In this age of information explosion, we are often caught in a sea of data, with no way to get started. Various data and documents are scattered all over the place, and it often takes a lot of time and effort to find specific information. Now, with deepin global search access to AI, we are able to quickly and accurately locate the information we need in the massive amount of data, thus greatly improving the efficiency of information retrieval.

The global search after accessing AI has two core highlight functions: natural language search and text search within images, which can provide users with a more convenient, accurate and practical search experience.

The natural language search function can help users enter keywords in natural language form, so as to quickly search the required content; while the text search function within the image is an innovative way of searching, which allows users to look for keywords or phrases in the image without having to manually enter textual information, providing users with a more convenient, accurate and practical search experience.

Natural Language Search

Functional FeaturesSmart Global Search supports powerful natural language search function, which means you can describe the information you want to retrieve in natural language without using specific keywords.Usage Scenarios

  • Fuzzy Search: When you can't remember the exact search keywords and only know a fuzzy description of what you need to find, you can input the fuzzy description to Smart Global Search and it will help you find the answer.
  • Question Answer: You can use natural language to describe a relevant question covered in a local document, and Intelligent Global Search will find the document containing the relevant answer.

In-image text search

FeaturesSmart Global Search recognizes the text in an image and converts it into searchable text. This means you can search for the text in the image directly to find that image.This feature is not only suitable for study and work, but can also help you quickly search for attraction information, recipes and more while traveling.

Usage Scenarios

  • Study and research: You can take a picture of the text in a textbook or paper, archive it, and then quickly search for related information through the intelligent global search to improve your study efficiency.
  • Traveling and exploring: You take a lot of photos during your travels, and you can quickly find the pictures of the attractions you want by searching the text on the attraction's introduction board.
  • Efficient office work: you can search the content in the screenshot to quickly find the information of the screenshot; you can search the information on the ID card and scanned pictures to quickly locate the picture files you need.

How is the local model implemented?

All the above AI capabilities are realized based on local models. Based on internationally recognized open source pre-trained models, we developed local models built into the global search, which are smaller in size, take up less arithmetic power, and have more stable performance.For the natural language search capability, it is realized by the built-in AI model of natural language parsing, which can better parse natural language, understand the user's search intent, and then find the file that best meets the user's intent in the file manager.In the future, we will further improve the accuracy of user intent prediction and enhance the performance of search. At the same time, we will also expand the scope of search (e.g., the search is not limited to the title of the file, but also the content of the file can be searched), to achieve the goal of "everything can be searched".

For image text search, it is realized by the built-in AI model of OCR recognition, which can better recognize the text information in the image, so that the global search can search for the corresponding information.

We will continue to optimize the recognition accuracy and add more recognition languages. At the same time, we also plan to introduce more image recognition models, such as: image content recognition. At that time, users can directly describe the picture content of the picture, and the corresponding picture can be searched, such as: help me find the picture of my kitten.

Meanwhile, the intelligent global search application has been open-sourced, and we welcome you to build it together and create more useful functions.

Address：https://github.com/linuxdeepin/dde-grand-search

News source: deepin community

Reprinted with attribution

Leave a Reply