Liferay Portal: Consultant (Canada, USA)

[Fuad Efendi, geändert vor 15 Jahren.]

My next availability date is October 01, 2008; curently implementing Liferay Portal migration project in Toronto, Ontario (Canada).

Just finished: Click Magazine at Lavalife.com

- extremely fast
- 3-months development
- SOA: integration with partner sites (weekly polls, horoscopes, dating articles)

416-993-2060(cell)


Another project: Shopping Price Engine - web crawler & search engine, extremely fast (600ms average response time including network roundtrip)

Third one: Natural Language Processing for collaboration networks, to be announced!

Thanks!

[Fuad Efendi, geändert vor 15 Jahren.]

Some thoughts about Liferay:
- it looks ugly at first, but it is not ugly
- ServiceBuilder is extemely good tool, but you have to tune database indexes manually (including existing ones which are weird)
- Performance: excellent in right hands. Most client-side AJAX-related issues resolved in 5.1

Development time: excellent!!!!!!!!

I like this not-so-popular approach: we have access to internals via static classes (most of them generated); we use JSPs for presentation & interaction (including data reads)
- extremely fast development.

[Fuad Efendi, geändert vor 15 Jahren.]

Here is my parttime hobby, Shopping Price Engine. I don't have personal BLOGs available at Liferay Portal, sorry for using this thread for marketing purposes. BTW, my profile here is SEO-optimized, and at LinkedIn too: Liferay Consultant at LinkedIn.
Thank you!

About Tokenizer: it does not currently use Liferay. However, I am going to build Liferay-powered Bambarbia Kirkudu!!! I was extremely lucky by registering both domains, it's very popular words in former USSR (taken from Russian comedy Кавказская пленница, или Новые приключения Шурика - thanks for UTF-8 support!

Tokenizer is built upon Apache Lucene, Nutch, SOLR, Hadoop, and HBase; "token" means a piece of text used for indexing, tokens are mapped to documents for full-text search. Liferay currently uses simplified WhitespaceTokenizer and still have a problem in extracting text from XML (which is internal representation of Journal Article Content): two different structure elements city and country will be extracted without additional whitespace between them, [TorontoCanada] instead of [Toronto Canada]. WhitespaceTokenizer can't deal with that! See LEP-4392 which is still open...

What about more sophisticated search? Yes, we need to use stemmers which will trim words to their common part:
telescopes, telescoping, telescope ==> telescop

We need lowercase tokenizer, we need synonyms, ... much more is needed for Shopping Tokenizer!

If you are interested in SOLR: it is extremely slow out-of-the-box (only if you are trying to do faceting search by tokenized field). Use fast caches instead of default SolrLRUCache (to give you an idea, browse sourse of LingPipe); use TermVectors instead of DocSet Intersections.

I'd be happy to post more details here. May be it's time to write WIKI page on Liferay Search...

Thanks!

[Fuad Efendi, geändert vor 14 Jahren.]

I am available again:

Finished brilliant project at casaGURU.com - Helping Homeowners to Find Renovation Professionals - over 50 additional portlets, over 80 additional tables in schema, heavy customization, private messaging, project submissions & distributions, billing, Email alerts, SOLR, search by postal code and service area, REST integration with partner sites, and more.

Very strong with SQL-92, PL/SQL, Java/J2EE/JEE, Performance Tuning, AJAX/HTML/CSS. + Expert on Oracle, WebLogic, ESB/SOA.

Expert in Vertical Search & Data Mining & NLP: Hadoop, Cascading, Lucene, SOLR, Nutch, Heritrix, BIXO, LingPipe, GATE.


416-993-2060(cell)
Fuad Efendi
Liferay Portal Blackbelt, Vertical Search, Data Mining