Re: Afghanistan War Diary as topic map in Maiana
Patrick Durusau <patrick-Q/[email protected]>
| Newsgroups | gmane.text.xml.xtm.general |
|---|---|
| Message-ID | <[email protected]> |
Benjamin, On Fri, 2010-10-29 at 10:49 +0200, Benjamin Bock wrote: > Hi Lars, > > The work presented is just a simple conversion after some minimal data > cleaning. While the short text occurrences where cleaned, the long > text passages where not cleaned up and are full of typos. Besides > "Taliban" you find also "Talibah", "Tali-ban", "Talib", "Taliwan", > "Talibjan" and so on. Without detailed domain knowledge, we don't know > which are same and which are different. There are also "Taliban Flag", > "Taliban Check Point", etc. in the texts - should these be connected > to the Taliban Topic? If so - how? > Also, these texts are full of military speech and names of people and > places. Both again, often misspelled, transliterated in different ways > - or are they actually different subjects? How should we know? Running > a generic spell checker on these doesn't make things better. Garbage > In, Garbage Out. > Certainly raising the potential for text mining tools to augment the authoring of additional topic map constructs. Noting that different users of the map may be interested in different topic map constructs for the same tokens in the text. > As pointed out by others: You can easily add subjects and data > yourself and use the predefined TMQL functions like the distance to > get more information out - it would be nice of you showed how to do > that in Excel with a CSV file. > If you feel like you're more of a military expert than we are, or if > you know Afghanistan, its localities, its inhabitants and how they're > called by different parties, you're free to produce a semantically > rich topic map. > True, although it would be even more interesting if this were used as a teaching moment for a discussion on how to develop and *exchange* merging tests that don't rely on simple string matching. The "Neighborhood Auditing Tool" concept of neighborhoods (http://tm.durusau.net/?p=3047) is one that I think has legs, particularly for topic maps. Beyond defining merging on the basis of roles, imagine defining merging based on more complex relationships. That is what NAT is doing without calling it merging or making it explicit for automated processing. I don't have a syntax or even a model for such tests but suggest this as a possible starting point for experimentation and discussion. > Our mission was just to show that you can convert data to Topic Maps > with barely any effort and that you can go on from there doing things > you care about with the data, like querying reports around cities. > > Many people still think "Topic Maps" is something like "Google Maps" > and this is just one way of showing them what it really is. > Well, what I am working for is for people to be asking "Google Maps" why they are not more like "Topic Maps" but that day is at least after the first of next year. ;-) Hope you are having a great day! Patrick > Benjamin > > On Fri, Oct 29, 2010 at 9:24 AM, Lars Heuer <[email protected]> wrote: > > Hi Marcel, > > > > [...] > >> In short: converting raw material into Topic Maps (or RDF) makes it easier > >> to apply ontologies. And with this enhanced data we can merge other sources > >> like "George W. Bush"-ish sources. Or? > > > > Yes, you can. And adding more useful topics/subjects to the WikiLeaks > > Afghanistan topic map was/is my point. I want to see "Taliban", > > "Kabul", "Mazari Sharif" and the relationships. I see "Kabul" and > > "Mazari Sharif" but no relationships. They are cities and they have > > coordinates... . Hmm ... > > > > Maybe I was too harsh/not specific enough with the criticism of the > > topic map, but having something as topic map is not a value of its > > own. If the topic map is developed further it may become more useful. > > Currently, it consists mainly of (IMO) not very useful occurrences > > which are a literally translation of CSV data. > > > > Best regards, > > Lars > > -- > > Semagia > > <http://www.semagia.com> > > > > <http://www.topicmaps.de/mailinglist/> German Topic Maps mailing list > > <http://tinytim.sourceforge.net/> Open Source Topic Maps engine > > <http://mappa.semagia.com/> Mappa - Python Topic Maps engine > > _______________________________________________ > > topicmapmail mailing list > > topicmapmail-Zo64W7twoUFWk0Htik3J/[email protected] > > http://www.infoloom.com/mailman/listinfo/topicmapmail > > > _______________________________________________ > topicmapmail mailing list > topicmapmail-Zo64W7twoUFWk0Htik3J/[email protected] > http://www.infoloom.com/mailman/listinfo/topicmapmail