Search your document like you search the web

(Originally posted May 19, 2020 By ~ Derek Stenerson and Maria River)

Hi Insiders! We’re Derik Stenerson from the Word team and Maria Kang from the Bing Search & AI team. We’re excited to give you a sneak peek at how Microsoft Search can make searching within your documents easier and more natural. We’ll show you some great examples of how we’ve improved the Find feature in Word. Microsoft AI at scale Powering the next generation of AI experiences. In this case, Office.

How many times have you searched for an exact word and found yourself searching for it? We know from your feedback that this happens often and that you expect more. We often hear, “Why can’t it work like searching on the web? Why can’t I just ask a question and not worry about spelling errors?”

We’re excited to share with you an insider’s look at how we’re addressing this feedback, starting with Word for the web and soon rolling out to all platforms.

Microsoft Search: Search like the web

We are leveraging well-established web search techniques like query and document understanding, and adding deep learning-based natural language models. This allows us to handle a much wider range of search queries beyond “exact matches.”

Search within documents using imprecise queries

• typo: Search can now show relevant matches when your query contains spelling errors. For example, technincian vs. technician.
• word form: When the form of a word is different in the document and query. For example: tech, technology, technologies”; or USA, USA, United States, United States of America; or newborn, new born, new-born, etc.
• synonym: For example: quote, citation, citation, reference are all possible candidates for the term you are searching for within the document.
• Queries consisting of multiple words: One word can result in too many search results. Modern semantic search can often yield better results. For example, if the query is oil price, content with oil price, oil price, oil cost can provide relevant matches within the document content.

Get answers with Q&A

Recent breakthroughs in deep learning technology have made it possible to go beyond standard query-based queries. The result is answers to questions based on the content of documents. This opens up a whole new way of finding knowledge. When looking at a water quality report, you can answer questions like, “Where does my city’s water come from? How can I reduce the amount of lead in my water?”

Ask a question to find information in the documentation

An insider’s view on technology

All this is possible using: Microsoft Turing Model for Natural Language RepresentationA large-scale AI model that brings the power of deep learning to your nearby search bar.

Microsoft developed two deep learning models that use public Bing web data to power the SmartFind feature.

• Encoder model: This feature outputs a vector that captures the semantic meaning of the text, so that semantically similar texts have similar numeric values, allowing for a better understanding of the language.
• Question Answering Model: Built using a state-of-the-art large-scale pre-trained Turing language model, the same model used for question answering and captioning in Bing Search.

effectiveness

These updates are being distributed as Word on the web to organizations and users who have signed up for the Targeted release in English. They will be available in Production soon.

We usually release features over a period of time to make sure everything is working smoothly. This also applies to Insiders. We release slowly to a larger number of Insiders, so we highlight features that you may not have. Sometimes we remove elements based on your feedback to make it even better. Although this is rare, we also reserve the option to remove features from the product entirely.

As an insider, you get a chance to try them out.

feedback

We want to hear from you! Go to: help > feedback To submit feedback on this feature, please contact Word.

Learn about Microsoft 365 Insider Program And join us Microsoft 365 Insider Newsletter Get the latest Insider features straight to your inbox once a month!