This is OOSUGI at ATL. I report on the Hackathon of the Natural Language Processing at Recruit, where we invited public participation.

There were total of 10 participants a long way from home (1) in a room for 5 days. We analyzed the access log of the news contents and the product information run by a service at a Recruit group business (2).

1hackathon_work

Although the room was too tiny, which I realized I should put more consideration, I think I provided a good environment where the participants could concentrate on the analysis very well.

関根聡

Dr. Satoshi SEKINE from New York University had a look around, too.

1hackathon_prezen

On the final day, all participants made their presentations of the analysis in front of the staff at Recruit, including the CEO of Recruit Technologies Co., Ltd. They talked about the stunning points of their ideas and the results of their own analysis.

1hackathon_ending

We ranked the result of the analysis because this was Hackathon; for the first place an e-book device was awarded, and for the second place (in technology category and in technique category) a 10,000 yen coupon book was awarded to each.

  1. Create the combination of keywords received by calculating the distance between the words in an article
  2. Discover the grammar structure in a received article from the titles in the articles with high page views.
  3. With expecting the page views from the content of the articles, study-observations of the characteristics of the title with low page views which are inexplicable just from the content.

There were a variety of results even in the same dataset like below.

What impressed me a lot was that the undergraduate students did well, who seemed more inferior (*3) because they only just learned Python and SQL this time.

The reason of that was mostly because they faced earnestly to the data by challenging what they could do, not what they could not do. How well the students handle the raw data is a main factor in analysis. That was a good event both for the participants and facility staff because they studied well and inspired each other. Thanks for your participation. We will be back by popular demand soon. So, see ya!

  1. We gave priority to the people who came from all over Japan like Kinki area and those who could not easily come to Tokyo so often. Recruit Technologies was generous enough to pay all transportation and hotel fees.↩
  2. Thanks for the cooperation: Recruit Sumai Company.↩
  3. It means, in comparison with the graduate students who are really into the Natural Language Processing. The undergraduate students were not the beginners at programming at all. They really developed their faculties in the meanwhile.↩