We (my wife and myself) are now proud members of the North Albert Field Archers club and the 3D Archery Association of Australia. Starting regular practice with my (ultra) traditional Grózer Nomad G 5 – Old Hungarian bow…
A significant problem of the dominant web search model is the lack of a realistic way to acquire user search context. Search engines use implicit feedback, which is extremely sparse and does not allow users to properly define what they want to know, or what they think of search results. In our proposed “web exploration engine”, which we implemented as a prototype, documents have been automatically pre-classified into a large number of categories representing a hierarchy of search contexts.
Users can browse this structure or search within a particular category (context) by explicitly selecting it. Keyword relevance is not global but specific to a category. The main innovation we propose is the “floating” query resulting from this feature: the original search query is re-evaluated and the importance of its features re-calculated for every context the user explores. This allows users to search or browse in a truly local (context-dependent) way with a minimum of effort on their part.
Article presented at The 14th Asia-Pacific Web Conference (APWeb) in Kunming, China (11-13 April 2012).
Download PDF version: Towards Real Intelligent Web Exploration (12 pages).
The Federal Government has announced it will withdraw tax benefits from families whose children are not immunised (news: http://www.abc.net.au/news/2011-11-25/immunise-or-lose-benefits-parents-told/3694236).
Immunisation is a (very invasive) medical procedure and as such requires informed consent by the patient. My son being less than 3 years old, it is my responsibility to care for his health and provide such consent. However, I have not been informed and the little I have learned about immunisation makes me believe it is more harmful than good.
What is often quoted by proponents of immunisation (the government included) are incidental studies confirming some benefits (X children were immunised, Y percent of them got a disease while we expected more of them to – hence it’s working). This is NOT convincing. People against immunisation cite opposite cases. Both of these approaches are based on anecdotal data. The FULL data is already available to the Australian government but is ignored . A simple correlation between various Medicare datasets could reveal how healthy immunised children are against how healthy not immunised children are (based on average yearly visits to the doctor resulting in prescribed treatment). This is a) not anecdotal evidence but based on the whole population, and b) revealing the whole picture (i.e. immunisation might save children from some diseases but weaken their organisms and open them up for others which are not studied in the cases supporting immunisation).
I requested the above information from Medicare but was denied it, on the basis that such a correlation would be too expensive to perform. If it has not been performed already (and made public), that means the government has neglected the facts and this new legislation is not based on fact but belief.
Furthermore, whatever the facts are, immunisation is a situation of balancing risks – definite risks from adverse side effects (which are admitted even by vaccine manufacturers) against hypothetical future risks of diseases. Decisions about such balance are personal and cannot be mandated by government. Putting financial pressure on me in order to accept one view point is an attempt to force me into a decision, whereas the law says I should be free to decide for myself.
Starting to get pissed off with how everyone just assumes it’s entirely OK to follow me around the web to see what I’m doing, what I’m searching for, who I talk to etc.
Last year my university email account was switched to GMail. I had some misgivings, but had no choice. Today, however, when I was searching for something on Google, I saw in the top right corner I am logged in as pavel.kalinov@my uni. Er… no, I am not – or rather, I should not be. I logged into my email account which is on a completely different domain. I specifically did not log into Google search, Google plus or whatever. What makes you think it’s OK to log me in and track me?
Furthermore, even if I log out at this point – I will still be tracked in a personally identifiable manner since logging out does not delete the cookie. So, emailed the letter below to IT support, will wait to see what they do and then send it to the Privacy Commissioner and/or the Telecommunications Industry Ombudsman (you need to resolve the issue with the offending company first before you go to them).
I am concerned about some features of my university email that have appeared recently (without my, and I believe the university’s consent).
Since the introduction of Google+ or whatever, I seem to be logged into Google all the time after I check my university email. I have a personal Google account, and now it seems another one was automatically created for me without my consent and it keeps “following” me whether I want it or not – I search for something on Google and it shows my uni account on top as being logged in. Well, I specifically logged into Griffith Mail, not Google Search. If I don’t want to be tracked there is nothing I can do. I can log out, but then I am logged out not only from search but from my mail as well, and I am still being tracked in a personally identifiable way even after that because the cookie in my browser still says who I am – it now just has a “logged out” status.
