Earlier this year Microsoft’s rolled out large-scale AI language processing across all of its apps and products. But now they have revealed key improvements to Bing search features by offering natural language models and real-time suggestions.
Here are some of these changes you can expect:
The aim of autosuggest technology is to improve the user’s search experience. It allows them to type less and find information faster. Autosuggest also reduces the chance of miss-spelling and enable users to find information that they may have not known the correct term for.
Previous versions of Bing’s autosuggest relied on users’ search history and were confined to completing a word that was already being typed. But with the new Microsoft Turing-NLG, the Bing AI is now able to generate autosuggest whole words and phrases in real-time using intelligent predictions.
This makes the autosuggest completions far more diverse, as they are no longer dependant on previous search history– a big improvement to user experience.
Natural language generation
Whilst computers are efficient in organising and processing data at speed, they lack the human capacity for judgement, creativity or empathy. Training AI to understand and generate natural language – the way in which humans talk, with all the flaws and imperfections – requires the AI to have a far more extensive and diverse understanding of human’s use of language.
Similar to Google’s BERT or OpenAI’s GPT-2, Microsoft’s AI has been trained on a dataset of millions of webpages. So it is able to independently predict what the next most likely word will be in a partially complete sentence, based on the initial words that they have typed.
People Also Ask (PAA)
When it comes to PAA, again Bing’s upgraded AI is no longer constrained by previous search history in order to generate question-answer pairs. The PPA feature allows users to expand their search by viewing other example questions that are similar to their original query. Rather than simply scrolling through the results pages, they can continually alter their question to bring up more exact results. As well as saving time, this can improve the search experience as users explore alternative avenues and broaden their search.
In the past, Bing’s PAA has been served by its extensive bank of previous searches to generate these alternate questions. However, they are now able to create alternate question-answer pairs from scratch, in case a user asks a question that has never been asked before.
In the past Bing’s natural language and intelligent answers have only been available in English. However, in recent years they have worked to make their search results more inclusive. So, having added several languages last year – including French and German – in this latest addition Microsoft have expanded intelligent answers into over 100 languages across 200 different regions.
Not only do the intelligent answers function in different languages, but they are also sensitive to the different markets they are operating in. This means the AI is trained in the nuances of culture and dialect between different regions, ensuring the generated answers are better suited to the user.
The final improvement in the new Bing AI is the introduction of semantic highlighting. This feature highlights captions within search result text, to enable the user to locate information faster as they browse through the results. Previous versions of Bing’s captions highlighting simply matched the keywords in the search query. But semantic search is a far more sophisticated AI process, understanding the question and identifying the likely answer within the text.
Benefits of Bing applying AI for autosuggest
Search engines are increasingly developing and improving the natural language sensitivities of their AI, and Microsoft is by no means alone in bringing intelligent predictive capabilities to its search engine. But there is no doubt these new AI features will greatly improve Bing’s user experience. The increased diversity and flexibility of the autosuggest will enable faster, more effective searches, with users quickly accessing the results they want. And users from across the world will be able to access the AI functions in their own language for the first time.