It could be because of the lack of good teachers in my life after I left school. It may be because I just love presenting ideas to people and seeing them feel inspired. For whatever reason, my ultimate and awesome goal in life is to be a good teacher. So the natural next step for me is to do a Masters course.
In my 7th Semester of B.Tech I took Information Retrieval as one of my optional subjects. It was the first (and only) course in the whole of my 4 years that I really enjoyed learning. The funny thing is that we didn’t have a teacher for it. We were 10 people taking the course and the teacher in charge said that she didn’t really know much about it. So we split the modules between ourselves and got cracking. It worked well and I know the subject fairly well. Also, classes were fun. (And we didn’t need to have a lot of them)
Little wonder then, that the I am going to do my masters in Information Retrieval Systems at the University of Glasgow. The Computer Science department is not the biggest. But they do some good research in the IR field. The leader of the pack is Prof. Keith van Rijsbergen, whose book on IR is one of the few books I followed in the earlier course I took.
There are a few areas in Information Retrieval that I am interested in:
- HCIR: Its been almost 20 years and we are still interacting with the searcher in the same way. There must be a better way for the user to communicate with the search system – both in terms of making a query as well as perusing the relevant results. Human Computer IR is a field where Human computer Interaction and IR combine to deal with this issue. I’m sure there are many fascinating things to be learnt here.
- Social IR: This is something which was inspired by a good friend from college, Cyriac Thomas, who made a social layer above the search engine which will pool the annotations made by your social network to the search results. His project aims to use the principle that if something is relevant to your friend, it is more likely relevant to you as well.
- Multimedia Retrieval: For me, this is probably that portion of IR that interests me the most as far as a final result goes. But I am also the least skilled in this area. My knowledge of how multimedia works is very poor from an indexing and retrieval PoV. But imagine a situation where a search can give you content based retrieval of user uploaded videos based on content and not just the (often horrible) naming and annotation based system we use now. (On a sad note, this is something I feel copyright sheriffs will be greatly interested in)
- Collaborative Search: This is different from Social IR. The concept of many people searching with the same objective ( or a same final goal – like a project) is a theme which is less explored. A good article to get you going on the topic is written by the chaps at the Augmented Social Cognition Research Group at the Palo Alto Research Center.
- Contextual Search: Fully satisfying a random guy’s everyday search needs to the fullest, needs the search engine to know the context in which the search is being made. Are you searching for something based on a project you are working on currently? Or is it a search about a local area you want to check out over the weekend? Based on the context, the relevancy changes. Most search engines nowadays have resorted to showing you all possible situations for a given search. But that isn’t exactly perfect. The results for each query should vary according to various other factors like weather, personal details like family/kids, likes/dislikes and provide pics, video and rich media to let the user make a decision as fast as possible.Is there a model for this kind of retrieval? How best can we index all this diverse data to actually get it at pace and relevantly? At Glasgow there is work going on in modeling this situation based on Quantum Theory math models. I need to catchup on Math and QT before I am at a level to understand the idea and appreciate the solutions.
- Web Search : I guess most of the points here would be best applied in this context. At the end of the day, for a common man, the Internet is the largest storehouse of knowledge (and trash) they will ever come across. Efficiently indexing and retrieving information from that using methods more evolved than being text based analysis is something I’d really like to see on a large scale. Even Google, at its heart, is essentially a text based searching and machine learning system.
- Medical IR : This is an interesting concept I was playing with. An intelligent system which could suggest possible diagnoses based on trained data. It can retrieve cases with relevance to the one you are currently on from the large past database. It can learn as the system is used more. I don’t have much idea about what it would entail, or how useful it would be for a random doctor or student of medicine. But imagine a situation in which doctors from across the world could have a super fast reference system about every case that is being dealt with at any given point of time and get a list of cases relevant to the current one based on tests, land of origin, course of treatment etc.
One wonders what will come of all this…
I have been accepted by UniGlasgow. Now to work on my Visa application. September it will be, I suppose.