SDL announced today that it was buying machine translation (MT) supplier Language Weaver for US$42.5 million. The MT supplier will become SDL Language Weaver as the entire management and technical team joins SDL as a business unit reporting directly to CEO Mark Lancaster.

Language Weaver CEO Mark Tapling briefed us on what was happening — and what changed since Language Weaver granted reseller rights to SDL in April 2009. Tapling offered a few reasons for selling to SDL: 1) a “sincere” offer to the market, focused on giving customers the real choice of a single vendor with both technology and service competencies; 2) Language Weaver’s buy-in to SDL’s vision  of global information management; and 3) lowering barriers to MT by limiting customer involvement to dealing with a single vendor, invoice, and service organization. This last point underscored commercial uncertainty about innovative small players versus larger ones with more traction and stability.

The big question for everyone is “why sell now just as MT seems to be taking off?” Tapling said that Language Weaver, which has been investigating its next-round financing options for the last year, didn’t have to be acquired by SDL. Rather, he claimed that the company had some good offers from the venture capitalists, but we suspect that they would have required a lot of preparation and work to understand and address the opportunity. Likewise, we figure that the current venture capitalists, In-Q-Tel (the CIA’s venture arm) and Palisades Ventures, were like many investors whose patience beyond a five-year investment horizon was getting thin. On the pro side of being acquired, Tapling saw immediate synergy in teaming up with SDL. The investors came away with a decent return, but nothing like the ROI that the huge MT opportunity would indicate.

What does SDL get from the deal? The company gets Language Weaver’s development team, its government business that we estimate to be around US$8 million annually, and US$5 million in the bank. Most importantly, it vaults from a second-tier MT position to first-rank potential. SDL’s current MT product is an aging rules-based MT engine that sits behind its KbTS service solution, but now it can field one of the leading statistical MT (SMT) engines that originated in the same academic lab as Google Translate and Asia Online.

What does this mean for the market?

• Consolidation and expanded connectivity continue. Fresh off its acquisition of XML editor company Xopus last month, SDL bought one of the leading MT suppliers. Earlier in the month, SDL also pledged to support Microsoft Translation Server with its TMS product. As SDL expands its software portfolio and partnerships, the importance of extending its OpenExchange initiative beyond desktop Trados becomes even more important. Language Weaver has been working on its own integration interfaces, so it will be interesting to see how committed SDL actually is to opening its ecosystem to both competing and complementary products. Otherwise, SDL leaves itself exposed to more open, more agile rivals for the role of global content backbone.

• MT will get exposed to more buyers. As SDL integrates Language Weaver into its service stack and sells the engine outright for on-premises and SaaS usage, more MT sales representatives will push the technology into more accounts. The previous reseller relationship gave SDL reseller rights à la Lionbridge’s use of IBM technology, but there was some speculation that SDL did the earlier deal just to hedge its bets. Owning the Language Weaver technology outright will ensure that it gets exposed to every SDL client and prospect.
  

• MT gets a faster horse in the global content race. As an innovative but small supplier, Language Weaver had to move more deliberately with some innovations such as its next-generation multilingual web portal. As part of SDL, Language Weaver gets plugged into a more comprehensive content strategy, including content and translation management, translation memory, terminology, and authoring. This enhanced portfolio of products could bring MT into more business functions such as customer care and more of the verticals where SDL’s services business plays. At the same time, it elevates Language Weaver’s TrustScore metrics to a broader audience where it could arguably be used in MT-to-MT competitions.

• Other MT players get a lift. As many language service providers (LSP) begin to provide post-edited MT as one of their service offerings, software-only suppliers such as AppTek, Asia Online, Lucy Software, PROMT, and SYSTRAN can look forward to increased inquiries as rival LSPs erase SDL Language Weaver from their short lists. Open-source offerings such as MOSES, new entrants such as LinguaSys, and LSPs like Pangeanic that own their MT technology should also benefit.

But the MT sector is not all beer and skittles as the market coalesces. We see some residual weakness, perhaps a “hangover” from the 2009 market and lingering concern about the ability of smaller suppliers to get the traction and revenue they need with corporate buyers, government agencies, and LSPs. We also wonder about the ability of all the MT suppliers to compete against free Google Translate, despite concerns about the integrity of the latter’s data, lack of multi-tenancy, and inability to train or otherwise refine the engine.

Finally, we also question whether recent improvements in MT quality are as substantive as they seem to be, whether the tolerance of information consumers for less-than-ideal translation may have increased (see “The Good Enough Revolution — When Cheap and Simple Is Just Fine“), and whether that quality matters at all. In every business sector that we review, business buyers and consumers regularly make trade-offs among price, longevity, and time to market. For global content, the vast ocean of content means that organizations have the choice of translating a small amount of content with human translators doubling or tripling that volume with post-edited MT for the same price with no loss of quality for the same amount of money – or dramatically pumping up the volume but with shakier levels of quality. For many, translation of any quality will always be preferable to zero translation. The question for SDL, other commercial MT suppliers, and post-editing LSPs will be how much of this content goes through their portals.