My Thoughts on Semantic Networks

I wasn’t sure if it was by some kind of intuition or just a wild guess that I discovered the Artist in Rufus Gurugulla, when I met him first at a painting competition. He was there to photograph the event. He is a crop artist and made several art works with seeds, including a portrait of George Bush. I am sure he would have completed Obama’s by this time.
During our conversation he asked me one simple question, “Do you know any plant producing blue coloured seeds?”. The moment `blue’ is mentioned, what flshed in my mind was `indigo’, soon followed by a train of doubts – is it the bark which imparts the blue colour, that also only after processing? Thenn what would be its seed colour? Could it be blue? Then leaves also could look little bluish green? Does it contain anthocyanin? Does anthocyanin impart blue colour? I was confused.
The question was addressed to me as I was perceived by him as an agricultural expert. But the question was received by the information professional in me. I took the earliest oppurtunityto look for it on internet.
The seemingly simple question was not all that simple. I did retrieve a few useful references but I definitely required a better way of searching, some search query like – (crops or trees or plants) and seed colour is blue. There it is! I am now convinced that the elaborate topic maps we were frantically trying to build are worth the effort. Hither to they seemed to be like elementary school English work books containing `is a’ and `has a’ kind of phrases. To be honest I had a feeling that people are unnecessarily complicating things. But of course technology advances only after several vain efforts have taken place.

That’s how the `blue seeds’ case has convinced me. But the `neem seeds’ case was far more than convincing. It was rather compelling. I so desperately needed a concept map that I got tempted to train someone and put on the job immediately. The story behind this `neem seeds’ case goes like this. One of our close relatives, a software engineer asked me if I can help him procure `neem seeds’. I tried my personal contacts and got a couple of useful links. But they were far from sufficient. My internet search yielded lists of addresses selling and buying various related products and raw materials. But I needed something more specific. Luckily I could manage to arrange a get together of the people I know, over the weekend and I remained a facilitator for the knowledge sharing to take place.
I was so glad that I could co-ordinate that mile stone event bringing people from different sectors together, like pieces of a jig-saw puzzle. A couple of them were software engineers looking for greener investment options, one was a dealer in agricultural inputs, another was agronomist, and the other person was a government official committed to the welfare of grass root level farmers and artisans. The others who followed the discussion with rapt attention represented the general public and expressed the concerns of the old, current and new generations.
The group was so heterogeneous that although the topic of discussion was the same, the viewpoints were very different. It was like the story of blind men and the Indian elephant. If only knowledge/information with each one of them is mapped and merged, a comprehensive picture would emerge, to which more and more content could be added like in snow-balling. Such a map, once developed would/should answer questions like, who is growing neem and who is selling neem and neem products, Are they growing in a small scale for local consumption or as a plantation, Or there is someone organizing neem seed collection and supplies them? If so, how much quantity and at what price? Such type of searching should be made possible by emerging technologies and initiatives. Building topic maps and concept maps with semantic relations is a step in this direction.
In these maps, each concept will be an entity and will be linked to all other related entities by means of semantic relations. For example, `neem’ is a concept, to be more specific – neem plant, neem seed, neem plantation etc. Biopesticides is another concept. Neem `isUsedAs’ a biopesticide. Biopesticides `areProducedBy’ `soandso company’. Mr. X `isDealerOf’ biopesticides `producedBy’ the company `Y’. Mr. X `has_degree’ in agriculture, `is a student of’ ANGRAU and `is_a_friend_of’ Mr.Z, who `isOwnerOf’ so and so company.

This is a very pragmatic approach and this is how people associate and discuss things during conversations. I saw the need for such mapping when there appeared to be some comedy of errors about an IIT guy (read it as person, the jargon and parlance have their own role in semantics) who turned to eco-friendly initiatives. They were in fact two different persons, one growing/gathering neem seeds and the other one needs them as raw material for a biopesticide unit. It is fine if they happen to know each other, but if they don’t there is a critical missing link.
I am not quite sure how searching could be based on such a map, but I would say that it is certainly a better format for `who is who’ in a particular domain at different levels. Such maps can be built collaboratively and can be updated during the `getting-to-know’ and `ice-breaking’ sessions of seminars and training programmes.

Coming back to the familiar literature search scenario, we once received a search request for `pests and diseases of neem’. It sounded very interesting to see if neem gets affected by pests and diseases. But it was not as easy as it was interesting. All we could do was to search the databse with something like (Neem OR Azadirachta) AND (Pests OR Diseases). We retrieved several thousand references and we didn’t definitely expect that many pests and diseases affecting neem. The thing is that the list includes references of pests and diseases of all crops wherever neem is used for plant protection. But somewhere tucked within this long list were a few tens of references where neem was the thing affected. Out of curiosity and also knowing that no other method would work, we ventured to scan titles of all the references, with our trained eye spotting the odd ones out. But this cannot be done every time and can no longer be the solution. We ought to have something like Neem `isUsedToPrevent’, `isUsedToControl’ so and so pest/disease or Neem `isAffectedBy’ so and so pest/disease.

This problem is no more an information professional’s problem. It is now everybody’s problem. The moment someone searches something on internet they face this situation. Many people and many companies are working towards a solution for this kind of `false drops’. Several projects are being funded like how in olden days kings used to support alchemists to convert iron into gold.
Semantic technology, however is an imperative in today’s context and breakthroughs are a must. In the domain of agriculture, AGROVOC has been taken as the starting point to develop an Agricultural ontology. There were several studies and attempts to convert AGROVOC into a concept server by employing some rules and models. But now FAO has decided to do it manually and assigned the task to ICRISAT. The task is to classify all the concepts into one (or sometimes more than one) Top Concepts such as Time, Measure, Activities, Substances, Location, Subjects, State, Objects, Organisms, Phenomena and entities. Some of these concepts are very much in tune with Dr. S.R. Ranganathan’s PMEST concepts. Later we discovered that this grouping is similar to the one that is followed by `Knowledge 2008’ portal.

Under each Top concept, the terms and concepts are organized into hierarchies. Existing hierarchies are reviewed and only those concepts that have either an `is a’ relation or an `is an instance of’ relation are retained in the hierarchy and others are placed under their respective super classes. Sometimes a third kind of relation `whole part’ is also given as a BT (Broader Term) / NT (Narrower Term) relation. Relations between hierarchies are defined by RT relations which are refined by means of semantic relations like `has synonym’, `has spelling variant’, `is used for’, `has goal’, `is object of activity’ etc.
There will be several terms describing a concept, of which one is the preferred form and is chosen to be the descriptor., while all others will be non-descriptors leading to a descriptor. But in the case of plant and animal names, both the taxonomic names and common names are given as descriptors in AGROVOC, and are linked by `is same as’ relation.