Bing推出AI智能搜索：机器阅读理解与多源验证重塑搜索体验

Today, Microsoft announced a significant evolution of its search engine with new Intelligent Search features for Bing. Powered by advanced artificial intelligence, these updates are designed to deliver faster answers, provide more comprehensive information, and enable more natural interactions with the search engine. This marks a step forward in making search a more intuitive and efficient tool for knowledge discovery.

今天，微软宣布了其搜索引擎 Bing 的重大进化，推出了全新的智能搜索功能。这些由先进人工智能驱动的更新，旨在提供更快的答案、更全面的信息，并实现与搜索引擎更自然的交互。这标志着在使搜索成为更直观、更高效的知识发现工具方面迈出了重要一步。

Intelligent Answers: Confidence Through AI Comprehension

At the core of the update are Intelligent Answers. This feature leverages state-of-the-art machine reading comprehension technology, powered by Project Brainwave running on Intel FPGAs. This system can read and analyze billions of web documents to understand content at scale, helping users find the answers they need more quickly and with greater confidence.

本次更新的核心是智能答案功能。该功能利用了最先进的机器阅读理解一种AI技术，使计算机能够理解和分析文本内容，提取关键信息并回答相关问题。技术，由运行在英特尔FPGA上的Project Brainwave提供支持。该系统能够阅读和分析数十亿的网络文档，以实现大规模的内容理解，帮助用户更快速、更自信地找到所需答案。

Multi-Source Validation and Aggregation

Bing now employs deep neural networks to validate answers by aggregating information across multiple reputable sources, rather than relying on a single page. This approach increases the reliability of the information presented. Furthermore, to save users from the tedious task of clicking through multiple links, Bing now synthesizes content from various sources into a single, comprehensive answer directly on the search results page.

Bing现在采用深度神经网络一种模仿人脑神经元结构的机器学习模型，用于复杂模式识别和数据处理。，通过聚合多个权威来源的信息来验证答案，而不是依赖单一页面。这种方法提高了所呈现信息的可靠性。此外，为了让用户免于点击多个链接的繁琐操作，Bing现在将来自不同来源的内容整合成一个单一的、全面的答案，直接显示在搜索结果页面上。

Presenting Diverse Perspectives and Comparisons

Recognizing that not every question has a single definitive answer, Bing intelligently surfaces different authoritative perspectives on a topic—such as benefits versus drawbacks—aggregating these viewpoints from reputable sources at the top of the page. For queries with multiple valid answers, users are presented with a carousel of intelligent answers. Bing is also expanding its comparison answers beyond products, using machine reading comprehension to generate easy-to-read tables that highlight key differences between two entities or topics, saving users from manually sifting through dense documents.

认识到并非每个问题都有唯一的明确答案，Bing会智能地呈现关于某个主题的不同权威观点——例如优点与缺点——并将这些来自可靠来源的观点聚合在页面顶部。对于有多个有效答案的查询，用户会看到一个智能答案轮播。Bing还在将其比较答案功能扩展到产品之外，利用机器阅读理解一种AI技术，使计算机能够理解和分析文本内容，提取关键信息并回答相关问题。生成易于阅读的表格，突出显示两个实体或主题之间的关键差异，使用户无需手动筛选繁杂的文档。

Making Data Relatable

Bing utilizes technology from Microsoft Research to help contextualize numerical data encountered online. It translates raw statistics into simple, relatable concepts. For example, it can express the population of another country in terms familiar to the user, making complex data easier to grasp at a glance.

Bing利用来自微软研究院的技术，帮助理解在线遇到的数字数据。它将原始统计数据转化为简单、易于理解的概念。例如，它可以用用户熟悉的方式表达另一个国家的人口数量，使复杂的数据一目了然。

Many of these Intelligent Answer features are available now, with a full rollout to US users expected within the next week, followed by expansion to other markets.

