First of all, it's important to understand that the boundary between "industry" and "academia" is a lot blurrier than may be obvious. It is certainly this way in the United States, and what I hear from colleagues in Europe leads me to believe it is similar over there as well (though the specific mechanisms may differ from country to country). As somebody who straddles that boundary myself, I see a lot of different people in industry publishing for a lot of different reasons. At the end of the day, however, it mostly all boils down to the same core reason as in academia: credibility.
Despite all of its flaws, scientific peer review is still the best established means that our society has come up with for judging the credibility of factual claims (side note: legal and political systems are primarily concerned with justice, rather than facts, which has subtle but significant differences). When a scientist working in industry wants to establish the credibility of their work, peer review is an excellent choice for doing so. The reasons for wanting to do this in industry are very similar to those in academia:
Scientific credibility from publications improves a researcher's ability to secure funding. This is true both for external funding (e.g., from government agencies), and in competition over priorities for the allocation of internal funding.
Establishing credibility increases the likelihood that others outside of an organization will choose adopt the ideas or products that you are advocating. For example, published scientific studies are critical to establishing claims of safety and efficacy in the process of drug development.
Publications also establish a researcher (and their organization as a whole) as a contributing member of the community: as they are sharing information and working openly, so will others be more likely to share information with them and work together in return.
No company exists in a vacuum: participating in the scientific discourse can help resolve the problems that a company is facing in its own work. This is particularly true regarding standards and instrumentation, but extends more broadly. No matter what organization you work for, most of the smartest people work somewhere else, and demonstrating the scientific value of your problems by means of publication can attract the interest of others to work on those problems for their own reasons.
Scientific credibility also can improve one's standing in an organization, leading to promotions in your personal position. Some companies even provide a tenure-like status (and I believe in some European countries you can actually get tenure in a company).
Personal pride plays an important role too: industrial researchers are no more immune to ego and vanity than academics. If you've got a point to make about an idea (or lots of ideas), the scientific literature is one good place to advocate for your view and obtain satisfaction when others respond to it.
Finally, industry/academia collaboration is very common, and even if the industrial partner might not have published on their own, the academic side of the collaboration will want to publish, and the industrial partner will rightfully be included in the author list.
As these apply in general, so to do they apply to machine learning, natural language processing, and data-mining. Certainly this is a good description of why the researchers in these areas in the company I work at publish quite strongly, and fits what I see from others in the area as well.