In short, I find this a violation of privacy which I think you should resolve between the university and Google as an email provider. The fact that they provide email (and through the university – I have had to accept that service as well) does not entitle them to follow me through their other online properties.
04/10/2011 update: got a phone call from a very helpful lady who suggested to try and de-link (or nuke altogether) my personal Gmail/Google+ account as some sort of mitigation. Tried it, turned out this was not the issue: (copy/paste)
after your phone call of a few minutes ago, I tried to go to Google and delete the linked account and any possible interference between my Griffith email and my private Gmail account. As it turned out:
- my private email does not interfere with this one (i.e. – they have not linked them)
- they have NOT created a Google+ account for me as I initially thought (though the issue coincided with Google+ becoming public)
- but, they have created an “ordinary” Google account for me, which still raises the original issue. Namely: it happened without my consent (and, more importantly – without your consent as a corporate customer – I don’t think you allowed Google to poach your clients and make them its own clients), and my web search/web browsing habits are now being tracked in a personally identifiable way through this created link between my Griffith email, Google search and Google Analytics (the tracking scripts of which are installed all over the place).
In short, they have misused your customers’ data in a very alarming way.
04/10/2001 next update, after an email basically advising me that Google changed stuff unilaterally and that’s all… (copy/paste)
as I said, the issue is not that I cannot access my other account – the issue is that I have a new Google search account at all (NOT apps, I mean I end up logged in to google.com while I am using the standard Google search feature; and, as you know, logging out does not delete the cookie but only changes the status to “logged out”). I haven’t investigated, but I guess I remain trackable all over the web through this cookie and Google Analytics scripts on millions of sites.
“In the past, Google Apps (@griffithuni.edu.au) accounts were cookied separately from Google Accounts (e.g. gmail.com)” – this was a correct behaviour.
“Your Google Apps account has undergone a transition, and now functions more like a full Google Account.” – this is overstepping the boundaries on the part of Google. Think of it as a “forced sale of a service” (the fact that it is “free” does not mean there is no price to pay – the price in this case is the user’s privacy, and Google make good money out of it through selling personalised ads on third-party sites).
Since I have no relationship with them but with you, I take the issue with you so you can then take it up with them as a corporate customer. As I said in my previous letter, you have outsourced a service with them and they are now poaching your clients and forcing additional services (at a cost to them – see above) with no opt out (not that it would make the practice acceptable, as services should be opt-in).
I recently sent a letter to Medicare Australia asking a simple question about the effectiveness of child immunisation, which I assumed would get an informative answer thanks to the Freedom of Information Act. However, I got a formal reply which basically told me I don’t need to know what asked, as other information was available online (with kind pointers to said other information). However, this other information is not relevant to what I wanted to know, namely:
- how many children in Australia are immunised and how many are not
- how many of both groups do get sick
So, this is my reply which I just emailed:
I am writing in response to your letter which I received today, refusing to supply the data I had requested re immunisation and other statistics for Australia.
I appreciate the effort you have made in pointing me to available resources, but they do not help me in finding what I want to know, namely: how effective immunisation is in actually preventing sickness.
- Immunisation statistics as such say nothing: knowing how many people in Australia are immunised means nothing if I don’t know how many of them later got sick, and I cannot compare it to the same figure for people who are not immunised.
- Immunisation studies are wonderful, anti-immunisation publications are also fascinating, but both of these are useless as they are not comprehensive; they are all “anecdotal” if they do not cover the whole population and for all cases. Being a researcher myself (although in a different field), I know what “selective reporting bias” means: if a method is successful in 40% of cases, I can conduct a set of 10 experiments and publish 4 (truthful!) studies saying how it is ALWAYS successful, and a further 6 truthful studies saying how it is NEVER successful. I believe this to be the case of immunisation and anti-immunisation publications (the latter of which do not get much official publicity).