许多智能答案功能现已可用，预计将在下周内全面向美国用户推出，随后将扩展到其他市场。

Integrating Collective Wisdom: Bing's Partnership with Reddit

A key aspect of Intelligent Search is synthesizing diverse knowledge sources, including the "wisdom of the crowd." To this end, Bing is launching a new partnership with Reddit, an online community with 330 million monthly active users generating 2.8 million comments daily. This integration will bring Reddit's unique conversational insights directly into Bing search results through three initial experiences:

智能搜索的一个关键方面是综合不同的知识来源，包括"群体的智慧"。为此，Bing正在与Reddit启动一项新的合作。Reddit是一个拥有3.3亿月活跃用户、每日产生280万条评论的在线社区。此次整合将通过三种初始体验，将Reddit独特的对话见解直接引入Bing搜索结果：

Direct Reddit Topic Searches: When searching for a specific Reddit topic or subreddit (e.g., "Reddit Aww"), Bing will surface a preview with the day's top conversations.
- 直接Reddit主题搜索：当搜索特定的Reddit主题或子版块（例如"Reddit Aww"）时，Bing将显示包含当日热门对话的预览。
Contextual Reddit Insights: For general topics best answered by community discussion, Bing will surface relevant Reddit conversation snippets at the top of the search page.
- 情境化Reddit见解：对于最适合通过社区讨论来回答的通用主题，Bing将在搜索页面顶部显示相关的Reddit对话片段。
Discovering Reddit AMAReddit平台的“问我任何事”活动，名人和普通人通过问答形式与社区互动。s: Bing becomes a hub for discovering Reddit's "Ask Me Anything" (AMA) sessions. Users can search a person's name to see an AMA snapshot or search "Reddit AMAReddit平台的“问我任何事”活动，名人和普通人通过问答形式与社区互动。s" for a carousel of popular sessions.
- 发现Reddit AMAReddit平台的“问我任何事”活动，名人和普通人通过问答形式与社区互动。：Bing成为发现Reddit"问我任何问题"（AMA）会话的中心。用户可以搜索人名查看AMA快照，或搜索"Reddit AMAReddit平台的“问我任何事”活动，名人和普通人通过问答形式与社区互动。s"查看热门AMA的轮播。

Towards a More Conversational Search Experience

A common search pain point is formulating the right query. Bing is addressing this by combining web-scale data mining with conversational AI to create a more interactive search experience. This new capability builds on previous searches and can proactively ask clarifying questions based on a user's initial query to refine the search and deliver the best possible answer on the first try. This experience will debut in health, technology, and sports queries, with more topics to be added over time. As a machine learning-based system, its performance will improve with increased user engagement.

搜索的一个常见难点是构建正确的查询。Bing通过将网络规模的数据挖掘与对话式AI能够进行自然语言对话的人工智能系统，理解上下文并做出相应回应。相结合，创造一种更具交互性的搜索体验来解决这个问题。这项新功能建立在之前的搜索基础上，可以根据用户的初始查询主动提出澄清性问题，以优化搜索并在第一次尝试时就提供最佳答案。该体验将首先在健康、技术和体育类查询中推出，未来会添加更多主题。作为一个基于机器学习的系统，其性能将随着用户参与度的提高而不断改善。

Intelligent Image Search: Beyond Visual Discovery

Bing is also enhancing its image search with advanced computer vision and object recognition. Users can now search within images to shop for fashion or home furniture. Bing can detect and highlight products in images, or users can click a magnifying glass icon on any image to search within it for related items. A previewed new feature allows users to explore landmarks: by using an image from Bing or their camera roll, Bing can identify the landmark and provide relevant trivia, such as its origins or construction materials (e.g., explaining why the India Gate was built and the type of stone used).

Bing还通过先进的计算机视觉AI field that gives machines the ability to 'see' and interpret visual information from the world.和物体识别技术增强了其图像搜索功能。用户现在可以在图片内进行搜索以购买时尚或家居家具。Bing可以检测并高亮显示图片中的产品，或者用户可以点击任何图片上的放大镜图标，在图片内搜索相关物品。一项预览的新功能允许用户探索地标：通过使用来自Bing或自己相册的图片，Bing可以识别地标并提供相关的趣闻，例如其起源或建筑材料（例如，解释印度门为何建造以及所使用的石材类型）。

Microsoft is committed to continuing the development of Intelligent Search features aimed at saving users time and effort. The Bing team encourages users to try the new features and stay tuned for future updates.

微软致力于继续开发旨在为用户节省时间和精力的智能搜索功能。Bing团队鼓励用户尝试新功能，并关注未来的更新。