mining & analytics, information retrieval and extraction. I have been working on several problems related to the above three areas with solutions that benefit the general research problems related to opinion mining, entity extraction, information retrieval etc. Second, I apply my research findings/outcomes to other research communities (inter-disciplinary research) such as education, software systems, security and mobile applications. These research communities work with large unstructured data and needs principled methodologies that can aid in converting unstructured data into human/machine usable forms. My works on these areas provide such domain-specific data/text mining solutions. My works have been accepted in high ranked conferences such as EMNLP, COLING, CIKM and ASE.
||18th Mar 2015
||4th may 2015
||Singapore||5th-9th July 2015
||17th Feb 2015
||4th May 2015
||Lisbon, Portugal||11th-13th June 2015
||27th Apr 2015
||20th May 2015
||Texas, USA||21st-24th October 2015
||26th Mar 2015
||11th may 2015
||Lisbon, Portugal||13th-16th July 2015
||Paper and Workshop|
||30th Apr 2015
||18th May 2015
||Dubai, UAE||18th-19th June 2015
||6th Oct - Abstract, 23rd Oct - Paper
||14th Dec 2015
||Au Dhabi, UAE||2016
Learning analytics deals with the development of methods that harness educational data sets to support the learning process. In this conference contribution, the authors propose a generic curriculum analytics framework that can act as a useful guide for understanding the key dimensions that have to be considered when applying analytics in curriculum analysis and evaluation.
Users' comments on political issues or technical forums and their social network together aid in detecting users profile data such as gender, age, political affiliation, technical interests, ideological beliefs etc. Such information benefits recommendation systems and advertising applications. Predicting political party is useful for election campaigning, target advertising or even vote prediction applications.
Opinion mining and analytics in social media has been receiving much attention from corporate bodies and government agencies. Much work in this direction has been focusing on uncovering consumer sentiments from online reviews, blogs, forums, etc., to help businesses improve their products and services. Recently, citizens rely on social media as a channel to voice out their concerns and sentiments on socio-political issues. Sentiment analysis in this domain is more challenging due to the opinion targets are not a fixed list and opinion words are more towards expressing the ideology of the user and little work has been done to tackle these challenges.
I have also worked on a number of other problems in the text analytics area. Linking mentions of entities such as persons or places or organizations within specific contexts to their corresponding entries in an existing knowledge base is challenging due to ambiguities. The entities can be ambiguous in the real world where the person names can refer to more than one individual. Hence, while reading an article when the user wishes to know more about the entity, he/she can be connected to the repository and the challenge is to connect to the correct entity. I proposed a language modeling approach and query expansion for disambiguating the entities.
Software forums provide valuable content to developers and the challenge we face is finding relevant answers due to noisy data. Finding experts, retrieving similar questions and ranking relevant answers are a cluster of linked tasks in community question answering services. To retrieve the relevant answers, the current search techniques fail to filter the noisy and irrelevant posts. The data from Stackoverflow.com which is a popular software forum provides voting provision for the users to evaluate the answers and questions. Such voting information can be exploited to estimate the expertise of the users who answer the questions.
In area of securities, I worked on problems related to user privacy in social media and focusing on Facebook data. In this work we raised two important questions, is public information vulnerable to users' privacy and is there any privacy information that can be mined to infer private attributes of the users of Online Social Networks (OSNs)?
Mobile applications are recently gaining attention towards context-aware systems. One challenging task is to find the best deals for a user from his/her preferences together with card promotions and deals promotions. The challenge we face is that the deals can be numeric such as 5% discount or non-numeric such as a free teddy bear for first 10 customers; finding value for non-numeric deals needs special techniques and process structured and unstructured data.