The most exciting thing about visual search is that it's becoming a highly accessible way for users to interpret the real world, in real time, as they see it. Rather than being a passive observer, camera phones are now a primary resource for knowledge and understanding in daily life. Users are searching with their own, unique photos to discover content. Though SEOs have little control over which photos people take, we can optimize our brand presentation to ensure we are easily discoverable by visual search tools. By prioritizing the presence of high impact visual search elements and coordinating online SEO with offline branding, businesses of all sizes can see results.

And, today, we can do online searches simply by speaking to Alexa or Google Assistant. So, if you want to stay ahead of the competition, voice search optimization is something you must consider in your marketing strategies. Voice search optimization is crucial for any business because the possibilities of capturing local customers are endless. There are a few more things to keep in mind that can boost the number of times you appear on voice search results. The more data Google and other search engines have about your business, your website and your target audience, the more likely you are to appear on the Knowledge Graph Panel.

It's no secret that marketers and search optimization pros can be a bit obsessed with Google's zoo of algorithm updates. We've weathered the Penguins and Pandas that thankfully stopped old hacks from working these past couple of years; but what about what the next algo Zebra or Walrus, or maybe that bigger Elephant in the room affecting search traffic – personalization? This one's even bigger and less cuddly than the two that came before. The thing we should really be aware of, however, is the fact that Google's latest animal doesn't have a name because personalization is invisible. Still, I'm going to call it Elephant because that just feels right to me.

Sequence-to-Sequence (seq2seq) modeling has rapidly become an important general-purpose NLP tool that has proven effective for many text-generation and sequence-labeling tasks. Seq2seq builds on deep neural language modeling and inherits its remarkable accuracy in estimating local, next-word distributions. In this work, we introduce a model and beam-search training scheme, based on the work of Daume III and Marcu (2005), that extends seq2seq to learn global sequence scores. This structured approach avoids classical biases associated with local training and unifies the training loss with the test-time usage, while preserving the proven model architecture of seq2seq and its efficient training approach. We show that our system outperforms a highly-optimized attention-based seq2seq system and other baselines on three different sequence to sequence tasks: word ordering, parsing, and machine translation.