Let AI truly become your productivity super assistant
The era of AI is coming, we need to actively embrace AI tools
In the past two months, AI represented by ChatGPT has become popular all over the world. With the continuous optimization of the GPT model, ChatGPT has shown a level comparable to experts in many fields. At present, many experts regard ChatGPT as a representative AI wave and the beginning of a new industrial revolution . In the face of ChatGPT, some people are ecstatic, while others are anxious. Regardless of our attitudes, the age of AI has arrived . Whether AI can replace humans may be too far away. Those who make good use of AI tools will increase their efficiency tenfold and hundreds of times, replacing those who do not use or are not good at AI tools.
Go beyond ChatGPT and let AI process our proprietary data
Taking ChatGPT as an example, OpenAI officially provides more than 40 usage scenarios. Every day people are trying to explore the possibilities and boundaries of AI usage . At the same time, ChatGPT is not omniscient . Take the 3.0 behind ChatGPT as an example, and its training data ends in 2021. If the training model is not updated, then ChatGPT does not have the ability to create new knowledge . In addition, for unfamiliar fields, ChatGPT may make seemingly logical and serious nonsense based on the characteristics of large-scale language models .
So, how to solve this problem?
New Bing , represented by Microsoft , provides solutions, integrates the advantages of search engines and AI models , provides more accurate answers to users' questions, and provides citation information related to the answers . This solution, to a large extent, is that New Bing feeds new data to GPT in real time, and then generates the results required by users.
If the data we want to process cannot be crawled by search engines or is difficult to process, the results of New Bing will be greatly reduced. Another way of thinking, if we as users use our own knowledge base to actively feed our own exclusive data to AI , then AI will become a super assistant to improve our productivity. The product based on this idea is ChatDOC.
What is ChatDOC?
ChatDOC is an AI reading aid based on ChatGPT, which allows ChatGPT to communicate with the document specified by the user and process the user's exclusive data.
Access address : https://chatdoc.com/
Features of ChatDOC
Simple operation: one-click upload, get started quickly
Upload documents with one click, and quickly let AI process document data. Dig deep into text structure and content through conversational learning with an AI assistant.
Intelligent text analysis: accurate and efficient text and form data processing
- Multi-scene reading aid . Based on the powerful capabilities of GPT, the summary, abstract, translation, knowledge point query and other needs of document content can be quickly completed with the assistance of AI.
- Support query function . It allows users to open a new sub-window for multiple rounds of questions and answers for a specific question and answer like Twitter Thread .
- Powerful text parsing capabilities . In ChatDOC, in addition to the powerful text analysis capabilities of GPT, and based on the rich technical accumulation of the development team in the field of content recognition , ChatDOC is still accurate and efficient in processing data content such as tables .
Content backtracking function: quickly view the original text based on the Q&A results
Different from ChatGPT's answers that are in a black box and cannot be traced back to the source, the characteristic of ChatDOC is that the results generated by AI are based on the proprietary data provided by the user, and there is evidence to follow . Below each answer, the AI provides a citation page related to that answer. Click to jump directly to the specific location of the page to view the specific expression of the original text.
Guarantee user data security: encrypted storage, data autonomy
All uploaded data is stored in the cloud in encrypted form. Users have full ownership of their own data and can download and delete any files at any time.
Use of ChaDOC
ChatDOC is very simple to use.
- After the user logs in, upload the papers, books or other files you need to process with one click .
- ChatDOC will soon conduct intelligent analysis of document content based on GPT ;
- Afterwards, the user can start a dialogue and exchange on the content of the uploaded document in the form of chat .
upload files
As shown in the figure below, click or drag to upload local files. Then, after intelligently parsing the literature, the system automatically generates example questions based on the literature .
way of asking questions
- Ask a question about the full text : Type your question directly into the dialog. Or, click on the default sample questions to start your AI-assisted reading journey.
- Ask questions about specific content . Select the content, click the bullseye icon to confirm the selection. Then, ask questions.
Applicable scene
Through AI-assisted document reading, ChatDOC can help users quickly analyze documents , understand content, inspire inspiration and expand horizons . It is suitable for office workers, educational user groups, and various groups interested in knowledge management and information management.