The only useful information to settle the issue of immunisation effectiveness is the information you hold and which is NOT published: i.e., the correlation between children immunised in Australia and children with health problems in Australia: how many of one group have problems, and how many of the other group have problems.
In conclusion, I believe your letter was not informative in the sense I was seeking information and I need to repeat my request:
Can I please have the statistics of:
- children immunised / not immunised in Australia;
- members of both groups having health problems, defined as “average visits to the doctor per year which ended with a prescribed treatment that Medicare had to pay”. Additionally, it may be interesting to see the average dollar cost to Medicare of both groups per year.
A significant problem of the dominant web search model is the lack of a realistic way to acquire user search context. Search engines use implicit feedback, which is extremely sparse and does not allow users to properly define what they want to know, or what they think of search results. In our proposed “web exploration engine” documents have been automatically pre-classified into a large number of categories representing a hierarchy of search contexts. Users can browse this structure or search within a particular category (context). Search is truly “local” as keyword relevance is not global but specific to the category. The main innovation we propose is the “floating” query resulting from this feature: the original search query is re-evaluated and the importance of its features re-calculated for every search context the user explores. Additionally, users can provide explicit feedback, which automatically modifies the search query.
This article was submitted to the 12th International Conference on Web Information System Engineering (WISE 2011) in Sydney, Australia (13-14 October 2011).
Download PDF version: Towards Real Intelligent Web Exploration (14 pages).
The current search engine model considers users not trustworthy, so no tools are provided to let them specify what they are looking for or in what context, which severely limits what they are able to achieve. Instead, search engines try to guess that, which is currently done using “implicit feedback”.
In this paper we propose a “web exploration engine” – a model where users can use the search engine as their tool and explicitly specify the context of their search. Information about the web has been pre-classified in a large number of categories; users can explore this hierarchy by providing relevance feedback or search within a particular category. Search is truly “local” in the sense that keyword relevance is not global, but specific to that category. In contrast to the existing search engines, users can explore the web without any keywords, guiding the exploration engine with relevance feedback alone.
Download short PDF version: Let’s Trust Users – It Is Their Search (4 pages, as published).
Download full PDF version: Let’s Trust Users – It Is Their Search (8 pages, as originally submitted).
Download presentation (ZIP format; unzip, open index.html in web browser and navigate as in a standard presentation): WI2010,
or view presentation online (navigate as in a standard presentation).
Due to the lack of in-built tools to navigate the web, people have to use external solutions to find information. The most popular of these are search engines and web directories. Search engines allow users to locate specific information about a particular topic, whereas web directories facilitate exploration over a wider topic. In the recent past, statistical machine learning methods have been successfully exploited in search engines. Web directories remained in their primitive state, which resulted in their decline. Exploration however is a task which answers a different information need of the user and should not be neglected. Web directories should provide a user experience of the same quality as search engines. Their development by machine learning methods however is hindered by the noisy nature of the web, which makes text classifiers unreliable when applied to web data. In this paper we propose Stochastic Prior Distribution Adjustment (SPDA) – a variation of the Multinomial Naive Bayes (MNB) classifier which makes it more suitable to classify real-world data. By stochastically adjusting class prior distributions we achieve a better overall success rate, but more importantly we also significantly improve error distribution across classes, making the classifier equally reliable for all classes and therefore more usable.
This article was published at the Twenty-First Australasian Database Conference (ADC2010), Brisbane, Australia, January 2010, part of the Australasian Computer Science Week 2010.
Download full PDF version: Building a Dynamic Classifier for Large Text Data Collections.
Download presentation (PDF, view in presentation mode): ADC2010.
Welcome to pavka.com.au!
This is both the personal site of Pavel Kalinov, and the company site for his web development company pavka, registered in Australia.
You will find personal information here, as well as some info on projects of the company.
My other sites
- Transdat - Global earthmoving equipment database.