Ever wish you had a personal assistant? Someone who could recommend a good restaurant when you're traveling to a new city? Someone who could give you a list of shops where you might find that perfect gift?

In a few years, that personal assistant just might be your own mobile phone.

Engineers at Palo Alto Research Center (PARC) have developed software that they say can make recommendations about local restaurants, concerts, shopping areas and other activities based on the time of day, the user's physical location and the user's personal tastes. The researchers aim to let the software, called Magitti, turn a cell phone into a personal assistant of sorts.

"We're trying to make [the cell phone] more like a human," said Victoria Bellotti, a principal scientist at PARC. "Instead of just directing stuff at you, it tries to make inferences about what kind of activity you're engaged in. On a Sunday afternoon, it might suggest going to a park or a gallery. It will learn patterns based on what you tend to show an interest in. It looks at things like where you go, what recommendations you liked. It's like having a companion with you."

PARC is researching and developing the software on behalf of a Japanese company, Dai Nippon Printing Co. Bellotti said PARC already tried the software out on users in the Palo Alto, California, area, and it will have its first official trial run in Japan next spring. She added that it is tentatively slated for release during 2009.

Bellotti said the software focuses on five different activity modes: eating, shopping, seeing, doing and reading. The software, in essence, learns from the user. What does he like to do on Friday nights? What types of restaurants does he prefer when he's out of town? Where does he like to shop on Sundays? Magitti will offer suggestions to users based on the answers to such questions.

"It doesn't directly know anything except the time and location and how you interact with it," explained Bellotti. "It provides a list of recommendations. It will give you a wide range of what would be in a city guide. You get a big list and then you can ask to see more details. Then it can show you where these things are on a map. If it realizes you are shopping, it will offer you information to make it easier."

Bellotti said the software uses artificial intelligence algorithms that make inferences about what the user is doing by comparing the GPS location of, say a restaurant he frequents for brunch, with a database of eateries. Based on the knowledge the Magitti accumulates, recommendations will change, and hopefully become more honed, over time.

Don't like what the software is giving you for options? Say no and redirect it by doing a search for something different, she added. The software also learns by what users dismiss.

"Our hope is that over time, it will get more and more clever using your past behaviors to know you better and know your preferences better," said Bellotti. "Eventually, it will know that you will generally want a restaurant recommendation at 9 o'clock at night. It will know you want to go hear music on a Friday night. You have a routine. There are times you go to the gym. There are times you want to do something different."