The demonstration documents in this article are not in the field that I am familiar with. With AI assistance, I can quickly obtain article summaries, dig into specific viewpoints and data, translate texts, summarize content ...help me quickly understand knowledge and information in a field that I am completely unfamiliar with.
For academic papers, the common structure is: introduction, literature review, theoretical framework, innovation points, research methods, research process, main viewpoints and conclusions, and further research directions .
If you are not proficient in English reading, it usually takes 15-30 minutes to complete the reading of the above structure. Now, with ChatDOC's AI assistance, I can quickly finish reading an article within 1 minute.
Prompt words : Please tell me the introduction, literature review, theoretical framework, innovations, research methods, research process, main viewpoints and conclusions, and further research directions of this article in turn.
The answer is as follows:
In order to better organize the reading notes in the future, I decided to refine the questions .
In various forms of notes, the outline list can facilitate users to quickly view the structure of notes and view the details of notes. In addition, the outline list can be quickly switched to a mind map with one click. Based on the above assumptions. My new question is as follows:
Note : The format description in the second half of the question is to prevent AI from sometimes not strictly following the format to generate content.
The generated result is shown in the figure:
Then, I copied and pasted the above content to the FlowUs page of the note-taking software I am using , and placed it under the node of the title SelfDoc of this article. Click on a content block to convert it into a mind map .
Content Block Transformation
The effect of converting the outline list into a mind map with one click is as follows:
Of course, if it is not necessary to subdivide the content into secondary sub-node content for splitting, it is also possible. As shown in the picture:
If you prefer, you can also directly let the AI output the results in a tabular form.
The above are mainly questions based on the full text. ChatDOC also allows users to interact with AI on specific content in documents. For example, let AI interpret tables, explain formulas, and explain concepts in the text.
As shown in the figure below, we can see that ChaDOC is very effective in selecting and analyzing tables .
The figure below shows the results of questions about formulas and specific concepts .
It should be noted that ChatDOC also supports the questioning function, which can satisfy users' deep mining of knowledge .
Click the two collapsed dialog buttons in the lower right corner of each answer to enter the Thread window. Users can conduct multiple rounds of in-depth questions and answers on a topic .
As shown in the figure, for the existing answers, I first asked "what is a self-supervised learning framework", and then in the new answer, I asked "what is cross-modal learning?"
ChatDOC's Pricing Mechanism
Registering a free ChatDOC account is currently supported. Among them, in the free version, you can upload up to 5 files within 24 hours, each file not exceeding 200 pages. Additionally, the free version allows up to 300 questions to be asked in the same 24-hour period.
In the future, ChatDOC should add paid functions and provide some value-added services.
Future development of ChatDOC
Although Chatdoc is currently very helpful for reading assistance, there is still a lot of room for optimization. The following is ChatDOC's official product planning roadmap.
- Supports chat conversation queries based on multiple files .
- Support more file types , such as scanned documents or pictures.
- Optimize citation results .
User advice for ChatDOC
- Support more content sources . For example, it supports content reading based on web link URL.
- One-click export of all conversations and generate reading notes.
- Support the traditional way of reading . AI-based chat dialogue is undoubtedly an efficient way of reading. However, many users still prefer the traditional reading method of reading and annotating page by page. Artificial active reading and AI-assisted reading are not mutually exclusive, but complement each other . It is recommended to allow users to perform operations such as highlighting and commenting on document content .
- Based on the user's highlighted content and annotation notes, relevant AI-assisted functions are provided . For example, generate question and answer cards based on highlights and notes . For example, I selected or highlighted several contents, AI generated questions for these contents , provided corresponding interpretation and analysis, and finally generated a knowledge card including questions, highlighted content, annotation notes, and AI interpretation.
- other directions. Provide stronger note-taking functions, provide stronger literature data mining capabilities , etc. For example, if you read a paper, the generated content may have multiple arguments. As a user, I would like arguments to be automatically bolded.
- For different types of documents, more preset Prompts are provided .
The above are some experiences of using ChatDOC. In the use cases, only part of the functions of ChatDOC are shown. More AI usage skills require users to explore according to their own needs.