German Hotel Search Engine Mashes Up Watson And IBM i

February 9, 2015 Timothy Prickett Morgan

Here is the kind of story that you can bet IBM wants to see more of as it tries to peddle its cognitive analytics expertise to customers using traditional transaction processing systems running on what many people call legacy systems.

An intrepid IT consultant based in Kaufering, Germany, named Rainer Ross, who has expertise in Power Systems machinery running the IBM i operating system and its integrated DB2 for i relational database management system, has created a new search engine for locating hotel rooms called myhofi. The hotel search engine is actually providing a new revenue stream for the consulting firm, which is called Rainer Ross IT-Beratung. Hopefully the new search engine will not be so successful that Ross decides to leave the IBM i consulting business. The IBM i market needs people who can mash up legacy and new technologies to create value, and importantly see the opportunity to do such extensions.

The story of the myhofi hotel search engine was actually put together by IBM Deutschland in a recent case study, which you can read here. And basically, Ross got the idea for the myhofi hotel search engine by being an unsatisfied customer.

“I found that many existing hotel search engines were slow, difficult to use, and full of advertisements,” Ross explains. “Users can typically filter using only specified fields, rather than typing in exactly what they are looking for. I recognized a gap in the market for a fast, easy-to-use hotel search engine with full-text search. To build one without having to hire extra staff or give up my consulting work, I needed a platform with very low administrative requirements on which to run it. High performance was critical, as slow response times repel users. Also, I needed a solution that could easily scale to support growing query numbers as the site’s popularity grew.”

The setup that is running the myhofi hotel search engine is fairly modest, a Power 720 machine, which is a typical box for most IBM i shops that are running on relatively modern hardware. The machine is equipped with the latest IBM i 7.2 operating system and its integrated DB2 for i database. The details of the search engine were not divulged, but the application back-ending the search engine is written in RPG using IBM’s Rational Developer for i tools and the font end is coded using a mix of HTML5, JSON, and XML documents with JavaScript glue and CSS style sheets. The neat bit is that Ross has integrated IBM’s Watson Content Analytics to do linguistic matching on search items and descriptions of hotels and their rooms that are encoded in the documents stored in the DB2 for i database. The Content Analysis software means that searches can be much more natural and specific, not limited to tick boxes that more or less correspond to fields in a back-end database. The Watson Content Analysis software links to the DB2 for i database using SQL statements that include searches based on keywords in the searches that hotel bookers use to find rooms.

Of course, the myhofi search engine can’t compensate for thin descriptions about hotel rooms. But this approach may compel hotels both large and small to provide more detailed information that can then be indexed. (Having stayed in my share of hotel rooms in the last couple of years of business travel, I would say that accurate descriptions might not yield as many bookings as one might surmise. . . .)

At the moment, the myhofi search engine is free and it is not clear how Ross is monetizing it, but it seems likely that he is getting a commission on hotel bookings. Even if the revenue stream is modest, the application took only nine months to develop, has run nearly nonstop without any issues for nearly a year, and most importantly is a showcase that Ross can use to demonstrate how the integration of various IBM technologies can provide an improved service.

The average response time of the searches on the myhofi site is around a half second or less, so it is pretty zippy, too. Ross says that he reckons that the IBM i hardware and software costs about a third less than “competing solutions,” and as you might expect, I would love to see that quantified a bit better.

The Watson Content Analyzer software is one of the underpinnings of the Watson question-answer system that beat the humans playing Jeopardy!. It is available on AIX, Linux, and Windows platforms. It is not clear how Ross integrated it with IBM i, but it is most likely that the Watson software is running in an AIX or Linux partition, with Linux being the obvious first choice these days.

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search