In 1989, 15 years after my study at University College London, my supervisor Bertram Brookes asked Eisuke Naito, one of my UCL classmate to visit and ask me to write to him. Foolishly, however, I did not write to him. Still I wonder what he would have talked to me. Anyway, it must be that something was still going on from my thesis. So I am proud of my Unfinished Symphony or my MSc thesis, which aimed to build up a citation-context database to help scholars perform citation tracking better than their manual way. They had to go to the library to know more about the cited document. With such a database, they could do without the library. For it contains the citation contexts, though they might often be not enough to know the full aboutness of the cited.
Anyway, my principle was realized as CiteSeer, now CiteSeerX, in 1998 by NEC Research Institute, now NEC Labs, in Princeton, New Jersey. Google Scholar, Microsoft Academic (later withdrawn) and many others followed. The validity of my principle of citation-context indexing was proved. Unfortunately, however, few know the fact, either conceptual or historical. Why not? After my MSc study, I had to serve for my employer for three years. Brookes and UCL might have feared that during that period another university, say, Stanford would like to recruit me, if I had been well known by citations. So he is supposed to have recommended them not to cite my thesis. No other explanations appear possible. No citation is no evidence of the poor quality of my thesis. Its quality was already proved as suggested above, regardless of citations.
What is hypertext at all? The webpage is a typical kind. The scholarly paper with citations is a primitive kind, which are real or potential hyperlinks. To build up a citation-context database is to let go to the cited easily. Therefore, my thesis cannot do without hypertext. The World Wide Web made CiteSeer be realized more easily. You cannot dare to conclude that the WWW was developed completely independently from my thesis. In 1989 when he sent Naito to me, Brookes might have wanted to tell me that the WWW was an offspring of my thesis, which anyway relies on hypertext like the Web.
On the other hand, why my thesis should be taken more seriously is elsewhere than citation-context indexing. To prove it, I built up its theoretical foundation quite thoroughly. This may be more worth than the practical aspect. To justify such indexing, I took both context and cognition most seriously. We will call it ”cognitive contextualism”, which lead to the cognitve turn in information science since the late 1970s. Meanwhile, quite recently, another contextualism made a great success in generative AI or LLM, probably since Ashish Vaswani, et al. (2017) “Attention is All You Need.” We will call this success ”attentive contextualism”. Here, human cognition is compared with Transformer attention. Both attend to context. We experienced two great paradigm shifts: one was in information science since the late 1970s, while another in AI since the late 2010s. Both are commonly rooted in contextualism, which appears the reality of the universe.
Information seeking (IS) is a human instinct. The more information comes out, the more difficult IS becomes. This is reflected on the emergence of information retrieval (IR) emphasizing the mechanization of its process using a massive database. Instead, it has alienated human cognition, hence machine-centered instead of human-centered. This has made the handling of context difficult. The human intermediary and so-called reference interview has nothing to do with that handling. Otherwise machine-centered IR, human-centered IS has carried on with academic citation tracking based on cognition of context.
We have to take seriously not onlt verbal context but also mental and enviromental contexts. The same verbal text may be accepted differently by the different mental and enviromental contexts. It is not an easy task to expose them in detail, though you spend a lot of time. Such may be the case with scholarly research works. Early LLM researchers for many decades might have paid little attention to vitality of verbal context. The state of the art may better be understood as scholarly mental and environmental contexts beyond the published verbal context as a whole.
To be informed is to become wise. Then the science concerning information is a science for wisdom. And it begins with context. LLM has mastered the whole context or universe of knowledge much more than any specialized scholar. We may need metascience or science of science in addition to philosohpy of science. And both AI and information science may better combine together into metascience. Context as the state of the art would be the main target of study.
