Never be afraid to get dirty, but be sufficiently sure-footed to avoid the abyss of contamination.


When I had friends over last month for a South Indian dinner, one of the items was sambar. I had difficulty describing what sambar was. Soup didn’t quite work because it wasn’t watery enough, but there were too many vegetables in it to make it a sauce, either. Trader Joe’s now sells sambar, which they refer to as an Indian stew.

Joseph Conrad purportedly described English, his third language, as the hardest to write in because no two words mean exactly the same thing. While several words have the same denotation, few of these have the same connotation. While information theorists are not bound to worry about meaning, reliable transmission of information is one of the principle problems of the theory. Shannon’s mutual information has the property that translations of data can never increase the amount of information. This property is called the data processing lemma, and a challenge for the information theorist is to avoid potential losses due to such translations.

In exchange for a “free” lunch and bookbag, I accepted an invitation to meet with the CEO and other high-level executives at Marvell. The opportunity forced me to translate my research into business-speak. I had to hear business-speak, too. The Marvell executives described the company’s success with optimistic and positive words. At the same time, they handed out a copy of a Forbes article on the company, which acknowledged the company’s financial success but translated the factors leading to it with words carrying a different connotation.

Not one to subject myself to such conversations without company, I went with a friend that was also invited. He’s been speaking English for just a few years. While his comprehension has improved tremendously, he still needed an English to English translation of the following conversation with one of the coordinators.

Coordinator: All right, we want to get people started with the lunch. It’s going to be a buffet, so if you two could find a seat– there’s a complimentary bookbag for you there– and start a line at the table, you would speed up the process. Others will follow as soon as they see a few people starting.

Me: Sure thing.

Friend: (to me) What did she say?

Me: “Put your stuff down and grab some food.”

I doubt much information was lost in translation.

5 responses

  1. Rich

    You lost the “free bag” part in your translation!! šŸ™‚

    October 1, 2006 at 4:44 am

  2. Krish

    Lucky for me, he had eyes.

    October 1, 2006 at 5:12 am

  3. Triya

    Translation is always tricky. I remember once trying to explain Gulab Jamun to my American friends and as soon as they saw it one said.. aah!! meat balls and another tasted them and said doughnut balls!!

    October 1, 2006 at 12:49 pm

  4. Lav

    Hey Krish, I just came across your blog. Very good stuff.

    Do you think translation is a point-to-point problem? My impression is that it is really a multiterminal problem, where the translator is a second source rather than just a noise source, and the destination is interested in both the original message and the translator’s message.

    For example describing the milk, the frying, the sugar syrup, and the rose flavoring gives useful “recipe side information” by time sharing the translator’s message with the original message.

    October 1, 2006 at 7:24 pm

  5. Krish

    Hey Lav,
    It’s great to hear from you. I thought about the multiterminal perspective after posting. Ignoring “dubbing” in which the untranslated message is unavailable, it depends on how much one assumes the decoder knows about the joint statistics of the side information with the underlying source and how one defines rate. If the rate is defined to be robust to arbitrary joint distributions between the untranslated side information and original source, then the side information can be thought of as useless.

    In the case of my friend, however, the original message of the coordinator probably was useful as side information since he understands English.

    The problem with all of these analogies is that the messages in question have meaning, and a rate-distortion perspective is insufficient to get at the heart of these semantic issues.

    October 1, 2006 at 7:47 pm

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s